| blob | 8907578cd33b730a3f3429ff02fbb683805c3779 |
1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
5 declare <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8, <16 x i8>, i16)
7 define <16 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_128(i8 %x0, <16 x i8> %x1, i16 %mask) {
8 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
9 ; X86: # %bb.0:
10 ; X86-NEXT: vpbroadcastb {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0x4c,0x24,0x04]
11 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
12 ; X86-NEXT: vmovdqu8 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0xc1]
13 ; X86-NEXT: vmovdqu8 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0xd1]
14 ; X86-NEXT: vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
15 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
16 ; X86-NEXT: retl # encoding: [0xc3]
17 ;
18 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
19 ; X64: # %bb.0:
20 ; X64-NEXT: vpbroadcastb %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xcf]
21 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
22 ; X64-NEXT: vpbroadcastb %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc7]
23 ; X64-NEXT: vpbroadcastb %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd7]
24 ; X64-NEXT: vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
25 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
26 ; X64-NEXT: retq # encoding: [0xc3]
27 %res = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 -1)
28 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 %mask)
29 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> zeroinitializer, i16 %mask)
30 %res3 = add <16 x i8> %res, %res1
31 %res4 = add <16 x i8> %res2, %res3
32 ret <16 x i8> %res4
33 }
36 declare <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16, <8 x i16>, i8)
38 define <8 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_128(i16 %x0, <8 x i16> %x1, i8 %mask) {
39 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
40 ; X86: # %bb.0:
41 ; X86-NEXT: vpbroadcastw {{[0-9]+}}(%esp), %xmm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0x4c,0x24,0x04]
42 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
43 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
44 ; X86-NEXT: vmovdqu16 %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0xc1]
45 ; X86-NEXT: vmovdqu16 %xmm1, %xmm2 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0xd1]
46 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
47 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
48 ; X86-NEXT: retl # encoding: [0xc3]
49 ;
50 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
51 ; X64: # %bb.0:
52 ; X64-NEXT: vpbroadcastw %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xcf]
53 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
54 ; X64-NEXT: vpbroadcastw %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc7]
55 ; X64-NEXT: vpbroadcastw %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd7]
56 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
57 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
58 ; X64-NEXT: retq # encoding: [0xc3]
59 %res = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 -1)
60 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 %mask)
61 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> zeroinitializer, i8 %mask)
62 %res3 = add <8 x i16> %res, %res1
63 %res4 = add <8 x i16> %res2, %res3
64 ret <8 x i16> %res4
65 }
68 declare <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8, <32 x i8>, i32)
70 define <32 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_256(i8 %x0, <32 x i8> %x1, i32 %mask) {
71 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
72 ; X86: # %bb.0:
73 ; X86-NEXT: vpbroadcastb {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0x4c,0x24,0x04]
74 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
75 ; X86-NEXT: vmovdqu8 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0xc1]
76 ; X86-NEXT: vmovdqu8 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0xd1]
77 ; X86-NEXT: vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
78 ; X86-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
79 ; X86-NEXT: retl # encoding: [0xc3]
80 ;
81 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
82 ; X64: # %bb.0:
83 ; X64-NEXT: vpbroadcastb %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xcf]
84 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
85 ; X64-NEXT: vpbroadcastb %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc7]
86 ; X64-NEXT: vpbroadcastb %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd7]
87 ; X64-NEXT: vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
88 ; X64-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
89 ; X64-NEXT: retq # encoding: [0xc3]
90 %res = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 -1)
91 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 %mask)
92 %res2 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> zeroinitializer, i32 %mask)
93 %res3 = add <32 x i8> %res, %res1
94 %res4 = add <32 x i8> %res2, %res3
95 ret <32 x i8> %res4
96 }
100 declare <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16, <16 x i16>, i16)
102 define <16 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_256(i16 %x0, <16 x i16> %x1, i16 %mask) {
103 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
104 ; X86: # %bb.0:
105 ; X86-NEXT: vpbroadcastw {{[0-9]+}}(%esp), %ymm1 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0x4c,0x24,0x04]
106 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
107 ; X86-NEXT: vmovdqu16 %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0xc1]
108 ; X86-NEXT: vmovdqu16 %ymm1, %ymm2 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0xd1]
109 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
110 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
111 ; X86-NEXT: retl # encoding: [0xc3]
112 ;
113 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
114 ; X64: # %bb.0:
115 ; X64-NEXT: vpbroadcastw %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xcf]
116 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
117 ; X64-NEXT: vpbroadcastw %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc7]
118 ; X64-NEXT: vpbroadcastw %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd7]
119 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
120 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
121 ; X64-NEXT: retq # encoding: [0xc3]
122 %res = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 -1)
123 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 %mask)
124 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> zeroinitializer, i16 %mask)
125 %res3 = add <16 x i16> %res, %res1
126 %res4 = add <16 x i16> %res2, %res3
127 ret <16 x i16> %res4
128 }
130 declare <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8>, <32 x i8>, i32)
132 define <32 x i8>@test_int_x86_avx512_pbroadcastb_256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask) {
133 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_256:
134 ; X86: # %bb.0:
135 ; X86-NEXT: vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
136 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
137 ; X86-NEXT: vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
138 ; X86-NEXT: vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
139 ; X86-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
140 ; X86-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
141 ; X86-NEXT: retl # encoding: [0xc3]
142 ;
143 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_256:
144 ; X64: # %bb.0:
145 ; X64-NEXT: vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
146 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
147 ; X64-NEXT: vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
148 ; X64-NEXT: vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
149 ; X64-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
150 ; X64-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
151 ; X64-NEXT: retq # encoding: [0xc3]
152 %res = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 -1)
153 %res1 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask)
154 %res2 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> zeroinitializer, i32 %mask)
155 %res3 = add <32 x i8> %res, %res1
156 %res4 = add <32 x i8> %res2, %res3
157 ret <32 x i8> %res4
158 }
160 declare <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8>, <16 x i8>, i16)
162 define <16 x i8>@test_int_x86_avx512_pbroadcastb_128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask) {
163 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_128:
164 ; X86: # %bb.0:
165 ; X86-NEXT: vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
166 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
167 ; X86-NEXT: vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
168 ; X86-NEXT: vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
169 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
170 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
171 ; X86-NEXT: retl # encoding: [0xc3]
172 ;
173 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_128:
174 ; X64: # %bb.0:
175 ; X64-NEXT: vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
176 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
177 ; X64-NEXT: vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
178 ; X64-NEXT: vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
179 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
180 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
181 ; X64-NEXT: retq # encoding: [0xc3]
182 %res = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
183 %res1 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask)
184 %res2 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> zeroinitializer, i16 %mask)
185 %res3 = add <16 x i8> %res, %res1
186 %res4 = add <16 x i8> %res2, %res3
187 ret <16 x i8> %res4
188 }
190 declare <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16>, <16 x i16>, i16)
192 define <16 x i16>@test_int_x86_avx512_pbroadcastw_256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask) {
193 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_256:
194 ; X86: # %bb.0:
195 ; X86-NEXT: vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
196 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
197 ; X86-NEXT: vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
198 ; X86-NEXT: vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
199 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
200 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
201 ; X86-NEXT: retl # encoding: [0xc3]
202 ;
203 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_256:
204 ; X64: # %bb.0:
205 ; X64-NEXT: vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
206 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
207 ; X64-NEXT: vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
208 ; X64-NEXT: vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
209 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
210 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
211 ; X64-NEXT: retq # encoding: [0xc3]
212 %res = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 -1)
213 %res1 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask)
214 %res2 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> zeroinitializer, i16 %mask)
215 %res3 = add <16 x i16> %res, %res1
216 %res4 = add <16 x i16> %res2, %res3
217 ret <16 x i16> %res4
218 }
220 declare <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16>, <8 x i16>, i8)
222 define <8 x i16>@test_int_x86_avx512_pbroadcastw_128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask) {
223 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_128:
224 ; X86: # %bb.0:
225 ; X86-NEXT: vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
226 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
227 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
228 ; X86-NEXT: vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
229 ; X86-NEXT: vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
230 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
231 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
232 ; X86-NEXT: retl # encoding: [0xc3]
233 ;
234 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_128:
235 ; X64: # %bb.0:
236 ; X64-NEXT: vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
237 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
238 ; X64-NEXT: vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
239 ; X64-NEXT: vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
240 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
241 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
242 ; X64-NEXT: retq # encoding: [0xc3]
243 %res = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
244 %res1 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask)
245 %res2 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> zeroinitializer, i8 %mask)
246 %res3 = add <8 x i16> %res, %res1
247 %res4 = add <8 x i16> %res2, %res3
248 ret <8 x i16> %res4
249 }
251 declare <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8>, <64 x i8>, i64)
253 define <64 x i8>@test_int_x86_avx512_pbroadcastb_512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask) {
254 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_512:
255 ; X86: # %bb.0:
256 ; X86-NEXT: vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
257 ; X86-NEXT: kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x04]
258 ; X86-NEXT: vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
259 ; X86-NEXT: vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
260 ; X86-NEXT: vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
261 ; X86-NEXT: vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
262 ; X86-NEXT: retl # encoding: [0xc3]
263 ;
264 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_512:
265 ; X64: # %bb.0:
266 ; X64-NEXT: vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
267 ; X64-NEXT: kmovq %rdi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xcf]
268 ; X64-NEXT: vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
269 ; X64-NEXT: vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
270 ; X64-NEXT: vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
271 ; X64-NEXT: vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
272 ; X64-NEXT: retq # encoding: [0xc3]
273 %res = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 -1)
274 %res1 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask)
275 %res2 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> zeroinitializer, i64 %mask)
276 %res3 = add <64 x i8> %res, %res1
277 %res4 = add <64 x i8> %res2, %res3
278 ret <64 x i8> %res4
279 }
281 declare <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16>, <32 x i16>, i32)
283 define <32 x i16>@test_int_x86_avx512_pbroadcastw_512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask) {
284 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_512:
285 ; X86: # %bb.0:
286 ; X86-NEXT: vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
287 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
288 ; X86-NEXT: vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
289 ; X86-NEXT: vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
290 ; X86-NEXT: vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
291 ; X86-NEXT: vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
292 ; X86-NEXT: retl # encoding: [0xc3]
293 ;
294 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_512:
295 ; X64: # %bb.0:
296 ; X64-NEXT: vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
297 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
298 ; X64-NEXT: vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
299 ; X64-NEXT: vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
300 ; X64-NEXT: vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
301 ; X64-NEXT: vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
302 ; X64-NEXT: retq # encoding: [0xc3]
303 %res = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 -1)
304 %res1 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask)
305 %res2 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> zeroinitializer, i32 %mask)
306 %res3 = add <32 x i16> %res, %res1
307 %res4 = add <32 x i16> %res2, %res3
308 ret <32 x i16> %res4
309 }
311 declare void @llvm.x86.avx512.mask.storeu.b.128(i8*, <16 x i8>, i16)
313 define void@test_int_x86_avx512_mask_storeu_b_128(i8* %ptr1, i8* %ptr2, <16 x i8> %x1, i16 %x2) {
314 ; X86-LABEL: test_int_x86_avx512_mask_storeu_b_128:
315 ; X86: # %bb.0:
316 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
317 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
318 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
319 ; X86-NEXT: vmovdqu8 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x01]
320 ; X86-NEXT: vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
321 ; X86-NEXT: retl # encoding: [0xc3]
322 ;
323 ; X64-LABEL: test_int_x86_avx512_mask_storeu_b_128:
324 ; X64: # %bb.0:
325 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
326 ; X64-NEXT: vmovdqu8 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x07]
327 ; X64-NEXT: vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
328 ; X64-NEXT: retq # encoding: [0xc3]
329 call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr1, <16 x i8> %x1, i16 %x2)
330 call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr2, <16 x i8> %x1, i16 -1)
331 ret void
332 }
334 declare void @llvm.x86.avx512.mask.storeu.b.256(i8*, <32 x i8>, i32)
336 define void@test_int_x86_avx512_mask_storeu_b_256(i8* %ptr1, i8* %ptr2, <32 x i8> %x1, i32 %x2) {
337 ; X86-LABEL: test_int_x86_avx512_mask_storeu_b_256:
338 ; X86: # %bb.0:
339 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
340 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
341 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
342 ; X86-NEXT: vmovdqu8 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x01]
343 ; X86-NEXT: vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
344 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
345 ; X86-NEXT: retl # encoding: [0xc3]
346 ;
347 ; X64-LABEL: test_int_x86_avx512_mask_storeu_b_256:
348 ; X64: # %bb.0:
349 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
350 ; X64-NEXT: vmovdqu8 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x07]
351 ; X64-NEXT: vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
352 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
353 ; X64-NEXT: retq # encoding: [0xc3]
354 call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr1, <32 x i8> %x1, i32 %x2)
355 call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr2, <32 x i8> %x1, i32 -1)
356 ret void
357 }
359 declare void @llvm.x86.avx512.mask.storeu.w.128(i8*, <8 x i16>, i8)
361 define void@test_int_x86_avx512_mask_storeu_w_128(i8* %ptr1, i8* %ptr2, <8 x i16> %x1, i8 %x2) {
362 ; X86-LABEL: test_int_x86_avx512_mask_storeu_w_128:
363 ; X86: # %bb.0:
364 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
365 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
366 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
367 ; X86-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
368 ; X86-NEXT: vmovdqu16 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x01]
369 ; X86-NEXT: vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
370 ; X86-NEXT: retl # encoding: [0xc3]
371 ;
372 ; X64-LABEL: test_int_x86_avx512_mask_storeu_w_128:
373 ; X64: # %bb.0:
374 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
375 ; X64-NEXT: vmovdqu16 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x07]
376 ; X64-NEXT: vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
377 ; X64-NEXT: retq # encoding: [0xc3]
378 call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr1, <8 x i16> %x1, i8 %x2)
379 call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr2, <8 x i16> %x1, i8 -1)
380 ret void
381 }
383 declare void @llvm.x86.avx512.mask.storeu.w.256(i8*, <16 x i16>, i16)
385 define void@test_int_x86_avx512_mask_storeu_w_256(i8* %ptr1, i8* %ptr2, <16 x i16> %x1, i16 %x2) {
386 ; X86-LABEL: test_int_x86_avx512_mask_storeu_w_256:
387 ; X86: # %bb.0:
388 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
389 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
390 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
391 ; X86-NEXT: vmovdqu16 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x01]
392 ; X86-NEXT: vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
393 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
394 ; X86-NEXT: retl # encoding: [0xc3]
395 ;
396 ; X64-LABEL: test_int_x86_avx512_mask_storeu_w_256:
397 ; X64: # %bb.0:
398 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
399 ; X64-NEXT: vmovdqu16 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x07]
400 ; X64-NEXT: vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
401 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
402 ; X64-NEXT: retq # encoding: [0xc3]
403 call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr1, <16 x i16> %x1, i16 %x2)
404 call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr2, <16 x i16> %x1, i16 -1)
405 ret void
406 }
408 declare <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8*, <8 x i16>, i8)
410 define <8 x i16>@test_int_x86_avx512_mask_loadu_w_128(i8* %ptr, i8* %ptr2, <8 x i16> %x1, i8 %mask) {
411 ; X86-LABEL: test_int_x86_avx512_mask_loadu_w_128:
412 ; X86: # %bb.0:
413 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
414 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
415 ; X86-NEXT: vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
416 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
417 ; X86-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
418 ; X86-NEXT: vmovdqu16 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x00]
419 ; X86-NEXT: vmovdqu16 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x09]
420 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
421 ; X86-NEXT: retl # encoding: [0xc3]
422 ;
423 ; X64-LABEL: test_int_x86_avx512_mask_loadu_w_128:
424 ; X64: # %bb.0:
425 ; X64-NEXT: vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
426 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
427 ; X64-NEXT: vmovdqu16 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x06]
428 ; X64-NEXT: vmovdqu16 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x0f]
429 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
430 ; X64-NEXT: retq # encoding: [0xc3]
431 %res0 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> %x1, i8 -1)
432 %res = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr2, <8 x i16> %res0, i8 %mask)
433 %res1 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> zeroinitializer, i8 %mask)
434 %res2 = add <8 x i16> %res, %res1
435 ret <8 x i16> %res2
436 }
438 declare <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8*, <16 x i16>, i16)
440 define <16 x i16>@test_int_x86_avx512_mask_loadu_w_256(i8* %ptr, i8* %ptr2, <16 x i16> %x1, i16 %mask) {
441 ; X86-LABEL: test_int_x86_avx512_mask_loadu_w_256:
442 ; X86: # %bb.0:
443 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
444 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
445 ; X86-NEXT: vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
446 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
447 ; X86-NEXT: vmovdqu16 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x00]
448 ; X86-NEXT: vmovdqu16 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x09]
449 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
450 ; X86-NEXT: retl # encoding: [0xc3]
451 ;
452 ; X64-LABEL: test_int_x86_avx512_mask_loadu_w_256:
453 ; X64: # %bb.0:
454 ; X64-NEXT: vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
455 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
456 ; X64-NEXT: vmovdqu16 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x06]
457 ; X64-NEXT: vmovdqu16 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x0f]
458 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
459 ; X64-NEXT: retq # encoding: [0xc3]
460 %res0 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> %x1, i16 -1)
461 %res = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr2, <16 x i16> %res0, i16 %mask)
462 %res1 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> zeroinitializer, i16 %mask)
463 %res2 = add <16 x i16> %res, %res1
464 ret <16 x i16> %res2
465 }
467 declare <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8*, <16 x i8>, i16)
469 define <16 x i8>@test_int_x86_avx512_mask_loadu_b_128(i8* %ptr, i8* %ptr2, <16 x i8> %x1, i16 %mask) {
470 ; X86-LABEL: test_int_x86_avx512_mask_loadu_b_128:
471 ; X86: # %bb.0:
472 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
473 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
474 ; X86-NEXT: vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
475 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
476 ; X86-NEXT: vmovdqu8 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x00]
477 ; X86-NEXT: vmovdqu8 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x09]
478 ; X86-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
479 ; X86-NEXT: retl # encoding: [0xc3]
480 ;
481 ; X64-LABEL: test_int_x86_avx512_mask_loadu_b_128:
482 ; X64: # %bb.0:
483 ; X64-NEXT: vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
484 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
485 ; X64-NEXT: vmovdqu8 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x06]
486 ; X64-NEXT: vmovdqu8 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x0f]
487 ; X64-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
488 ; X64-NEXT: retq # encoding: [0xc3]
489 %res0 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> %x1, i16 -1)
490 %res = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr2, <16 x i8> %res0, i16 %mask)
491 %res1 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> zeroinitializer, i16 %mask)
492 %res2 = add <16 x i8> %res, %res1
493 ret <16 x i8> %res2
494 }
496 declare <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8*, <32 x i8>, i32)
498 define <32 x i8>@test_int_x86_avx512_mask_loadu_b_256(i8* %ptr, i8* %ptr2, <32 x i8> %x1, i32 %mask) {
499 ; X86-LABEL: test_int_x86_avx512_mask_loadu_b_256:
500 ; X86: # %bb.0:
501 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
502 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
503 ; X86-NEXT: vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
504 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
505 ; X86-NEXT: vmovdqu8 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x00]
506 ; X86-NEXT: vmovdqu8 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x09]
507 ; X86-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
508 ; X86-NEXT: retl # encoding: [0xc3]
509 ;
510 ; X64-LABEL: test_int_x86_avx512_mask_loadu_b_256:
511 ; X64: # %bb.0:
512 ; X64-NEXT: vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
513 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
514 ; X64-NEXT: vmovdqu8 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x06]
515 ; X64-NEXT: vmovdqu8 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x0f]
516 ; X64-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
517 ; X64-NEXT: retq # encoding: [0xc3]
518 %res0 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> %x1, i32 -1)
519 %res = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr2, <32 x i8> %res0, i32 %mask)
520 %res1 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> zeroinitializer, i32 %mask)
521 %res2 = add <32 x i8> %res, %res1
522 ret <32 x i8> %res2
523 }
525 declare <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8>, <16 x i8>, i32, <16 x i8>, i16)
527 define <16 x i8>@test_int_x86_avx512_mask_palignr_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x3, i16 %x4) {
528 ; X86-LABEL: test_int_x86_avx512_mask_palignr_128:
529 ; X86: # %bb.0:
530 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
531 ; X86-NEXT: # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
532 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
533 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
534 ; X86-NEXT: # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
535 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
536 ; X86-NEXT: # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
537 ; X86-NEXT: vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
538 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
539 ; X86-NEXT: retl # encoding: [0xc3]
540 ;
541 ; X64-LABEL: test_int_x86_avx512_mask_palignr_128:
542 ; X64: # %bb.0:
543 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
544 ; X64-NEXT: # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
545 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
546 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
547 ; X64-NEXT: # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
548 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
549 ; X64-NEXT: # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
550 ; X64-NEXT: vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
551 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
552 ; X64-NEXT: retq # encoding: [0xc3]
553 %res = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 %x4)
554 %res1 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> zeroinitializer, i16 %x4)
555 %res2 = call <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <16 x i8> %x3, i16 -1)
556 %res3 = add <16 x i8> %res, %res1
557 %res4 = add <16 x i8> %res3, %res2
558 ret <16 x i8> %res4
559 }
561 declare <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8>, <32 x i8>, i32, <32 x i8>, i32)
563 define <32 x i8>@test_int_x86_avx512_mask_palignr_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x3, i32 %x4) {
564 ; X86-LABEL: test_int_x86_avx512_mask_palignr_256:
565 ; X86: # %bb.0:
566 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
567 ; X86-NEXT: # ymm3 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
568 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
569 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
570 ; X86-NEXT: # ymm2 {%k1} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
571 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
572 ; X86-NEXT: # ymm0 {%k1} {z} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
573 ; X86-NEXT: vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
574 ; X86-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
575 ; X86-NEXT: retl # encoding: [0xc3]
576 ;
577 ; X64-LABEL: test_int_x86_avx512_mask_palignr_256:
578 ; X64: # %bb.0:
579 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
580 ; X64-NEXT: # ymm3 = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
581 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
582 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
583 ; X64-NEXT: # ymm2 {%k1} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
584 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
585 ; X64-NEXT: # ymm0 {%k1} {z} = ymm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],ymm0[0,1],ymm1[18,19,20,21,22,23,24,25,26,27,28,29,30,31],ymm0[16,17]
586 ; X64-NEXT: vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
587 ; X64-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
588 ; X64-NEXT: retq # encoding: [0xc3]
589 %res = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 %x4)
590 %res1 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> zeroinitializer, i32 %x4)
591 %res2 = call <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <32 x i8> %x3, i32 -1)
592 %res3 = add <32 x i8> %res, %res1
593 %res4 = add <32 x i8> %res3, %res2
594 ret <32 x i8> %res4
595 }
597 declare <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16>, i32, <8 x i16>, i8)
599 define <8 x i16>@test_int_x86_avx512_mask_pshufh_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
600 ; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
601 ; X86: # %bb.0:
602 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
603 ; X86-NEXT: # xmm2 = xmm0[0,1,2,3,7,4,4,4]
604 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
605 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
606 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
607 ; X86-NEXT: # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
608 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
609 ; X86-NEXT: # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
610 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
611 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
612 ; X86-NEXT: retl # encoding: [0xc3]
613 ;
614 ; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
615 ; X64: # %bb.0:
616 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
617 ; X64-NEXT: # xmm2 = xmm0[0,1,2,3,7,4,4,4]
618 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
619 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
620 ; X64-NEXT: # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
621 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
622 ; X64-NEXT: # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
623 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
624 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
625 ; X64-NEXT: retq # encoding: [0xc3]
626 %res = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
627 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
628 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
629 %res3 = add <8 x i16> %res, %res1
630 %res4 = add <8 x i16> %res3, %res2
631 ret <8 x i16> %res4
632 }
634 declare <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16>, i32, <16 x i16>, i16)
636 define <16 x i16>@test_int_x86_avx512_mask_pshufh_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
637 ; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
638 ; X86: # %bb.0:
639 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
640 ; X86-NEXT: # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
641 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
642 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
643 ; X86-NEXT: # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
644 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
645 ; X86-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
646 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
647 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
648 ; X86-NEXT: retl # encoding: [0xc3]
649 ;
650 ; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
651 ; X64: # %bb.0:
652 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
653 ; X64-NEXT: # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
654 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
655 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
656 ; X64-NEXT: # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
657 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
658 ; X64-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
659 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
660 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
661 ; X64-NEXT: retq # encoding: [0xc3]
662 %res = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
663 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
664 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
665 %res3 = add <16 x i16> %res, %res1
666 %res4 = add <16 x i16> %res3, %res2
667 ret <16 x i16> %res4
668 }
670 declare <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16>, i32, <8 x i16>, i8)
672 define <8 x i16>@test_int_x86_avx512_mask_pshufl_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
673 ; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
674 ; X86: # %bb.0:
675 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
676 ; X86-NEXT: # xmm2 = xmm0[3,0,0,0,4,5,6,7]
677 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
678 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
679 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
680 ; X86-NEXT: # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
681 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
682 ; X86-NEXT: # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
683 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
684 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
685 ; X86-NEXT: retl # encoding: [0xc3]
686 ;
687 ; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
688 ; X64: # %bb.0:
689 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
690 ; X64-NEXT: # xmm2 = xmm0[3,0,0,0,4,5,6,7]
691 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
692 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
693 ; X64-NEXT: # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
694 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
695 ; X64-NEXT: # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
696 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
697 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
698 ; X64-NEXT: retq # encoding: [0xc3]
699 %res = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
700 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
701 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
702 %res3 = add <8 x i16> %res, %res1
703 %res4 = add <8 x i16> %res3, %res2
704 ret <8 x i16> %res4
705 }
707 declare <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16>, i32, <16 x i16>, i16)
709 define <16 x i16>@test_int_x86_avx512_mask_pshufl_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
710 ; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
711 ; X86: # %bb.0:
712 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
713 ; X86-NEXT: # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
714 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
715 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
716 ; X86-NEXT: # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
717 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
718 ; X86-NEXT: # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
719 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
720 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
721 ; X86-NEXT: retl # encoding: [0xc3]
722 ;
723 ; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
724 ; X64: # %bb.0:
725 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
726 ; X64-NEXT: # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
727 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
728 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
729 ; X64-NEXT: # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
730 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
731 ; X64-NEXT: # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
732 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
733 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
734 ; X64-NEXT: retq # encoding: [0xc3]
735 %res = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
736 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
737 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
738 %res3 = add <16 x i16> %res, %res1
739 %res4 = add <16 x i16> %res3, %res2
740 ret <16 x i16> %res4
741 }
743 define i32 @test_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b) {
744 ; CHECK-LABEL: test_pcmpeq_b_256:
745 ; CHECK: # %bb.0:
746 ; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
747 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
748 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
749 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
750 %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
751 ret i32 %res
752 }
754 define i32 @test_mask_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
755 ; X86-LABEL: test_mask_pcmpeq_b_256:
756 ; X86: # %bb.0:
757 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
758 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
759 ; X86-NEXT: andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
760 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
761 ; X86-NEXT: retl # encoding: [0xc3]
762 ;
763 ; X64-LABEL: test_mask_pcmpeq_b_256:
764 ; X64: # %bb.0:
765 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
766 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
767 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
768 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
769 ; X64-NEXT: retq # encoding: [0xc3]
770 %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
771 ret i32 %res
772 }
774 declare i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8>, <32 x i8>, i32)
776 define i16 @test_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b) {
777 ; CHECK-LABEL: test_pcmpeq_w_256:
778 ; CHECK: # %bb.0:
779 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
780 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
781 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
782 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
783 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
784 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
785 ret i16 %res
786 }
788 define i16 @test_mask_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
789 ; X86-LABEL: test_mask_pcmpeq_w_256:
790 ; X86: # %bb.0:
791 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
792 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
793 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
794 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
795 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
796 ; X86-NEXT: retl # encoding: [0xc3]
797 ;
798 ; X64-LABEL: test_mask_pcmpeq_w_256:
799 ; X64: # %bb.0:
800 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
801 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
802 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
803 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
804 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
805 ; X64-NEXT: retq # encoding: [0xc3]
806 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
807 ret i16 %res
808 }
810 declare i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16>, <16 x i16>, i16)
812 define i32 @test_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b) {
813 ; CHECK-LABEL: test_pcmpgt_b_256:
814 ; CHECK: # %bb.0:
815 ; CHECK-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
816 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
817 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
818 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
819 %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
820 ret i32 %res
821 }
823 define i32 @test_mask_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
824 ; X86-LABEL: test_mask_pcmpgt_b_256:
825 ; X86: # %bb.0:
826 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
827 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
828 ; X86-NEXT: andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
829 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
830 ; X86-NEXT: retl # encoding: [0xc3]
831 ;
832 ; X64-LABEL: test_mask_pcmpgt_b_256:
833 ; X64: # %bb.0:
834 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
835 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
836 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
837 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
838 ; X64-NEXT: retq # encoding: [0xc3]
839 %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
840 ret i32 %res
841 }
843 declare i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8>, <32 x i8>, i32)
845 define i16 @test_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b) {
846 ; CHECK-LABEL: test_pcmpgt_w_256:
847 ; CHECK: # %bb.0:
848 ; CHECK-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
849 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
850 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
851 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
852 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
853 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
854 ret i16 %res
855 }
857 define i16 @test_mask_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
858 ; X86-LABEL: test_mask_pcmpgt_w_256:
859 ; X86: # %bb.0:
860 ; X86-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
861 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
862 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
863 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
864 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
865 ; X86-NEXT: retl # encoding: [0xc3]
866 ;
867 ; X64-LABEL: test_mask_pcmpgt_w_256:
868 ; X64: # %bb.0:
869 ; X64-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
870 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
871 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
872 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
873 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
874 ; X64-NEXT: retq # encoding: [0xc3]
875 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
876 ret i16 %res
877 }
879 declare i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16>, <16 x i16>, i16)
881 define i16 @test_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b) {
882 ; CHECK-LABEL: test_pcmpeq_b_128:
883 ; CHECK: # %bb.0:
884 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
885 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
886 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
887 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
888 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
889 ret i16 %res
890 }
892 define i16 @test_mask_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
893 ; X86-LABEL: test_mask_pcmpeq_b_128:
894 ; X86: # %bb.0:
895 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
896 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
897 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
898 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
899 ; X86-NEXT: retl # encoding: [0xc3]
900 ;
901 ; X64-LABEL: test_mask_pcmpeq_b_128:
902 ; X64: # %bb.0:
903 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
904 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
905 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
906 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
907 ; X64-NEXT: retq # encoding: [0xc3]
908 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
909 ret i16 %res
910 }
912 declare i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8>, <16 x i8>, i16)
914 define i8 @test_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b) {
915 ; CHECK-LABEL: test_pcmpeq_w_128:
916 ; CHECK: # %bb.0:
917 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
918 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
919 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
920 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
921 %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
922 ret i8 %res
923 }
925 define i8 @test_mask_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
926 ; X86-LABEL: test_mask_pcmpeq_w_128:
927 ; X86: # %bb.0:
928 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
929 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
930 ; X86-NEXT: andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
931 ; X86-NEXT: # kill: def $al killed $al killed $eax
932 ; X86-NEXT: retl # encoding: [0xc3]
933 ;
934 ; X64-LABEL: test_mask_pcmpeq_w_128:
935 ; X64: # %bb.0:
936 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
937 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
938 ; X64-NEXT: andb %dil, %al # encoding: [0x40,0x20,0xf8]
939 ; X64-NEXT: # kill: def $al killed $al killed $eax
940 ; X64-NEXT: retq # encoding: [0xc3]
941 %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
942 ret i8 %res
943 }
945 declare i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16>, <8 x i16>, i8)
947 define i16 @test_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b) {
948 ; CHECK-LABEL: test_pcmpgt_b_128:
949 ; CHECK: # %bb.0:
950 ; CHECK-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
951 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
952 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
953 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
954 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
955 ret i16 %res
956 }
958 define i16 @test_mask_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
959 ; X86-LABEL: test_mask_pcmpgt_b_128:
960 ; X86: # %bb.0:
961 ; X86-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
962 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
963 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
964 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
965 ; X86-NEXT: retl # encoding: [0xc3]
966 ;
967 ; X64-LABEL: test_mask_pcmpgt_b_128:
968 ; X64: # %bb.0:
969 ; X64-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
970 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
971 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
972 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
973 ; X64-NEXT: retq # encoding: [0xc3]
974 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
975 ret i16 %res
976 }
978 declare i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8>, <16 x i8>, i16)
980 define i8 @test_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b) {
981 ; CHECK-LABEL: test_pcmpgt_w_128:
982 ; CHECK: # %bb.0:
983 ; CHECK-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
984 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
985 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
986 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
987 %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
988 ret i8 %res
989 }
991 define i8 @test_mask_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
992 ; X86-LABEL: test_mask_pcmpgt_w_128:
993 ; X86: # %bb.0:
994 ; X86-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
995 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
996 ; X86-NEXT: andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
997 ; X86-NEXT: # kill: def $al killed $al killed $eax
998 ; X86-NEXT: retl # encoding: [0xc3]
999 ;
1000 ; X64-LABEL: test_mask_pcmpgt_w_128:
1001 ; X64: # %bb.0:
1002 ; X64-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
1003 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
1004 ; X64-NEXT: andb %dil, %al # encoding: [0x40,0x20,0xf8]
1005 ; X64-NEXT: # kill: def $al killed $al killed $eax
1006 ; X64-NEXT: retq # encoding: [0xc3]
1007 %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
1008 ret i8 %res
1009 }
1011 declare i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16>, <8 x i16>, i8)
1013 declare <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
1015 define <16 x i8>@test_int_x86_avx512_mask_punpckhb_w_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
1016 ; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
1017 ; X86: # %bb.0:
1018 ; X86-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
1019 ; X86-NEXT: # xmm3 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
1020 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1021 ; X86-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
1022 ; X86-NEXT: # xmm2 {%k1} = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
