| blob | 0c20b2679848382306fad892c47210f94e681d02 |
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: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
11 ; X86-NEXT: vpbroadcastb %eax, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xc8]
12 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
13 ; X86-NEXT: vpbroadcastb %eax, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc0]
14 ; X86-NEXT: vpbroadcastb %eax, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd0]
15 ; X86-NEXT: vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
16 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
17 ; X86-NEXT: retl # encoding: [0xc3]
18 ;
19 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_128:
20 ; X64: # %bb.0:
21 ; X64-NEXT: vpbroadcastb %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7a,0xcf]
22 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
23 ; X64-NEXT: vpbroadcastb %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7a,0xc7]
24 ; X64-NEXT: vpbroadcastb %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7a,0xd7]
25 ; X64-NEXT: vpaddb %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc2]
26 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
27 ; X64-NEXT: retq # encoding: [0xc3]
28 %res = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 -1)
29 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> %x1, i16 %mask)
30 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.128(i8 %x0, <16 x i8> zeroinitializer, i16 %mask)
31 %res3 = add <16 x i8> %res, %res1
32 %res4 = add <16 x i8> %res2, %res3
33 ret <16 x i8> %res4
34 }
37 declare <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16, <8 x i16>, i8)
39 define <8 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_128(i16 %x0, <8 x i16> %x1, i8 %mask) {
40 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
41 ; X86: # %bb.0:
42 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
43 ; X86-NEXT: vpbroadcastw %eax, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xc8]
44 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
45 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
46 ; X86-NEXT: vpbroadcastw %eax, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc0]
47 ; X86-NEXT: vpbroadcastw %eax, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd0]
48 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
49 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
50 ; X86-NEXT: retl # encoding: [0xc3]
51 ;
52 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_128:
53 ; X64: # %bb.0:
54 ; X64-NEXT: vpbroadcastw %edi, %xmm1 # encoding: [0x62,0xf2,0x7d,0x08,0x7b,0xcf]
55 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
56 ; X64-NEXT: vpbroadcastw %edi, %xmm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x7b,0xc7]
57 ; X64-NEXT: vpbroadcastw %edi, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x7b,0xd7]
58 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
59 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
60 ; X64-NEXT: retq # encoding: [0xc3]
61 %res = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 -1)
62 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> %x1, i8 %mask)
63 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.128(i16 %x0, <8 x i16> zeroinitializer, i8 %mask)
64 %res3 = add <8 x i16> %res, %res1
65 %res4 = add <8 x i16> %res2, %res3
66 ret <8 x i16> %res4
67 }
70 declare <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8, <32 x i8>, i32)
72 define <32 x i8>@test_int_x86_avx512_mask_pbroadcast_b_gpr_256(i8 %x0, <32 x i8> %x1, i32 %mask) {
73 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
74 ; X86: # %bb.0:
75 ; X86-NEXT: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
76 ; X86-NEXT: vpbroadcastb %eax, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xc8]
77 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
78 ; X86-NEXT: vpbroadcastb %eax, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc0]
79 ; X86-NEXT: vpbroadcastb %eax, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd0]
80 ; X86-NEXT: vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
81 ; X86-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
82 ; X86-NEXT: retl # encoding: [0xc3]
83 ;
84 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_b_gpr_256:
85 ; X64: # %bb.0:
86 ; X64-NEXT: vpbroadcastb %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7a,0xcf]
87 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
88 ; X64-NEXT: vpbroadcastb %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7a,0xc7]
89 ; X64-NEXT: vpbroadcastb %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7a,0xd7]
90 ; X64-NEXT: vpaddb %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc2]
91 ; X64-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
92 ; X64-NEXT: retq # encoding: [0xc3]
93 %res = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 -1)
94 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> %x1, i32 %mask)
95 %res2 = call <32 x i8> @llvm.x86.avx512.mask.pbroadcast.b.gpr.256(i8 %x0, <32 x i8> zeroinitializer, i32 %mask)
96 %res3 = add <32 x i8> %res, %res1
97 %res4 = add <32 x i8> %res2, %res3
98 ret <32 x i8> %res4
99 }
103 declare <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16, <16 x i16>, i16)
105 define <16 x i16>@test_int_x86_avx512_mask_pbroadcast_w_gpr_256(i16 %x0, <16 x i16> %x1, i16 %mask) {
106 ; X86-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