1023 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1024 ; X86-NEXT: retl # encoding: [0xc3]
1025 ;
1026 ; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
1027 ; X64: # %bb.0:
1028 ; X64-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
1029 ; X64-NEXT: # xmm3 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
1030 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1031 ; X64-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
1032 ; X64-NEXT: # xmm2 {%k1} = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
1033 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1034 ; X64-NEXT: retq # encoding: [0xc3]
1035 %res = call <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
1036 %res1 = call <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
1037 %res2 = add <16 x i8> %res, %res1
1038 ret <16 x i8> %res2
1039 }
1041 declare <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
1043 define <16 x i8>@test_int_x86_avx512_mask_punpcklb_w_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
1044 ; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
1045 ; X86: # %bb.0:
1046 ; X86-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
1047 ; X86-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1048 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1049 ; X86-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
1050 ; X86-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1051 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1052 ; X86-NEXT: retl # encoding: [0xc3]
1053 ;
1054 ; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
1055 ; X64: # %bb.0:
1056 ; X64-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
1057 ; X64-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1058 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1059 ; X64-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
1060 ; X64-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3],xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1061 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1062 ; X64-NEXT: retq # encoding: [0xc3]
1063 %res = call <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
1064 %res1 = call <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
1065 %res2 = add <16 x i8> %res, %res1
1066 ret <16 x i8> %res2
1067 }
1069 declare <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
1071 define <32 x i8>@test_int_x86_avx512_mask_punpckhb_w_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
1072 ; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
1073 ; X86: # %bb.0:
1074 ; X86-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
1075 ; X86-NEXT: # ymm3 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
1076 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1077 ; X86-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
1078 ; X86-NEXT: # ymm2 {%k1} = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
1079 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1080 ; X86-NEXT: retl # encoding: [0xc3]
1081 ;
1082 ; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
1083 ; X64: # %bb.0:
1084 ; X64-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
1085 ; X64-NEXT: # ymm3 = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
1086 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1087 ; X64-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
1088 ; X64-NEXT: # ymm2 {%k1} = ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15],ymm0[24],ymm1[24],ymm0[25],ymm1[25],ymm0[26],ymm1[26],ymm0[27],ymm1[27],ymm0[28],ymm1[28],ymm0[29],ymm1[29],ymm0[30],ymm1[30],ymm0[31],ymm1[31]
1089 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1090 ; X64-NEXT: retq # encoding: [0xc3]
1091 %res = call <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
1092 %res1 = call <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
1093 %res2 = add <32 x i8> %res, %res1
1094 ret <32 x i8> %res2
1095 }
1097 declare <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
1099 define <32 x i8>@test_int_x86_avx512_mask_punpcklb_w_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
1100 ; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
1101 ; X86: # %bb.0:
1102 ; X86-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
1103 ; X86-NEXT: # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
1104 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1105 ; X86-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
1106 ; X86-NEXT: # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
1107 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1108 ; X86-NEXT: retl # encoding: [0xc3]
1109 ;
1110 ; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
1111 ; X64: # %bb.0:
1112 ; X64-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
1113 ; X64-NEXT: # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
1114 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1115 ; X64-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
1116 ; X64-NEXT: # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[16],ymm1[16],ymm0[17],ymm1[17],ymm0[18],ymm1[18],ymm0[19],ymm1[19],ymm0[20],ymm1[20],ymm0[21],ymm1[21],ymm0[22],ymm1[22],ymm0[23],ymm1[23]
1117 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1118 ; X64-NEXT: retq # encoding: [0xc3]
1119 %res = call <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
1120 %res1 = call <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
1121 %res2 = add <32 x i8> %res, %res1
1122 ret <32 x i8> %res2
1123 }
1125 declare <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1127 define <8 x i16>@test_int_x86_avx512_mask_punpcklw_d_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
1128 ; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
1129 ; X86: # %bb.0:
1130 ; X86-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
1131 ; X86-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1132 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1133 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1134 ; X86-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
1135 ; X86-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1136 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1137 ; X86-NEXT: retl # encoding: [0xc3]
1138 ;
1139 ; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
1140 ; X64: # %bb.0:
1141 ; X64-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
1142 ; X64-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1143 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1144 ; X64-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
1145 ; X64-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1146 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1147 ; X64-NEXT: retq # encoding: [0xc3]
1148 %res = call <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
1149 %res1 = call <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
1150 %res2 = add <8 x i16> %res, %res1
1151 ret <8 x i16> %res2
1152 }
1154 declare <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1156 define <8 x i16>@test_int_x86_avx512_mask_punpckhw_d_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
1157 ; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
1158 ; X86: # %bb.0:
1159 ; X86-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
1160 ; X86-NEXT: # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1161 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1162 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1163 ; X86-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
1164 ; X86-NEXT: # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1165 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1166 ; X86-NEXT: retl # encoding: [0xc3]
1167 ;
1168 ; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
1169 ; X64: # %bb.0:
1170 ; X64-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
1171 ; X64-NEXT: # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1172 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1173 ; X64-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
1174 ; X64-NEXT: # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1175 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1176 ; X64-NEXT: retq # encoding: [0xc3]
1177 %res = call <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
1178 %res1 = call <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
1179 %res2 = add <8 x i16> %res, %res1
1180 ret <8 x i16> %res2
1181 }
1183 declare <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1185 define <16 x i16>@test_int_x86_avx512_mask_punpcklw_d_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
1186 ; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
1187 ; X86: # %bb.0:
1188 ; X86-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
1189 ; X86-NEXT: # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
1190 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1191 ; X86-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
1192 ; X86-NEXT: # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
1193 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1194 ; X86-NEXT: retl # encoding: [0xc3]
1195 ;
1196 ; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
1197 ; X64: # %bb.0:
1198 ; X64-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
1199 ; X64-NEXT: # ymm3 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
1200 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1201 ; X64-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
1202 ; X64-NEXT: # ymm2 {%k1} = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[8],ymm1[8],ymm0[9],ymm1[9],ymm0[10],ymm1[10],ymm0[11],ymm1[11]
1203 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1204 ; X64-NEXT: retq # encoding: [0xc3]
1205 %res = call <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
1206 %res1 = call <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
1207 %res2 = add <16 x i16> %res, %res1
1208 ret <16 x i16> %res2
1209 }
1211 declare <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1213 define <16 x i16>@test_int_x86_avx512_mask_punpckhw_d_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
1214 ; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
1215 ; X86: # %bb.0:
1216 ; X86-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
1217 ; X86-NEXT: # ymm3 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
1218 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1219 ; X86-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
1220 ; X86-NEXT: # ymm2 {%k1} = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
1221 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1222 ; X86-NEXT: retl # encoding: [0xc3]
1223 ;
1224 ; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
1225 ; X64: # %bb.0:
1226 ; X64-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
1227 ; X64-NEXT: # ymm3 = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
1228 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1229 ; X64-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
1230 ; X64-NEXT: # ymm2 {%k1} = ymm0[4],ymm1[4],ymm0[5],ymm1[5],ymm0[6],ymm1[6],ymm0[7],ymm1[7],ymm0[12],ymm1[12],ymm0[13],ymm1[13],ymm0[14],ymm1[14],ymm0[15],ymm1[15]
1231 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1232 ; X64-NEXT: retq # encoding: [0xc3]
1233 %res = call <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
1234 %res1 = call <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
1235 %res2 = add <16 x i16> %res, %res1
1236 ret <16 x i16> %res2
1237 }
1239 define <8 x i16> @test_mask_add_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1240 ; CHECK-LABEL: test_mask_add_epi16_rr_128:
1241 ; CHECK: # %bb.0:
1242 ; CHECK-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
1243 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1244 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1245 ret <8 x i16> %res
1246 }
1248 define <8 x i16> @test_mask_add_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1249 ; X86-LABEL: test_mask_add_epi16_rrk_128:
1250 ; X86: # %bb.0:
1251 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1252 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1253 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
1254 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1255 ; X86-NEXT: retl # encoding: [0xc3]
1256 ;
1257 ; X64-LABEL: test_mask_add_epi16_rrk_128:
1258 ; X64: # %bb.0:
1259 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1260 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
1261 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1262 ; X64-NEXT: retq # encoding: [0xc3]
1263 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
1264 ret <8 x i16> %res
1265 }
1267 define <8 x i16> @test_mask_add_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1268 ; X86-LABEL: test_mask_add_epi16_rrkz_128:
1269 ; X86: # %bb.0:
1270 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1271 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1272 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
1273 ; X86-NEXT: retl # encoding: [0xc3]
1274 ;
1275 ; X64-LABEL: test_mask_add_epi16_rrkz_128:
1276 ; X64: # %bb.0:
1277 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1278 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
1279 ; X64-NEXT: retq # encoding: [0xc3]
1280 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
1281 ret <8 x i16> %res
1282 }
1284 define <8 x i16> @test_mask_add_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1285 ; X86-LABEL: test_mask_add_epi16_rm_128:
1286 ; X86: # %bb.0:
1287 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1288 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x00]
1289 ; X86-NEXT: retl # encoding: [0xc3]
1290 ;
1291 ; X64-LABEL: test_mask_add_epi16_rm_128:
1292 ; X64: # %bb.0:
1293 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x07]
1294 ; X64-NEXT: retq # encoding: [0xc3]
1295 %b = load <8 x i16>, <8 x i16>* %ptr_b
1296 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1297 ret <8 x i16> %res
1298 }
1300 define <8 x i16> @test_mask_add_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
1301 ; X86-LABEL: test_mask_add_epi16_rmk_128:
1302 ; X86: # %bb.0:
1303 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1304 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1305 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1306 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x08]
1307 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1308 ; X86-NEXT: retl # encoding: [0xc3]
1309 ;
1310 ; X64-LABEL: test_mask_add_epi16_rmk_128:
1311 ; X64: # %bb.0:
1312 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1313 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x0f]
1314 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1315 ; X64-NEXT: retq # encoding: [0xc3]
1316 %b = load <8 x i16>, <8 x i16>* %ptr_b
1317 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
1318 ret <8 x i16> %res
1319 }
1321 define <8 x i16> @test_mask_add_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
1322 ; X86-LABEL: test_mask_add_epi16_rmkz_128:
1323 ; X86: # %bb.0:
1324 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1325 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1326 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1327 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x00]
1328 ; X86-NEXT: retl # encoding: [0xc3]
1329 ;
1330 ; X64-LABEL: test_mask_add_epi16_rmkz_128:
1331 ; X64: # %bb.0:
1332 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1333 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x07]
1334 ; X64-NEXT: retq # encoding: [0xc3]
1335 %b = load <8 x i16>, <8 x i16>* %ptr_b
1336 %res = call <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
1337 ret <8 x i16> %res
1338 }
1340 declare <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1342 define <16 x i16> @test_mask_add_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
1343 ; CHECK-LABEL: test_mask_add_epi16_rr_256:
1344 ; CHECK: # %bb.0:
1345 ; CHECK-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
1346 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1347 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
1348 ret <16 x i16> %res
1349 }
1351 define <16 x i16> @test_mask_add_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
1352 ; X86-LABEL: test_mask_add_epi16_rrk_256:
1353 ; X86: # %bb.0:
1354 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1355 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
1356 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1357 ; X86-NEXT: retl # encoding: [0xc3]
1358 ;
1359 ; X64-LABEL: test_mask_add_epi16_rrk_256:
1360 ; X64: # %bb.0:
1361 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1362 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
1363 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1364 ; X64-NEXT: retq # encoding: [0xc3]
1365 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
1366 ret <16 x i16> %res
1367 }
1369 define <16 x i16> @test_mask_add_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
1370 ; X86-LABEL: test_mask_add_epi16_rrkz_256:
1371 ; X86: # %bb.0:
1372 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1373 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
1374 ; X86-NEXT: retl # encoding: [0xc3]
1375 ;
1376 ; X64-LABEL: test_mask_add_epi16_rrkz_256:
1377 ; X64: # %bb.0:
1378 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1379 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
1380 ; X64-NEXT: retq # encoding: [0xc3]
1381 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
1382 ret <16 x i16> %res
1383 }
1385 define <16 x i16> @test_mask_add_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
1386 ; X86-LABEL: test_mask_add_epi16_rm_256:
1387 ; X86: # %bb.0:
1388 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1389 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x00]
1390 ; X86-NEXT: retl # encoding: [0xc3]
1391 ;
1392 ; X64-LABEL: test_mask_add_epi16_rm_256:
1393 ; X64: # %bb.0:
1394 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x07]
1395 ; X64-NEXT: retq # encoding: [0xc3]
1396 %b = load <16 x i16>, <16 x i16>* %ptr_b
1397 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
1398 ret <16 x i16> %res
1399 }
1401 define <16 x i16> @test_mask_add_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
1402 ; X86-LABEL: test_mask_add_epi16_rmk_256:
1403 ; X86: # %bb.0:
1404 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1405 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1406 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x08]
1407 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1408 ; X86-NEXT: retl # encoding: [0xc3]
1409 ;
1410 ; X64-LABEL: test_mask_add_epi16_rmk_256:
1411 ; X64: # %bb.0:
1412 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1413 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x0f]
1414 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1415 ; X64-NEXT: retq # encoding: [0xc3]
1416 %b = load <16 x i16>, <16 x i16>* %ptr_b
1417 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
1418 ret <16 x i16> %res
1419 }
1421 define <16 x i16> @test_mask_add_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
1422 ; X86-LABEL: test_mask_add_epi16_rmkz_256:
1423 ; X86: # %bb.0:
1424 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1425 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1426 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x00]
1427 ; X86-NEXT: retl # encoding: [0xc3]
1428 ;
1429 ; X64-LABEL: test_mask_add_epi16_rmkz_256:
1430 ; X64: # %bb.0:
1431 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1432 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x07]
1433 ; X64-NEXT: retq # encoding: [0xc3]
1434 %b = load <16 x i16>, <16 x i16>* %ptr_b
1435 %res = call <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
1436 ret <16 x i16> %res
1437 }
1439 declare <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1441 define <8 x i16> @test_mask_sub_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1442 ; CHECK-LABEL: test_mask_sub_epi16_rr_128:
1443 ; CHECK: # %bb.0:
1444 ; CHECK-NEXT: vpsubw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0xc1]
1445 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1446 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1447 ret <8 x i16> %res
1448 }
1450 define <8 x i16> @test_mask_sub_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1451 ; X86-LABEL: test_mask_sub_epi16_rrk_128:
1452 ; X86: # %bb.0:
1453 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1454 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1455 ; X86-NEXT: vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
1456 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1457 ; X86-NEXT: retl # encoding: [0xc3]
1458 ;
1459 ; X64-LABEL: test_mask_sub_epi16_rrk_128:
1460 ; X64: # %bb.0:
1461 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1462 ; X64-NEXT: vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
1463 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1464 ; X64-NEXT: retq # encoding: [0xc3]
1465 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
1466 ret <8 x i16> %res
1467 }
1469 define <8 x i16> @test_mask_sub_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1470 ; X86-LABEL: test_mask_sub_epi16_rrkz_128:
1471 ; X86: # %bb.0:
1472 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1473 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1474 ; X86-NEXT: vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
1475 ; X86-NEXT: retl # encoding: [0xc3]
1476 ;
1477 ; X64-LABEL: test_mask_sub_epi16_rrkz_128:
1478 ; X64: # %bb.0:
1479 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1480 ; X64-NEXT: vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
1481 ; X64-NEXT: retq # encoding: [0xc3]
1482 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
1483 ret <8 x i16> %res
1484 }
1486 define <8 x i16> @test_mask_sub_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1487 ; X86-LABEL: test_mask_sub_epi16_rm_128:
1488 ; X86: # %bb.0:
1489 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1490 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x00]
1491 ; X86-NEXT: retl # encoding: [0xc3]
1492 ;
1493 ; X64-LABEL: test_mask_sub_epi16_rm_128:
1494 ; X64: # %bb.0:
1495 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x07]
1496 ; X64-NEXT: retq # encoding: [0xc3]
1497 %b = load <8 x i16>, <8 x i16>* %ptr_b
1498 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1499 ret <8 x i16> %res
1500 }
1502 define <8 x i16> @test_mask_sub_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
1503 ; X86-LABEL: test_mask_sub_epi16_rmk_128:
1504 ; X86: # %bb.0:
1505 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1506 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1507 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1508 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x08]
1509 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1510 ; X86-NEXT: retl # encoding: [0xc3]
1511 ;
1512 ; X64-LABEL: test_mask_sub_epi16_rmk_128:
1513 ; X64: # %bb.0:
1514 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1515 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x0f]
1516 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1517 ; X64-NEXT: retq # encoding: [0xc3]
1518 %b = load <8 x i16>, <8 x i16>* %ptr_b
1519 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
1520 ret <8 x i16> %res
1521 }
1523 define <8 x i16> @test_mask_sub_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
1524 ; X86-LABEL: test_mask_sub_epi16_rmkz_128:
1525 ; X86: # %bb.0:
1526 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1527 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1528 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1529 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x00]
1530 ; X86-NEXT: retl # encoding: [0xc3]
1531 ;
1532 ; X64-LABEL: test_mask_sub_epi16_rmkz_128:
1533 ; X64: # %bb.0:
1534 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1535 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x07]
1536 ; X64-NEXT: retq # encoding: [0xc3]
1537 %b = load <8 x i16>, <8 x i16>* %ptr_b
1538 %res = call <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
1539 ret <8 x i16> %res
1540 }
1542 declare <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1544 define <16 x i16> @test_mask_sub_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
1545 ; CHECK-LABEL: test_mask_sub_epi16_rr_256:
1546 ; CHECK: # %bb.0:
1547 ; CHECK-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0xc1]
1548 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1549 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
1550 ret <16 x i16> %res
1551 }
1553 define <16 x i16> @test_mask_sub_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
1554 ; X86-LABEL: test_mask_sub_epi16_rrk_256:
1555 ; X86: # %bb.0:
1556 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1557 ; X86-NEXT: vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
1558 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1559 ; X86-NEXT: retl # encoding: [0xc3]
1560 ;
1561 ; X64-LABEL: test_mask_sub_epi16_rrk_256:
1562 ; X64: # %bb.0:
1563 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1564 ; X64-NEXT: vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
1565 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1566 ; X64-NEXT: retq # encoding: [0xc3]
1567 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
1568 ret <16 x i16> %res
1569 }
1571 define <16 x i16> @test_mask_sub_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
1572 ; X86-LABEL: test_mask_sub_epi16_rrkz_256:
1573 ; X86: # %bb.0:
1574 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1575 ; X86-NEXT: vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
1576 ; X86-NEXT: retl # encoding: [0xc3]
1577 ;
1578 ; X64-LABEL: test_mask_sub_epi16_rrkz_256:
1579 ; X64: # %bb.0:
1580 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1581 ; X64-NEXT: vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
1582 ; X64-NEXT: retq # encoding: [0xc3]
1583 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
1584 ret <16 x i16> %res
1585 }
1587 define <16 x i16> @test_mask_sub_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
1588 ; X86-LABEL: test_mask_sub_epi16_rm_256:
1589 ; X86: # %bb.0:
1590 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1591 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x00]
1592 ; X86-NEXT: retl # encoding: [0xc3]
1593 ;
1594 ; X64-LABEL: test_mask_sub_epi16_rm_256:
1595 ; X64: # %bb.0:
1596 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x07]
1597 ; X64-NEXT: retq # encoding: [0xc3]
1598 %b = load <16 x i16>, <16 x i16>* %ptr_b
1599 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
1600 ret <16 x i16> %res
1601 }
1603 define <16 x i16> @test_mask_sub_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
1604 ; X86-LABEL: test_mask_sub_epi16_rmk_256:
1605 ; X86: # %bb.0:
1606 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1607 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1608 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x08]
1609 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1610 ; X86-NEXT: retl # encoding: [0xc3]
1611 ;
1612 ; X64-LABEL: test_mask_sub_epi16_rmk_256:
1613 ; X64: # %bb.0:
1614 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1615 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x0f]
1616 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1617 ; X64-NEXT: retq # encoding: [0xc3]
1618 %b = load <16 x i16>, <16 x i16>* %ptr_b
1619 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
1620 ret <16 x i16> %res
1621 }
1623 define <16 x i16> @test_mask_sub_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
1624 ; X86-LABEL: test_mask_sub_epi16_rmkz_256:
1625 ; X86: # %bb.0:
1626 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1627 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1628 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x00]
1629 ; X86-NEXT: retl # encoding: [0xc3]
1630 ;
1631 ; X64-LABEL: test_mask_sub_epi16_rmkz_256:
1632 ; X64: # %bb.0:
1633 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1634 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x07]
1635 ; X64-NEXT: retq # encoding: [0xc3]
1636 %b = load <16 x i16>, <16 x i16>* %ptr_b
1637 %res = call <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
1638 ret <16 x i16> %res
1639 }
1641 declare <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1643 define <32 x i16> @test_mask_add_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1644 ; CHECK-LABEL: test_mask_add_epi16_rr_512:
1645 ; CHECK: # %bb.0:
1646 ; CHECK-NEXT: vpaddw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0xc1]
1647 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1648 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1649 ret <32 x i16> %res
1650 }
1652 define <32 x i16> @test_mask_add_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1653 ; X86-LABEL: test_mask_add_epi16_rrk_512:
1654 ; X86: # %bb.0:
1655 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1656 ; X86-NEXT: vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
1657 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1658 ; X86-NEXT: retl # encoding: [0xc3]
1659 ;
1660 ; X64-LABEL: test_mask_add_epi16_rrk_512:
1661 ; X64: # %bb.0:
1662 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1663 ; X64-NEXT: vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
1664 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1665 ; X64-NEXT: retq # encoding: [0xc3]
1666 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1667 ret <32 x i16> %res
1668 }
1670 define <32 x i16> @test_mask_add_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1671 ; X86-LABEL: test_mask_add_epi16_rrkz_512:
1672 ; X86: # %bb.0:
1673 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1674 ; X86-NEXT: vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
1675 ; X86-NEXT: retl # encoding: [0xc3]
1676 ;
1677 ; X64-LABEL: test_mask_add_epi16_rrkz_512:
1678 ; X64: # %bb.0:
1679 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1680 ; X64-NEXT: vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
1681 ; X64-NEXT: retq # encoding: [0xc3]
1682 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1683 ret <32 x i16> %res
1684 }
1686 define <32 x i16> @test_mask_add_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1687 ; X86-LABEL: test_mask_add_epi16_rm_512:
1688 ; X86: # %bb.0:
1689 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1690 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x00]
1691 ; X86-NEXT: retl # encoding: [0xc3]
1692 ;
1693 ; X64-LABEL: test_mask_add_epi16_rm_512:
1694 ; X64: # %bb.0:
1695 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x07]
1696 ; X64-NEXT: retq # encoding: [0xc3]
1697 %b = load <32 x i16>, <32 x i16>* %ptr_b
1698 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1699 ret <32 x i16> %res
1700 }
1702 define <32 x i16> @test_mask_add_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
1703 ; X86-LABEL: test_mask_add_epi16_rmk_512:
1704 ; X86: # %bb.0:
1705 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1706 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1707 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x08]
1708 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1709 ; X86-NEXT: retl # encoding: [0xc3]
1710 ;
1711 ; X64-LABEL: test_mask_add_epi16_rmk_512:
1712 ; X64: # %bb.0:
1713 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1714 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x0f]
1715 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1716 ; X64-NEXT: retq # encoding: [0xc3]
1717 %b = load <32 x i16>, <32 x i16>* %ptr_b
1718 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1719 ret <32 x i16> %res
1720 }
1722 define <32 x i16> @test_mask_add_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1723 ; X86-LABEL: test_mask_add_epi16_rmkz_512:
1724 ; X86: # %bb.0:
1725 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1726 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1727 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x00]
1728 ; X86-NEXT: retl # encoding: [0xc3]
1729 ;
1730 ; X64-LABEL: test_mask_add_epi16_rmkz_512:
1731 ; X64: # %bb.0:
1732 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1733 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x07]
1734 ; X64-NEXT: retq # encoding: [0xc3]
1735 %b = load <32 x i16>, <32 x i16>* %ptr_b
1736 %res = call <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1737 ret <32 x i16> %res
1738 }
1740 declare <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1742 define <32 x i16> @test_mask_sub_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1743 ; CHECK-LABEL: test_mask_sub_epi16_rr_512:
1744 ; CHECK: # %bb.0:
1745 ; CHECK-NEXT: vpsubw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0xc1]
1746 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1747 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1748 ret <32 x i16> %res
1749 }
1751 define <32 x i16> @test_mask_sub_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1752 ; X86-LABEL: test_mask_sub_epi16_rrk_512:
1753 ; X86: # %bb.0:
1754 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1755 ; X86-NEXT: vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
1756 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1757 ; X86-NEXT: retl # encoding: [0xc3]
1758 ;
1759 ; X64-LABEL: test_mask_sub_epi16_rrk_512:
1760 ; X64: # %bb.0:
1761 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1762 ; X64-NEXT: vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
1763 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1764 ; X64-NEXT: retq # encoding: [0xc3]
1765 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1766 ret <32 x i16> %res
1767 }
1769 define <32 x i16> @test_mask_sub_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1770 ; X86-LABEL: test_mask_sub_epi16_rrkz_512:
1771 ; X86: # %bb.0:
1772 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1773 ; X86-NEXT: vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
1774 ; X86-NEXT: retl # encoding: [0xc3]
1775 ;
1776 ; X64-LABEL: test_mask_sub_epi16_rrkz_512:
1777 ; X64: # %bb.0:
1778 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1779 ; X64-NEXT: vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
1780 ; X64-NEXT: retq # encoding: [0xc3]
1781 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1782 ret <32 x i16> %res
1783 }
1785 define <32 x i16> @test_mask_sub_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1786 ; X86-LABEL: test_mask_sub_epi16_rm_512:
1787 ; X86: # %bb.0:
1788 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1789 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x00]
1790 ; X86-NEXT: retl # encoding: [0xc3]
1791 ;
1792 ; X64-LABEL: test_mask_sub_epi16_rm_512:
1793 ; X64: # %bb.0:
1794 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x07]
1795 ; X64-NEXT: retq # encoding: [0xc3]
1796 %b = load <32 x i16>, <32 x i16>* %ptr_b
1797 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1798 ret <32 x i16> %res
1799 }
1801 define <32 x i16> @test_mask_sub_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
1802 ; X86-LABEL: test_mask_sub_epi16_rmk_512:
1803 ; X86: # %bb.0:
1804 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1805 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1806 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x08]
1807 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1808 ; X86-NEXT: retl # encoding: [0xc3]
1809 ;
1810 ; X64-LABEL: test_mask_sub_epi16_rmk_512:
1811 ; X64: # %bb.0:
1812 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1813 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x0f]
1814 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1815 ; X64-NEXT: retq # encoding: [0xc3]
1816 %b = load <32 x i16>, <32 x i16>* %ptr_b
1817 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1818 ret <32 x i16> %res
1819 }
1821 define <32 x i16> @test_mask_sub_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1822 ; X86-LABEL: test_mask_sub_epi16_rmkz_512:
1823 ; X86: # %bb.0:
1824 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1825 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1826 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x00]
1827 ; X86-NEXT: retl # encoding: [0xc3]
1828 ;
1829 ; X64-LABEL: test_mask_sub_epi16_rmkz_512:
1830 ; X64: # %bb.0:
1831 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1832 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x07]
1833 ; X64-NEXT: retq # encoding: [0xc3]
1834 %b = load <32 x i16>, <32 x i16>* %ptr_b
1835 %res = call <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1836 ret <32 x i16> %res
1837 }
1839 declare <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1841 define <32 x i16> @test_mask_mullo_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1842 ; CHECK-LABEL: test_mask_mullo_epi16_rr_512:
1843 ; CHECK: # %bb.0:
1844 ; CHECK-NEXT: vpmullw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0xc1]
1845 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1846 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1847 ret <32 x i16> %res
1848 }
1850 define <32 x i16> @test_mask_mullo_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1851 ; X86-LABEL: test_mask_mullo_epi16_rrk_512:
1852 ; X86: # %bb.0:
1853 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1854 ; X86-NEXT: vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
1855 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1856 ; X86-NEXT: retl # encoding: [0xc3]
1857 ;
1858 ; X64-LABEL: test_mask_mullo_epi16_rrk_512:
1859 ; X64: # %bb.0:
1860 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1861 ; X64-NEXT: vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
1862 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1863 ; X64-NEXT: retq # encoding: [0xc3]
1864 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1865 ret <32 x i16> %res
1866 }
1868 define <32 x i16> @test_mask_mullo_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1869 ; X86-LABEL: test_mask_mullo_epi16_rrkz_512:
1870 ; X86: # %bb.0:
1871 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1872 ; X86-NEXT: vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
1873 ; X86-NEXT: retl # encoding: [0xc3]
1874 ;
1875 ; X64-LABEL: test_mask_mullo_epi16_rrkz_512:
1876 ; X64: # %bb.0:
1877 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1878 ; X64-NEXT: vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
1879 ; X64-NEXT: retq # encoding: [0xc3]
1880 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1881 ret <32 x i16> %res
1882 }
1884 define <32 x i16> @test_mask_mullo_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1885 ; X86-LABEL: test_mask_mullo_epi16_rm_512:
1886 ; X86: # %bb.0:
1887 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1888 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x00]
1889 ; X86-NEXT: retl # encoding: [0xc3]
1890 ;
1891 ; X64-LABEL: test_mask_mullo_epi16_rm_512:
1892 ; X64: # %bb.0:
1893 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x07]
1894 ; X64-NEXT: retq # encoding: [0xc3]
1895 %b = load <32 x i16>, <32 x i16>* %ptr_b
1896 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 -1)
1897 ret <32 x i16> %res
1898 }
1900 define <32 x i16> @test_mask_mullo_epi16_rmk_512(<32 x i16> %a, <32 x i16>* %ptr_b, <32 x i16> %passThru, i32 %mask) {
1901 ; X86-LABEL: test_mask_mullo_epi16_rmk_512:
1902 ; X86: # %bb.0:
1903 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1904 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1905 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x08]
1906 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1907 ; X86-NEXT: retl # encoding: [0xc3]
1908 ;
1909 ; X64-LABEL: test_mask_mullo_epi16_rmk_512:
1910 ; X64: # %bb.0:
1911 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1912 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x0f]
1913 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1914 ; X64-NEXT: retq # encoding: [0xc3]
1915 %b = load <32 x i16>, <32 x i16>* %ptr_b
1916 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask)
1917 ret <32 x i16> %res
1918 }
1920 define <32 x i16> @test_mask_mullo_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1921 ; X86-LABEL: test_mask_mullo_epi16_rmkz_512:
1922 ; X86: # %bb.0:
1923 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1924 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1925 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x00]
1926 ; X86-NEXT: retl # encoding: [0xc3]
1927 ;
1928 ; X64-LABEL: test_mask_mullo_epi16_rmkz_512:
1929 ; X64: # %bb.0:
1930 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1931 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x07]
1932 ; X64-NEXT: retq # encoding: [0xc3]
1933 %b = load <32 x i16>, <32 x i16>* %ptr_b
1934 %res = call <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16> %a, <32 x i16> %b, <32 x i16> zeroinitializer, i32 %mask)
1935 ret <32 x i16> %res
1936 }
1938 declare <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1940 define <8 x i16> @test_mask_mullo_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1941 ; CHECK-LABEL: test_mask_mullo_epi16_rr_128:
1942 ; CHECK: # %bb.0:
1943 ; CHECK-NEXT: vpmullw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0xc1]
1944 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1945 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1946 ret <8 x i16> %res
1947 }
1949 define <8 x i16> @test_mask_mullo_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1950 ; X86-LABEL: test_mask_mullo_epi16_rrk_128:
1951 ; X86: # %bb.0:
1952 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1953 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1954 ; X86-NEXT: vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
1955 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1956 ; X86-NEXT: retl # encoding: [0xc3]
1957 ;
1958 ; X64-LABEL: test_mask_mullo_epi16_rrk_128:
1959 ; X64: # %bb.0:
1960 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1961 ; X64-NEXT: vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
1962 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1963 ; X64-NEXT: retq # encoding: [0xc3]
1964 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
1965 ret <8 x i16> %res
1966 }
1968 define <8 x i16> @test_mask_mullo_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1969 ; X86-LABEL: test_mask_mullo_epi16_rrkz_128:
1970 ; X86: # %bb.0:
1971 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1972 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1973 ; X86-NEXT: vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
1974 ; X86-NEXT: retl # encoding: [0xc3]
1975 ;
1976 ; X64-LABEL: test_mask_mullo_epi16_rrkz_128:
1977 ; X64: # %bb.0:
1978 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1979 ; X64-NEXT: vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
1980 ; X64-NEXT: retq # encoding: [0xc3]
1981 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
1982 ret <8 x i16> %res
1983 }
1985 define <8 x i16> @test_mask_mullo_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1986 ; X86-LABEL: test_mask_mullo_epi16_rm_128:
1987 ; X86: # %bb.0:
1988 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1989 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x00]
1990 ; X86-NEXT: retl # encoding: [0xc3]
1991 ;
1992 ; X64-LABEL: test_mask_mullo_epi16_rm_128:
1993 ; X64: # %bb.0:
1994 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x07]
1995 ; X64-NEXT: retq # encoding: [0xc3]
1996 %b = load <8 x i16>, <8 x i16>* %ptr_b
1997 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
1998 ret <8 x i16> %res
1999 }
2001 define <8 x i16> @test_mask_mullo_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
2002 ; X86-LABEL: test_mask_mullo_epi16_rmk_128:
2003 ; X86: # %bb.0:
2004 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2005 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
2006 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
2007 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x08]
2008 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
2009 ; X86-NEXT: retl # encoding: [0xc3]
2010 ;
2011 ; X64-LABEL: test_mask_mullo_epi16_rmk_128:
2012 ; X64: # %bb.0:
2013 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2014 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x0f]
2015 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
2016 ; X64-NEXT: retq # encoding: [0xc3]
2017 %b = load <8 x i16>, <8 x i16>* %ptr_b
2018 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
2019 ret <8 x i16> %res
2020 }
2022 define <8 x i16> @test_mask_mullo_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
2023 ; X86-LABEL: test_mask_mullo_epi16_rmkz_128:
2024 ; X86: # %bb.0:
2025 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2026 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
2027 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
2028 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x00]
2029 ; X86-NEXT: retl # encoding: [0xc3]
2030 ;
2031 ; X64-LABEL: test_mask_mullo_epi16_rmkz_128:
2032 ; X64: # %bb.0:
2033 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2034 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x07]
2035 ; X64-NEXT: retq # encoding: [0xc3]
2036 %b = load <8 x i16>, <8 x i16>* %ptr_b
2037 %res = call <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
2038 ret <8 x i16> %res
2039 }
2041 declare <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2043 define <16 x i16> @test_mask_mullo_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
2044 ; CHECK-LABEL: test_mask_mullo_epi16_rr_256:
2045 ; CHECK: # %bb.0:
2046 ; CHECK-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0xc1]
2047 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2048 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
2049 ret <16 x i16> %res
2050 }
2052 define <16 x i16> @test_mask_mullo_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
2053 ; X86-LABEL: test_mask_mullo_epi16_rrk_256:
2054 ; X86: # %bb.0:
2055 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2056 ; X86-NEXT: vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
2057 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2058 ; X86-NEXT: retl # encoding: [0xc3]
2059 ;
2060 ; X64-LABEL: test_mask_mullo_epi16_rrk_256:
2061 ; X64: # %bb.0:
2062 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2063 ; X64-NEXT: vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
2064 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2065 ; X64-NEXT: retq # encoding: [0xc3]
2066 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
2067 ret <16 x i16> %res
2068 }
2070 define <16 x i16> @test_mask_mullo_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
2071 ; X86-LABEL: test_mask_mullo_epi16_rrkz_256:
2072 ; X86: # %bb.0:
2073 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2074 ; X86-NEXT: vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
2075 ; X86-NEXT: retl # encoding: [0xc3]
2076 ;
2077 ; X64-LABEL: test_mask_mullo_epi16_rrkz_256:
2078 ; X64: # %bb.0:
2079 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2080 ; X64-NEXT: vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
2081 ; X64-NEXT: retq # encoding: [0xc3]
2082 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
2083 ret <16 x i16> %res
2084 }
2086 define <16 x i16> @test_mask_mullo_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
2087 ; X86-LABEL: test_mask_mullo_epi16_rm_256:
2088 ; X86: # %bb.0:
2089 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2090 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x00]
2091 ; X86-NEXT: retl # encoding: [0xc3]
2092 ;
2093 ; X64-LABEL: test_mask_mullo_epi16_rm_256:
2094 ; X64: # %bb.0:
2095 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x07]
2096 ; X64-NEXT: retq # encoding: [0xc3]
2097 %b = load <16 x i16>, <16 x i16>* %ptr_b
2098 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
2099 ret <16 x i16> %res
2100 }
2102 define <16 x i16> @test_mask_mullo_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
2103 ; X86-LABEL: test_mask_mullo_epi16_rmk_256:
2104 ; X86: # %bb.0:
2105 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2106 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2107 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x08]
2108 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
2109 ; X86-NEXT: retl # encoding: [0xc3]
2110 ;
2111 ; X64-LABEL: test_mask_mullo_epi16_rmk_256:
2112 ; X64: # %bb.0:
2113 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2114 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x0f]
2115 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
2116 ; X64-NEXT: retq # encoding: [0xc3]
2117 %b = load <16 x i16>, <16 x i16>* %ptr_b
2118 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
2119 ret <16 x i16> %res
2120 }
2122 define <16 x i16> @test_mask_mullo_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
2123 ; X86-LABEL: test_mask_mullo_epi16_rmkz_256:
2124 ; X86: # %bb.0:
2125 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2126 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2127 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x00]
2128 ; X86-NEXT: retl # encoding: [0xc3]
2129 ;
2130 ; X64-LABEL: test_mask_mullo_epi16_rmkz_256:
2131 ; X64: # %bb.0:
2132 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2133 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x07]
2134 ; X64-NEXT: retq # encoding: [0xc3]
2135 %b = load <16 x i16>, <16 x i16>* %ptr_b
2136 %res = call <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
2137 ret <16 x i16> %res
2138 }
2140 declare <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2142 declare <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2144 define <16 x i8>@test_int_x86_avx512_mask_pmaxs_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
2145 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
2146 ; X86: # %bb.0:
2147 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2148 ; X86-NEXT: vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
2149 ; X86-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
2150 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2151 ; X86-NEXT: retl # encoding: [0xc3]
2152 ;
2153 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
2154 ; X64: # %bb.0:
2155 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2156 ; X64-NEXT: vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
2157 ; X64-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
2158 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2159 ; X64-NEXT: retq # encoding: [0xc3]
2160 %res = call <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2 ,i16 %mask)
2161 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
2162 %res2 = add <16 x i8> %res, %res1
2163 ret <16 x i8> %res2
2164 }
2166 declare <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2168 define <32 x i8>@test_int_x86_avx512_mask_pmaxs_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
2169 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
2170 ; X86: # %bb.0:
2171 ; X86-NEXT: vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
2172 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2173 ; X86-NEXT: vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
2174 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2175 ; X86-NEXT: retl # encoding: [0xc3]
2176 ;
2177 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
2178 ; X64: # %bb.0:
2179 ; X64-NEXT: vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
2180 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2181 ; X64-NEXT: vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
2182 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2183 ; X64-NEXT: retq # encoding: [0xc3]
2184 %res = call <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
2185 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
2186 %res2 = add <32 x i8> %res, %res1
2187 ret <32 x i8> %res2
2188 }
2190 declare <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2192 define <8 x i16>@test_int_x86_avx512_mask_pmaxs_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2193 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
2194 ; X86: # %bb.0:
2195 ; X86-NEXT: vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
2196 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2197 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2198 ; X86-NEXT: vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
2199 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2200 ; X86-NEXT: retl # encoding: [0xc3]
2201 ;
2202 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
2203 ; X64: # %bb.0:
2204 ; X64-NEXT: vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
2205 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2206 ; X64-NEXT: vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
2207 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2208 ; X64-NEXT: retq # encoding: [0xc3]
2209 %res = call <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2210 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2211 %res2 = add <8 x i16> %res, %res1
2212 ret <8 x i16> %res2
2213 }
2215 declare <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2217 define <16 x i16>@test_int_x86_avx512_mask_pmaxs_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
2218 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
2219 ; X86: # %bb.0:
2220 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2221 ; X86-NEXT: vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
2222 ; X86-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
2223 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2224 ; X86-NEXT: retl # encoding: [0xc3]
2225 ;
2226 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
2227 ; X64: # %bb.0:
2228 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2229 ; X64-NEXT: vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
2230 ; X64-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
2231 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2232 ; X64-NEXT: retq # encoding: [0xc3]
2233 %res = call <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
2234 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
2235 %res2 = add <16 x i16> %res, %res1
2236 ret <16 x i16> %res2
2237 }
2239 declare <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2241 define <16 x i8>@test_int_x86_avx512_mask_pmaxu_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2,i16 %mask) {
2242 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
2243 ; X86: # %bb.0:
2244 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2245 ; X86-NEXT: vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
2246 ; X86-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
2247 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2248 ; X86-NEXT: retl # encoding: [0xc3]
2249 ;
2250 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
2251 ; X64: # %bb.0:
2252 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2253 ; X64-NEXT: vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
2254 ; X64-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
2255 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2256 ; X64-NEXT: retq # encoding: [0xc3]
2257 %res = call <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
2258 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
2259 %res2 = add <16 x i8> %res, %res1
2260 ret <16 x i8> %res2
2261 }
2263 declare <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2265 define <32 x i8>@test_int_x86_avx512_mask_pmaxu_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
2266 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
2267 ; X86: # %bb.0:
2268 ; X86-NEXT: vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
2269 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2270 ; X86-NEXT: vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
2271 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2272 ; X86-NEXT: retl # encoding: [0xc3]
2273 ;
2274 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
2275 ; X64: # %bb.0:
2276 ; X64-NEXT: vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
2277 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2278 ; X64-NEXT: vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
2279 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2280 ; X64-NEXT: retq # encoding: [0xc3]
2281 %res = call <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
2282 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
2283 %res2 = add <32 x i8> %res, %res1
2284 ret <32 x i8> %res2
2285 }
2287 declare <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2289 define <8 x i16>@test_int_x86_avx512_mask_pmaxu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2290 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
2291 ; X86: # %bb.0:
2292 ; X86-NEXT: vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
2293 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2294 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2295 ; X86-NEXT: vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
2296 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2297 ; X86-NEXT: retl # encoding: [0xc3]
2298 ;
2299 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
2300 ; X64: # %bb.0:
2301 ; X64-NEXT: vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
2302 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2303 ; X64-NEXT: vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
2304 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2305 ; X64-NEXT: retq # encoding: [0xc3]
2306 %res = call <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2307 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2308 %res2 = add <8 x i16> %res, %res1
2309 ret <8 x i16> %res2
2310 }
2312 declare <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2314 define <16 x i16>@test_int_x86_avx512_mask_pmaxu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
2315 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
2316 ; X86: # %bb.0:
2317 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2318 ; X86-NEXT: vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
2319 ; X86-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
2320 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2321 ; X86-NEXT: retl # encoding: [0xc3]
2322 ;
2323 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
2324 ; X64: # %bb.0:
2325 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2326 ; X64-NEXT: vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
2327 ; X64-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
2328 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2329 ; X64-NEXT: retq # encoding: [0xc3]
2330 %res = call <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
2331 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
2332 %res2 = add <16 x i16> %res, %res1
2333 ret <16 x i16> %res2
2334 }
2336 declare <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2338 define <16 x i8>@test_int_x86_avx512_mask_pmins_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
2339 ; X86-LABEL: test_int_x86_avx512_mask_pmins_b_128:
2340 ; X86: # %bb.0:
2341 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2342 ; X86-NEXT: vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
2343 ; X86-NEXT: vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
2344 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2345 ; X86-NEXT: retl # encoding: [0xc3]
2346 ;
2347 ; X64-LABEL: test_int_x86_avx512_mask_pmins_b_128:
2348 ; X64: # %bb.0:
2349 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2350 ; X64-NEXT: vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
2351 ; X64-NEXT: vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
2352 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2353 ; X64-NEXT: retq # encoding: [0xc3]
2354 %res = call <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
2355 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
2356 %res2 = add <16 x i8> %res, %res1
2357 ret <16 x i8> %res2
2358 }
2360 declare <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2362 define <32 x i8>@test_int_x86_avx512_mask_pmins_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
2363 ; X86-LABEL: test_int_x86_avx512_mask_pmins_b_256:
2364 ; X86: # %bb.0:
2365 ; X86-NEXT: vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
2366 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2367 ; X86-NEXT: vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
2368 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2369 ; X86-NEXT: retl # encoding: [0xc3]
2370 ;
2371 ; X64-LABEL: test_int_x86_avx512_mask_pmins_b_256:
2372 ; X64: # %bb.0:
2373 ; X64-NEXT: vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
2374 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2375 ; X64-NEXT: vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
2376 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2377 ; X64-NEXT: retq # encoding: [0xc3]
2378 %res = call <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
2379 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
2380 %res2 = add <32 x i8> %res, %res1
2381 ret <32 x i8> %res2
2382 }
2384 declare <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2386 define <8 x i16>@test_int_x86_avx512_mask_pmins_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2387 ; X86-LABEL: test_int_x86_avx512_mask_pmins_w_128:
2388 ; X86: # %bb.0:
2389 ; X86-NEXT: vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
2390 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2391 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2392 ; X86-NEXT: vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
2393 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2394 ; X86-NEXT: retl # encoding: [0xc3]
2395 ;
2396 ; X64-LABEL: test_int_x86_avx512_mask_pmins_w_128:
2397 ; X64: # %bb.0:
2398 ; X64-NEXT: vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
2399 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2400 ; X64-NEXT: vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
2401 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2402 ; X64-NEXT: retq # encoding: [0xc3]
2403 %res = call <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2404 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2405 %res2 = add <8 x i16> %res, %res1
2406 ret <8 x i16> %res2
2407 }
2409 declare <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2411 define <16 x i16>@test_int_x86_avx512_mask_pmins_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
2412 ; X86-LABEL: test_int_x86_avx512_mask_pmins_w_256:
2413 ; X86: # %bb.0:
2414 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2415 ; X86-NEXT: vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
2416 ; X86-NEXT: vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
2417 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2418 ; X86-NEXT: retl # encoding: [0xc3]
2419 ;
2420 ; X64-LABEL: test_int_x86_avx512_mask_pmins_w_256:
2421 ; X64: # %bb.0:
2422 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2423 ; X64-NEXT: vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
2424 ; X64-NEXT: vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
2425 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2426 ; X64-NEXT: retq # encoding: [0xc3]
2427 %res = call <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
2428 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
2429 %res2 = add <16 x i16> %res, %res1
2430 ret <16 x i16> %res2
2431 }
2433 declare <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2435 define <16 x i8>@test_int_x86_avx512_mask_pminu_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask) {
2436 ; X86-LABEL: test_int_x86_avx512_mask_pminu_b_128:
2437 ; X86: # %bb.0:
2438 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2439 ; X86-NEXT: vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
2440 ; X86-NEXT: vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
2441 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2442 ; X86-NEXT: retl # encoding: [0xc3]
2443 ;
2444 ; X64-LABEL: test_int_x86_avx512_mask_pminu_b_128:
2445 ; X64: # %bb.0:
2446 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2447 ; X64-NEXT: vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
2448 ; X64-NEXT: vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
2449 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2450 ; X64-NEXT: retq # encoding: [0xc3]
2451 %res = call <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %mask)
2452 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %mask)
2453 %res2 = add <16 x i8> %res, %res1
2454 ret <16 x i8> %res2
2455 }
2457 declare <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2459 define <32 x i8>@test_int_x86_avx512_mask_pminu_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
2460 ; X86-LABEL: test_int_x86_avx512_mask_pminu_b_256:
2461 ; X86: # %bb.0:
2462 ; X86-NEXT: vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
2463 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2464 ; X86-NEXT: vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
2465 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2466 ; X86-NEXT: retl # encoding: [0xc3]
2467 ;
2468 ; X64-LABEL: test_int_x86_avx512_mask_pminu_b_256:
2469 ; X64: # %bb.0:
2470 ; X64-NEXT: vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
2471 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2472 ; X64-NEXT: vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
2473 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2474 ; X64-NEXT: retq # encoding: [0xc3]
2475 %res = call <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
2476 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
2477 %res2 = add <32 x i8> %res, %res1
2478 ret <32 x i8> %res2
2479 }
2481 declare <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2483 define <8 x i16>@test_int_x86_avx512_mask_pminu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2484 ; X86-LABEL: test_int_x86_avx512_mask_pminu_w_128:
2485 ; X86: # %bb.0:
2486 ; X86-NEXT: vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
2487 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2488 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2489 ; X86-NEXT: vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
2490 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2491 ; X86-NEXT: retl # encoding: [0xc3]
2492 ;
2493 ; X64-LABEL: test_int_x86_avx512_mask_pminu_w_128:
2494 ; X64: # %bb.0:
2495 ; X64-NEXT: vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
2496 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2497 ; X64-NEXT: vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
2498 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2499 ; X64-NEXT: retq # encoding: [0xc3]
2500 %res = call <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2501 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2502 %res2 = add <8 x i16> %res, %res1
2503 ret <8 x i16> %res2
2504 }
2506 declare <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2508 define <16 x i16>@test_int_x86_avx512_mask_pminu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask) {
2509 ; X86-LABEL: test_int_x86_avx512_mask_pminu_w_256:
2510 ; X86: # %bb.0:
2511 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2512 ; X86-NEXT: vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
2513 ; X86-NEXT: vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
2514 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2515 ; X86-NEXT: retl # encoding: [0xc3]
2516 ;
2517 ; X64-LABEL: test_int_x86_avx512_mask_pminu_w_256:
2518 ; X64: # %bb.0:
2519 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2520 ; X64-NEXT: vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
2521 ; X64-NEXT: vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
2522 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2523 ; X64-NEXT: retq # encoding: [0xc3]
2524 %res = call <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %mask)
2525 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %mask)
2526 %res2 = add <16 x i16> %res, %res1
2527 ret <16 x i16> %res2
2528 }
2530 declare <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2532 define <8 x i16>@test_int_x86_avx512_mask_psrl_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2533 ; X86-LABEL: test_int_x86_avx512_mask_psrl_w_128:
2534 ; X86: # %bb.0:
2535 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
2536 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2537 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2538 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
2539 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
2540 ; X86-NEXT: vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
2541 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2542 ; X86-NEXT: retl # encoding: [0xc3]
2543 ;
2544 ; X64-LABEL: test_int_x86_avx512_mask_psrl_w_128:
2545 ; X64: # %bb.0:
2546 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
2547 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2548 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
2549 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
2550 ; X64-NEXT: vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
2551 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2552 ; X64-NEXT: retq # encoding: [0xc3]
2553 %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2554 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2555 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
2556 %res3 = add <8 x i16> %res, %res1
2557 %res4 = add <8 x i16> %res2, %res3
2558 ret <8 x i16> %res4
2559 }
2561 declare <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2563 define <16 x i16>@test_int_x86_avx512_mask_psrl_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
2564 ; X86-LABEL: test_int_x86_avx512_mask_psrl_w_256:
2565 ; X86: # %bb.0:
2566 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
2567 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2568 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
2569 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
2570 ; X86-NEXT: vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
2571 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2572 ; X86-NEXT: retl # encoding: [0xc3]
2573 ;
2574 ; X64-LABEL: test_int_x86_avx512_mask_psrl_w_256:
2575 ; X64: # %bb.0:
2576 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
2577 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2578 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
2579 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
2580 ; X64-NEXT: vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
2581 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2582 ; X64-NEXT: retq # encoding: [0xc3]
2583 %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
2584 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
2585 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
2586 %res3 = add <16 x i16> %res, %res1
2587 %res4 = add <16 x i16> %res3, %res2
2588 ret <16 x i16> %res4
2589 }
2591 declare <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2593 define <8 x i16>@test_int_x86_avx512_mask_psra_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2594 ; X86-LABEL: test_int_x86_avx512_mask_psra_w_128:
2595 ; X86: # %bb.0:
2596 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
2597 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2598 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2599 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
2600 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
2601 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2602 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2603 ; X86-NEXT: retl # encoding: [0xc3]
2604 ;
2605 ; X64-LABEL: test_int_x86_avx512_mask_psra_w_128:
2606 ; X64: # %bb.0:
2607 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
2608 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2609 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
2610 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
2611 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2612 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2613 ; X64-NEXT: retq # encoding: [0xc3]
2614 %res = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2615 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
2616 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2617 %res3 = add <8 x i16> %res, %res1
2618 %res4 = add <8 x i16> %res3, %res2
2619 ret <8 x i16> %res4
2620 }
2622 declare <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2624 define <16 x i16>@test_int_x86_avx512_mask_psra_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
2625 ; X86-LABEL: test_int_x86_avx512_mask_psra_w_256:
2626 ; X86: # %bb.0:
2627 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
2628 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2629 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
2630 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
2631 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2632 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2633 ; X86-NEXT: retl # encoding: [0xc3]
2634 ;
2635 ; X64-LABEL: test_int_x86_avx512_mask_psra_w_256:
2636 ; X64: # %bb.0:
2637 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
2638 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2639 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
2640 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
2641 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2642 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2643 ; X64-NEXT: retq # encoding: [0xc3]
2644 %res = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
2645 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
2646 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
2647 %res3 = add <16 x i16> %res, %res1
2648 %res4 = add <16 x i16> %res3, %res2
2649 ret <16 x i16> %res4
2650 }
2652 declare <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2654 define <8 x i16>@test_int_x86_avx512_mask_psll_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
2655 ; X86-LABEL: test_int_x86_avx512_mask_psll_w_128:
2656 ; X86: # %bb.0:
2657 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
2658 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2659 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2660 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
2661 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
2662 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2663 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2664 ; X86-NEXT: retl # encoding: [0xc3]
2665 ;
2666 ; X64-LABEL: test_int_x86_avx512_mask_psll_w_128:
2667 ; X64: # %bb.0:
2668 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
2669 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2670 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
2671 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
2672 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2673 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2674 ; X64-NEXT: retq # encoding: [0xc3]
2675 %res = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
2676 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
2677 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2678 %res3 = add <8 x i16> %res, %res1
2679 %res4 = add <8 x i16> %res3, %res2
2680 ret <8 x i16> %res4
2681 }
2683 declare <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2685 define <16 x i16>@test_int_x86_avx512_mask_psll_w_256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3) {
2686 ; X86-LABEL: test_int_x86_avx512_mask_psll_w_256:
2687 ; X86: # %bb.0:
2688 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
2689 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2690 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
2691 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
2692 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2693 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2694 ; X86-NEXT: retl # encoding: [0xc3]
2695 ;
2696 ; X64-LABEL: test_int_x86_avx512_mask_psll_w_256:
2697 ; X64: # %bb.0:
2698 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
2699 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2700 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
2701 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
2702 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2703 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2704 ; X64-NEXT: retq # encoding: [0xc3]
2705 %res = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 %x3)
2706 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
2707 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16> %x0, <8 x i16> %x1, <16 x i16> %x2, i16 -1)
2708 %res3 = add <16 x i16> %res, %res1
2709 %res4 = add <16 x i16> %res3, %res2
2710 ret <16 x i16> %res4
2711 }
2713 declare <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2715 define <8 x i16>@test_int_x86_avx512_mask_psrl_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2716 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
2717 ; X86: # %bb.0:
2718 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
2719 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2720 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2721 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
2722 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
2723 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2724 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2725 ; X86-NEXT: retl # encoding: [0xc3]
2726 ;
2727 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
2728 ; X64: # %bb.0:
2729 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
2730 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2731 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
2732 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
2733 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2734 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2735 ; X64-NEXT: retq # encoding: [0xc3]
2736 %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2737 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2738 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2739 %res3 = add <8 x i16> %res, %res1
2740 %res4 = add <8 x i16> %res2, %res3
2741 ret <8 x i16> %res4
2742 }
2744 declare <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2746 define <16 x i16>@test_int_x86_avx512_mask_psrl_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2747 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
2748 ; X86: # %bb.0:
2749 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
2750 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2751 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
2752 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
2753 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2754 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2755 ; X86-NEXT: retl # encoding: [0xc3]
2756 ;
2757 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
2758 ; X64: # %bb.0:
2759 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
2760 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2761 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
2762 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
2763 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2764 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2765 ; X64-NEXT: retq # encoding: [0xc3]
2766 %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2767 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2768 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2769 %res3 = add <16 x i16> %res, %res1
2770 %res4 = add <16 x i16> %res3, %res2
2771 ret <16 x i16> %res4
2772 }
2774 declare <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2776 define <8 x i16>@test_int_x86_avx512_mask_psra_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2777 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_128:
2778 ; X86: # %bb.0:
2779 ; X86-NEXT: vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
2780 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2781 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2782 ; X86-NEXT: vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
2783 ; X86-NEXT: vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
2784 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2785 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2786 ; X86-NEXT: retl # encoding: [0xc3]
2787 ;
2788 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_128:
2789 ; X64: # %bb.0:
2790 ; X64-NEXT: vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
2791 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2792 ; X64-NEXT: vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
2793 ; X64-NEXT: vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
2794 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2795 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2796 ; X64-NEXT: retq # encoding: [0xc3]
2797 %res = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2798 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2799 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2800 %res3 = add <8 x i16> %res, %res1
2801 %res4 = add <8 x i16> %res3, %res2
2802 ret <8 x i16> %res4
2803 }
2805 declare <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2807 define <16 x i16>@test_int_x86_avx512_mask_psra_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2808 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_256:
2809 ; X86: # %bb.0:
2810 ; X86-NEXT: vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
2811 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2812 ; X86-NEXT: vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
2813 ; X86-NEXT: vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
2814 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2815 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2816 ; X86-NEXT: retl # encoding: [0xc3]
2817 ;
2818 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_256:
2819 ; X64: # %bb.0:
2820 ; X64-NEXT: vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
2821 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2822 ; X64-NEXT: vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
2823 ; X64-NEXT: vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
2824 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2825 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2826 ; X64-NEXT: retq # encoding: [0xc3]
2827 %res = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2828 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2829 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2830 %res3 = add <16 x i16> %res, %res1
2831 %res4 = add <16 x i16> %res3, %res2
2832 ret <16 x i16> %res4
2833 }
2835 declare <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2837 define <8 x i16>@test_int_x86_avx512_mask_psll_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2838 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_128:
2839 ; X86: # %bb.0:
2840 ; X86-NEXT: vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
2841 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2842 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2843 ; X86-NEXT: vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
2844 ; X86-NEXT: vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
2845 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2846 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2847 ; X86-NEXT: retl # encoding: [0xc3]
2848 ;
2849 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_128:
2850 ; X64: # %bb.0:
2851 ; X64-NEXT: vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
2852 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2853 ; X64-NEXT: vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
2854 ; X64-NEXT: vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
2855 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2856 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2857 ; X64-NEXT: retq # encoding: [0xc3]
2858 %res = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2859 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2860 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2861 %res3 = add <8 x i16> %res, %res1
2862 %res4 = add <8 x i16> %res3, %res2
2863 ret <8 x i16> %res4
2864 }
2866 declare <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2868 define <16 x i16>@test_int_x86_avx512_mask_psll_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2869 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_256:
2870 ; X86: # %bb.0:
2871 ; X86-NEXT: vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
2872 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2873 ; X86-NEXT: vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
2874 ; X86-NEXT: vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
2875 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2876 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2877 ; X86-NEXT: retl # encoding: [0xc3]
2878 ;
2879 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_256:
2880 ; X64: # %bb.0:
2881 ; X64-NEXT: vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
2882 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2883 ; X64-NEXT: vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
2884 ; X64-NEXT: vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
2885 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2886 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2887 ; X64-NEXT: retq # encoding: [0xc3]
2888 %res = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2889 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2890 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2891 %res3 = add <16 x i16> %res, %res1
2892 %res4 = add <16 x i16> %res3, %res2
2893 ret <16 x i16> %res4
2894 }
2896 declare <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2898 define <16 x i8>@test_int_x86_avx512_mask_pshuf_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
2899 ; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
2900 ; X86: # %bb.0:
2901 ; X86-NEXT: vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
2902 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2903 ; X86-NEXT: vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
2904 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
2905 ; X86-NEXT: retl # encoding: [0xc3]
2906 ;
2907 ; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
2908 ; X64: # %bb.0:
2909 ; X64-NEXT: vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
2910 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2911 ; X64-NEXT: vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
2912 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
2913 ; X64-NEXT: retq # encoding: [0xc3]
2914 %res = call <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
2915 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
2916 %res2 = add <16 x i8> %res, %res1
2917 ret <16 x i8> %res2
2918 }
2920 declare <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2922 define <32 x i8>@test_int_x86_avx512_mask_pshuf_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
2923 ; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
2924 ; X86: # %bb.0:
2925 ; X86-NEXT: vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
2926 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2927 ; X86-NEXT: vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
2928 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2929 ; X86-NEXT: retl # encoding: [0xc3]
2930 ;
2931 ; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
2932 ; X64: # %bb.0:
2933 ; X64-NEXT: vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
2934 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2935 ; X64-NEXT: vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
2936 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2937 ; X64-NEXT: retq # encoding: [0xc3]
2938 %res = call <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
2939 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
2940 %res2 = add <32 x i8> %res, %res1
2941 ret <32 x i8> %res2
2942 }
2944 declare <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8>, <8 x i16>, i8)
2946 define <8 x i16>@test_int_x86_avx512_mask_pmovzxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
2947 ; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
2948 ; X86: # %bb.0:
2949 ; X86-NEXT: vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
2950 ; X86-NEXT: # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2951 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2952 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2953 ; X86-NEXT: vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
2954 ; X86-NEXT: # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2955 ; X86-NEXT: vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
2956 ; X86-NEXT: # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2957 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2958 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2959 ; X86-NEXT: retl # encoding: [0xc3]
2960 ;
2961 ; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
2962 ; X64: # %bb.0:
2963 ; X64-NEXT: vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
2964 ; X64-NEXT: # xmm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2965 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2966 ; X64-NEXT: vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
2967 ; X64-NEXT: # xmm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2968 ; X64-NEXT: vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
2969 ; X64-NEXT: # xmm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
2970 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2971 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2972 ; X64-NEXT: retq # encoding: [0xc3]
2973 %res = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
2974 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
2975 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
2976 %res3 = add <8 x i16> %res, %res1
2977 %res4 = add <8 x i16> %res3, %res2
2978 ret <8 x i16> %res4
2979 }
2981 declare <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8>, <16 x i16>, i16)
2983 define <16 x i16>@test_int_x86_avx512_mask_pmovzxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
2984 ; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
2985 ; X86: # %bb.0:
2986 ; X86-NEXT: vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
2987 ; X86-NEXT: # ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
2988 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2989 ; X86-NEXT: vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
2990 ; X86-NEXT: # ymm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
2991 ; X86-NEXT: vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
2992 ; X86-NEXT: # ymm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
2993 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2994 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2995 ; X86-NEXT: retl # encoding: [0xc3]
2996 ;
2997 ; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
2998 ; X64: # %bb.0:
2999 ; X64-NEXT: vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
3000 ; X64-NEXT: # ymm2 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
3001 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3002 ; X64-NEXT: vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
3003 ; X64-NEXT: # ymm1 {%k1} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
3004 ; X64-NEXT: vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
3005 ; X64-NEXT: # ymm0 {%k1} {z} = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero,xmm0[8],zero,xmm0[9],zero,xmm0[10],zero,xmm0[11],zero,xmm0[12],zero,xmm0[13],zero,xmm0[14],zero,xmm0[15],zero
3006 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3007 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3008 ; X64-NEXT: retq # encoding: [0xc3]
3009 %res = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
3010 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
3011 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
3012 %res3 = add <16 x i16> %res, %res1
3013 %res4 = add <16 x i16> %res3, %res2
3014 ret <16 x i16> %res4
3015 }
3018 declare <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8>, <8 x i16>, i8)
3020 define <8 x i16>@test_int_x86_avx512_mask_pmovsxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
3021 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
3022 ; X86: # %bb.0:
3023 ; X86-NEXT: vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
3024 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3025 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3026 ; X86-NEXT: vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
3027 ; X86-NEXT: vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
3028 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
3029 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
3030 ; X86-NEXT: retl # encoding: [0xc3]
3031 ;
3032 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
3033 ; X64: # %bb.0:
3034 ; X64-NEXT: vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
3035 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3036 ; X64-NEXT: vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
3037 ; X64-NEXT: vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
3038 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
3039 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
3040 ; X64-NEXT: retq # encoding: [0xc3]
3041 %res = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
3042 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
3043 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
3044 %res3 = add <8 x i16> %res, %res1
3045 %res4 = add <8 x i16> %res3, %res2
3046 ret <8 x i16> %res4
3047 }
3049 declare <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8>, <16 x i16>, i16)
3051 define <16 x i16>@test_int_x86_avx512_mask_pmovsxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
3052 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
3053 ; X86: # %bb.0:
3054 ; X86-NEXT: vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
3055 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3056 ; X86-NEXT: vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
3057 ; X86-NEXT: vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
3058 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3059 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3060 ; X86-NEXT: retl # encoding: [0xc3]
3061 ;
3062 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
3063 ; X64: # %bb.0:
3064 ; X64-NEXT: vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
3065 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3066 ; X64-NEXT: vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
3067 ; X64-NEXT: vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
3068 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3069 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3070 ; X64-NEXT: retq # encoding: [0xc3]
3071 %res = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
3072 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
3073 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
3074 %res3 = add <16 x i16> %res, %res1
3075 %res4 = add <16 x i16> %res3, %res2
3076 ret <16 x i16> %res4
3077 }
3079 declare <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32>, <2 x i64>, i8)
3081 define <2 x i64>@test_int_x86_avx512_mask_pmovsxd_q_128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2) {
3082 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
3083 ; X86: # %bb.0:
3084 ; X86-NEXT: vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
3085 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3086 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3087 ; X86-NEXT: vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
3088 ; X86-NEXT: vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
3089 ; X86-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
3090 ; X86-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
3091 ; X86-NEXT: retl # encoding: [0xc3]
3092 ;
3093 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
3094 ; X64: # %bb.0:
3095 ; X64-NEXT: vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
3096 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3097 ; X64-NEXT: vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
3098 ; X64-NEXT: vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
3099 ; X64-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
3100 ; X64-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
3101 ; X64-NEXT: retq # encoding: [0xc3]
3102 %res = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2)
3103 %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> zeroinitializer, i8 %x2)
3104 %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 -1)
3105 %res3 = add <2 x i64> %res, %res1
3106 %res4 = add <2 x i64> %res3, %res2
3107 ret <2 x i64> %res4
3108 }
3110 declare <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32>, <4 x i64>, i8)
3112 define <4 x i64>@test_int_x86_avx512_mask_pmovsxd_q_256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2) {
3113 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
3114 ; X86: # %bb.0:
3115 ; X86-NEXT: vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
3116 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3117 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3118 ; X86-NEXT: vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
3119 ; X86-NEXT: vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
3120 ; X86-NEXT: vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
3121 ; X86-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
3122 ; X86-NEXT: retl # encoding: [0xc3]
3123 ;
3124 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
3125 ; X64: # %bb.0:
3126 ; X64-NEXT: vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
3127 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3128 ; X64-NEXT: vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
3129 ; X64-NEXT: vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
3130 ; X64-NEXT: vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
3131 ; X64-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
3132 ; X64-NEXT: retq # encoding: [0xc3]
3133 %res = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2)
3134 %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> zeroinitializer, i8 %x2)
3135 %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 -1)
3136 %res3 = add <4 x i64> %res, %res1
3137 %res4 = add <4 x i64> %res3, %res2
3138 ret <4 x i64> %res4
3139 }
3142 declare <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16)
3144 define <16 x i8>@test_int_x86_avx512_cvtmask2b_128(i16 %x0) {
3145 ; X86-LABEL: test_int_x86_avx512_cvtmask2b_128:
3146 ; X86: # %bb.0:
3147 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
3148 ; X86-NEXT: vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
3149 ; X86-NEXT: retl # encoding: [0xc3]
3150 ;
3151 ; X64-LABEL: test_int_x86_avx512_cvtmask2b_128:
3152 ; X64: # %bb.0:
3153 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3154 ; X64-NEXT: vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
3155 ; X64-NEXT: retq # encoding: [0xc3]
3156 %res = call <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16 %x0)
3157 ret <16 x i8> %res
3158 }
3160 declare <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32)
3162 define <32 x i8>@test_int_x86_avx512_cvtmask2b_256(i32 %x0) {
3163 ; X86-LABEL: test_int_x86_avx512_cvtmask2b_256:
3164 ; X86: # %bb.0:
3165 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0 # encoding: [0xc4,0xe1,0xf9,0x90,0x44,0x24,0x04]
3166 ; X86-NEXT: vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
3167 ; X86-NEXT: retl # encoding: [0xc3]
3168 ;
3169 ; X64-LABEL: test_int_x86_avx512_cvtmask2b_256:
3170 ; X64: # %bb.0:
3171 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3172 ; X64-NEXT: vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
3173 ; X64-NEXT: retq # encoding: [0xc3]
3174 %res = call <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32 %x0)
3175 ret <32 x i8> %res
3176 }
3178 declare <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8)
3180 define <8 x i16>@test_int_x86_avx512_cvtmask2w_128(i8 %x0) {
3181 ; X86-LABEL: test_int_x86_avx512_cvtmask2w_128:
3182 ; X86: # %bb.0:
3183 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3184 ; X86-NEXT: kmovd %eax, %k0 # encoding: [0xc5,0xfb,0x92,0xc0]
3185 ; X86-NEXT: vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
3186 ; X86-NEXT: retl # encoding: [0xc3]
3187 ;
3188 ; X64-LABEL: test_int_x86_avx512_cvtmask2w_128:
3189 ; X64: # %bb.0:
3190 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3191 ; X64-NEXT: vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
3192 ; X64-NEXT: retq # encoding: [0xc3]
3193 %res = call <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8 %x0)
3194 ret <8 x i16> %res
3195 }
3197 declare <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16)
3199 define <16 x i16>@test_int_x86_avx512_cvtmask2w_256(i16 %x0) {
3200 ; X86-LABEL: test_int_x86_avx512_cvtmask2w_256:
3201 ; X86: # %bb.0:
3202 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
3203 ; X86-NEXT: vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
3204 ; X86-NEXT: retl # encoding: [0xc3]
3205 ;
3206 ; X64-LABEL: test_int_x86_avx512_cvtmask2w_256:
3207 ; X64: # %bb.0:
3208 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3209 ; X64-NEXT: vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
3210 ; X64-NEXT: retq # encoding: [0xc3]
3211 %res = call <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16 %x0)
3212 ret <16 x i16> %res
3213 }
3214 define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
3215 ; CHECK-LABEL: test_mask_packs_epi32_rr_128:
3216 ; CHECK: # %bb.0:
3217 ; CHECK-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
3218 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3219 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3220 ret <8 x i16> %res
3221 }
3223 define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
3224 ; X86-LABEL: test_mask_packs_epi32_rrk_128:
3225 ; X86: # %bb.0:
3226 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3227 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3228 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
3229 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3230 ; X86-NEXT: retl # encoding: [0xc3]