107 ; X86: # %bb.0:
108 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
109 ; X86-NEXT: vpbroadcastw %eax, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xc8]
110 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
111 ; X86-NEXT: vpbroadcastw %eax, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc0]
112 ; X86-NEXT: vpbroadcastw %eax, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd0]
113 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
114 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
115 ; X86-NEXT: retl # encoding: [0xc3]
116 ;
117 ; X64-LABEL: test_int_x86_avx512_mask_pbroadcast_w_gpr_256:
118 ; X64: # %bb.0:
119 ; X64-NEXT: vpbroadcastw %edi, %ymm1 # encoding: [0x62,0xf2,0x7d,0x28,0x7b,0xcf]
120 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
121 ; X64-NEXT: vpbroadcastw %edi, %ymm0 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x7b,0xc7]
122 ; X64-NEXT: vpbroadcastw %edi, %ymm2 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x7b,0xd7]
123 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
124 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
125 ; X64-NEXT: retq # encoding: [0xc3]
126 %res = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 -1)
127 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> %x1, i16 %mask)
128 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pbroadcast.w.gpr.256(i16 %x0, <16 x i16> zeroinitializer, i16 %mask)
129 %res3 = add <16 x i16> %res, %res1
130 %res4 = add <16 x i16> %res2, %res3
131 ret <16 x i16> %res4
132 }
134 declare <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8>, <32 x i8>, i32)
136 define <32 x i8>@test_int_x86_avx512_pbroadcastb_256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask) {
137 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_256:
138 ; X86: # %bb.0:
139 ; X86-NEXT: vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
140 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
141 ; X86-NEXT: vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
142 ; X86-NEXT: vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
143 ; X86-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
144 ; X86-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
145 ; X86-NEXT: retl # encoding: [0xc3]
146 ;
147 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_256:
148 ; X64: # %bb.0:
149 ; X64-NEXT: vpbroadcastb %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x78,0xd0]
150 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
151 ; X64-NEXT: vpbroadcastb %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x78,0xc8]
152 ; X64-NEXT: vpbroadcastb %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x78,0xc0]
153 ; X64-NEXT: vpaddb %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc0]
154 ; X64-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
155 ; X64-NEXT: retq # encoding: [0xc3]
156 %res = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 -1)
157 %res1 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> %x1, i32 %mask)
158 %res2 = call <32 x i8> @llvm.x86.avx512.pbroadcastb.256(<16 x i8> %x0, <32 x i8> zeroinitializer, i32 %mask)
159 %res3 = add <32 x i8> %res, %res1
160 %res4 = add <32 x i8> %res2, %res3
161 ret <32 x i8> %res4
162 }
164 declare <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8>, <16 x i8>, i16)
166 define <16 x i8>@test_int_x86_avx512_pbroadcastb_128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask) {
167 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_128:
168 ; X86: # %bb.0:
169 ; X86-NEXT: vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
170 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
171 ; X86-NEXT: vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
172 ; X86-NEXT: vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
173 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
174 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
175 ; X86-NEXT: retl # encoding: [0xc3]
176 ;
177 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_128:
178 ; X64: # %bb.0:
179 ; X64-NEXT: vpbroadcastb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x78,0xd0]
180 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
181 ; X64-NEXT: vpbroadcastb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x78,0xc8]
182 ; X64-NEXT: vpbroadcastb %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x78,0xc0]
183 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
184 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
185 ; X64-NEXT: retq # encoding: [0xc3]
186 %res = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
187 %res1 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> %x1, i16 %mask)
188 %res2 = call <16 x i8> @llvm.x86.avx512.pbroadcastb.128(<16 x i8> %x0, <16 x i8> zeroinitializer, i16 %mask)
189 %res3 = add <16 x i8> %res, %res1
190 %res4 = add <16 x i8> %res2, %res3
191 ret <16 x i8> %res4
192 }
194 declare <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16>, <16 x i16>, i16)
196 define <16 x i16>@test_int_x86_avx512_pbroadcastw_256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask) {
197 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_256:
198 ; X86: # %bb.0:
199 ; X86-NEXT: vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
200 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
201 ; X86-NEXT: vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
202 ; X86-NEXT: vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