3231 ;
3232 ; X64-LABEL: test_mask_packs_epi32_rrk_128:
3233 ; X64: # %bb.0:
3234 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3235 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
3236 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3237 ; X64-NEXT: retq # encoding: [0xc3]
3238 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3239 ret <8 x i16> %res
3240 }
3242 define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
3243 ; X86-LABEL: test_mask_packs_epi32_rrkz_128:
3244 ; X86: # %bb.0:
3245 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3246 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3247 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
3248 ; X86-NEXT: retl # encoding: [0xc3]
3249 ;
3250 ; X64-LABEL: test_mask_packs_epi32_rrkz_128:
3251 ; X64: # %bb.0:
3252 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3253 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
3254 ; X64-NEXT: retq # encoding: [0xc3]
3255 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3256 ret <8 x i16> %res
3257 }
3259 define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
3260 ; X86-LABEL: test_mask_packs_epi32_rm_128:
3261 ; X86: # %bb.0:
3262 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3263 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x00]
3264 ; X86-NEXT: retl # encoding: [0xc3]
3265 ;
3266 ; X64-LABEL: test_mask_packs_epi32_rm_128:
3267 ; X64: # %bb.0:
3268 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
3269 ; X64-NEXT: retq # encoding: [0xc3]
3270 %b = load <4 x i32>, <4 x i32>* %ptr_b
3271 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3272 ret <8 x i16> %res
3273 }
3275 define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3276 ; X86-LABEL: test_mask_packs_epi32_rmk_128:
3277 ; X86: # %bb.0:
3278 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3279 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3280 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3281 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x08]
3282 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3283 ; X86-NEXT: retl # encoding: [0xc3]
3284 ;
3285 ; X64-LABEL: test_mask_packs_epi32_rmk_128:
3286 ; X64: # %bb.0:
3287 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3288 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
3289 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3290 ; X64-NEXT: retq # encoding: [0xc3]
3291 %b = load <4 x i32>, <4 x i32>* %ptr_b
3292 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3293 ret <8 x i16> %res
3294 }
3296 define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
3297 ; X86-LABEL: test_mask_packs_epi32_rmkz_128:
3298 ; X86: # %bb.0:
3299 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3300 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3301 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3302 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x00]
3303 ; X86-NEXT: retl # encoding: [0xc3]
3304 ;
3305 ; X64-LABEL: test_mask_packs_epi32_rmkz_128:
3306 ; X64: # %bb.0:
3307 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3308 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
3309 ; X64-NEXT: retq # encoding: [0xc3]
3310 %b = load <4 x i32>, <4 x i32>* %ptr_b
3311 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3312 ret <8 x i16> %res
3313 }
3315 define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
3316 ; X86-LABEL: test_mask_packs_epi32_rmb_128:
3317 ; X86: # %bb.0:
3318 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3319 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x00]
3320 ; X86-NEXT: retl # encoding: [0xc3]
3321 ;
3322 ; X64-LABEL: test_mask_packs_epi32_rmb_128:
3323 ; X64: # %bb.0:
3324 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
3325 ; X64-NEXT: retq # encoding: [0xc3]
3326 %q = load i32, i32* %ptr_b
3327 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3328 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3329 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3330 ret <8 x i16> %res
3331 }
3333 define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3334 ; X86-LABEL: test_mask_packs_epi32_rmbk_128:
3335 ; X86: # %bb.0:
3336 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3337 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3338 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3339 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x08]
3340 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3341 ; X86-NEXT: retl # encoding: [0xc3]
3342 ;
3343 ; X64-LABEL: test_mask_packs_epi32_rmbk_128:
3344 ; X64: # %bb.0:
3345 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3346 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
3347 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3348 ; X64-NEXT: retq # encoding: [0xc3]
3349 %q = load i32, i32* %ptr_b
3350 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3351 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3352 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3353 ret <8 x i16> %res
3354 }
3356 define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
3357 ; X86-LABEL: test_mask_packs_epi32_rmbkz_128:
3358 ; X86: # %bb.0:
3359 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3360 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3361 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3362 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x00]
3363 ; X86-NEXT: retl # encoding: [0xc3]
3364 ;
3365 ; X64-LABEL: test_mask_packs_epi32_rmbkz_128:
3366 ; X64: # %bb.0:
3367 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3368 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
3369 ; X64-NEXT: retq # encoding: [0xc3]
3370 %q = load i32, i32* %ptr_b
3371 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3372 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3373 %res = call <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3374 ret <8 x i16> %res
3375 }
3377 declare <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
3379 define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
3380 ; CHECK-LABEL: test_mask_packs_epi32_rr_256:
3381 ; CHECK: # %bb.0:
3382 ; CHECK-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
3383 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3384 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
3385 ret <16 x i16> %res
3386 }
3388 define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
3389 ; X86-LABEL: test_mask_packs_epi32_rrk_256:
3390 ; X86: # %bb.0:
3391 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3392 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
3393 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3394 ; X86-NEXT: retl # encoding: [0xc3]
3395 ;
3396 ; X64-LABEL: test_mask_packs_epi32_rrk_256:
3397 ; X64: # %bb.0:
3398 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3399 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
3400 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3401 ; X64-NEXT: retq # encoding: [0xc3]
3402 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
3403 ret <16 x i16> %res
3404 }
3406 define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
3407 ; X86-LABEL: test_mask_packs_epi32_rrkz_256:
3408 ; X86: # %bb.0:
3409 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3410 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
3411 ; X86-NEXT: retl # encoding: [0xc3]
3412 ;
3413 ; X64-LABEL: test_mask_packs_epi32_rrkz_256:
3414 ; X64: # %bb.0:
3415 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3416 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
3417 ; X64-NEXT: retq # encoding: [0xc3]
3418 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
3419 ret <16 x i16> %res
3420 }
3422 define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
3423 ; X86-LABEL: test_mask_packs_epi32_rm_256:
3424 ; X86: # %bb.0:
3425 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3426 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x00]
3427 ; X86-NEXT: retl # encoding: [0xc3]
3428 ;
3429 ; X64-LABEL: test_mask_packs_epi32_rm_256:
3430 ; X64: # %bb.0:
3431 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
3432 ; X64-NEXT: retq # encoding: [0xc3]
3433 %b = load <8 x i32>, <8 x i32>* %ptr_b
3434 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
3435 ret <16 x i16> %res
3436 }
3438 define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
3439 ; X86-LABEL: test_mask_packs_epi32_rmk_256:
3440 ; X86: # %bb.0:
3441 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3442 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3443 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x08]
3444 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3445 ; X86-NEXT: retl # encoding: [0xc3]
3446 ;
3447 ; X64-LABEL: test_mask_packs_epi32_rmk_256:
3448 ; X64: # %bb.0:
3449 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3450 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
3451 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3452 ; X64-NEXT: retq # encoding: [0xc3]
3453 %b = load <8 x i32>, <8 x i32>* %ptr_b
3454 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
3455 ret <16 x i16> %res
3456 }
3458 define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
3459 ; X86-LABEL: test_mask_packs_epi32_rmkz_256:
3460 ; X86: # %bb.0:
3461 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3462 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3463 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x00]
3464 ; X86-NEXT: retl # encoding: [0xc3]
3465 ;
3466 ; X64-LABEL: test_mask_packs_epi32_rmkz_256:
3467 ; X64: # %bb.0:
3468 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3469 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
3470 ; X64-NEXT: retq # encoding: [0xc3]
3471 %b = load <8 x i32>, <8 x i32>* %ptr_b
3472 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
3473 ret <16 x i16> %res
3474 }
3476 define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
3477 ; X86-LABEL: test_mask_packs_epi32_rmb_256:
3478 ; X86: # %bb.0:
3479 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3480 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x00]
3481 ; X86-NEXT: retl # encoding: [0xc3]
3482 ;
3483 ; X64-LABEL: test_mask_packs_epi32_rmb_256:
3484 ; X64: # %bb.0:
3485 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
3486 ; X64-NEXT: retq # encoding: [0xc3]
3487 %q = load i32, i32* %ptr_b
3488 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3489 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3490 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
3491 ret <16 x i16> %res
3492 }
3494 define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
3495 ; X86-LABEL: test_mask_packs_epi32_rmbk_256:
3496 ; X86: # %bb.0:
3497 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3498 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3499 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x08]
3500 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3501 ; X86-NEXT: retl # encoding: [0xc3]
3502 ;
3503 ; X64-LABEL: test_mask_packs_epi32_rmbk_256:
3504 ; X64: # %bb.0:
3505 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3506 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
3507 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3508 ; X64-NEXT: retq # encoding: [0xc3]
3509 %q = load i32, i32* %ptr_b
3510 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3511 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3512 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
3513 ret <16 x i16> %res
3514 }
3516 define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
3517 ; X86-LABEL: test_mask_packs_epi32_rmbkz_256:
3518 ; X86: # %bb.0:
3519 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3520 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3521 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x00]
3522 ; X86-NEXT: retl # encoding: [0xc3]
3523 ;
3524 ; X64-LABEL: test_mask_packs_epi32_rmbkz_256:
3525 ; X64: # %bb.0:
3526 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3527 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
3528 ; X64-NEXT: retq # encoding: [0xc3]
3529 %q = load i32, i32* %ptr_b
3530 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3531 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3532 %res = call <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
3533 ret <16 x i16> %res
3534 }
3536 declare <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
3538 define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
3539 ; CHECK-LABEL: test_mask_packs_epi16_rr_128:
3540 ; CHECK: # %bb.0:
3541 ; CHECK-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
3542 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3543 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
3544 ret <16 x i8> %res
3545 }
3547 define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
3548 ; X86-LABEL: test_mask_packs_epi16_rrk_128:
3549 ; X86: # %bb.0:
3550 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3551 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
3552 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3553 ; X86-NEXT: retl # encoding: [0xc3]
3554 ;
3555 ; X64-LABEL: test_mask_packs_epi16_rrk_128:
3556 ; X64: # %bb.0:
3557 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3558 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
3559 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3560 ; X64-NEXT: retq # encoding: [0xc3]
3561 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
3562 ret <16 x i8> %res
3563 }
3565 define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
3566 ; X86-LABEL: test_mask_packs_epi16_rrkz_128:
3567 ; X86: # %bb.0:
3568 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3569 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
3570 ; X86-NEXT: retl # encoding: [0xc3]
3571 ;
3572 ; X64-LABEL: test_mask_packs_epi16_rrkz_128:
3573 ; X64: # %bb.0:
3574 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3575 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
3576 ; X64-NEXT: retq # encoding: [0xc3]
3577 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
3578 ret <16 x i8> %res
3579 }
3581 define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
3582 ; X86-LABEL: test_mask_packs_epi16_rm_128:
3583 ; X86: # %bb.0:
3584 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3585 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x00]
3586 ; X86-NEXT: retl # encoding: [0xc3]
3587 ;
3588 ; X64-LABEL: test_mask_packs_epi16_rm_128:
3589 ; X64: # %bb.0:
3590 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
3591 ; X64-NEXT: retq # encoding: [0xc3]
3592 %b = load <8 x i16>, <8 x i16>* %ptr_b
3593 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
3594 ret <16 x i8> %res
3595 }
3597 define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
3598 ; X86-LABEL: test_mask_packs_epi16_rmk_128:
3599 ; X86: # %bb.0:
3600 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3601 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3602 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x08]
3603 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3604 ; X86-NEXT: retl # encoding: [0xc3]
3605 ;
3606 ; X64-LABEL: test_mask_packs_epi16_rmk_128:
3607 ; X64: # %bb.0:
3608 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3609 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
3610 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3611 ; X64-NEXT: retq # encoding: [0xc3]
3612 %b = load <8 x i16>, <8 x i16>* %ptr_b
3613 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
3614 ret <16 x i8> %res
3615 }
3617 define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
3618 ; X86-LABEL: test_mask_packs_epi16_rmkz_128:
3619 ; X86: # %bb.0:
3620 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3621 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3622 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x00]
3623 ; X86-NEXT: retl # encoding: [0xc3]
3624 ;
3625 ; X64-LABEL: test_mask_packs_epi16_rmkz_128:
3626 ; X64: # %bb.0:
3627 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3628 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
3629 ; X64-NEXT: retq # encoding: [0xc3]
3630 %b = load <8 x i16>, <8 x i16>* %ptr_b
3631 %res = call <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
3632 ret <16 x i8> %res
3633 }
3635 declare <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
3637 define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
3638 ; CHECK-LABEL: test_mask_packs_epi16_rr_256:
3639 ; CHECK: # %bb.0:
3640 ; CHECK-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
3641 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3642 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
3643 ret <32 x i8> %res
3644 }
3646 define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
3647 ; X86-LABEL: test_mask_packs_epi16_rrk_256:
3648 ; X86: # %bb.0:
3649 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
3650 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
3651 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3652 ; X86-NEXT: retl # encoding: [0xc3]
3653 ;
3654 ; X64-LABEL: test_mask_packs_epi16_rrk_256:
3655 ; X64: # %bb.0:
3656 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3657 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
3658 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3659 ; X64-NEXT: retq # encoding: [0xc3]
3660 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
3661 ret <32 x i8> %res
3662 }
3664 define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
3665 ; X86-LABEL: test_mask_packs_epi16_rrkz_256:
3666 ; X86: # %bb.0:
3667 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
3668 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
3669 ; X86-NEXT: retl # encoding: [0xc3]
3670 ;
3671 ; X64-LABEL: test_mask_packs_epi16_rrkz_256:
3672 ; X64: # %bb.0:
3673 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3674 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
3675 ; X64-NEXT: retq # encoding: [0xc3]
3676 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
3677 ret <32 x i8> %res
3678 }
3680 define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
3681 ; X86-LABEL: test_mask_packs_epi16_rm_256:
3682 ; X86: # %bb.0:
3683 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3684 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x00]
3685 ; X86-NEXT: retl # encoding: [0xc3]
3686 ;
3687 ; X64-LABEL: test_mask_packs_epi16_rm_256:
3688 ; X64: # %bb.0:
3689 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
3690 ; X64-NEXT: retq # encoding: [0xc3]
3691 %b = load <16 x i16>, <16 x i16>* %ptr_b
3692 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
3693 ret <32 x i8> %res
3694 }
3696 define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
3697 ; X86-LABEL: test_mask_packs_epi16_rmk_256:
3698 ; X86: # %bb.0:
3699 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3700 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
3701 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x08]
3702 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3703 ; X86-NEXT: retl # encoding: [0xc3]
3704 ;
3705 ; X64-LABEL: test_mask_packs_epi16_rmk_256:
3706 ; X64: # %bb.0:
3707 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3708 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
3709 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3710 ; X64-NEXT: retq # encoding: [0xc3]
3711 %b = load <16 x i16>, <16 x i16>* %ptr_b
3712 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
3713 ret <32 x i8> %res
3714 }
3716 define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
3717 ; X86-LABEL: test_mask_packs_epi16_rmkz_256:
3718 ; X86: # %bb.0:
3719 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3720 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
3721 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x00]
3722 ; X86-NEXT: retl # encoding: [0xc3]
3723 ;
3724 ; X64-LABEL: test_mask_packs_epi16_rmkz_256:
3725 ; X64: # %bb.0:
3726 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3727 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
3728 ; X64-NEXT: retq # encoding: [0xc3]
3729 %b = load <16 x i16>, <16 x i16>* %ptr_b
3730 %res = call <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
3731 ret <32 x i8> %res
3732 }
3734 declare <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
3737 define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
3738 ; CHECK-LABEL: test_mask_packus_epi32_rr_128:
3739 ; CHECK: # %bb.0:
3740 ; CHECK-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
3741 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3742 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3743 ret <8 x i16> %res
3744 }
3746 define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
3747 ; X86-LABEL: test_mask_packus_epi32_rrk_128:
3748 ; X86: # %bb.0:
3749 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3750 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3751 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
3752 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3753 ; X86-NEXT: retl # encoding: [0xc3]
3754 ;
3755 ; X64-LABEL: test_mask_packus_epi32_rrk_128:
3756 ; X64: # %bb.0:
3757 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3758 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
3759 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3760 ; X64-NEXT: retq # encoding: [0xc3]
3761 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3762 ret <8 x i16> %res
3763 }
3765 define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
3766 ; X86-LABEL: test_mask_packus_epi32_rrkz_128:
3767 ; X86: # %bb.0:
3768 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3769 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3770 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
3771 ; X86-NEXT: retl # encoding: [0xc3]
3772 ;
3773 ; X64-LABEL: test_mask_packus_epi32_rrkz_128:
3774 ; X64: # %bb.0:
3775 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3776 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
3777 ; X64-NEXT: retq # encoding: [0xc3]
3778 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3779 ret <8 x i16> %res
3780 }
3782 define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
3783 ; X86-LABEL: test_mask_packus_epi32_rm_128:
3784 ; X86: # %bb.0:
3785 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3786 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x00]
3787 ; X86-NEXT: retl # encoding: [0xc3]
3788 ;
3789 ; X64-LABEL: test_mask_packus_epi32_rm_128:
3790 ; X64: # %bb.0:
3791 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
3792 ; X64-NEXT: retq # encoding: [0xc3]
3793 %b = load <4 x i32>, <4 x i32>* %ptr_b
3794 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3795 ret <8 x i16> %res
3796 }
3798 define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, <4 x i32>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3799 ; X86-LABEL: test_mask_packus_epi32_rmk_128:
3800 ; X86: # %bb.0:
3801 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3802 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3803 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3804 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x08]
3805 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3806 ; X86-NEXT: retl # encoding: [0xc3]
3807 ;
3808 ; X64-LABEL: test_mask_packus_epi32_rmk_128:
3809 ; X64: # %bb.0:
3810 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3811 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
3812 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3813 ; X64-NEXT: retq # encoding: [0xc3]
3814 %b = load <4 x i32>, <4 x i32>* %ptr_b
3815 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3816 ret <8 x i16> %res
3817 }
3819 define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
3820 ; X86-LABEL: test_mask_packus_epi32_rmkz_128:
3821 ; X86: # %bb.0:
3822 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3823 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3824 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3825 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x00]
3826 ; X86-NEXT: retl # encoding: [0xc3]
3827 ;
3828 ; X64-LABEL: test_mask_packus_epi32_rmkz_128:
3829 ; X64: # %bb.0:
3830 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3831 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
3832 ; X64-NEXT: retq # encoding: [0xc3]
3833 %b = load <4 x i32>, <4 x i32>* %ptr_b
3834 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3835 ret <8 x i16> %res
3836 }
3838 define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
3839 ; X86-LABEL: test_mask_packus_epi32_rmb_128:
3840 ; X86: # %bb.0:
3841 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3842 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x00]
3843 ; X86-NEXT: retl # encoding: [0xc3]
3844 ;
3845 ; X64-LABEL: test_mask_packus_epi32_rmb_128:
3846 ; X64: # %bb.0:
3847 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
3848 ; X64-NEXT: retq # encoding: [0xc3]
3849 %q = load i32, i32* %ptr_b
3850 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3851 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3852 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 -1)
3853 ret <8 x i16> %res
3854 }
3856 define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3857 ; X86-LABEL: test_mask_packus_epi32_rmbk_128:
3858 ; X86: # %bb.0:
3859 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3860 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3861 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3862 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x08]
3863 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3864 ; X86-NEXT: retl # encoding: [0xc3]
3865 ;
3866 ; X64-LABEL: test_mask_packus_epi32_rmbk_128:
3867 ; X64: # %bb.0:
3868 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3869 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
3870 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3871 ; X64-NEXT: retq # encoding: [0xc3]
3872 %q = load i32, i32* %ptr_b
3873 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3874 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3875 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask)
3876 ret <8 x i16> %res
3877 }
3879 define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
3880 ; X86-LABEL: test_mask_packus_epi32_rmbkz_128:
3881 ; X86: # %bb.0:
3882 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3883 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3884 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3885 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x00]
3886 ; X86-NEXT: retl # encoding: [0xc3]
3887 ;
3888 ; X64-LABEL: test_mask_packus_epi32_rmbkz_128:
3889 ; X64: # %bb.0:
3890 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3891 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
3892 ; X64-NEXT: retq # encoding: [0xc3]
3893 %q = load i32, i32* %ptr_b
3894 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3895 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3896 %res = call <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32> %a, <4 x i32> %b, <8 x i16> zeroinitializer, i8 %mask)
3897 ret <8 x i16> %res
3898 }
3900 declare <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
3902 define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
3903 ; CHECK-LABEL: test_mask_packus_epi32_rr_256:
3904 ; CHECK: # %bb.0:
3905 ; CHECK-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
3906 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3907 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
3908 ret <16 x i16> %res
3909 }
3911 define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
3912 ; X86-LABEL: test_mask_packus_epi32_rrk_256:
3913 ; X86: # %bb.0:
3914 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3915 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
3916 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3917 ; X86-NEXT: retl # encoding: [0xc3]
3918 ;
3919 ; X64-LABEL: test_mask_packus_epi32_rrk_256:
3920 ; X64: # %bb.0:
3921 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3922 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
3923 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3924 ; X64-NEXT: retq # encoding: [0xc3]
3925 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
3926 ret <16 x i16> %res
3927 }
3929 define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
3930 ; X86-LABEL: test_mask_packus_epi32_rrkz_256:
3931 ; X86: # %bb.0:
3932 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3933 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
3934 ; X86-NEXT: retl # encoding: [0xc3]
3935 ;
3936 ; X64-LABEL: test_mask_packus_epi32_rrkz_256:
3937 ; X64: # %bb.0:
3938 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3939 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
3940 ; X64-NEXT: retq # encoding: [0xc3]
3941 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
3942 ret <16 x i16> %res
3943 }
3945 define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
3946 ; X86-LABEL: test_mask_packus_epi32_rm_256:
3947 ; X86: # %bb.0:
3948 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3949 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x00]
3950 ; X86-NEXT: retl # encoding: [0xc3]
3951 ;
3952 ; X64-LABEL: test_mask_packus_epi32_rm_256:
3953 ; X64: # %bb.0:
3954 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
3955 ; X64-NEXT: retq # encoding: [0xc3]
3956 %b = load <8 x i32>, <8 x i32>* %ptr_b
3957 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
3958 ret <16 x i16> %res
3959 }
3961 define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, <8 x i32>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
3962 ; X86-LABEL: test_mask_packus_epi32_rmk_256:
3963 ; X86: # %bb.0:
3964 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3965 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3966 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x08]
3967 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3968 ; X86-NEXT: retl # encoding: [0xc3]
3969 ;
3970 ; X64-LABEL: test_mask_packus_epi32_rmk_256:
3971 ; X64: # %bb.0:
3972 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3973 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
3974 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3975 ; X64-NEXT: retq # encoding: [0xc3]
3976 %b = load <8 x i32>, <8 x i32>* %ptr_b
3977 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
3978 ret <16 x i16> %res
3979 }
3981 define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
3982 ; X86-LABEL: test_mask_packus_epi32_rmkz_256:
3983 ; X86: # %bb.0:
3984 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3985 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3986 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x00]
3987 ; X86-NEXT: retl # encoding: [0xc3]
3988 ;
3989 ; X64-LABEL: test_mask_packus_epi32_rmkz_256:
3990 ; X64: # %bb.0:
3991 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3992 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
3993 ; X64-NEXT: retq # encoding: [0xc3]
3994 %b = load <8 x i32>, <8 x i32>* %ptr_b
3995 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
3996 ret <16 x i16> %res
3997 }
3999 define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
4000 ; X86-LABEL: test_mask_packus_epi32_rmb_256:
4001 ; X86: # %bb.0:
4002 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4003 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x00]
4004 ; X86-NEXT: retl # encoding: [0xc3]
4005 ;
4006 ; X64-LABEL: test_mask_packus_epi32_rmb_256:
4007 ; X64: # %bb.0:
4008 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
4009 ; X64-NEXT: retq # encoding: [0xc3]
4010 %q = load i32, i32* %ptr_b
4011 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4012 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4013 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 -1)
4014 ret <16 x i16> %res
4015 }
4017 define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
4018 ; X86-LABEL: test_mask_packus_epi32_rmbk_256:
4019 ; X86: # %bb.0:
4020 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4021 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4022 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x08]
4023 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4024 ; X86-NEXT: retl # encoding: [0xc3]
4025 ;
4026 ; X64-LABEL: test_mask_packus_epi32_rmbk_256:
4027 ; X64: # %bb.0:
4028 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4029 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
4030 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4031 ; X64-NEXT: retq # encoding: [0xc3]
4032 %q = load i32, i32* %ptr_b
4033 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4034 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4035 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask)
4036 ret <16 x i16> %res
4037 }
4039 define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
4040 ; X86-LABEL: test_mask_packus_epi32_rmbkz_256:
4041 ; X86: # %bb.0:
4042 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4043 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4044 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x00]
4045 ; X86-NEXT: retl # encoding: [0xc3]
4046 ;
4047 ; X64-LABEL: test_mask_packus_epi32_rmbkz_256:
4048 ; X64: # %bb.0:
4049 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4050 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
4051 ; X64-NEXT: retq # encoding: [0xc3]
4052 %q = load i32, i32* %ptr_b
4053 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4054 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4055 %res = call <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32> %a, <8 x i32> %b, <16 x i16> zeroinitializer, i16 %mask)
4056 ret <16 x i16> %res
4057 }
4059 declare <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
4061 define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
4062 ; CHECK-LABEL: test_mask_packus_epi16_rr_128:
4063 ; CHECK: # %bb.0:
4064 ; CHECK-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
4065 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4066 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
4067 ret <16 x i8> %res
4068 }
4070 define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
4071 ; X86-LABEL: test_mask_packus_epi16_rrk_128:
4072 ; X86: # %bb.0:
4073 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
4074 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
4075 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
4076 ; X86-NEXT: retl # encoding: [0xc3]
4077 ;
4078 ; X64-LABEL: test_mask_packus_epi16_rrk_128:
4079 ; X64: # %bb.0:
4080 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4081 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
4082 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
4083 ; X64-NEXT: retq # encoding: [0xc3]
4084 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
4085 ret <16 x i8> %res
4086 }
4088 define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
4089 ; X86-LABEL: test_mask_packus_epi16_rrkz_128:
4090 ; X86: # %bb.0:
4091 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
4092 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
4093 ; X86-NEXT: retl # encoding: [0xc3]
4094 ;
4095 ; X64-LABEL: test_mask_packus_epi16_rrkz_128:
4096 ; X64: # %bb.0:
4097 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4098 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
4099 ; X64-NEXT: retq # encoding: [0xc3]
4100 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
4101 ret <16 x i8> %res
4102 }
4104 define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
4105 ; X86-LABEL: test_mask_packus_epi16_rm_128:
4106 ; X86: # %bb.0:
4107 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4108 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x00]
4109 ; X86-NEXT: retl # encoding: [0xc3]
4110 ;
4111 ; X64-LABEL: test_mask_packus_epi16_rm_128:
4112 ; X64: # %bb.0:
4113 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
4114 ; X64-NEXT: retq # encoding: [0xc3]
4115 %b = load <8 x i16>, <8 x i16>* %ptr_b
4116 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 -1)
4117 ret <16 x i8> %res
4118 }
4120 define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
4121 ; X86-LABEL: test_mask_packus_epi16_rmk_128:
4122 ; X86: # %bb.0:
4123 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4124 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4125 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x08]
4126 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
4127 ; X86-NEXT: retl # encoding: [0xc3]
4128 ;
4129 ; X64-LABEL: test_mask_packus_epi16_rmk_128:
4130 ; X64: # %bb.0:
4131 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4132 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
4133 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
4134 ; X64-NEXT: retq # encoding: [0xc3]
4135 %b = load <8 x i16>, <8 x i16>* %ptr_b
4136 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask)
4137 ret <16 x i8> %res
4138 }
4140 define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
4141 ; X86-LABEL: test_mask_packus_epi16_rmkz_128:
4142 ; X86: # %bb.0:
4143 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4144 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4145 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x00]
4146 ; X86-NEXT: retl # encoding: [0xc3]
4147 ;
4148 ; X64-LABEL: test_mask_packus_epi16_rmkz_128:
4149 ; X64: # %bb.0:
4150 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4151 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
4152 ; X64-NEXT: retq # encoding: [0xc3]
4153 %b = load <8 x i16>, <8 x i16>* %ptr_b
4154 %res = call <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16> %a, <8 x i16> %b, <16 x i8> zeroinitializer, i16 %mask)
4155 ret <16 x i8> %res
4156 }
4158 declare <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
4160 define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
4161 ; CHECK-LABEL: test_mask_packus_epi16_rr_256:
4162 ; CHECK: # %bb.0:
4163 ; CHECK-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
4164 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4165 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
4166 ret <32 x i8> %res
4167 }
4169 define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
4170 ; X86-LABEL: test_mask_packus_epi16_rrk_256:
4171 ; X86: # %bb.0:
4172 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
4173 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
4174 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
4175 ; X86-NEXT: retl # encoding: [0xc3]
4176 ;
4177 ; X64-LABEL: test_mask_packus_epi16_rrk_256:
4178 ; X64: # %bb.0:
4179 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4180 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
4181 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
4182 ; X64-NEXT: retq # encoding: [0xc3]
4183 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
4184 ret <32 x i8> %res
4185 }
4187 define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
4188 ; X86-LABEL: test_mask_packus_epi16_rrkz_256:
4189 ; X86: # %bb.0:
4190 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
4191 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
4192 ; X86-NEXT: retl # encoding: [0xc3]
4193 ;
4194 ; X64-LABEL: test_mask_packus_epi16_rrkz_256:
4195 ; X64: # %bb.0:
4196 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4197 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
4198 ; X64-NEXT: retq # encoding: [0xc3]
4199 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
4200 ret <32 x i8> %res
4201 }
4203 define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
4204 ; X86-LABEL: test_mask_packus_epi16_rm_256:
4205 ; X86: # %bb.0:
4206 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4207 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x00]
4208 ; X86-NEXT: retl # encoding: [0xc3]
4209 ;
4210 ; X64-LABEL: test_mask_packus_epi16_rm_256:
4211 ; X64: # %bb.0:
4212 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
4213 ; X64-NEXT: retq # encoding: [0xc3]
4214 %b = load <16 x i16>, <16 x i16>* %ptr_b
4215 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 -1)
4216 ret <32 x i8> %res
4217 }
4219 define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
4220 ; X86-LABEL: test_mask_packus_epi16_rmk_256:
4221 ; X86: # %bb.0:
4222 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4223 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
4224 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x08]
4225 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4226 ; X86-NEXT: retl # encoding: [0xc3]
4227 ;
4228 ; X64-LABEL: test_mask_packus_epi16_rmk_256:
4229 ; X64: # %bb.0:
4230 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4231 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
4232 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4233 ; X64-NEXT: retq # encoding: [0xc3]
4234 %b = load <16 x i16>, <16 x i16>* %ptr_b
4235 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask)
4236 ret <32 x i8> %res
4237 }
4239 define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
4240 ; X86-LABEL: test_mask_packus_epi16_rmkz_256:
4241 ; X86: # %bb.0:
4242 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4243 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
4244 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x00]
4245 ; X86-NEXT: retl # encoding: [0xc3]
4246 ;
4247 ; X64-LABEL: test_mask_packus_epi16_rmkz_256:
4248 ; X64: # %bb.0:
4249 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4250 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
4251 ; X64-NEXT: retq # encoding: [0xc3]
4252 %b = load <16 x i16>, <16 x i16>* %ptr_b
4253 %res = call <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16> %a, <16 x i16> %b, <32 x i8> zeroinitializer, i32 %mask)
4254 ret <32 x i8> %res
4255 }
4257 declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
4259 define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
4260 ; X86-LABEL: test_cmp_b_256:
4261 ; X86: # %bb.0:
4262 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4263 ; X86-NEXT: .cfi_def_cfa_offset 8
4264 ; X86-NEXT: pushl %edi # encoding: [0x57]
4265 ; X86-NEXT: .cfi_def_cfa_offset 12
4266 ; X86-NEXT: pushl %esi # encoding: [0x56]
4267 ; X86-NEXT: .cfi_def_cfa_offset 16
4268 ; X86-NEXT: .cfi_offset %esi, -16
4269 ; X86-NEXT: .cfi_offset %edi, -12
4270 ; X86-NEXT: .cfi_offset %ebx, -8
4271 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4272 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4273 ; X86-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
4274 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4275 ; X86-NEXT: vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
4276 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4277 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4278 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4279 ; X86-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
4280 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4281 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
4282 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4283 ; X86-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4284 ; X86-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4285 ; X86-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
4286 ; X86-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4287 ; X86-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4288 ; X86-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4289 ; X86-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4290 ; X86-NEXT: vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
4291 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4292 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4293 ; X86-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4294 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4295 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4296 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4297 ; X86-NEXT: popl %esi # encoding: [0x5e]
4298 ; X86-NEXT: .cfi_def_cfa_offset 12
4299 ; X86-NEXT: popl %edi # encoding: [0x5f]
4300 ; X86-NEXT: .cfi_def_cfa_offset 8
4301 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4302 ; X86-NEXT: .cfi_def_cfa_offset 4
4303 ; X86-NEXT: retl # encoding: [0xc3]
4304 ;
4305 ; X64-LABEL: test_cmp_b_256:
4306 ; X64: # %bb.0:
4307 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4308 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4309 ; X64-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
4310 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4311 ; X64-NEXT: vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
4312 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4313 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4314 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4315 ; X64-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
4316 ; X64-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4317 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
4318 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4319 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4320 ; X64-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4321 ; X64-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
4322 ; X64-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4323 ; X64-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4324 ; X64-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4325 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4326 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4327 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4328 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4329 ; X64-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4330 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4331 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4332 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4333 ; X64-NEXT: retq # encoding: [0xc3]
4334 %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
4335 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4336 %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
4337 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4338 %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
4339 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4340 %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
4341 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4342 %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
4343 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4344 %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
4345 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4346 %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
4347 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4348 %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
4349 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4350 ret <8 x i32> %vec7
4351 }
4353 define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
4354 ; X86-LABEL: test_mask_cmp_b_256:
4355 ; X86: # %bb.0:
4356 ; X86-NEXT: pushl %ebp # encoding: [0x55]
4357 ; X86-NEXT: .cfi_def_cfa_offset 8
4358 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4359 ; X86-NEXT: .cfi_def_cfa_offset 12
4360 ; X86-NEXT: pushl %edi # encoding: [0x57]
4361 ; X86-NEXT: .cfi_def_cfa_offset 16
4362 ; X86-NEXT: pushl %esi # encoding: [0x56]
4363 ; X86-NEXT: .cfi_def_cfa_offset 20
4364 ; X86-NEXT: .cfi_offset %esi, -20
4365 ; X86-NEXT: .cfi_offset %edi, -16
4366 ; X86-NEXT: .cfi_offset %ebx, -12
4367 ; X86-NEXT: .cfi_offset %ebp, -8
4368 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
4369 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4370 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4371 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4372 ; X86-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
4373 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4374 ; X86-NEXT: vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
4375 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4376 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4377 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4378 ; X86-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
4379 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4380 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
4381 ; X86-NEXT: kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
4382 ; X86-NEXT: vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
4383 ; X86-NEXT: vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
4384 ; X86-NEXT: vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
4385 ; X86-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
4386 ; X86-NEXT: vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
4387 ; X86-NEXT: vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
4388 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4389 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4390 ; X86-NEXT: vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
4391 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4392 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4393 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4394 ; X86-NEXT: popl %esi # encoding: [0x5e]
4395 ; X86-NEXT: .cfi_def_cfa_offset 16
4396 ; X86-NEXT: popl %edi # encoding: [0x5f]
4397 ; X86-NEXT: .cfi_def_cfa_offset 12
4398 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4399 ; X86-NEXT: .cfi_def_cfa_offset 8
4400 ; X86-NEXT: popl %ebp # encoding: [0x5d]
4401 ; X86-NEXT: .cfi_def_cfa_offset 4
4402 ; X86-NEXT: retl # encoding: [0xc3]
4403 ;
4404 ; X64-LABEL: test_mask_cmp_b_256:
4405 ; X64: # %bb.0:
4406 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4407 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4408 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4409 ; X64-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
4410 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4411 ; X64-NEXT: vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
4412 ; X64-NEXT: kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
4413 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4414 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4415 ; X64-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
4416 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4417 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
4418 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4419 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4420 ; X64-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
4421 ; X64-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
4422 ; X64-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
4423 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4424 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4425 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4426 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4427 ; X64-NEXT: vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
4428 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4429 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4430 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4431 ; X64-NEXT: retq # encoding: [0xc3]
4432 %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
4433 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4434 %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
4435 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4436 %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
4437 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4438 %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
4439 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4440 %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
4441 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4442 %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
4443 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4444 %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
4445 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4446 %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
4447 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4448 ret <8 x i32> %vec7
4449 }
4451 declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
4453 define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
4454 ; X86-LABEL: test_ucmp_b_256:
4455 ; X86: # %bb.0:
4456 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4457 ; X86-NEXT: .cfi_def_cfa_offset 8
4458 ; X86-NEXT: pushl %edi # encoding: [0x57]
4459 ; X86-NEXT: .cfi_def_cfa_offset 12
4460 ; X86-NEXT: pushl %esi # encoding: [0x56]
4461 ; X86-NEXT: .cfi_def_cfa_offset 16
4462 ; X86-NEXT: .cfi_offset %esi, -16
4463 ; X86-NEXT: .cfi_offset %edi, -12
4464 ; X86-NEXT: .cfi_offset %ebx, -8
4465 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4466 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4467 ; X86-NEXT: vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
4468 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4469 ; X86-NEXT: vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
4470 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4471 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4472 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4473 ; X86-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
4474 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4475 ; X86-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
4476 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4477 ; X86-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4478 ; X86-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4479 ; X86-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
4480 ; X86-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4481 ; X86-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4482 ; X86-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4483 ; X86-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4484 ; X86-NEXT: vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
4485 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4486 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4487 ; X86-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4488 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4489 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4490 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4491 ; X86-NEXT: popl %esi # encoding: [0x5e]
4492 ; X86-NEXT: .cfi_def_cfa_offset 12
4493 ; X86-NEXT: popl %edi # encoding: [0x5f]
4494 ; X86-NEXT: .cfi_def_cfa_offset 8
4495 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4496 ; X86-NEXT: .cfi_def_cfa_offset 4
4497 ; X86-NEXT: retl # encoding: [0xc3]
4498 ;
4499 ; X64-LABEL: test_ucmp_b_256:
4500 ; X64: # %bb.0:
4501 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4502 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4503 ; X64-NEXT: vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
4504 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4505 ; X64-NEXT: vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
4506 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4507 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4508 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4509 ; X64-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
4510 ; X64-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4511 ; X64-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
4512 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4513 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4514 ; X64-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4515 ; X64-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
4516 ; X64-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4517 ; X64-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4518 ; X64-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4519 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4520 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4521 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4522 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4523 ; X64-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4524 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4525 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4526 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4527 ; X64-NEXT: retq # encoding: [0xc3]
4528 %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
4529 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4530 %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
4531 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4532 %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
4533 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4534 %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
4535 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4536 %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
4537 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4538 %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
4539 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4540 %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
4541 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4542 %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
4543 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4544 ret <8 x i32> %vec7
4545 }
4547 define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
4548 ; X86-LABEL: test_mask_ucmp_b_256:
4549 ; X86: # %bb.0:
4550 ; X86-NEXT: pushl %ebp # encoding: [0x55]
4551 ; X86-NEXT: .cfi_def_cfa_offset 8
4552 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4553 ; X86-NEXT: .cfi_def_cfa_offset 12
4554 ; X86-NEXT: pushl %edi # encoding: [0x57]
4555 ; X86-NEXT: .cfi_def_cfa_offset 16
4556 ; X86-NEXT: pushl %esi # encoding: [0x56]
4557 ; X86-NEXT: .cfi_def_cfa_offset 20
4558 ; X86-NEXT: .cfi_offset %esi, -20
4559 ; X86-NEXT: .cfi_offset %edi, -16
4560 ; X86-NEXT: .cfi_offset %ebx, -12
4561 ; X86-NEXT: .cfi_offset %ebp, -8
4562 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
4563 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4564 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4565 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4566 ; X86-NEXT: vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
4567 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4568 ; X86-NEXT: vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
4569 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4570 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4571 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4572 ; X86-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
4573 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4574 ; X86-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
4575 ; X86-NEXT: kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
4576 ; X86-NEXT: vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
4577 ; X86-NEXT: vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
4578 ; X86-NEXT: vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
4579 ; X86-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
4580 ; X86-NEXT: vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
4581 ; X86-NEXT: vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
4582 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4583 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4584 ; X86-NEXT: vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
4585 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4586 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4587 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4588 ; X86-NEXT: popl %esi # encoding: [0x5e]
4589 ; X86-NEXT: .cfi_def_cfa_offset 16
4590 ; X86-NEXT: popl %edi # encoding: [0x5f]
4591 ; X86-NEXT: .cfi_def_cfa_offset 12
4592 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4593 ; X86-NEXT: .cfi_def_cfa_offset 8
4594 ; X86-NEXT: popl %ebp # encoding: [0x5d]
4595 ; X86-NEXT: .cfi_def_cfa_offset 4
4596 ; X86-NEXT: retl # encoding: [0xc3]
4597 ;
4598 ; X64-LABEL: test_mask_ucmp_b_256:
4599 ; X64: # %bb.0:
4600 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4601 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4602 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4603 ; X64-NEXT: vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
4604 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4605 ; X64-NEXT: vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
4606 ; X64-NEXT: kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
4607 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4608 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4609 ; X64-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
4610 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4611 ; X64-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
4612 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4613 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4614 ; X64-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
4615 ; X64-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
4616 ; X64-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
4617 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4618 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4619 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4620 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4621 ; X64-NEXT: vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
4622 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4623 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4624 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4625 ; X64-NEXT: retq # encoding: [0xc3]
4626 %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
4627 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4628 %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
4629 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4630 %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
4631 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4632 %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
4633 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4634 %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
4635 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4636 %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
4637 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4638 %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
4639 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4640 %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
4641 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4642 ret <8 x i32> %vec7
4643 }
4645 declare i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
4647 define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
4648 ; CHECK-LABEL: test_cmp_w_256:
4649 ; CHECK: # %bb.0:
4650 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
4651 ; CHECK-NEXT: vpcmpgtw %ymm0, %ymm1, %k1 # encoding: [0x62,0xf1,0x75,0x28,0x65,0xc8]
4652 ; CHECK-NEXT: vpcmplew %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
4653 ; CHECK-NEXT: vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
4654 ; CHECK-NEXT: vpcmpnltw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe1,0x05]
4655 ; CHECK-NEXT: vpcmpgtw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xe9]
4656 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4657 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4658 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4659 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4660 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4661 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4662 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4663 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4664 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4665 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4666 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4667 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4668 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4669 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4670 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4671 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4672 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4673 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4674 %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
4675 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4676 %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
4677 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4678 %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
4679 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4680 %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
4681 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4682 %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
4683 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4684 %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
4685 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4686 %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
4687 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4688 %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
4689 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4690 ret <8 x i16> %vec7
4691 }
4693 define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
4694 ; X86-LABEL: test_mask_cmp_w_256:
4695 ; X86: # %bb.0:
4696 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4697 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4698 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4699 ; X86-NEXT: vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
4700 ; X86-NEXT: vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
4701 ; X86-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4702 ; X86-NEXT: vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
4703 ; X86-NEXT: vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
4704 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4705 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4706 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4707 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4708 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4709 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4710 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4711 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4712 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4713 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4714 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4715 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4716 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4717 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4718 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4719 ; X86-NEXT: retl # encoding: [0xc3]
4720 ;
4721 ; X64-LABEL: test_mask_cmp_w_256:
4722 ; X64: # %bb.0:
4723 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4724 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4725 ; X64-NEXT: vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
4726 ; X64-NEXT: vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
4727 ; X64-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4728 ; X64-NEXT: vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
4729 ; X64-NEXT: vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
4730 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4731 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4732 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4733 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4734 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4735 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4736 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4737 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4738 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4739 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4740 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4741 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4742 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4743 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4744 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4745 ; X64-NEXT: retq # encoding: [0xc3]
4746 %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
4747 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4748 %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
4749 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4750 %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
4751 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4752 %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
4753 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4754 %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
4755 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4756 %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
4757 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4758 %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
4759 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4760 %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
4761 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4762 ret <8 x i16> %vec7
4763 }
4765 declare i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
4767 define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
4768 ; CHECK-LABEL: test_ucmp_w_256:
4769 ; CHECK: # %bb.0:
4770 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
4771 ; CHECK-NEXT: vpcmpltuw %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc9,0x01]
4772 ; CHECK-NEXT: vpcmpleuw %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd1,0x02]
4773 ; CHECK-NEXT: vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
4774 ; CHECK-NEXT: vpcmpnltuw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x05]
4775 ; CHECK-NEXT: vpcmpnleuw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x06]
4776 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4777 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4778 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4779 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4780 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4781 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4782 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4783 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4784 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4785 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4786 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4787 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4788 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4789 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4790 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4791 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4792 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4793 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4794 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
4795 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4796 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
4797 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4798 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
4799 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4800 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
4801 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4802 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
4803 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4804 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
4805 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4806 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
4807 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4808 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
4809 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4810 ret <8 x i16> %vec7
4811 }
4813 define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
4814 ; X86-LABEL: test_mask_ucmp_w_256:
4815 ; X86: # %bb.0:
4816 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4817 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4818 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4819 ; X86-NEXT: vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
4820 ; X86-NEXT: vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
4821 ; X86-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4822 ; X86-NEXT: vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
4823 ; X86-NEXT: vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
4824 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4825 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4826 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4827 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4828 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4829 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4830 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4831 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4832 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4833 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4834 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4835 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4836 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4837 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4838 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4839 ; X86-NEXT: retl # encoding: [0xc3]
4840 ;
4841 ; X64-LABEL: test_mask_ucmp_w_256:
4842 ; X64: # %bb.0:
4843 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4844 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4845 ; X64-NEXT: vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
4846 ; X64-NEXT: vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
4847 ; X64-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4848 ; X64-NEXT: vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
4849 ; X64-NEXT: vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
4850 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4851 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4852 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4853 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4854 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4855 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4856 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4857 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4858 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4859 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4860 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4861 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4862 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4863 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4864 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4865 ; X64-NEXT: retq # encoding: [0xc3]
4866 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
4867 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4868 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
4869 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4870 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
4871 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4872 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
4873 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4874 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
4875 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4876 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
4877 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4878 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
4879 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4880 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
4881 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4882 ret <8 x i16> %vec7
4883 }
4885 declare i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
4887 define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
4888 ; CHECK-LABEL: test_cmp_b_128:
4889 ; CHECK: # %bb.0:
4890 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
4891 ; CHECK-NEXT: vpcmpgtb %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x64,0xc8]
4892 ; CHECK-NEXT: vpcmpleb %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
4893 ; CHECK-NEXT: vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
4894 ; CHECK-NEXT: vpcmpnltb %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe1,0x05]
4895 ; CHECK-NEXT: vpcmpgtb %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xe9]
4896 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4897 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4898 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4899 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4900 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4901 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4902 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4903 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4904 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4905 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4906 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4907 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4908 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4909 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4910 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4911 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4912 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4913 %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
4914 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4915 %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
4916 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4917 %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
4918 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4919 %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
4920 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4921 %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
4922 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4923 %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
4924 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4925 %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
4926 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4927 %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
4928 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4929 ret <8 x i16> %vec7
4930 }
4932 define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
4933 ; X86-LABEL: test_mask_cmp_b_128:
4934 ; X86: # %bb.0:
4935 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4936 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4937 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
4938 ; X86-NEXT: vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
4939 ; X86-NEXT: vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
4940 ; X86-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
4941 ; X86-NEXT: vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
4942 ; X86-NEXT: vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
4943 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4944 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4945 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4946 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4947 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4948 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4949 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4950 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4951 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4952 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4953 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4954 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4955 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4956 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4957 ; X86-NEXT: retl # encoding: [0xc3]
4958 ;
4959 ; X64-LABEL: test_mask_cmp_b_128:
4960 ; X64: # %bb.0:
4961 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4962 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
4963 ; X64-NEXT: vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
4964 ; X64-NEXT: vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
4965 ; X64-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
4966 ; X64-NEXT: vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
4967 ; X64-NEXT: vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
4968 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4969 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4970 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4971 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4972 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4973 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4974 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4975 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4976 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4977 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4978 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4979 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4980 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4981 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4982 ; X64-NEXT: retq # encoding: [0xc3]
4983 %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
4984 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4985 %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
4986 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4987 %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
4988 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4989 %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
4990 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4991 %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
4992 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4993 %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
4994 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4995 %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
4996 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4997 %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
4998 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4999 ret <8 x i16> %vec7
5000 }
5002 declare i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
5004 define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
5005 ; CHECK-LABEL: test_ucmp_b_128:
5006 ; CHECK: # %bb.0:
5007 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
5008 ; CHECK-NEXT: vpcmpltub %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc9,0x01]
5009 ; CHECK-NEXT: vpcmpleub %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd1,0x02]
5010 ; CHECK-NEXT: vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
5011 ; CHECK-NEXT: vpcmpnltub %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x05]
5012 ; CHECK-NEXT: vpcmpnleub %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x06]
5013 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5014 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5015 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5016 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5017 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5018 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5019 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5020 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5021 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5022 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5023 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5024 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5025 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5026 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
5027 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
5028 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
5029 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5030 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
5031 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
5032 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
5033 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
5034 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
5035 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
5036 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
5037 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
5038 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
5039 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
5040 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
5041 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
5042 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
5043 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
5044 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
5045 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
5046 ret <8 x i16> %vec7
5047 }
5049 define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
5050 ; X86-LABEL: test_mask_ucmp_b_128:
5051 ; X86: # %bb.0:
5052 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5053 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5054 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
5055 ; X86-NEXT: vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
5056 ; X86-NEXT: vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
5057 ; X86-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
5058 ; X86-NEXT: vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
5059 ; X86-NEXT: vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
5060 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5061 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5062 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
5063 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5064 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
5065 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5066 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
5067 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5068 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
5069 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5070 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
5071 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5072 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
5073 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5074 ; X86-NEXT: retl # encoding: [0xc3]
5075 ;
5076 ; X64-LABEL: test_mask_ucmp_b_128:
5077 ; X64: # %bb.0:
5078 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5079 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
5080 ; X64-NEXT: vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
5081 ; X64-NEXT: vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
5082 ; X64-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
5083 ; X64-NEXT: vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
5084 ; X64-NEXT: vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
5085 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5086 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5087 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5088 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5089 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5090 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5091 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5092 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5093 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5094 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5095 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5096 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5097 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5098 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
5099 ; X64-NEXT: retq # encoding: [0xc3]
5100 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
5101 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
5102 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
5103 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
5104 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
5105 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
5106 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
5107 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
5108 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
5109 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
5110 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
5111 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
5112 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
5113 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
5114 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
5115 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
5116 ret <8 x i16> %vec7
5117 }
5119 declare i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
5121 define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
5122 ; CHECK-LABEL: test_cmp_w_128:
5123 ; CHECK: # %bb.0:
5124 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
5125 ; CHECK-NEXT: vpcmpgtw %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x65,0xc8]
5126 ; CHECK-NEXT: vpcmplew %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
5127 ; CHECK-NEXT: vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
5128 ; CHECK-NEXT: vpcmpnltw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe1,0x05]
5129 ; CHECK-NEXT: vpcmpgtw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xe9]
5130 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5131 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5132 ; CHECK-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
5133 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5134 ; CHECK-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
5135 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5136 ; CHECK-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
5137 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5138 ; CHECK-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
5139 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5140 ; CHECK-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
5141 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5142 ; CHECK-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
5143 ; CHECK-NEXT: movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
5144 ; CHECK-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
5145 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5146 %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
5147 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5148 %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
5149 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5150 %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
5151 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5152 %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
5153 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5154 %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
5155 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5156 %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
5157 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5158 %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
5159 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5160 %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
5161 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5162 ret <8 x i8> %vec7
5163 }
5165 define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
5166 ; X86-LABEL: test_mask_cmp_w_128:
5167 ; X86: # %bb.0:
5168 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5169 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5170 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5171 ; X86-NEXT: vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
5172 ; X86-NEXT: vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
5173 ; X86-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5174 ; X86-NEXT: vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
5175 ; X86-NEXT: vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
5176 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5177 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5178 ; X86-NEXT: vpinsrb $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x00]
5179 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5180 ; X86-NEXT: vpinsrb $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
5181 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5182 ; X86-NEXT: vpinsrb $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
5183 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5184 ; X86-NEXT: vpinsrb $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
5185 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5186 ; X86-NEXT: vpinsrb $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
5187 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5188 ; X86-NEXT: vpinsrb $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
5189 ; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
5190 ; X86-NEXT: retl # encoding: [0xc3]
5191 ;
5192 ; X64-LABEL: test_mask_cmp_w_128:
5193 ; X64: # %bb.0:
5194 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5195 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5196 ; X64-NEXT: vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
5197 ; X64-NEXT: vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
5198 ; X64-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5199 ; X64-NEXT: vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
5200 ; X64-NEXT: vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
5201 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5202 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5203 ; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
5204 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5205 ; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
5206 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5207 ; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
5208 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5209 ; X64-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
5210 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5211 ; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
5212 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5213 ; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
5214 ; X64-NEXT: vpinsrb $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
5215 ; X64-NEXT: retq # encoding: [0xc3]
5216 %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
5217 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5218 %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
5219 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5220 %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
5221 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5222 %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
5223 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5224 %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
5225 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5226 %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
5227 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5228 %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
5229 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5230 %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
5231 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5232 ret <8 x i8> %vec7
5233 }
5235 declare i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
5237 define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
5238 ; CHECK-LABEL: test_ucmp_w_128:
5239 ; CHECK: # %bb.0:
5240 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
5241 ; CHECK-NEXT: vpcmpltuw %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc9,0x01]
5242 ; CHECK-NEXT: vpcmpleuw %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd1,0x02]
5243 ; CHECK-NEXT: vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
5244 ; CHECK-NEXT: vpcmpnltuw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x05]
5245 ; CHECK-NEXT: vpcmpnleuw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x06]
5246 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5247 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5248 ; CHECK-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
5249 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5250 ; CHECK-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
5251 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5252 ; CHECK-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
5253 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5254 ; CHECK-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
5255 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5256 ; CHECK-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
5257 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5258 ; CHECK-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
5259 ; CHECK-NEXT: movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
5260 ; CHECK-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
5261 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5262 %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
5263 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5264 %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
5265 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5266 %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
5267 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5268 %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
5269 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5270 %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
5271 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5272 %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
5273 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5274 %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
5275 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5276 %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
5277 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5278 ret <8 x i8> %vec7
5279 }
5281 define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
5282 ; X86-LABEL: test_mask_ucmp_w_128:
5283 ; X86: # %bb.0:
5284 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5285 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5286 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5287 ; X86-NEXT: vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
5288 ; X86-NEXT: vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
5289 ; X86-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5290 ; X86-NEXT: vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
5291 ; X86-NEXT: vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
5292 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5293 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5294 ; X86-NEXT: vpinsrb $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x00]
5295 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5296 ; X86-NEXT: vpinsrb $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x01]
5297 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5298 ; X86-NEXT: vpinsrb $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x02]
5299 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5300 ; X86-NEXT: vpinsrb $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x04]
5301 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5302 ; X86-NEXT: vpinsrb $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x05]
5303 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5304 ; X86-NEXT: vpinsrb $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc1,0x06]
5305 ; X86-NEXT: vpinsrb $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x07]
5306 ; X86-NEXT: retl # encoding: [0xc3]
5307 ;
5308 ; X64-LABEL: test_mask_ucmp_w_128:
5309 ; X64: # %bb.0:
5310 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5311 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5312 ; X64-NEXT: vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
5313 ; X64-NEXT: vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
5314 ; X64-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5315 ; X64-NEXT: vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
5316 ; X64-NEXT: vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
5317 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5318 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5319 ; X64-NEXT: vpinsrb $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x00]
5320 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5321 ; X64-NEXT: vpinsrb $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x01]
5322 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5323 ; X64-NEXT: vpinsrb $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x02]
5324 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5325 ; X64-NEXT: vpinsrb $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x04]
5326 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5327 ; X64-NEXT: vpinsrb $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x05]
5328 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5329 ; X64-NEXT: vpinsrb $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc0,0x06]
5330 ; X64-NEXT: vpinsrb $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x20,0xc7,0x07]
5331 ; X64-NEXT: retq # encoding: [0xc3]
5332 %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
5333 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5334 %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
5335 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5336 %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
5337 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5338 %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
5339 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5340 %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
5341 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5342 %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
5343 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5344 %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
5345 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5346 %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
5347 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5348 ret <8 x i8> %vec7
5349 }
5351 declare i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
5353 define <16 x i8>@mm_mask_avg_epu8(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
5354 ; X86-LABEL: mm_mask_avg_epu8:
5355 ; X86: # %bb.0:
5356 ; X86-NEXT: vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
5357 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5358 ; X86-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
5359 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
5360 ; X86-NEXT: retl # encoding: [0xc3]
5361 ;
5362 ; X64-LABEL: mm_mask_avg_epu8:
5363 ; X64: # %bb.0:
5364 ; X64-NEXT: vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
5365 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5366 ; X64-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
5367 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
5368 ; X64-NEXT: retq # encoding: [0xc3]
5369 %res = call <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
5370 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
5371 %res2 = add <16 x i8> %res, %res1
5372 ret <16 x i8> %res2
5373 }
5375 declare <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8>, <16 x i8>, i16)
5377 define <16 x i8>@test_int_x86_avx512_mask_pabs_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5378 ; X86-LABEL: test_int_x86_avx512_mask_pabs_b_128:
5379 ; X86: # %bb.0:
5380 ; X86-NEXT: vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
5381 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5382 ; X86-NEXT: vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
5383 ; X86-NEXT: vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
5384 ; X86-NEXT: retl # encoding: [0xc3]
5385 ;
5386 ; X64-LABEL: test_int_x86_avx512_mask_pabs_b_128:
5387 ; X64: # %bb.0:
5388 ; X64-NEXT: vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
5389 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5390 ; X64-NEXT: vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
5391 ; X64-NEXT: vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
5392 ; X64-NEXT: retq # encoding: [0xc3]
5393 %res = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5394 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
5395 %res2 = add <16 x i8> %res, %res1
5396 ret <16 x i8> %res2
5397 }
5399 declare <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
5401 define <32 x i8>@mm256_mask_avg_epu8(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
5402 ; X86-LABEL: mm256_mask_avg_epu8:
5403 ; X86: # %bb.0:
5404 ; X86-NEXT: vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
5405 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
5406 ; X86-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
5407 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
5408 ; X86-NEXT: retl # encoding: [0xc3]
5409 ;
5410 ; X64-LABEL: mm256_mask_avg_epu8:
5411 ; X64: # %bb.0:
5412 ; X64-NEXT: vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
5413 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5414 ; X64-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
5415 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
5416 ; X64-NEXT: retq # encoding: [0xc3]
5417 %res = call <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
5418 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
5419 %res2 = add <32 x i8> %res, %res1
5420 ret <32 x i8> %res2
5421 }
5423 declare <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8>, <32 x i8>, i32)
5425 define <32 x i8>@test_int_x86_avx512_mask_pabs_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5426 ; X86-LABEL: test_int_x86_avx512_mask_pabs_b_256:
5427 ; X86: # %bb.0:
5428 ; X86-NEXT: vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
5429 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
5430 ; X86-NEXT: vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
5431 ; X86-NEXT: vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
5432 ; X86-NEXT: retl # encoding: [0xc3]
5433 ;
5434 ; X64-LABEL: test_int_x86_avx512_mask_pabs_b_256:
5435 ; X64: # %bb.0:
5436 ; X64-NEXT: vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
5437 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5438 ; X64-NEXT: vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
5439 ; X64-NEXT: vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
5440 ; X64-NEXT: retq # encoding: [0xc3]
5441 %res = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5442 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 -1)
5443 %res2 = add <32 x i8> %res, %res1
5444 ret <32 x i8> %res2
5445 }
5447 declare <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
5449 define <8 x i16>@mm_mask_avg_epu16(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
5450 ; X86-LABEL: mm_mask_avg_epu16:
5451 ; X86: # %bb.0:
5452 ; X86-NEXT: vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
5453 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5454 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5455 ; X86-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
5456 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5457 ; X86-NEXT: retl # encoding: [0xc3]
5458 ;
5459 ; X64-LABEL: mm_mask_avg_epu16:
5460 ; X64: # %bb.0:
5461 ; X64-NEXT: vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
5462 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5463 ; X64-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
5464 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5465 ; X64-NEXT: retq # encoding: [0xc3]
5466 %res = call <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
5467 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
5468 %res2 = add <8 x i16> %res, %res1
5469 ret <8 x i16> %res2
5470 }
5472 declare <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16>, <8 x i16>, i8)
5474 define <8 x i16>@test_int_x86_avx512_mask_pabs_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5475 ; X86-LABEL: test_int_x86_avx512_mask_pabs_w_128:
5476 ; X86: # %bb.0:
5477 ; X86-NEXT: vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
5478 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5479 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5480 ; X86-NEXT: vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
5481 ; X86-NEXT: vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
5482 ; X86-NEXT: retl # encoding: [0xc3]
5483 ;
5484 ; X64-LABEL: test_int_x86_avx512_mask_pabs_w_128:
5485 ; X64: # %bb.0:
5486 ; X64-NEXT: vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
5487 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5488 ; X64-NEXT: vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
5489 ; X64-NEXT: vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
5490 ; X64-NEXT: retq # encoding: [0xc3]
5491 %res = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5492 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
5493 %res2 = add <8 x i16> %res, %res1
5494 ret <8 x i16> %res2
5495 }
5497 declare <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5499 define <16 x i16>@mm256_mask_avg_epu16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
5500 ; X86-LABEL: mm256_mask_avg_epu16:
5501 ; X86: # %bb.0:
5502 ; X86-NEXT: vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
5503 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5504 ; X86-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
5505 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5506 ; X86-NEXT: retl # encoding: [0xc3]
5507 ;
5508 ; X64-LABEL: mm256_mask_avg_epu16:
5509 ; X64: # %bb.0:
5510 ; X64-NEXT: vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
5511 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5512 ; X64-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
5513 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5514 ; X64-NEXT: retq # encoding: [0xc3]
5515 %res = call <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
5516 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
5517 %res2 = add <16 x i16> %res, %res1
5518 ret <16 x i16> %res2
5519 }
5521 declare <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16>, <16 x i16>, i16)
5523 define <16 x i16>@test_int_x86_avx512_mask_pabs_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5524 ; X86-LABEL: test_int_x86_avx512_mask_pabs_w_256:
5525 ; X86: # %bb.0:
5526 ; X86-NEXT: vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
5527 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5528 ; X86-NEXT: vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
5529 ; X86-NEXT: vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
5530 ; X86-NEXT: retl # encoding: [0xc3]
5531 ;
5532 ; X64-LABEL: test_int_x86_avx512_mask_pabs_w_256:
5533 ; X64: # %bb.0:
5534 ; X64-NEXT: vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
5535 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5536 ; X64-NEXT: vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
5537 ; X64-NEXT: vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
5538 ; X64-NEXT: retq # encoding: [0xc3]
5539 %res = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5540 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 -1)
5541 %res2 = add <16 x i16> %res, %res1
5542 ret <16 x i16> %res2
5543 }
5545 declare <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5547 declare i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8>, <16 x i8>, i16)
5549 define i16@test_int_x86_avx512_ptestm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5550 ; X86-LABEL: test_int_x86_avx512_ptestm_b_128:
5551 ; X86: # %bb.0:
5552 ; X86-NEXT: vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
5553 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5554 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5555 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5556 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5557 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5558 ; X86-NEXT: retl # encoding: [0xc3]
5559 ;
5560 ; X64-LABEL: test_int_x86_avx512_ptestm_b_128:
5561 ; X64: # %bb.0:
5562 ; X64-NEXT: vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
5563 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5564 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5565 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5566 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5567 ; X64-NEXT: retq # encoding: [0xc3]
5568 %res = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5569 %res1 = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
5570 %res2 = add i16 %res, %res1
5571 ret i16 %res2
5572 }
5574 declare i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8>, <32 x i8>, i32)
5576 define i32@test_int_x86_avx512_ptestm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5577 ; X86-LABEL: test_int_x86_avx512_ptestm_b_256:
5578 ; X86: # %bb.0:
5579 ; X86-NEXT: vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
5580 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5581 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5582 ; X86-NEXT: andl %ecx, %eax # encoding: [0x21,0xc8]
5583 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5584 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5585 ; X86-NEXT: retl # encoding: [0xc3]
5586 ;
5587 ; X64-LABEL: test_int_x86_avx512_ptestm_b_256:
5588 ; X64: # %bb.0:
5589 ; X64-NEXT: vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
5590 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5591 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5592 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5593 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5594 ; X64-NEXT: retq # encoding: [0xc3]
5595 %res = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5596 %res1 = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
5597 %res2 = add i32 %res, %res1
5598 ret i32 %res2
5599 }
5601 declare i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16>, <8 x i16>, i8)
5603 define i8@test_int_x86_avx512_ptestm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5604 ; X86-LABEL: test_int_x86_avx512_ptestm_w_128:
5605 ; X86: # %bb.0:
5606 ; X86-NEXT: vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
5607 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5608 ; X86-NEXT: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
5609 ; X86-NEXT: andb %cl, %al # encoding: [0x20,0xc8]
5610 ; X86-NEXT: addb %cl, %al # encoding: [0x00,0xc8]
5611 ; X86-NEXT: retl # encoding: [0xc3]
5612 ;
5613 ; X64-LABEL: test_int_x86_avx512_ptestm_w_128:
5614 ; X64: # %bb.0:
5615 ; X64-NEXT: vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
5616 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5617 ; X64-NEXT: andb %al, %dil # encoding: [0x40,0x20,0xc7]
5618 ; X64-NEXT: addb %dil, %al # encoding: [0x40,0x00,0xf8]
5619 ; X64-NEXT: # kill: def $al killed $al killed $eax
5620 ; X64-NEXT: retq # encoding: [0xc3]
5621 %res = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5622 %res1 = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
5623 %res2 = add i8 %res, %res1
5624 ret i8 %res2
5625 }
5627 declare i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16>, <16 x i16>, i16)
5629 define i16@test_int_x86_avx512_ptestm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5630 ; X86-LABEL: test_int_x86_avx512_ptestm_w_256:
5631 ; X86: # %bb.0:
5632 ; X86-NEXT: vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
5633 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5634 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5635 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5636 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5637 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5638 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5639 ; X86-NEXT: retl # encoding: [0xc3]
5640 ;
5641 ; X64-LABEL: test_int_x86_avx512_ptestm_w_256:
5642 ; X64: # %bb.0:
5643 ; X64-NEXT: vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
5644 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5645 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5646 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5647 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5648 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5649 ; X64-NEXT: retq # encoding: [0xc3]
5650 %res = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5651 %res1 = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
5652 %res2 = add i16 %res, %res1
5653 ret i16 %res2
5654 }
5656 declare i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8>, <16 x i8>, i16)
5658 define i16@test_int_x86_avx512_ptestnm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5659 ; X86-LABEL: test_int_x86_avx512_ptestnm_b_128:
5660 ; X86: # %bb.0:
5661 ; X86-NEXT: vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
5662 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5663 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5664 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5665 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5666 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5667 ; X86-NEXT: retl # encoding: [0xc3]
5668 ;
5669 ; X64-LABEL: test_int_x86_avx512_ptestnm_b_128:
5670 ; X64: # %bb.0:
5671 ; X64-NEXT: vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
5672 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5673 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5674 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5675 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5676 ; X64-NEXT: retq # encoding: [0xc3]
5677 %res = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5678 %res1 = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
5679 %res2 = add i16 %res, %res1
5680 ret i16 %res2
5681 }
5683 declare i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8>, <32 x i8>, i32)
5685 define i32@test_int_x86_avx512_ptestnm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5686 ; X86-LABEL: test_int_x86_avx512_ptestnm_b_256:
5687 ; X86: # %bb.0:
5688 ; X86-NEXT: vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
5689 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5690 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5691 ; X86-NEXT: andl %ecx, %eax # encoding: [0x21,0xc8]
5692 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5693 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5694 ; X86-NEXT: retl # encoding: [0xc3]
5695 ;
5696 ; X64-LABEL: test_int_x86_avx512_ptestnm_b_256:
5697 ; X64: # %bb.0:
5698 ; X64-NEXT: vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
5699 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5700 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5701 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5702 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5703 ; X64-NEXT: retq # encoding: [0xc3]
5704 %res = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5705 %res1 = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
5706 %res2 = add i32 %res, %res1
5707 ret i32 %res2
5708 }
5710 declare i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16>, <8 x i16>, i8 %x2)
5712 define i8@test_int_x86_avx512_ptestnm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5713 ; X86-LABEL: test_int_x86_avx512_ptestnm_w_128:
5714 ; X86: # %bb.0:
5715 ; X86-NEXT: vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
5716 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5717 ; X86-NEXT: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
5718 ; X86-NEXT: andb %cl, %al # encoding: [0x20,0xc8]
5719 ; X86-NEXT: addb %cl, %al # encoding: [0x00,0xc8]
5720 ; X86-NEXT: retl # encoding: [0xc3]
5721 ;
5722 ; X64-LABEL: test_int_x86_avx512_ptestnm_w_128:
5723 ; X64: # %bb.0:
5724 ; X64-NEXT: vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
5725 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5726 ; X64-NEXT: andb %al, %dil # encoding: [0x40,0x20,0xc7]
5727 ; X64-NEXT: addb %dil, %al # encoding: [0x40,0x00,0xf8]
5728 ; X64-NEXT: # kill: def $al killed $al killed $eax
5729 ; X64-NEXT: retq # encoding: [0xc3]
5730 %res = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5731 %res1 = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
5732 %res2 = add i8 %res, %res1
5733 ret i8 %res2
5734 }
5736 declare i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16>, <16 x i16>, i16 %x2)
5738 define i16@test_int_x86_avx512_ptestnm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5739 ; X86-LABEL: test_int_x86_avx512_ptestnm_w_256:
5740 ; X86: # %bb.0:
5741 ; X86-NEXT: vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
5742 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5743 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5744 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5745 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5746 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5747 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5748 ; X86-NEXT: retl # encoding: [0xc3]
5749 ;
5750 ; X64-LABEL: test_int_x86_avx512_ptestnm_w_256:
5751 ; X64: # %bb.0:
5752 ; X64-NEXT: vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
5753 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5754 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5755 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5756 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5757 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5758 ; X64-NEXT: retq # encoding: [0xc3]
5759 %res = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5760 %res1 = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
5761 %res2 = add i16 %res, %res1
5762 ret i16 %res2
5763 }
5765 declare i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8>)
5767 define i16@test_int_x86_avx512_cvtb2mask_128(<16 x i8> %x0) {
5768 ; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_128:
5769 ; CHECK: # %bb.0:
5770 ; CHECK-NEXT: vpmovb2m %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x29,0xc0]
5771 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5772 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
5773 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5774 %res = call i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8> %x0)
5775 ret i16 %res
5776 }
5778 declare i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8>)
5780 define i32@test_int_x86_avx512_cvtb2mask_256(<32 x i8> %x0) {
5781 ; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_256:
5782 ; CHECK: # %bb.0:
5783 ; CHECK-NEXT: vpmovb2m %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x29,0xc0]
5784 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5785 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5786 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5787 %res = call i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8> %x0)
5788 ret i32 %res
5789 }
5791 declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>)
5793 define i8@test_int_x86_avx512_cvtw2mask_128(<8 x i16> %x0) {
5794 ; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_128:
5795 ; CHECK: # %bb.0:
5796 ; CHECK-NEXT: vpmovw2m %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x29,0xc0]
5797 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5798 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
5799 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5800 %res = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %x0)
5801 ret i8 %res
5802 }
5804 declare i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16>)
5806 define i16@test_int_x86_avx512_cvtw2mask_256(<16 x i16> %x0) {
5807 ; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_256:
5808 ; CHECK: # %bb.0:
5809 ; CHECK-NEXT: vpmovw2m %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x29,0xc0]
5810 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5811 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
5812 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5813 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5814 %res = call i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16> %x0)
5815 ret i16 %res
5816 }
5818 declare <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5820 define <8 x i16>@test_int_x86_avx512_mask_pmulhu_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
5821 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
5822 ; X86: # %bb.0:
5823 ; X86-NEXT: vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
5824 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5825 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5826 ; X86-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
5827 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5828 ; X86-NEXT: retl # encoding: [0xc3]
5829 ;
5830 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
5831 ; X64: # %bb.0:
5832 ; X64-NEXT: vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
5833 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5834 ; X64-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
5835 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5836 ; X64-NEXT: retq # encoding: [0xc3]
5837 %res = call <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
5838 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
5839 %res2 = add <8 x i16> %res, %res1
5840 ret <8 x i16> %res2
5841 }
5843 declare <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5845 define <16 x i16>@test_int_x86_avx512_mask_pmulhu_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
5846 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
5847 ; X86: # %bb.0:
5848 ; X86-NEXT: vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
5849 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5850 ; X86-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
5851 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5852 ; X86-NEXT: retl # encoding: [0xc3]
5853 ;
5854 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
5855 ; X64: # %bb.0:
5856 ; X64-NEXT: vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
5857 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5858 ; X64-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
5859 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5860 ; X64-NEXT: retq # encoding: [0xc3]
5861 %res = call <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
5862 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
5863 %res2 = add <16 x i16> %res, %res1
5864 ret <16 x i16> %res2
5865 }
5867 declare <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5869 define <8 x i16>@test_int_x86_avx512_mask_pmulh_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
5870 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
5871 ; X86: # %bb.0:
5872 ; X86-NEXT: vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
5873 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5874 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5875 ; X86-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
5876 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5877 ; X86-NEXT: retl # encoding: [0xc3]
5878 ;
5879 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
5880 ; X64: # %bb.0:
5881 ; X64-NEXT: vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
5882 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5883 ; X64-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
5884 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5885 ; X64-NEXT: retq # encoding: [0xc3]
5886 %res = call <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
5887 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
5888 %res2 = add <8 x i16> %res, %res1
5889 ret <8 x i16> %res2
5890 }
5892 declare <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5894 define <16 x i16>@test_int_x86_avx512_mask_pmulh_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
5895 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
5896 ; X86: # %bb.0:
5897 ; X86-NEXT: vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
5898 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5899 ; X86-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
5900 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5901 ; X86-NEXT: retl # encoding: [0xc3]
5902 ;
5903 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
5904 ; X64: # %bb.0:
5905 ; X64-NEXT: vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
5906 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5907 ; X64-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
5908 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5909 ; X64-NEXT: retq # encoding: [0xc3]
5910 %res = call <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
5911 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
5912 %res2 = add <16 x i16> %res, %res1
5913 ret <16 x i16> %res2
5914 }
5916 declare <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5918 define <8 x i16>@test_int_x86_avx512_mask_pmulhr_sw_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
5919 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
5920 ; X86: # %bb.0:
5921 ; X86-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
5922 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5923 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5924 ; X86-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
5925 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5926 ; X86-NEXT: retl # encoding: [0xc3]
5927 ;
5928 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
5929 ; X64: # %bb.0:
5930 ; X64-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
5931 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5932 ; X64-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
5933 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5934 ; X64-NEXT: retq # encoding: [0xc3]
5935 %res = call <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
5936 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
5937 %res2 = add <8 x i16> %res, %res1
5938 ret <8 x i16> %res2
5939 }
5941 declare <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5943 define <16 x i16>@test_int_x86_avx512_mask_pmulhr_sw_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
5944 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
5945 ; X86: # %bb.0:
5946 ; X86-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
5947 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5948 ; X86-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
5949 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5950 ; X86-NEXT: retl # encoding: [0xc3]
5951 ;
5952 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
5953 ; X64: # %bb.0:
5954 ; X64-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
5955 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5956 ; X64-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
5957 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5958 ; X64-NEXT: retq # encoding: [0xc3]
5959 %res = call <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
5960 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
5961 %res2 = add <16 x i16> %res, %res1
5962 ret <16 x i16> %res2
5963 }
5965 declare <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8>, <16 x i8>, <8 x i16>, i8)
5967 define <8 x i16>@test_int_x86_avx512_mask_pmaddubs_w_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 %x3) {
5968 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
5969 ; X86: # %bb.0:
5970 ; X86-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
5971 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5972 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5973 ; X86-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
5974 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5975 ; X86-NEXT: retl # encoding: [0xc3]
5976 ;
5977 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
5978 ; X64: # %bb.0:
5979 ; X64-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
5980 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5981 ; X64-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
5982 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5983 ; X64-NEXT: retq # encoding: [0xc3]
5984 %res = call <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 %x3)
5985 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x2, i8 -1)
5986 %res2 = add <8 x i16> %res, %res1
5987 ret <8 x i16> %res2
5988 }
5990 declare <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8>, <32 x i8>, <16 x i16>, i16)
5992 define <16 x i16>@test_int_x86_avx512_mask_pmaddubs_w_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 %x3) {
5993 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
5994 ; X86: # %bb.0:
5995 ; X86-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
5996 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5997 ; X86-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
5998 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5999 ; X86-NEXT: retl # encoding: [0xc3]
6000 ;
6001 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
6002 ; X64: # %bb.0:
6003 ; X64-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
6004 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6005 ; X64-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
6006 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
6007 ; X64-NEXT: retq # encoding: [0xc3]
6008 %res = call <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 %x3)
6009 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x2, i16 -1)
6010 %res2 = add <16 x i16> %res, %res1
6011 ret <16 x i16> %res2
6012 }
6014 declare <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16>, <8 x i16>, <4 x i32>, i8)
6016 define <4 x i32>@test_int_x86_avx512_mask_pmaddw_d_128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 %x3) {
6017 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
6018 ; X86: # %bb.0:
6019 ; X86-NEXT: vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
6020 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6021 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6022 ; X86-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
6023 ; X86-NEXT: vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
6024 ; X86-NEXT: retl # encoding: [0xc3]
6025 ;
6026 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
6027 ; X64: # %bb.0:
6028 ; X64-NEXT: vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
6029 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6030 ; X64-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
6031 ; X64-NEXT: vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
6032 ; X64-NEXT: retq # encoding: [0xc3]
6033 %res = call <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 %x3)
6034 %res1 = call <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16> %x0, <8 x i16> %x1, <4 x i32> %x2, i8 -1)
6035 %res2 = add <4 x i32> %res, %res1
6036 ret <4 x i32> %res2
6037 }
6039 declare <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16>, <16 x i16>, <8 x i32>, i8)
6041 define <8 x i32>@test_int_x86_avx512_mask_pmaddw_d_256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 %x3) {
6042 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
6043 ; X86: # %bb.0:
6044 ; X86-NEXT: vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
6045 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6046 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6047 ; X86-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
6048 ; X86-NEXT: vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
6049 ; X86-NEXT: retl # encoding: [0xc3]
6050 ;
6051 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
6052 ; X64: # %bb.0:
6053 ; X64-NEXT: vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
6054 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6055 ; X64-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
6056 ; X64-NEXT: vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
6057 ; X64-NEXT: retq # encoding: [0xc3]
6058 %res = call <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 %x3)
6059 %res1 = call <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16> %x0, <16 x i16> %x1, <8 x i32> %x2, i8 -1)
6060 %res2 = add <8 x i32> %res, %res1
6061 ret <8 x i32> %res2
6062 }
6064 declare <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6066 define <8 x i16>@test_int_x86_avx512_mask_permvar_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
6067 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
6068 ; X86: # %bb.0:
6069 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
6070 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6071 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6072 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
6073 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
6074 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6075 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6076 ; X86-NEXT: retl # encoding: [0xc3]
6077 ;
6078 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
6079 ; X64: # %bb.0:
6080 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
6081 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6082 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
6083 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
6084 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6085 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6086 ; X64-NEXT: retq # encoding: [0xc3]
6087 %res = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
6088 %res1 = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
6089 %res2 = call <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
6090 %res3 = add <8 x i16> %res, %res1
6091 %res4 = add <8 x i16> %res3, %res2
6092 ret <8 x i16> %res4
6093 }
6095 declare <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6097 define <16 x i16>@test_int_x86_avx512_mask_permvar_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
6098 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
6099 ; X86: # %bb.0:
6100 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
6101 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6102 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
6103 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
6104 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6105 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6106 ; X86-NEXT: retl # encoding: [0xc3]
6107 ;
6108 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
6109 ; X64: # %bb.0:
6110 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
6111 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6112 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
6113 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
6114 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6115 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6116 ; X64-NEXT: retq # encoding: [0xc3]
6117 %res = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
6118 %res1 = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
6119 %res2 = call <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
6120 %res3 = add <16 x i16> %res, %res1
6121 %res4 = add <16 x i16> %res3, %res2
6122 ret <16 x i16> %res4
6123 }
6125 declare <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6127 define <8 x i16>@test_int_x86_avx512_mask_vpermt2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
6128 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
6129 ; X86: # %bb.0:
6130 ; X86-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6131 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6132 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6133 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6134 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
6135 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6136 ; X86-NEXT: retl # encoding: [0xc3]
6137 ;
6138 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
6139 ; X64: # %bb.0:
6140 ; X64-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6141 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6142 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6143 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
6144 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6145 ; X64-NEXT: retq # encoding: [0xc3]
6146 %res = call <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
6147 %res1 = call <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
6148 %res2 = add <8 x i16> %res, %res1
6149 ret <8 x i16> %res2
6150 }
6152 declare <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6154 define <8 x i16>@test_int_x86_avx512_maskz_vpermt2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
6155 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
6156 ; X86: # %bb.0:
6157 ; X86-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6158 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6159 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6160 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6161 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
6162 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6163 ; X86-NEXT: retl # encoding: [0xc3]
6164 ;
6165 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
6166 ; X64: # %bb.0:
6167 ; X64-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6168 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6169 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6170 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
6171 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6172 ; X64-NEXT: retq # encoding: [0xc3]
6173 %res = call <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
6174 %res1 = call <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
6175 %res2 = add <8 x i16> %res, %res1
6176 ret <8 x i16> %res2
6177 }
6179 declare <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6181 define <16 x i16>@test_int_x86_avx512_mask_vpermt2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
6182 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
6183 ; X86: # %bb.0:
6184 ; X86-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6185 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6186 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6187 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
6188 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6189 ; X86-NEXT: retl # encoding: [0xc3]
6190 ;
6191 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
6192 ; X64: # %bb.0:
6193 ; X64-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6194 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6195 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6196 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
6197 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6198 ; X64-NEXT: retq # encoding: [0xc3]
6199 %res = call <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
6200 %res1 = call <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
6201 %res2 = add <16 x i16> %res, %res1
6202 ret <16 x i16> %res2
6203 }
6205 declare <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6207 define <16 x i16>@test_int_x86_avx512_maskz_vpermt2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
6208 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
6209 ; X86: # %bb.0:
6210 ; X86-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6211 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6212 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6213 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
6214 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6215 ; X86-NEXT: retl # encoding: [0xc3]
6216 ;
6217 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
6218 ; X64: # %bb.0:
6219 ; X64-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6220 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6221 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6222 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
6223 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6224 ; X64-NEXT: retq # encoding: [0xc3]
6225 %res = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
6226 %res1 = call <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
6227 %res2 = add <16 x i16> %res, %res1
6228 ret <16 x i16> %res2
6229 }
6231 declare <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6233 define <8 x i16>@test_int_x86_avx512_mask_vpermi2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
6234 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
6235 ; X86: # %bb.0:
6236 ; X86-NEXT: vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
6237 ; X86-NEXT: vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
6238 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6239 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6240 ; X86-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
6241 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6242 ; X86-NEXT: retl # encoding: [0xc3]
6243 ;
6244 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
6245 ; X64: # %bb.0:
6246 ; X64-NEXT: vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
6247 ; X64-NEXT: vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
6248 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6249 ; X64-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
6250 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6251 ; X64-NEXT: retq # encoding: [0xc3]
6252 %res = call <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
6253 %res1 = call <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
6254 %res2 = add <8 x i16> %res, %res1
6255 ret <8 x i16> %res2
6256 }
6258 declare <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6260 define <16 x i16>@test_int_x86_avx512_mask_vpermi2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
6261 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
6262 ; X86: # %bb.0:
6263 ; X86-NEXT: vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
6264 ; X86-NEXT: vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
6265 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6266 ; X86-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
6267 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6268 ; X86-NEXT: retl # encoding: [0xc3]
6269 ;
6270 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
6271 ; X64: # %bb.0:
6272 ; X64-NEXT: vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
6273 ; X64-NEXT: vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
6274 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6275 ; X64-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
6276 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6277 ; X64-NEXT: retq # encoding: [0xc3]
6278 %res = call <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
6279 %res1 = call <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
6280 %res2 = add <16 x i16> %res, %res1
6281 ret <16 x i16> %res2
6282 }
6284 declare <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8>, <16 x i8>, i32, <8 x i16>, i8)
6286 define <8 x i16>@test_int_x86_avx512_mask_dbpsadbw_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x3, i8 %x4) {
6287 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
6288 ; X86: # %bb.0:
6289 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
6290 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6291 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6292 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
6293 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
6294 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6295 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6296 ; X86-NEXT: retl # encoding: [0xc3]
6297 ;
6298 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
6299 ; X64: # %bb.0:
6300 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
6301 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6302 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
6303 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
6304 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6305 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6306 ; X64-NEXT: retq # encoding: [0xc3]
6307 %res = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> %x3, i8 %x4)
6308 %res1 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> zeroinitializer, i8 %x4)
6309 %res2 = call <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2, <8 x i16> %x3, i8 -1)
6310 %res3 = add <8 x i16> %res, %res1
6311 %res4 = add <8 x i16> %res2, %res3
6312 ret <8 x i16> %res4
6313 }
6315 declare <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8>, <32 x i8>, i32, <16 x i16>, i16)
6317 define <16 x i16>@test_int_x86_avx512_mask_dbpsadbw_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x3, i16 %x4) {
6318 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
6319 ; X86: # %bb.0:
6320 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
6321 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6322 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
6323 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
6324 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6325 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6326 ; X86-NEXT: retl # encoding: [0xc3]
6327 ;
6328 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
6329 ; X64: # %bb.0:
6330 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
6331 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6332 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
6333 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
6334 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6335 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6336 ; X64-NEXT: retq # encoding: [0xc3]
6337 %res = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> %x3, i16 %x4)
6338 %res1 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> zeroinitializer, i16 %x4)
6339 %res2 = call <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2, <16 x i16> %x3, i16 -1)
6340 %res3 = add <16 x i16> %res, %res1
6341 %res4 = add <16 x i16> %res3, %res2
6342 ret <16 x i16> %res4
6343 }
6345 define <8 x i16> @test_mask_adds_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
6346 ; CHECK-LABEL: test_mask_adds_epu16_rr_128:
6347 ; CHECK: # %bb.0:
6348 ; CHECK-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0xc1]
6349 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6350 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
6351 ret <8 x i16> %res
6352 }
6354 define <8 x i16> @test_mask_adds_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
6355 ; X86-LABEL: test_mask_adds_epu16_rrk_128:
6356 ; X86: # %bb.0:
6357 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6358 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6359 ; X86-NEXT: vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
6360 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6361 ; X86-NEXT: retl # encoding: [0xc3]
6362 ;
6363 ; X64-LABEL: test_mask_adds_epu16_rrk_128:
6364 ; X64: # %bb.0:
6365 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6366 ; X64-NEXT: vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
6367 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6368 ; X64-NEXT: retq # encoding: [0xc3]
6369 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
6370 ret <8 x i16> %res
6371 }
6373 define <8 x i16> @test_mask_adds_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
6374 ; X86-LABEL: test_mask_adds_epu16_rrkz_128:
6375 ; X86: # %bb.0:
6376 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6377 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6378 ; X86-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
6379 ; X86-NEXT: retl # encoding: [0xc3]
6380 ;
6381 ; X64-LABEL: test_mask_adds_epu16_rrkz_128:
6382 ; X64: # %bb.0:
6383 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6384 ; X64-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
6385 ; X64-NEXT: retq # encoding: [0xc3]
6386 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
6387 ret <8 x i16> %res
6388 }
6390 define <8 x i16> @test_mask_adds_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
6391 ; X86-LABEL: test_mask_adds_epu16_rm_128:
6392 ; X86: # %bb.0:
6393 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6394 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x00]
6395 ; X86-NEXT: retl # encoding: [0xc3]
6396 ;
6397 ; X64-LABEL: test_mask_adds_epu16_rm_128:
6398 ; X64: # %bb.0:
6399 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x07]
6400 ; X64-NEXT: retq # encoding: [0xc3]
6401 %b = load <8 x i16>, <8 x i16>* %ptr_b
6402 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
6403 ret <8 x i16> %res
6404 }
6406 define <8 x i16> @test_mask_adds_epu16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
6407 ; X86-LABEL: test_mask_adds_epu16_rmk_128:
6408 ; X86: # %bb.0:
6409 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6410 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6411 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6412 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x08]
6413 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6414 ; X86-NEXT: retl # encoding: [0xc3]
6415 ;
6416 ; X64-LABEL: test_mask_adds_epu16_rmk_128:
6417 ; X64: # %bb.0:
6418 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6419 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x0f]
6420 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6421 ; X64-NEXT: retq # encoding: [0xc3]
6422 %b = load <8 x i16>, <8 x i16>* %ptr_b
6423 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
6424 ret <8 x i16> %res
6425 }
6427 define <8 x i16> @test_mask_adds_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
6428 ; X86-LABEL: test_mask_adds_epu16_rmkz_128:
6429 ; X86: # %bb.0:
6430 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6431 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6432 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6433 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x00]
6434 ; X86-NEXT: retl # encoding: [0xc3]
6435 ;
6436 ; X64-LABEL: test_mask_adds_epu16_rmkz_128:
6437 ; X64: # %bb.0:
6438 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6439 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x07]
6440 ; X64-NEXT: retq # encoding: [0xc3]
6441 %b = load <8 x i16>, <8 x i16>* %ptr_b
6442 %res = call <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
6443 ret <8 x i16> %res
6444 }
6446 declare <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6448 define <16 x i16> @test_mask_adds_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
6449 ; CHECK-LABEL: test_mask_adds_epu16_rr_256:
6450 ; CHECK: # %bb.0:
6451 ; CHECK-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0xc1]
6452 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6453 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
6454 ret <16 x i16> %res
6455 }
6457 define <16 x i16> @test_mask_adds_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
6458 ; X86-LABEL: test_mask_adds_epu16_rrk_256:
6459 ; X86: # %bb.0:
6460 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6461 ; X86-NEXT: vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
6462 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6463 ; X86-NEXT: retl # encoding: [0xc3]
6464 ;
6465 ; X64-LABEL: test_mask_adds_epu16_rrk_256:
6466 ; X64: # %bb.0:
6467 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6468 ; X64-NEXT: vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
6469 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6470 ; X64-NEXT: retq # encoding: [0xc3]
6471 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
6472 ret <16 x i16> %res
6473 }
6475 define <16 x i16> @test_mask_adds_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
6476 ; X86-LABEL: test_mask_adds_epu16_rrkz_256:
6477 ; X86: # %bb.0:
6478 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6479 ; X86-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
6480 ; X86-NEXT: retl # encoding: [0xc3]
6481 ;
6482 ; X64-LABEL: test_mask_adds_epu16_rrkz_256:
6483 ; X64: # %bb.0:
6484 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6485 ; X64-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
6486 ; X64-NEXT: retq # encoding: [0xc3]
6487 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
6488 ret <16 x i16> %res
6489 }
6491 define <16 x i16> @test_mask_adds_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
6492 ; X86-LABEL: test_mask_adds_epu16_rm_256:
6493 ; X86: # %bb.0:
6494 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6495 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x00]
6496 ; X86-NEXT: retl # encoding: [0xc3]
6497 ;
6498 ; X64-LABEL: test_mask_adds_epu16_rm_256:
6499 ; X64: # %bb.0:
6500 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x07]
6501 ; X64-NEXT: retq # encoding: [0xc3]
6502 %b = load <16 x i16>, <16 x i16>* %ptr_b
6503 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
6504 ret <16 x i16> %res
6505 }
6507 define <16 x i16> @test_mask_adds_epu16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
6508 ; X86-LABEL: test_mask_adds_epu16_rmk_256:
6509 ; X86: # %bb.0:
6510 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6511 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6512 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x08]
6513 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6514 ; X86-NEXT: retl # encoding: [0xc3]
6515 ;
6516 ; X64-LABEL: test_mask_adds_epu16_rmk_256:
6517 ; X64: # %bb.0:
6518 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6519 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x0f]
6520 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6521 ; X64-NEXT: retq # encoding: [0xc3]
6522 %b = load <16 x i16>, <16 x i16>* %ptr_b
6523 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
6524 ret <16 x i16> %res
6525 }
6527 define <16 x i16> @test_mask_adds_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
6528 ; X86-LABEL: test_mask_adds_epu16_rmkz_256:
6529 ; X86: # %bb.0:
6530 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6531 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6532 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x00]
6533 ; X86-NEXT: retl # encoding: [0xc3]
6534 ;
6535 ; X64-LABEL: test_mask_adds_epu16_rmkz_256:
6536 ; X64: # %bb.0:
6537 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6538 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x07]
6539 ; X64-NEXT: retq # encoding: [0xc3]
6540 %b = load <16 x i16>, <16 x i16>* %ptr_b
6541 %res = call <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
6542 ret <16 x i16> %res
6543 }
6545 declare <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6547 define <8 x i16> @test_mask_subs_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
6548 ; CHECK-LABEL: test_mask_subs_epu16_rr_128:
6549 ; CHECK: # %bb.0:
6550 ; CHECK-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0xc1]
6551 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6552 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
6553 ret <8 x i16> %res
6554 }
6556 define <8 x i16> @test_mask_subs_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
6557 ; X86-LABEL: test_mask_subs_epu16_rrk_128:
6558 ; X86: # %bb.0:
6559 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6560 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6561 ; X86-NEXT: vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
6562 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6563 ; X86-NEXT: retl # encoding: [0xc3]
6564 ;
6565 ; X64-LABEL: test_mask_subs_epu16_rrk_128:
6566 ; X64: # %bb.0:
6567 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6568 ; X64-NEXT: vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
6569 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6570 ; X64-NEXT: retq # encoding: [0xc3]
6571 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
6572 ret <8 x i16> %res
6573 }
6575 define <8 x i16> @test_mask_subs_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
6576 ; X86-LABEL: test_mask_subs_epu16_rrkz_128:
6577 ; X86: # %bb.0:
6578 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6579 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6580 ; X86-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
6581 ; X86-NEXT: retl # encoding: [0xc3]
6582 ;
6583 ; X64-LABEL: test_mask_subs_epu16_rrkz_128:
6584 ; X64: # %bb.0:
6585 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6586 ; X64-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
6587 ; X64-NEXT: retq # encoding: [0xc3]
6588 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
6589 ret <8 x i16> %res
6590 }
6592 define <8 x i16> @test_mask_subs_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
6593 ; X86-LABEL: test_mask_subs_epu16_rm_128:
6594 ; X86: # %bb.0:
6595 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6596 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x00]
6597 ; X86-NEXT: retl # encoding: [0xc3]
6598 ;
6599 ; X64-LABEL: test_mask_subs_epu16_rm_128:
6600 ; X64: # %bb.0:
6601 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x07]
6602 ; X64-NEXT: retq # encoding: [0xc3]
6603 %b = load <8 x i16>, <8 x i16>* %ptr_b
6604 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
6605 ret <8 x i16> %res
6606 }
6608 define <8 x i16> @test_mask_subs_epu16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
6609 ; X86-LABEL: test_mask_subs_epu16_rmk_128:
6610 ; X86: # %bb.0:
6611 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6612 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6613 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6614 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x08]
6615 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6616 ; X86-NEXT: retl # encoding: [0xc3]
6617 ;
6618 ; X64-LABEL: test_mask_subs_epu16_rmk_128:
6619 ; X64: # %bb.0:
6620 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6621 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x0f]
6622 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6623 ; X64-NEXT: retq # encoding: [0xc3]
6624 %b = load <8 x i16>, <8 x i16>* %ptr_b
6625 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
6626 ret <8 x i16> %res
6627 }
6629 define <8 x i16> @test_mask_subs_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
6630 ; X86-LABEL: test_mask_subs_epu16_rmkz_128:
6631 ; X86: # %bb.0:
6632 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6633 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6634 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6635 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x00]
6636 ; X86-NEXT: retl # encoding: [0xc3]
6637 ;
6638 ; X64-LABEL: test_mask_subs_epu16_rmkz_128:
6639 ; X64: # %bb.0:
6640 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6641 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x07]
6642 ; X64-NEXT: retq # encoding: [0xc3]
6643 %b = load <8 x i16>, <8 x i16>* %ptr_b
6644 %res = call <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
6645 ret <8 x i16> %res
6646 }
6648 declare <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6650 define <16 x i16> @test_mask_subs_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
6651 ; CHECK-LABEL: test_mask_subs_epu16_rr_256:
6652 ; CHECK: # %bb.0:
6653 ; CHECK-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0xc1]
6654 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6655 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
6656 ret <16 x i16> %res
6657 }
6659 define <16 x i16> @test_mask_subs_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
6660 ; X86-LABEL: test_mask_subs_epu16_rrk_256:
6661 ; X86: # %bb.0:
6662 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6663 ; X86-NEXT: vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
6664 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6665 ; X86-NEXT: retl # encoding: [0xc3]
6666 ;
6667 ; X64-LABEL: test_mask_subs_epu16_rrk_256:
6668 ; X64: # %bb.0:
6669 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6670 ; X64-NEXT: vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
6671 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6672 ; X64-NEXT: retq # encoding: [0xc3]
6673 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
6674 ret <16 x i16> %res
6675 }
6677 define <16 x i16> @test_mask_subs_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
6678 ; X86-LABEL: test_mask_subs_epu16_rrkz_256:
6679 ; X86: # %bb.0:
6680 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6681 ; X86-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
6682 ; X86-NEXT: retl # encoding: [0xc3]
6683 ;
6684 ; X64-LABEL: test_mask_subs_epu16_rrkz_256:
6685 ; X64: # %bb.0:
6686 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6687 ; X64-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
6688 ; X64-NEXT: retq # encoding: [0xc3]
6689 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
6690 ret <16 x i16> %res
6691 }
6693 define <16 x i16> @test_mask_subs_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
6694 ; X86-LABEL: test_mask_subs_epu16_rm_256:
6695 ; X86: # %bb.0:
6696 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6697 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x00]
6698 ; X86-NEXT: retl # encoding: [0xc3]
6699 ;
6700 ; X64-LABEL: test_mask_subs_epu16_rm_256:
6701 ; X64: # %bb.0:
6702 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x07]
6703 ; X64-NEXT: retq # encoding: [0xc3]
6704 %b = load <16 x i16>, <16 x i16>* %ptr_b
6705 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
6706 ret <16 x i16> %res
6707 }
6709 define <16 x i16> @test_mask_subs_epu16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
6710 ; X86-LABEL: test_mask_subs_epu16_rmk_256:
6711 ; X86: # %bb.0:
6712 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6713 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6714 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x08]
6715 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6716 ; X86-NEXT: retl # encoding: [0xc3]
6717 ;
6718 ; X64-LABEL: test_mask_subs_epu16_rmk_256:
6719 ; X64: # %bb.0:
6720 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6721 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x0f]
6722 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6723 ; X64-NEXT: retq # encoding: [0xc3]
6724 %b = load <16 x i16>, <16 x i16>* %ptr_b
6725 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
6726 ret <16 x i16> %res
6727 }
6729 define <16 x i16> @test_mask_subs_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
6730 ; X86-LABEL: test_mask_subs_epu16_rmkz_256:
6731 ; X86: # %bb.0:
6732 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6733 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6734 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x00]
6735 ; X86-NEXT: retl # encoding: [0xc3]
6736 ;
6737 ; X64-LABEL: test_mask_subs_epu16_rmkz_256:
6738 ; X64: # %bb.0:
6739 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6740 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x07]
6741 ; X64-NEXT: retq # encoding: [0xc3]
6742 %b = load <16 x i16>, <16 x i16>* %ptr_b
6743 %res = call <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
6744 ret <16 x i16> %res
6745 }
6747 declare <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6749 define <16 x i8> @test_mask_adds_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
6750 ; CHECK-LABEL: test_mask_adds_epu8_rr_128:
6751 ; CHECK: # %bb.0:
6752 ; CHECK-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0xc1]
6753 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6754 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
6755 ret <16 x i8> %res
6756 }
6758 define <16 x i8> @test_mask_adds_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
6759 ; X86-LABEL: test_mask_adds_epu8_rrk_128:
6760 ; X86: # %bb.0:
6761 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6762 ; X86-NEXT: vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
6763 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6764 ; X86-NEXT: retl # encoding: [0xc3]
6765 ;
6766 ; X64-LABEL: test_mask_adds_epu8_rrk_128:
6767 ; X64: # %bb.0:
6768 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6769 ; X64-NEXT: vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
6770 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6771 ; X64-NEXT: retq # encoding: [0xc3]
6772 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
6773 ret <16 x i8> %res
6774 }
6776 define <16 x i8> @test_mask_adds_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
6777 ; X86-LABEL: test_mask_adds_epu8_rrkz_128:
6778 ; X86: # %bb.0:
6779 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6780 ; X86-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
6781 ; X86-NEXT: retl # encoding: [0xc3]
6782 ;
6783 ; X64-LABEL: test_mask_adds_epu8_rrkz_128:
6784 ; X64: # %bb.0:
6785 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6786 ; X64-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
6787 ; X64-NEXT: retq # encoding: [0xc3]
6788 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
6789 ret <16 x i8> %res
6790 }
6792 define <16 x i8> @test_mask_adds_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
6793 ; X86-LABEL: test_mask_adds_epu8_rm_128:
6794 ; X86: # %bb.0:
6795 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6796 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x00]
6797 ; X86-NEXT: retl # encoding: [0xc3]
6798 ;
6799 ; X64-LABEL: test_mask_adds_epu8_rm_128:
6800 ; X64: # %bb.0:
6801 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x07]
6802 ; X64-NEXT: retq # encoding: [0xc3]
6803 %b = load <16 x i8>, <16 x i8>* %ptr_b
6804 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
6805 ret <16 x i8> %res
6806 }
6808 define <16 x i8> @test_mask_adds_epu8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
6809 ; X86-LABEL: test_mask_adds_epu8_rmk_128:
6810 ; X86: # %bb.0:
6811 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6812 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6813 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x08]
6814 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6815 ; X86-NEXT: retl # encoding: [0xc3]
6816 ;
6817 ; X64-LABEL: test_mask_adds_epu8_rmk_128:
6818 ; X64: # %bb.0:
6819 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6820 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x0f]
6821 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6822 ; X64-NEXT: retq # encoding: [0xc3]
6823 %b = load <16 x i8>, <16 x i8>* %ptr_b
6824 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
6825 ret <16 x i8> %res
6826 }
6828 define <16 x i8> @test_mask_adds_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
6829 ; X86-LABEL: test_mask_adds_epu8_rmkz_128:
6830 ; X86: # %bb.0:
6831 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6832 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6833 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x00]
6834 ; X86-NEXT: retl # encoding: [0xc3]
6835 ;
6836 ; X64-LABEL: test_mask_adds_epu8_rmkz_128:
6837 ; X64: # %bb.0:
6838 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6839 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x07]
6840 ; X64-NEXT: retq # encoding: [0xc3]
6841 %b = load <16 x i8>, <16 x i8>* %ptr_b
6842 %res = call <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
6843 ret <16 x i8> %res
6844 }
6846 declare <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
6848 define <32 x i8> @test_mask_adds_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
6849 ; CHECK-LABEL: test_mask_adds_epu8_rr_256:
6850 ; CHECK: # %bb.0:
6851 ; CHECK-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0xc1]
6852 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6853 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
6854 ret <32 x i8> %res
6855 }
6857 define <32 x i8> @test_mask_adds_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
6858 ; X86-LABEL: test_mask_adds_epu8_rrk_256:
6859 ; X86: # %bb.0:
6860 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
6861 ; X86-NEXT: vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
6862 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6863 ; X86-NEXT: retl # encoding: [0xc3]
6864 ;
6865 ; X64-LABEL: test_mask_adds_epu8_rrk_256:
6866 ; X64: # %bb.0:
6867 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6868 ; X64-NEXT: vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
6869 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6870 ; X64-NEXT: retq # encoding: [0xc3]
6871 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
6872 ret <32 x i8> %res
6873 }
6875 define <32 x i8> @test_mask_adds_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
6876 ; X86-LABEL: test_mask_adds_epu8_rrkz_256:
6877 ; X86: # %bb.0:
6878 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
6879 ; X86-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
6880 ; X86-NEXT: retl # encoding: [0xc3]
6881 ;
6882 ; X64-LABEL: test_mask_adds_epu8_rrkz_256:
6883 ; X64: # %bb.0:
6884 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6885 ; X64-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
6886 ; X64-NEXT: retq # encoding: [0xc3]
6887 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
6888 ret <32 x i8> %res
6889 }
6891 define <32 x i8> @test_mask_adds_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
6892 ; X86-LABEL: test_mask_adds_epu8_rm_256:
6893 ; X86: # %bb.0:
6894 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6895 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x00]
6896 ; X86-NEXT: retl # encoding: [0xc3]
6897 ;
6898 ; X64-LABEL: test_mask_adds_epu8_rm_256:
6899 ; X64: # %bb.0:
6900 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x07]
6901 ; X64-NEXT: retq # encoding: [0xc3]
6902 %b = load <32 x i8>, <32 x i8>* %ptr_b
6903 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
6904 ret <32 x i8> %res
6905 }
6907 define <32 x i8> @test_mask_adds_epu8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
6908 ; X86-LABEL: test_mask_adds_epu8_rmk_256:
6909 ; X86: # %bb.0:
6910 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6911 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
6912 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x08]
6913 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6914 ; X86-NEXT: retl # encoding: [0xc3]
6915 ;
6916 ; X64-LABEL: test_mask_adds_epu8_rmk_256:
6917 ; X64: # %bb.0:
6918 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6919 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x0f]
6920 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6921 ; X64-NEXT: retq # encoding: [0xc3]
6922 %b = load <32 x i8>, <32 x i8>* %ptr_b
6923 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
6924 ret <32 x i8> %res
6925 }
6927 define <32 x i8> @test_mask_adds_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
6928 ; X86-LABEL: test_mask_adds_epu8_rmkz_256:
6929 ; X86: # %bb.0:
6930 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6931 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
6932 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x00]
6933 ; X86-NEXT: retl # encoding: [0xc3]
6934 ;
6935 ; X64-LABEL: test_mask_adds_epu8_rmkz_256:
6936 ; X64: # %bb.0:
6937 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6938 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x07]
6939 ; X64-NEXT: retq # encoding: [0xc3]
6940 %b = load <32 x i8>, <32 x i8>* %ptr_b
6941 %res = call <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
6942 ret <32 x i8> %res
6943 }
6945 declare <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
6947 define <16 x i8> @test_mask_subs_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
6948 ; CHECK-LABEL: test_mask_subs_epu8_rr_128:
6949 ; CHECK: # %bb.0:
6950 ; CHECK-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0xc1]
6951 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6952 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
6953 ret <16 x i8> %res
6954 }
6956 define <16 x i8> @test_mask_subs_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
6957 ; X86-LABEL: test_mask_subs_epu8_rrk_128:
6958 ; X86: # %bb.0:
6959 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6960 ; X86-NEXT: vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
6961 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6962 ; X86-NEXT: retl # encoding: [0xc3]
6963 ;
6964 ; X64-LABEL: test_mask_subs_epu8_rrk_128:
6965 ; X64: # %bb.0:
6966 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6967 ; X64-NEXT: vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
6968 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6969 ; X64-NEXT: retq # encoding: [0xc3]
6970 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
6971 ret <16 x i8> %res
6972 }
6974 define <16 x i8> @test_mask_subs_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
6975 ; X86-LABEL: test_mask_subs_epu8_rrkz_128:
6976 ; X86: # %bb.0:
6977 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6978 ; X86-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
6979 ; X86-NEXT: retl # encoding: [0xc3]
6980 ;
6981 ; X64-LABEL: test_mask_subs_epu8_rrkz_128:
6982 ; X64: # %bb.0:
6983 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6984 ; X64-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
6985 ; X64-NEXT: retq # encoding: [0xc3]
6986 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
6987 ret <16 x i8> %res
6988 }
6990 define <16 x i8> @test_mask_subs_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
6991 ; X86-LABEL: test_mask_subs_epu8_rm_128:
6992 ; X86: # %bb.0:
6993 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6994 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x00]
6995 ; X86-NEXT: retl # encoding: [0xc3]
6996 ;
6997 ; X64-LABEL: test_mask_subs_epu8_rm_128:
6998 ; X64: # %bb.0:
6999 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x07]
7000 ; X64-NEXT: retq # encoding: [0xc3]
7001 %b = load <16 x i8>, <16 x i8>* %ptr_b
7002 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
7003 ret <16 x i8> %res
7004 }
7006 define <16 x i8> @test_mask_subs_epu8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
7007 ; X86-LABEL: test_mask_subs_epu8_rmk_128:
7008 ; X86: # %bb.0:
7009 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7010 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7011 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x08]
7012 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7013 ; X86-NEXT: retl # encoding: [0xc3]
7014 ;
7015 ; X64-LABEL: test_mask_subs_epu8_rmk_128:
7016 ; X64: # %bb.0:
7017 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7018 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x0f]
7019 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7020 ; X64-NEXT: retq # encoding: [0xc3]
7021 %b = load <16 x i8>, <16 x i8>* %ptr_b
7022 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
7023 ret <16 x i8> %res
7024 }
7026 define <16 x i8> @test_mask_subs_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
7027 ; X86-LABEL: test_mask_subs_epu8_rmkz_128:
7028 ; X86: # %bb.0:
7029 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7030 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7031 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x00]
7032 ; X86-NEXT: retl # encoding: [0xc3]
7033 ;
7034 ; X64-LABEL: test_mask_subs_epu8_rmkz_128:
7035 ; X64: # %bb.0:
7036 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7037 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x07]
7038 ; X64-NEXT: retq # encoding: [0xc3]
7039 %b = load <16 x i8>, <16 x i8>* %ptr_b
7040 %res = call <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
7041 ret <16 x i8> %res
7042 }
7044 declare <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
7046 define <32 x i8> @test_mask_subs_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
7047 ; CHECK-LABEL: test_mask_subs_epu8_rr_256:
7048 ; CHECK: # %bb.0:
7049 ; CHECK-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0xc1]
7050 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7051 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
7052 ret <32 x i8> %res
7053 }
7055 define <32 x i8> @test_mask_subs_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
7056 ; X86-LABEL: test_mask_subs_epu8_rrk_256:
7057 ; X86: # %bb.0:
7058 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7059 ; X86-NEXT: vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
7060 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7061 ; X86-NEXT: retl # encoding: [0xc3]
7062 ;
7063 ; X64-LABEL: test_mask_subs_epu8_rrk_256:
7064 ; X64: # %bb.0:
7065 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7066 ; X64-NEXT: vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
7067 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7068 ; X64-NEXT: retq # encoding: [0xc3]
7069 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
7070 ret <32 x i8> %res
7071 }
7073 define <32 x i8> @test_mask_subs_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
7074 ; X86-LABEL: test_mask_subs_epu8_rrkz_256:
7075 ; X86: # %bb.0:
7076 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7077 ; X86-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
7078 ; X86-NEXT: retl # encoding: [0xc3]
7079 ;
7080 ; X64-LABEL: test_mask_subs_epu8_rrkz_256:
7081 ; X64: # %bb.0:
7082 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7083 ; X64-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
7084 ; X64-NEXT: retq # encoding: [0xc3]
7085 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
7086 ret <32 x i8> %res
7087 }
7089 define <32 x i8> @test_mask_subs_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
7090 ; X86-LABEL: test_mask_subs_epu8_rm_256:
7091 ; X86: # %bb.0:
7092 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7093 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x00]
7094 ; X86-NEXT: retl # encoding: [0xc3]
7095 ;
7096 ; X64-LABEL: test_mask_subs_epu8_rm_256:
7097 ; X64: # %bb.0:
7098 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x07]
7099 ; X64-NEXT: retq # encoding: [0xc3]
7100 %b = load <32 x i8>, <32 x i8>* %ptr_b
7101 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
7102 ret <32 x i8> %res
7103 }
7105 define <32 x i8> @test_mask_subs_epu8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
7106 ; X86-LABEL: test_mask_subs_epu8_rmk_256:
7107 ; X86: # %bb.0:
7108 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7109 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7110 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x08]
7111 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7112 ; X86-NEXT: retl # encoding: [0xc3]
7113 ;
7114 ; X64-LABEL: test_mask_subs_epu8_rmk_256:
7115 ; X64: # %bb.0:
7116 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7117 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x0f]
7118 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7119 ; X64-NEXT: retq # encoding: [0xc3]
7120 %b = load <32 x i8>, <32 x i8>* %ptr_b
7121 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
7122 ret <32 x i8> %res
7123 }
7125 define <32 x i8> @test_mask_subs_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
7126 ; X86-LABEL: test_mask_subs_epu8_rmkz_256:
7127 ; X86: # %bb.0:
7128 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7129 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7130 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x00]
7131 ; X86-NEXT: retl # encoding: [0xc3]
7132 ;
7133 ; X64-LABEL: test_mask_subs_epu8_rmkz_256:
7134 ; X64: # %bb.0:
7135 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7136 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x07]
7137 ; X64-NEXT: retq # encoding: [0xc3]
7138 %b = load <32 x i8>, <32 x i8>* %ptr_b
7139 %res = call <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
7140 ret <32 x i8> %res
7141 }
7143 declare <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
7145 define <8 x i16> @test_mask_adds_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
7146 ; CHECK-LABEL: test_mask_adds_epi16_rr_128:
7147 ; CHECK: # %bb.0:
7148 ; CHECK-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0xc1]
7149 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7150 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
7151 ret <8 x i16> %res
7152 }
7154 define <8 x i16> @test_mask_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7155 ; X86-LABEL: test_mask_adds_epi16_rrk_128:
7156 ; X86: # %bb.0:
7157 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7158 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7159 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7160 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7161 ; X86-NEXT: retl # encoding: [0xc3]
7162 ;
7163 ; X64-LABEL: test_mask_adds_epi16_rrk_128:
7164 ; X64: # %bb.0:
7165 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7166 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7167 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7168 ; X64-NEXT: retq # encoding: [0xc3]
7169 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
7170 ret <8 x i16> %res
7171 }
7173 define <8 x i16> @test_mask_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7174 ; X86-LABEL: test_mask_adds_epi16_rrkz_128:
7175 ; X86: # %bb.0:
7176 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7177 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7178 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7179 ; X86-NEXT: retl # encoding: [0xc3]
7180 ;
7181 ; X64-LABEL: test_mask_adds_epi16_rrkz_128:
7182 ; X64: # %bb.0:
7183 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7184 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7185 ; X64-NEXT: retq # encoding: [0xc3]
7186 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
7187 ret <8 x i16> %res
7188 }
7190 define <8 x i16> @test_mask_adds_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
7191 ; X86-LABEL: test_mask_adds_epi16_rm_128:
7192 ; X86: # %bb.0:
7193 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7194 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x00]
7195 ; X86-NEXT: retl # encoding: [0xc3]
7196 ;
7197 ; X64-LABEL: test_mask_adds_epi16_rm_128:
7198 ; X64: # %bb.0:
7199 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x07]
7200 ; X64-NEXT: retq # encoding: [0xc3]
7201 %b = load <8 x i16>, <8 x i16>* %ptr_b
7202 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
7203 ret <8 x i16> %res
7204 }
7206 define <8 x i16> @test_mask_adds_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
7207 ; X86-LABEL: test_mask_adds_epi16_rmk_128:
7208 ; X86: # %bb.0:
7209 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7210 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7211 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7212 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
7213 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7214 ; X86-NEXT: retl # encoding: [0xc3]
7215 ;
7216 ; X64-LABEL: test_mask_adds_epi16_rmk_128:
7217 ; X64: # %bb.0:
7218 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7219 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
7220 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7221 ; X64-NEXT: retq # encoding: [0xc3]
7222 %b = load <8 x i16>, <8 x i16>* %ptr_b
7223 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
7224 ret <8 x i16> %res
7225 }
7227 define <8 x i16> @test_mask_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7228 ; X86-LABEL: test_mask_adds_epi16_rmkz_128:
7229 ; X86: # %bb.0:
7230 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7231 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7232 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7233 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
7234 ; X86-NEXT: retl # encoding: [0xc3]
7235 ;
7236 ; X64-LABEL: test_mask_adds_epi16_rmkz_128:
7237 ; X64: # %bb.0:
7238 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7239 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
7240 ; X64-NEXT: retq # encoding: [0xc3]
7241 %b = load <8 x i16>, <8 x i16>* %ptr_b
7242 %res = call <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
7243 ret <8 x i16> %res
7244 }
7246 declare <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
7248 define <16 x i16> @test_mask_adds_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
7249 ; CHECK-LABEL: test_mask_adds_epi16_rr_256:
7250 ; CHECK: # %bb.0:
7251 ; CHECK-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0xc1]
7252 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7253 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
7254 ret <16 x i16> %res
7255 }
7257 define <16 x i16> @test_mask_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7258 ; X86-LABEL: test_mask_adds_epi16_rrk_256:
7259 ; X86: # %bb.0:
7260 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7261 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7262 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7263 ; X86-NEXT: retl # encoding: [0xc3]
7264 ;
7265 ; X64-LABEL: test_mask_adds_epi16_rrk_256:
7266 ; X64: # %bb.0:
7267 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7268 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7269 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7270 ; X64-NEXT: retq # encoding: [0xc3]
7271 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
7272 ret <16 x i16> %res
7273 }
7275 define <16 x i16> @test_mask_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7276 ; X86-LABEL: test_mask_adds_epi16_rrkz_256:
7277 ; X86: # %bb.0:
7278 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7279 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7280 ; X86-NEXT: retl # encoding: [0xc3]
7281 ;
7282 ; X64-LABEL: test_mask_adds_epi16_rrkz_256:
7283 ; X64: # %bb.0:
7284 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7285 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7286 ; X64-NEXT: retq # encoding: [0xc3]
7287 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
7288 ret <16 x i16> %res
7289 }
7291 define <16 x i16> @test_mask_adds_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
7292 ; X86-LABEL: test_mask_adds_epi16_rm_256:
7293 ; X86: # %bb.0:
7294 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7295 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x00]
7296 ; X86-NEXT: retl # encoding: [0xc3]
7297 ;
7298 ; X64-LABEL: test_mask_adds_epi16_rm_256:
7299 ; X64: # %bb.0:
7300 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x07]
7301 ; X64-NEXT: retq # encoding: [0xc3]
7302 %b = load <16 x i16>, <16 x i16>* %ptr_b
7303 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
7304 ret <16 x i16> %res
7305 }
7307 define <16 x i16> @test_mask_adds_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
7308 ; X86-LABEL: test_mask_adds_epi16_rmk_256:
7309 ; X86: # %bb.0:
7310 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7311 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7312 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
7313 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7314 ; X86-NEXT: retl # encoding: [0xc3]
7315 ;
7316 ; X64-LABEL: test_mask_adds_epi16_rmk_256:
7317 ; X64: # %bb.0:
7318 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7319 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
7320 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7321 ; X64-NEXT: retq # encoding: [0xc3]
7322 %b = load <16 x i16>, <16 x i16>* %ptr_b
7323 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
7324 ret <16 x i16> %res
7325 }
7327 define <16 x i16> @test_mask_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7328 ; X86-LABEL: test_mask_adds_epi16_rmkz_256:
7329 ; X86: # %bb.0:
7330 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7331 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7332 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
7333 ; X86-NEXT: retl # encoding: [0xc3]
7334 ;
7335 ; X64-LABEL: test_mask_adds_epi16_rmkz_256:
7336 ; X64: # %bb.0:
7337 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7338 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
7339 ; X64-NEXT: retq # encoding: [0xc3]
7340 %b = load <16 x i16>, <16 x i16>* %ptr_b
7341 %res = call <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
7342 ret <16 x i16> %res
7343 }
7345 declare <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
7347 declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) #0
7349 define <8 x i16> @test_test_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7350 ; X86-LABEL: test_test_subs_epi16_rrk_128:
7351 ; X86: # %bb.0:
7352 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7353 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7354 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7355 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7356 ; X86-NEXT: retl # encoding: [0xc3]
7357 ;
7358 ; X64-LABEL: test_test_subs_epi16_rrk_128:
7359 ; X64: # %bb.0:
7360 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7361 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7362 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7363 ; X64-NEXT: retq # encoding: [0xc3]
7364 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7365 %2 = bitcast i8 %mask to <8 x i1>
7366 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7367 ret <8 x i16> %3
7368 }
7370 define <8 x i16> @test_test_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7371 ; X86-LABEL: test_test_subs_epi16_rrkz_128:
7372 ; X86: # %bb.0:
7373 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7374 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7375 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7376 ; X86-NEXT: retl # encoding: [0xc3]
7377 ;
7378 ; X64-LABEL: test_test_subs_epi16_rrkz_128:
7379 ; X64: # %bb.0:
7380 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7381 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7382 ; X64-NEXT: retq # encoding: [0xc3]
7383 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7384 %2 = bitcast i8 %mask to <8 x i1>
7385 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7386 ret <8 x i16> %3
7387 }
7389 define <8 x i16> @test_test_subs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
7390 ; X86-LABEL: test_test_subs_epi16_rmk_128:
7391 ; X86: # %bb.0:
7392 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7393 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7394 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7395 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
7396 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7397 ; X86-NEXT: retl # encoding: [0xc3]
7398 ;
7399 ; X64-LABEL: test_test_subs_epi16_rmk_128:
7400 ; X64: # %bb.0:
7401 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7402 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
7403 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7404 ; X64-NEXT: retq # encoding: [0xc3]
7405 %b = load <8 x i16>, <8 x i16>* %ptr_b
7406 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7407 %2 = bitcast i8 %mask to <8 x i1>
7408 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7409 ret <8 x i16> %3
7410 }
7412 define <8 x i16> @test_test_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7413 ; X86-LABEL: test_test_subs_epi16_rmkz_128:
7414 ; X86: # %bb.0:
7415 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7416 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7417 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7418 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
7419 ; X86-NEXT: retl # encoding: [0xc3]
7420 ;
7421 ; X64-LABEL: test_test_subs_epi16_rmkz_128:
7422 ; X64: # %bb.0:
7423 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7424 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
7425 ; X64-NEXT: retq # encoding: [0xc3]
7426 %b = load <8 x i16>, <8 x i16>* %ptr_b
7427 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7428 %2 = bitcast i8 %mask to <8 x i1>
7429 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7430 ret <8 x i16> %3
7431 }
7433 declare <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16>, <16 x i16>) #0
7435 define <16 x i16> @test_test_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7436 ; X86-LABEL: test_test_subs_epi16_rrk_256:
7437 ; X86: # %bb.0:
7438 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7439 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7440 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7441 ; X86-NEXT: retl # encoding: [0xc3]
7442 ;
7443 ; X64-LABEL: test_test_subs_epi16_rrk_256:
7444 ; X64: # %bb.0:
7445 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7446 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7447 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7448 ; X64-NEXT: retq # encoding: [0xc3]
7449 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7450 %2 = bitcast i16 %mask to <16 x i1>
7451 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7452 ret <16 x i16> %3
7453 }
7455 define <16 x i16> @test_test_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7456 ; X86-LABEL: test_test_subs_epi16_rrkz_256:
7457 ; X86: # %bb.0:
7458 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7459 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7460 ; X86-NEXT: retl # encoding: [0xc3]
7461 ;
7462 ; X64-LABEL: test_test_subs_epi16_rrkz_256:
7463 ; X64: # %bb.0:
7464 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7465 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7466 ; X64-NEXT: retq # encoding: [0xc3]
7467 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7468 %2 = bitcast i16 %mask to <16 x i1>
7469 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
7470 ret <16 x i16> %3
7471 }
7473 define <16 x i16> @test_test_subs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
7474 ; X86-LABEL: test_test_subs_epi16_rmk_256:
7475 ; X86: # %bb.0:
7476 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7477 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7478 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
7479 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7480 ; X86-NEXT: retl # encoding: [0xc3]
7481 ;
7482 ; X64-LABEL: test_test_subs_epi16_rmk_256:
7483 ; X64: # %bb.0:
7484 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7485 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
7486 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7487 ; X64-NEXT: retq # encoding: [0xc3]
7488 %b = load <16 x i16>, <16 x i16>* %ptr_b
7489 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7490 %2 = bitcast i16 %mask to <16 x i1>
7491 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7492 ret <16 x i16> %3
7493 }
7495 define <16 x i16> @test_test_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7496 ; X86-LABEL: test_test_subs_epi16_rmkz_256:
7497 ; X86: # %bb.0:
7498 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7499 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7500 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
7501 ; X86-NEXT: retl # encoding: [0xc3]
7502 ;
7503 ; X64-LABEL: test_test_subs_epi16_rmkz_256:
7504 ; X64: # %bb.0:
7505 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7506 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
7507 ; X64-NEXT: retq # encoding: [0xc3]
7508 %b = load <16 x i16>, <16 x i16>* %ptr_b
7509 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7510 %2 = bitcast i16 %mask to <16 x i1>
7511 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
7512 ret <16 x i16> %3
7513 }
7515 declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) #0
7517 define <16 x i8> @test_test_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
7518 ; X86-LABEL: test_test_subs_epi8_rrk_128:
7519 ; X86: # %bb.0:
7520 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7521 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
7522 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7523 ; X86-NEXT: retl # encoding: [0xc3]
7524 ;
7525 ; X64-LABEL: test_test_subs_epi8_rrk_128:
7526 ; X64: # %bb.0:
7527 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7528 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
7529 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7530 ; X64-NEXT: retq # encoding: [0xc3]
7531 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7532 %2 = bitcast i16 %mask to <16 x i1>
7533 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
7534 ret <16 x i8> %3
7535 }
7537 define <16 x i8> @test_test_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
7538 ; X86-LABEL: test_test_subs_epi8_rrkz_128:
7539 ; X86: # %bb.0:
7540 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7541 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
7542 ; X86-NEXT: retl # encoding: [0xc3]
7543 ;
7544 ; X64-LABEL: test_test_subs_epi8_rrkz_128:
7545 ; X64: # %bb.0:
7546 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7547 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
7548 ; X64-NEXT: retq # encoding: [0xc3]
7549 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7550 %2 = bitcast i16 %mask to <16 x i1>
7551 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
7552 ret <16 x i8> %3
7553 }
7555 define <16 x i8> @test_test_subs_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
7556 ; X86-LABEL: test_test_subs_epi8_rmk_128:
7557 ; X86: # %bb.0:
7558 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7559 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7560 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
7561 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7562 ; X86-NEXT: retl # encoding: [0xc3]
7563 ;
7564 ; X64-LABEL: test_test_subs_epi8_rmk_128:
7565 ; X64: # %bb.0:
7566 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7567 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
7568 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7569 ; X64-NEXT: retq # encoding: [0xc3]
7570 %b = load <16 x i8>, <16 x i8>* %ptr_b
7571 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7572 %2 = bitcast i16 %mask to <16 x i1>
7573 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
7574 ret <16 x i8> %3
7575 }
7577 define <16 x i8> @test_test_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
7578 ; X86-LABEL: test_test_subs_epi8_rmkz_128:
7579 ; X86: # %bb.0:
7580 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7581 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7582 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
7583 ; X86-NEXT: retl # encoding: [0xc3]
7584 ;
7585 ; X64-LABEL: test_test_subs_epi8_rmkz_128:
7586 ; X64: # %bb.0:
7587 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7588 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
7589 ; X64-NEXT: retq # encoding: [0xc3]
7590 %b = load <16 x i8>, <16 x i8>* %ptr_b
7591 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7592 %2 = bitcast i16 %mask to <16 x i1>
7593 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
7594 ret <16 x i8> %3
7595 }
7597 declare <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8>, <32 x i8>) #0
7599 define <32 x i8> @test_test_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
7600 ; X86-LABEL: test_test_subs_epi8_rrk_256:
7601 ; X86: # %bb.0:
7602 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7603 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
7604 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7605 ; X86-NEXT: retl # encoding: [0xc3]
7606 ;
7607 ; X64-LABEL: test_test_subs_epi8_rrk_256:
7608 ; X64: # %bb.0:
7609 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7610 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
7611 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7612 ; X64-NEXT: retq # encoding: [0xc3]
7613 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7614 %2 = bitcast i32 %mask to <32 x i1>
7615 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
7616 ret <32 x i8> %3
7617 }
7619 define <32 x i8> @test_test_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
7620 ; X86-LABEL: test_test_subs_epi8_rrkz_256:
7621 ; X86: # %bb.0:
7622 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7623 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
7624 ; X86-NEXT: retl # encoding: [0xc3]
7625 ;
7626 ; X64-LABEL: test_test_subs_epi8_rrkz_256:
7627 ; X64: # %bb.0:
7628 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7629 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
7630 ; X64-NEXT: retq # encoding: [0xc3]
7631 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7632 %2 = bitcast i32 %mask to <32 x i1>
7633 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
7634 ret <32 x i8> %3
7635 }
7637 define <32 x i8> @test_test_subs_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
7638 ; X86-LABEL: test_test_subs_epi8_rmk_256:
7639 ; X86: # %bb.0:
7640 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7641 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7642 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
7643 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7644 ; X86-NEXT: retl # encoding: [0xc3]
7645 ;
7646 ; X64-LABEL: test_test_subs_epi8_rmk_256:
7647 ; X64: # %bb.0:
7648 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7649 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
7650 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7651 ; X64-NEXT: retq # encoding: [0xc3]
7652 %b = load <32 x i8>, <32 x i8>* %ptr_b
7653 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7654 %2 = bitcast i32 %mask to <32 x i1>
7655 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
7656 ret <32 x i8> %3
7657 }
7659 define <32 x i8> @test_test_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
7660 ; X86-LABEL: test_test_subs_epi8_rmkz_256:
7661 ; X86: # %bb.0:
7662 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7663 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7664 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
7665 ; X86-NEXT: retl # encoding: [0xc3]
7666 ;
7667 ; X64-LABEL: test_test_subs_epi8_rmkz_256:
7668 ; X64: # %bb.0:
7669 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7670 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
7671 ; X64-NEXT: retq # encoding: [0xc3]
7672 %b = load <32 x i8>, <32 x i8>* %ptr_b
7673 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7674 %2 = bitcast i32 %mask to <32 x i1>
7675 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
7676 ret <32 x i8> %3
7677 }
7679 define <8 x i16> @test_mask_subs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
7680 ; CHECK-LABEL: test_mask_subs_epi16_rr_128:
7681 ; CHECK: # %bb.0:
7682 ; CHECK-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0xc1]
7683 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7684 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
7685 ret <8 x i16> %res
7686 }
7688 define <8 x i16> @test_mask_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7689 ; X86-LABEL: test_mask_subs_epi16_rrk_128:
7690 ; X86: # %bb.0:
7691 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7692 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7693 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7694 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7695 ; X86-NEXT: retl # encoding: [0xc3]
7696 ;
7697 ; X64-LABEL: test_mask_subs_epi16_rrk_128:
7698 ; X64: # %bb.0:
7699 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7700 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7701 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7702 ; X64-NEXT: retq # encoding: [0xc3]
7703 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
7704 ret <8 x i16> %res
7705 }
7707 define <8 x i16> @test_mask_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7708 ; X86-LABEL: test_mask_subs_epi16_rrkz_128:
7709 ; X86: # %bb.0:
7710 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7711 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7712 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7713 ; X86-NEXT: retl # encoding: [0xc3]
7714 ;
7715 ; X64-LABEL: test_mask_subs_epi16_rrkz_128:
7716 ; X64: # %bb.0:
7717 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7718 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7719 ; X64-NEXT: retq # encoding: [0xc3]
7720 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
7721 ret <8 x i16> %res
7722 }
7724 define <8 x i16> @test_mask_subs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
7725 ; X86-LABEL: test_mask_subs_epi16_rm_128:
7726 ; X86: # %bb.0:
7727 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7728 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x00]
7729 ; X86-NEXT: retl # encoding: [0xc3]
7730 ;
7731 ; X64-LABEL: test_mask_subs_epi16_rm_128:
7732 ; X64: # %bb.0:
7733 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x07]
7734 ; X64-NEXT: retq # encoding: [0xc3]
7735 %b = load <8 x i16>, <8 x i16>* %ptr_b
7736 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 -1)
7737 ret <8 x i16> %res
7738 }
7740 define <8 x i16> @test_mask_subs_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
7741 ; X86-LABEL: test_mask_subs_epi16_rmk_128:
7742 ; X86: # %bb.0:
7743 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7744 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7745 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7746 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
7747 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7748 ; X86-NEXT: retl # encoding: [0xc3]
7749 ;
7750 ; X64-LABEL: test_mask_subs_epi16_rmk_128:
7751 ; X64: # %bb.0:
7752 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7753 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
7754 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7755 ; X64-NEXT: retq # encoding: [0xc3]
7756 %b = load <8 x i16>, <8 x i16>* %ptr_b
7757 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask)
7758 ret <8 x i16> %res
7759 }
7761 define <8 x i16> @test_mask_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7762 ; X86-LABEL: test_mask_subs_epi16_rmkz_128:
7763 ; X86: # %bb.0:
7764 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7765 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7766 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7767 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
7768 ; X86-NEXT: retl # encoding: [0xc3]
7769 ;
7770 ; X64-LABEL: test_mask_subs_epi16_rmkz_128:
7771 ; X64: # %bb.0:
7772 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7773 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
7774 ; X64-NEXT: retq # encoding: [0xc3]
7775 %b = load <8 x i16>, <8 x i16>* %ptr_b
7776 %res = call <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16> %a, <8 x i16> %b, <8 x i16> zeroinitializer, i8 %mask)
7777 ret <8 x i16> %res
7778 }
7780 declare <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
7782 define <16 x i16> @test_mask_subs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
7783 ; CHECK-LABEL: test_mask_subs_epi16_rr_256:
7784 ; CHECK: # %bb.0:
7785 ; CHECK-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0xc1]
7786 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7787 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
7788 ret <16 x i16> %res
7789 }
7791 define <16 x i16> @test_mask_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7792 ; X86-LABEL: test_mask_subs_epi16_rrk_256:
7793 ; X86: # %bb.0:
7794 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7795 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7796 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7797 ; X86-NEXT: retl # encoding: [0xc3]
7798 ;
7799 ; X64-LABEL: test_mask_subs_epi16_rrk_256:
7800 ; X64: # %bb.0:
7801 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7802 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7803 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7804 ; X64-NEXT: retq # encoding: [0xc3]
7805 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
7806 ret <16 x i16> %res
7807 }
7809 define <16 x i16> @test_mask_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7810 ; X86-LABEL: test_mask_subs_epi16_rrkz_256:
7811 ; X86: # %bb.0:
7812 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7813 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7814 ; X86-NEXT: retl # encoding: [0xc3]
7815 ;
7816 ; X64-LABEL: test_mask_subs_epi16_rrkz_256:
7817 ; X64: # %bb.0:
7818 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7819 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7820 ; X64-NEXT: retq # encoding: [0xc3]
7821 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
7822 ret <16 x i16> %res
7823 }
7825 define <16 x i16> @test_mask_subs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
7826 ; X86-LABEL: test_mask_subs_epi16_rm_256:
7827 ; X86: # %bb.0:
7828 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7829 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x00]
7830 ; X86-NEXT: retl # encoding: [0xc3]
7831 ;
7832 ; X64-LABEL: test_mask_subs_epi16_rm_256:
7833 ; X64: # %bb.0:
7834 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x07]
7835 ; X64-NEXT: retq # encoding: [0xc3]
7836 %b = load <16 x i16>, <16 x i16>* %ptr_b
7837 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 -1)
7838 ret <16 x i16> %res
7839 }
7841 define <16 x i16> @test_mask_subs_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
7842 ; X86-LABEL: test_mask_subs_epi16_rmk_256:
7843 ; X86: # %bb.0:
7844 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7845 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7846 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
7847 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7848 ; X86-NEXT: retl # encoding: [0xc3]
7849 ;
7850 ; X64-LABEL: test_mask_subs_epi16_rmk_256:
7851 ; X64: # %bb.0:
7852 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7853 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
7854 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7855 ; X64-NEXT: retq # encoding: [0xc3]
7856 %b = load <16 x i16>, <16 x i16>* %ptr_b
7857 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask)
7858 ret <16 x i16> %res
7859 }
7861 define <16 x i16> @test_mask_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7862 ; X86-LABEL: test_mask_subs_epi16_rmkz_256:
7863 ; X86: # %bb.0:
7864 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7865 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7866 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
7867 ; X86-NEXT: retl # encoding: [0xc3]
7868 ;
7869 ; X64-LABEL: test_mask_subs_epi16_rmkz_256:
7870 ; X64: # %bb.0:
7871 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7872 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
7873 ; X64-NEXT: retq # encoding: [0xc3]
7874 %b = load <16 x i16>, <16 x i16>* %ptr_b
7875 %res = call <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16> %a, <16 x i16> %b, <16 x i16> zeroinitializer, i16 %mask)
7876 ret <16 x i16> %res
7877 }
7879 declare <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
7881 declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) #0
7883 define <8 x i16> @test_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7884 ; X86-LABEL: test_adds_epi16_rrk_128:
7885 ; X86: # %bb.0:
7886 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7887 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7888 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7889 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7890 ; X86-NEXT: retl # encoding: [0xc3]
7891 ;
7892 ; X64-LABEL: test_adds_epi16_rrk_128:
7893 ; X64: # %bb.0:
7894 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7895 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7896 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7897 ; X64-NEXT: retq # encoding: [0xc3]
7898 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7899 %2 = bitcast i8 %mask to <8 x i1>
7900 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7901 ret <8 x i16> %3
7902 }
7904 define <8 x i16> @test_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7905 ; X86-LABEL: test_adds_epi16_rrkz_128:
7906 ; X86: # %bb.0:
7907 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7908 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7909 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7910 ; X86-NEXT: retl # encoding: [0xc3]
7911 ;
7912 ; X64-LABEL: test_adds_epi16_rrkz_128:
7913 ; X64: # %bb.0:
7914 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7915 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7916 ; X64-NEXT: retq # encoding: [0xc3]
7917 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7918 %2 = bitcast i8 %mask to <8 x i1>
7919 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7920 ret <8 x i16> %3
7921 }
7923 define <8 x i16> @test_adds_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
7924 ; X86-LABEL: test_adds_epi16_rmk_128:
7925 ; X86: # %bb.0:
7926 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7927 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7928 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7929 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
7930 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7931 ; X86-NEXT: retl # encoding: [0xc3]
7932 ;
7933 ; X64-LABEL: test_adds_epi16_rmk_128:
7934 ; X64: # %bb.0:
7935 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7936 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
7937 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7938 ; X64-NEXT: retq # encoding: [0xc3]
7939 %b = load <8 x i16>, <8 x i16>* %ptr_b
7940 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7941 %2 = bitcast i8 %mask to <8 x i1>
7942 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7943 ret <8 x i16> %3
7944 }
7946 define <8 x i16> @test_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7947 ; X86-LABEL: test_adds_epi16_rmkz_128:
7948 ; X86: # %bb.0:
7949 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7950 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7951 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7952 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
7953 ; X86-NEXT: retl # encoding: [0xc3]
7954 ;
7955 ; X64-LABEL: test_adds_epi16_rmkz_128:
7956 ; X64: # %bb.0:
7957 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7958 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
7959 ; X64-NEXT: retq # encoding: [0xc3]
7960 %b = load <8 x i16>, <8 x i16>* %ptr_b
7961 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7962 %2 = bitcast i8 %mask to <8 x i1>
7963 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7964 ret <8 x i16> %3
7965 }
7967 declare <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16>, <16 x i16>) #0
7969 define <16 x i16> @test_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7970 ; X86-LABEL: test_adds_epi16_rrk_256:
7971 ; X86: # %bb.0:
7972 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7973 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7974 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7975 ; X86-NEXT: retl # encoding: [0xc3]
7976 ;
7977 ; X64-LABEL: test_adds_epi16_rrk_256:
7978 ; X64: # %bb.0:
7979 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7980 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7981 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7982 ; X64-NEXT: retq # encoding: [0xc3]
7983 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
7984 %2 = bitcast i16 %mask to <16 x i1>
7985 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7986 ret <16 x i16> %3
7987 }
7989 define <16 x i16> @test_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7990 ; X86-LABEL: test_adds_epi16_rrkz_256:
7991 ; X86: # %bb.0:
7992 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7993 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7994 ; X86-NEXT: retl # encoding: [0xc3]
7995 ;
7996 ; X64-LABEL: test_adds_epi16_rrkz_256:
7997 ; X64: # %bb.0:
7998 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7999 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
8000 ; X64-NEXT: retq # encoding: [0xc3]
8001 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8002 %2 = bitcast i16 %mask to <16 x i1>
8003 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
8004 ret <16 x i16> %3
8005 }
8007 define <16 x i16> @test_adds_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
8008 ; X86-LABEL: test_adds_epi16_rmk_256:
8009 ; X86: # %bb.0:
8010 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8011 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8012 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
8013 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8014 ; X86-NEXT: retl # encoding: [0xc3]
8015 ;
8016 ; X64-LABEL: test_adds_epi16_rmk_256:
8017 ; X64: # %bb.0:
8018 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8019 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
8020 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8021 ; X64-NEXT: retq # encoding: [0xc3]
8022 %b = load <16 x i16>, <16 x i16>* %ptr_b
8023 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8024 %2 = bitcast i16 %mask to <16 x i1>
8025 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
8026 ret <16 x i16> %3
8027 }
8029 define <16 x i16> @test_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
8030 ; X86-LABEL: test_adds_epi16_rmkz_256:
8031 ; X86: # %bb.0:
8032 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8033 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8034 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
8035 ; X86-NEXT: retl # encoding: [0xc3]
8036 ;
8037 ; X64-LABEL: test_adds_epi16_rmkz_256:
8038 ; X64: # %bb.0:
8039 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8040 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
8041 ; X64-NEXT: retq # encoding: [0xc3]
8042 %b = load <16 x i16>, <16 x i16>* %ptr_b
8043 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8044 %2 = bitcast i16 %mask to <16 x i1>
8045 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
8046 ret <16 x i16> %3
8047 }
8049 declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) #0
8051 define <16 x i8> @test_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8052 ; X86-LABEL: test_adds_epi8_rrk_128:
8053 ; X86: # %bb.0:
8054 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8055 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8056 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8057 ; X86-NEXT: retl # encoding: [0xc3]
8058 ;
8059 ; X64-LABEL: test_adds_epi8_rrk_128:
8060 ; X64: # %bb.0:
8061 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8062 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8063 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8064 ; X64-NEXT: retq # encoding: [0xc3]
8065 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8066 %2 = bitcast i16 %mask to <16 x i1>
8067 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
8068 ret <16 x i8> %3
8069 }
8071 define <16 x i8> @test_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8072 ; X86-LABEL: test_adds_epi8_rrkz_128:
8073 ; X86: # %bb.0:
8074 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8075 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8076 ; X86-NEXT: retl # encoding: [0xc3]
8077 ;
8078 ; X64-LABEL: test_adds_epi8_rrkz_128:
8079 ; X64: # %bb.0:
8080 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8081 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8082 ; X64-NEXT: retq # encoding: [0xc3]
8083 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8084 %2 = bitcast i16 %mask to <16 x i1>
8085 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
8086 ret <16 x i8> %3
8087 }
8089 define <16 x i8> @test_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8090 ; X86-LABEL: test_adds_epi8_rm_128:
8091 ; X86: # %bb.0:
8092 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8093 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
8094 ; X86-NEXT: retl # encoding: [0xc3]
8095 ;
8096 ; X64-LABEL: test_adds_epi8_rm_128:
8097 ; X64: # %bb.0:
8098 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
8099 ; X64-NEXT: retq # encoding: [0xc3]
8100 %b = load <16 x i8>, <16 x i8>* %ptr_b
8101 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8102 ret <16 x i8> %1
8103 }
8105 define <16 x i8> @test_adds_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
8106 ; X86-LABEL: test_adds_epi8_rmk_128:
8107 ; X86: # %bb.0:
8108 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8109 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8110 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
8111 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8112 ; X86-NEXT: retl # encoding: [0xc3]
8113 ;
8114 ; X64-LABEL: test_adds_epi8_rmk_128:
8115 ; X64: # %bb.0:
8116 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8117 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
8118 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8119 ; X64-NEXT: retq # encoding: [0xc3]
8120 %b = load <16 x i8>, <16 x i8>* %ptr_b
8121 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8122 %2 = bitcast i16 %mask to <16 x i1>
8123 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
8124 ret <16 x i8> %3
8125 }
8127 define <16 x i8> @test_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8128 ; X86-LABEL: test_adds_epi8_rmkz_128:
8129 ; X86: # %bb.0:
8130 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8131 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8132 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
8133 ; X86-NEXT: retl # encoding: [0xc3]
8134 ;
8135 ; X64-LABEL: test_adds_epi8_rmkz_128:
8136 ; X64: # %bb.0:
8137 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8138 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
8139 ; X64-NEXT: retq # encoding: [0xc3]
8140 %b = load <16 x i8>, <16 x i8>* %ptr_b
8141 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8142 %2 = bitcast i16 %mask to <16 x i1>
8143 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
8144 ret <16 x i8> %3
8145 }
8147 declare <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8>, <32 x i8>) #0
8149 define <32 x i8> @test_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8150 ; X86-LABEL: test_adds_epi8_rrk_256:
8151 ; X86: # %bb.0:
8152 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8153 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8154 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8155 ; X86-NEXT: retl # encoding: [0xc3]
8156 ;
8157 ; X64-LABEL: test_adds_epi8_rrk_256:
8158 ; X64: # %bb.0:
8159 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8160 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8161 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8162 ; X64-NEXT: retq # encoding: [0xc3]
8163 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8164 %2 = bitcast i32 %mask to <32 x i1>
8165 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
8166 ret <32 x i8> %3
8167 }
8169 define <32 x i8> @test_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8170 ; X86-LABEL: test_adds_epi8_rrkz_256:
8171 ; X86: # %bb.0:
8172 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8173 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8174 ; X86-NEXT: retl # encoding: [0xc3]
8175 ;
8176 ; X64-LABEL: test_adds_epi8_rrkz_256:
8177 ; X64: # %bb.0:
8178 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8179 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8180 ; X64-NEXT: retq # encoding: [0xc3]
8181 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8182 %2 = bitcast i32 %mask to <32 x i1>
8183 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
8184 ret <32 x i8> %3
8185 }
8187 define <32 x i8> @test_adds_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
8188 ; X86-LABEL: test_adds_epi8_rmk_256:
8189 ; X86: # %bb.0:
8190 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8191 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8192 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
8193 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8194 ; X86-NEXT: retl # encoding: [0xc3]
8195 ;
8196 ; X64-LABEL: test_adds_epi8_rmk_256:
8197 ; X64: # %bb.0:
8198 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8199 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
8200 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8201 ; X64-NEXT: retq # encoding: [0xc3]
8202 %b = load <32 x i8>, <32 x i8>* %ptr_b
8203 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8204 %2 = bitcast i32 %mask to <32 x i1>
8205 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
8206 ret <32 x i8> %3
8207 }
8209 define <32 x i8> @test_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8210 ; X86-LABEL: test_adds_epi8_rmkz_256:
8211 ; X86: # %bb.0:
8212 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8213 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8214 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
8215 ; X86-NEXT: retl # encoding: [0xc3]
8216 ;
8217 ; X64-LABEL: test_adds_epi8_rmkz_256:
8218 ; X64: # %bb.0:
8219 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8220 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
8221 ; X64-NEXT: retq # encoding: [0xc3]
8222 %b = load <32 x i8>, <32 x i8>* %ptr_b
8223 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8224 %2 = bitcast i32 %mask to <32 x i1>
8225 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
8226 ret <32 x i8> %3
8227 }
8229 define <16 x i8> @test_mask_adds_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
8230 ; CHECK-LABEL: test_mask_adds_epi8_rr_128:
8231 ; CHECK: # %bb.0:
8232 ; CHECK-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0xc1]
8233 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8234 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
8235 ret <16 x i8> %res
8236 }
8238 define <16 x i8> @test_mask_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8239 ; X86-LABEL: test_mask_adds_epi8_rrk_128:
8240 ; X86: # %bb.0:
8241 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8242 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8243 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8244 ; X86-NEXT: retl # encoding: [0xc3]
8245 ;
8246 ; X64-LABEL: test_mask_adds_epi8_rrk_128:
8247 ; X64: # %bb.0:
8248 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8249 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8250 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8251 ; X64-NEXT: retq # encoding: [0xc3]
8252 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
8253 ret <16 x i8> %res
8254 }
8256 define <16 x i8> @test_mask_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8257 ; X86-LABEL: test_mask_adds_epi8_rrkz_128:
8258 ; X86: # %bb.0:
8259 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8260 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8261 ; X86-NEXT: retl # encoding: [0xc3]
8262 ;
8263 ; X64-LABEL: test_mask_adds_epi8_rrkz_128:
8264 ; X64: # %bb.0:
8265 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8266 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8267 ; X64-NEXT: retq # encoding: [0xc3]
8268 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
8269 ret <16 x i8> %res
8270 }
8272 define <16 x i8> @test_mask_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8273 ; X86-LABEL: test_mask_adds_epi8_rm_128:
8274 ; X86: # %bb.0:
8275 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8276 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
8277 ; X86-NEXT: retl # encoding: [0xc3]
8278 ;
8279 ; X64-LABEL: test_mask_adds_epi8_rm_128:
8280 ; X64: # %bb.0:
8281 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
8282 ; X64-NEXT: retq # encoding: [0xc3]
8283 %b = load <16 x i8>, <16 x i8>* %ptr_b
8284 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
8285 ret <16 x i8> %res
8286 }
8288 define <16 x i8> @test_mask_adds_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
8289 ; X86-LABEL: test_mask_adds_epi8_rmk_128:
8290 ; X86: # %bb.0:
8291 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8292 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8293 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
8294 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8295 ; X86-NEXT: retl # encoding: [0xc3]
8296 ;
8297 ; X64-LABEL: test_mask_adds_epi8_rmk_128:
8298 ; X64: # %bb.0:
8299 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8300 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
8301 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8302 ; X64-NEXT: retq # encoding: [0xc3]
8303 %b = load <16 x i8>, <16 x i8>* %ptr_b
8304 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
8305 ret <16 x i8> %res
8306 }
8308 define <16 x i8> @test_mask_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8309 ; X86-LABEL: test_mask_adds_epi8_rmkz_128:
8310 ; X86: # %bb.0:
8311 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8312 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8313 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
8314 ; X86-NEXT: retl # encoding: [0xc3]
8315 ;
8316 ; X64-LABEL: test_mask_adds_epi8_rmkz_128:
8317 ; X64: # %bb.0:
8318 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8319 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
8320 ; X64-NEXT: retq # encoding: [0xc3]
8321 %b = load <16 x i8>, <16 x i8>* %ptr_b
8322 %res = call <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
8323 ret <16 x i8> %res
8324 }
8326 declare <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
8328 define <32 x i8> @test_mask_adds_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
8329 ; CHECK-LABEL: test_mask_adds_epi8_rr_256:
8330 ; CHECK: # %bb.0:
8331 ; CHECK-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0xc1]
8332 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8333 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
8334 ret <32 x i8> %res
8335 }
8337 define <32 x i8> @test_mask_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8338 ; X86-LABEL: test_mask_adds_epi8_rrk_256:
8339 ; X86: # %bb.0:
8340 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8341 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8342 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8343 ; X86-NEXT: retl # encoding: [0xc3]
8344 ;
8345 ; X64-LABEL: test_mask_adds_epi8_rrk_256:
8346 ; X64: # %bb.0:
8347 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8348 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8349 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8350 ; X64-NEXT: retq # encoding: [0xc3]
8351 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
8352 ret <32 x i8> %res
8353 }
8355 define <32 x i8> @test_mask_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8356 ; X86-LABEL: test_mask_adds_epi8_rrkz_256:
8357 ; X86: # %bb.0:
8358 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8359 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8360 ; X86-NEXT: retl # encoding: [0xc3]
8361 ;
8362 ; X64-LABEL: test_mask_adds_epi8_rrkz_256:
8363 ; X64: # %bb.0:
8364 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8365 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8366 ; X64-NEXT: retq # encoding: [0xc3]
8367 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
8368 ret <32 x i8> %res
8369 }
8371 define <32 x i8> @test_mask_adds_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
8372 ; X86-LABEL: test_mask_adds_epi8_rm_256:
8373 ; X86: # %bb.0:
8374 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8375 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x00]
8376 ; X86-NEXT: retl # encoding: [0xc3]
8377 ;
8378 ; X64-LABEL: test_mask_adds_epi8_rm_256:
8379 ; X64: # %bb.0:
8380 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x07]
8381 ; X64-NEXT: retq # encoding: [0xc3]
8382 %b = load <32 x i8>, <32 x i8>* %ptr_b
8383 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
8384 ret <32 x i8> %res
8385 }
8387 define <32 x i8> @test_mask_adds_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
8388 ; X86-LABEL: test_mask_adds_epi8_rmk_256:
8389 ; X86: # %bb.0:
8390 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8391 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8392 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
8393 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8394 ; X86-NEXT: retl # encoding: [0xc3]
8395 ;
8396 ; X64-LABEL: test_mask_adds_epi8_rmk_256:
8397 ; X64: # %bb.0:
8398 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8399 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
8400 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8401 ; X64-NEXT: retq # encoding: [0xc3]
8402 %b = load <32 x i8>, <32 x i8>* %ptr_b
8403 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
8404 ret <32 x i8> %res
8405 }
8407 define <32 x i8> @test_mask_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8408 ; X86-LABEL: test_mask_adds_epi8_rmkz_256:
8409 ; X86: # %bb.0:
8410 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8411 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8412 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
8413 ; X86-NEXT: retl # encoding: [0xc3]
8414 ;
8415 ; X64-LABEL: test_mask_adds_epi8_rmkz_256:
8416 ; X64: # %bb.0:
8417 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8418 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
8419 ; X64-NEXT: retq # encoding: [0xc3]
8420 %b = load <32 x i8>, <32 x i8>* %ptr_b
8421 %res = call <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
8422 ret <32 x i8> %res
8423 }
8425 declare <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
8427 define <16 x i8> @test_mask_subs_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
8428 ; CHECK-LABEL: test_mask_subs_epi8_rr_128:
8429 ; CHECK: # %bb.0:
8430 ; CHECK-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0xc1]
8431 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8432 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
8433 ret <16 x i8> %res
8434 }
8436 define <16 x i8> @test_mask_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8437 ; X86-LABEL: test_mask_subs_epi8_rrk_128:
8438 ; X86: # %bb.0:
8439 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8440 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
8441 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8442 ; X86-NEXT: retl # encoding: [0xc3]
8443 ;
8444 ; X64-LABEL: test_mask_subs_epi8_rrk_128:
8445 ; X64: # %bb.0:
8446 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8447 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
8448 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8449 ; X64-NEXT: retq # encoding: [0xc3]
8450 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
8451 ret <16 x i8> %res
8452 }
8454 define <16 x i8> @test_mask_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8455 ; X86-LABEL: test_mask_subs_epi8_rrkz_128:
8456 ; X86: # %bb.0:
8457 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8458 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
8459 ; X86-NEXT: retl # encoding: [0xc3]
8460 ;
8461 ; X64-LABEL: test_mask_subs_epi8_rrkz_128:
8462 ; X64: # %bb.0:
8463 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8464 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
8465 ; X64-NEXT: retq # encoding: [0xc3]
8466 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
8467 ret <16 x i8> %res
8468 }
8470 define <16 x i8> @test_mask_subs_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8471 ; X86-LABEL: test_mask_subs_epi8_rm_128:
8472 ; X86: # %bb.0:
8473 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8474 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x00]
8475 ; X86-NEXT: retl # encoding: [0xc3]
8476 ;
8477 ; X64-LABEL: test_mask_subs_epi8_rm_128:
8478 ; X64: # %bb.0:
8479 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x07]
8480 ; X64-NEXT: retq # encoding: [0xc3]
8481 %b = load <16 x i8>, <16 x i8>* %ptr_b
8482 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 -1)
8483 ret <16 x i8> %res
8484 }
8486 define <16 x i8> @test_mask_subs_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
8487 ; X86-LABEL: test_mask_subs_epi8_rmk_128:
8488 ; X86: # %bb.0:
8489 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8490 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8491 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
8492 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8493 ; X86-NEXT: retl # encoding: [0xc3]
8494 ;
8495 ; X64-LABEL: test_mask_subs_epi8_rmk_128:
8496 ; X64: # %bb.0:
8497 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8498 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
8499 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8500 ; X64-NEXT: retq # encoding: [0xc3]
8501 %b = load <16 x i8>, <16 x i8>* %ptr_b
8502 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask)
8503 ret <16 x i8> %res
8504 }
8506 define <16 x i8> @test_mask_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8507 ; X86-LABEL: test_mask_subs_epi8_rmkz_128:
8508 ; X86: # %bb.0:
8509 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8510 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8511 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
8512 ; X86-NEXT: retl # encoding: [0xc3]
8513 ;
8514 ; X64-LABEL: test_mask_subs_epi8_rmkz_128:
8515 ; X64: # %bb.0:
8516 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8517 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
8518 ; X64-NEXT: retq # encoding: [0xc3]
8519 %b = load <16 x i8>, <16 x i8>* %ptr_b
8520 %res = call <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8> %a, <16 x i8> %b, <16 x i8> zeroinitializer, i16 %mask)
8521 ret <16 x i8> %res
8522 }
8524 declare <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
8526 define <32 x i8> @test_mask_subs_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
8527 ; CHECK-LABEL: test_mask_subs_epi8_rr_256:
8528 ; CHECK: # %bb.0:
8529 ; CHECK-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0xc1]
8530 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8531 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
8532 ret <32 x i8> %res
8533 }
8535 define <32 x i8> @test_mask_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8536 ; X86-LABEL: test_mask_subs_epi8_rrk_256:
8537 ; X86: # %bb.0:
8538 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8539 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
8540 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8541 ; X86-NEXT: retl # encoding: [0xc3]
8542 ;
8543 ; X64-LABEL: test_mask_subs_epi8_rrk_256:
8544 ; X64: # %bb.0:
8545 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8546 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
8547 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8548 ; X64-NEXT: retq # encoding: [0xc3]
8549 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
8550 ret <32 x i8> %res
8551 }
8553 define <32 x i8> @test_mask_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8554 ; X86-LABEL: test_mask_subs_epi8_rrkz_256:
8555 ; X86: # %bb.0:
8556 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8557 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
8558 ; X86-NEXT: retl # encoding: [0xc3]
8559 ;
8560 ; X64-LABEL: test_mask_subs_epi8_rrkz_256:
8561 ; X64: # %bb.0:
8562 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8563 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
8564 ; X64-NEXT: retq # encoding: [0xc3]
8565 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
8566 ret <32 x i8> %res
8567 }
8569 define <32 x i8> @test_mask_subs_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
8570 ; X86-LABEL: test_mask_subs_epi8_rm_256:
8571 ; X86: # %bb.0:
8572 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8573 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x00]
8574 ; X86-NEXT: retl # encoding: [0xc3]
8575 ;
8576 ; X64-LABEL: test_mask_subs_epi8_rm_256:
8577 ; X64: # %bb.0:
8578 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x07]
8579 ; X64-NEXT: retq # encoding: [0xc3]
8580 %b = load <32 x i8>, <32 x i8>* %ptr_b
8581 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 -1)
8582 ret <32 x i8> %res
8583 }
8585 define <32 x i8> @test_mask_subs_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
8586 ; X86-LABEL: test_mask_subs_epi8_rmk_256:
8587 ; X86: # %bb.0:
8588 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8589 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8590 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
8591 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8592 ; X86-NEXT: retl # encoding: [0xc3]
8593 ;
8594 ; X64-LABEL: test_mask_subs_epi8_rmk_256:
8595 ; X64: # %bb.0:
8596 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8597 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
8598 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8599 ; X64-NEXT: retq # encoding: [0xc3]
8600 %b = load <32 x i8>, <32 x i8>* %ptr_b
8601 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask)
8602 ret <32 x i8> %res
8603 }
8605 define <32 x i8> @test_mask_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8606 ; X86-LABEL: test_mask_subs_epi8_rmkz_256:
8607 ; X86: # %bb.0:
8608 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8609 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8610 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
8611 ; X86-NEXT: retl # encoding: [0xc3]
8612 ;
8613 ; X64-LABEL: test_mask_subs_epi8_rmkz_256:
8614 ; X64: # %bb.0:
8615 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8616 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
8617 ; X64-NEXT: retq # encoding: [0xc3]
8618 %b = load <32 x i8>, <32 x i8>* %ptr_b
8619 %res = call <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8> %a, <32 x i8> %b, <32 x i8> zeroinitializer, i32 %mask)
8620 ret <32 x i8> %res
8621 }
8623 declare <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
8625 declare <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8627 define <16 x i16>@test_int_x86_avx512_mask_psrav16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
8628 ; X86-LABEL: test_int_x86_avx512_mask_psrav16_hi:
8629 ; X86: # %bb.0:
8630 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x11,0xd9]
8631 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8632 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
8633 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
8634 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8635 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8636 ; X86-NEXT: retl # encoding: [0xc3]
8637 ;
8638 ; X64-LABEL: test_int_x86_avx512_mask_psrav16_hi:
8639 ; X64: # %bb.0:
8640 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x11,0xd9]
8641 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8642 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
8643 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
8644 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8645 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8646 ; X64-NEXT: retq # encoding: [0xc3]
8647 %res = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
8648 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
8649 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
8650 %res3 = add <16 x i16> %res, %res1
8651 %res4 = add <16 x i16> %res3, %res2
8652 ret <16 x i16> %res4
8653 }
8655 declare <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8657 define <8 x i16>@test_int_x86_avx512_mask_psrav8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
8658 ; X86-LABEL: test_int_x86_avx512_mask_psrav8_hi:
8659 ; X86: # %bb.0:
8660 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x11,0xd9]
8661 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8662 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8663 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
8664 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
8665 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8666 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8667 ; X86-NEXT: retl # encoding: [0xc3]
8668 ;
8669 ; X64-LABEL: test_int_x86_avx512_mask_psrav8_hi:
8670 ; X64: # %bb.0:
8671 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x11,0xd9]
8672 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8673 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
8674 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
8675 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8676 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8677 ; X64-NEXT: retq # encoding: [0xc3]
8678 %res = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
8679 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
8680 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
8681 %res3 = add <8 x i16> %res, %res1
8682 %res4 = add <8 x i16> %res3, %res2
8683 ret <8 x i16> %res4
8684 }
8686 declare <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8688 define <16 x i16>@test_int_x86_avx512_mask_psllv16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
8689 ; X86-LABEL: test_int_x86_avx512_mask_psllv16_hi:
8690 ; X86: # %bb.0:
8691 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0xd9]
8692 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8693 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
8694 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
8695 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8696 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8697 ; X86-NEXT: retl # encoding: [0xc3]
8698 ;
8699 ; X64-LABEL: test_int_x86_avx512_mask_psllv16_hi:
8700 ; X64: # %bb.0:
8701 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0xd9]
8702 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8703 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
8704 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
8705 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8706 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8707 ; X64-NEXT: retq # encoding: [0xc3]
8708 %res = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
8709 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
8710 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
8711 %res3 = add <16 x i16> %res, %res1
8712 %res4 = add <16 x i16> %res3, %res2
8713 ret <16 x i16> %res4
8714 }
8716 declare <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8718 define <8 x i16>@test_int_x86_avx512_mask_psllv8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
8719 ; X86-LABEL: test_int_x86_avx512_mask_psllv8_hi:
8720 ; X86: # %bb.0:
8721 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0xd9]
8722 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8723 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8724 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
8725 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
8726 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8727 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8728 ; X86-NEXT: retl # encoding: [0xc3]
8729 ;
8730 ; X64-LABEL: test_int_x86_avx512_mask_psllv8_hi:
8731 ; X64: # %bb.0:
8732 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0xd9]
8733 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8734 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
8735 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
8736 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8737 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8738 ; X64-NEXT: retq # encoding: [0xc3]
8739 %res = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
8740 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
8741 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
8742 %res3 = add <8 x i16> %res, %res1
8743 %res4 = add <8 x i16> %res3, %res2
8744 ret <8 x i16> %res4
8745 }
8747 declare <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8749 define <16 x i16>@test_int_x86_avx512_mask_psrlv16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
8750 ; X86-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
8751 ; X86: # %bb.0:
8752 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0xd9]
8753 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8754 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
8755 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
8756 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8757 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8758 ; X86-NEXT: retl # encoding: [0xc3]
8759 ;
8760 ; X64-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
8761 ; X64: # %bb.0:
8762 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0xd9]
8763 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8764 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
8765 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
8766 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8767 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8768 ; X64-NEXT: retq # encoding: [0xc3]
8769 %res = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3)
8770 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
8771 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
8772 %res3 = add <16 x i16> %res, %res1
8773 %res4 = add <16 x i16> %res3, %res2
8774 ret <16 x i16> %res4
8775 }
8777 declare <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8779 define <8 x i16>@test_int_x86_avx512_mask_psrlv8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
8780 ; X86-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
8781 ; X86: # %bb.0:
8782 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0xd9]
8783 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8784 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8785 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
8786 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
8787 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8788 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8789 ; X86-NEXT: retl # encoding: [0xc3]
8790 ;
8791 ; X64-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
8792 ; X64: # %bb.0:
8793 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0xd9]
8794 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8795 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
8796 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
8797 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8798 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8799 ; X64-NEXT: retq # encoding: [0xc3]
8800 %res = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3)
8801 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
8802 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
8803 %res3 = add <8 x i16> %res, %res1
8804 %res4 = add <8 x i16> %res3, %res2
8805 ret <8 x i16> %res4
8806 }
8808 declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16>, <16 x i8>, i16)
8810 define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
8811 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
8812 ; X86: # %bb.0:
8813 ; X86-NEXT: vpmovwb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc2]
8814 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8815 ; X86-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
8816 ; X86-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
8817 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
8818 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
8819 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
8820 ; X86-NEXT: retl # encoding: [0xc3]
8821 ;
8822 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
8823 ; X64: # %bb.0:
8824 ; X64-NEXT: vpmovwb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc2]
8825 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8826 ; X64-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
8827 ; X64-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
8828 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
8829 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
8830 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
8831 ; X64-NEXT: retq # encoding: [0xc3]
8832 %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
8833 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
8834 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
8835 %res3 = add <16 x i8> %res0, %res1
8836 %res4 = add <16 x i8> %res3, %res2
8837 ret <16 x i8> %res4
8838 }