203 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
204 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
205 ; X86-NEXT: retl # encoding: [0xc3]
206 ;
207 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_256:
208 ; X64: # %bb.0:
209 ; X64-NEXT: vpbroadcastw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x79,0xd0]
210 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
211 ; X64-NEXT: vpbroadcastw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x79,0xc8]
212 ; X64-NEXT: vpbroadcastw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x79,0xc0]
213 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
214 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
215 ; X64-NEXT: retq # encoding: [0xc3]
216 %res = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 -1)
217 %res1 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> %x1, i16 %mask)
218 %res2 = call <16 x i16> @llvm.x86.avx512.pbroadcastw.256(<8 x i16> %x0, <16 x i16> zeroinitializer, i16 %mask)
219 %res3 = add <16 x i16> %res, %res1
220 %res4 = add <16 x i16> %res2, %res3
221 ret <16 x i16> %res4
222 }
224 declare <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16>, <8 x i16>, i8)
226 define <8 x i16>@test_int_x86_avx512_pbroadcastw_128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask) {
227 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_128:
228 ; X86: # %bb.0:
229 ; X86-NEXT: vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
230 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
231 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
232 ; X86-NEXT: vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
233 ; X86-NEXT: vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
234 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
235 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
236 ; X86-NEXT: retl # encoding: [0xc3]
237 ;
238 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_128:
239 ; X64: # %bb.0:
240 ; X64-NEXT: vpbroadcastw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x79,0xd0]
241 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
242 ; X64-NEXT: vpbroadcastw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x79,0xc8]
243 ; X64-NEXT: vpbroadcastw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x79,0xc0]
244 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
245 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
246 ; X64-NEXT: retq # encoding: [0xc3]
247 %res = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
248 %res1 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> %x1, i8 %mask)
249 %res2 = call <8 x i16> @llvm.x86.avx512.pbroadcastw.128(<8 x i16> %x0, <8 x i16> zeroinitializer, i8 %mask)
250 %res3 = add <8 x i16> %res, %res1
251 %res4 = add <8 x i16> %res2, %res3
252 ret <8 x i16> %res4
253 }
255 declare <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8>, <64 x i8>, i64)
257 define <64 x i8>@test_int_x86_avx512_pbroadcastb_512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask) {
258 ; X86-LABEL: test_int_x86_avx512_pbroadcastb_512:
259 ; X86: # %bb.0:
260 ; X86-NEXT: vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
261 ; X86-NEXT: kmovq {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf8,0x90,0x4c,0x24,0x04]
262 ; X86-NEXT: vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
263 ; X86-NEXT: vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
264 ; X86-NEXT: vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
265 ; X86-NEXT: vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
266 ; X86-NEXT: retl # encoding: [0xc3]
267 ;
268 ; X64-LABEL: test_int_x86_avx512_pbroadcastb_512:
269 ; X64: # %bb.0:
270 ; X64-NEXT: vpbroadcastb %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x78,0xd0]
271 ; X64-NEXT: kmovq %rdi, %k1 # encoding: [0xc4,0xe1,0xfb,0x92,0xcf]
272 ; X64-NEXT: vpbroadcastb %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x78,0xc8]
273 ; X64-NEXT: vpbroadcastb %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x78,0xc0]
274 ; X64-NEXT: vpaddb %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfc,0xc0]
275 ; X64-NEXT: vpaddb %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfc,0xc0]
276 ; X64-NEXT: retq # encoding: [0xc3]
277 %res = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 -1)
278 %res1 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> %x1, i64 %mask)
279 %res2 = call <64 x i8> @llvm.x86.avx512.pbroadcastb.512(<16 x i8> %x0, <64 x i8> zeroinitializer, i64 %mask)
280 %res3 = add <64 x i8> %res, %res1
281 %res4 = add <64 x i8> %res2, %res3
282 ret <64 x i8> %res4
283 }
285 declare <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16>, <32 x i16>, i32)
287 define <32 x i16>@test_int_x86_avx512_pbroadcastw_512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask) {
288 ; X86-LABEL: test_int_x86_avx512_pbroadcastw_512:
289 ; X86: # %bb.0:
290 ; X86-NEXT: vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
291 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
292 ; X86-NEXT: vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
293 ; X86-NEXT: vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
294 ; X86-NEXT: vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
295 ; X86-NEXT: vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
296 ; X86-NEXT: retl # encoding: [0xc3]
297 ;
298 ; X64-LABEL: test_int_x86_avx512_pbroadcastw_512:
299 ; X64: # %bb.0:
300 ; X64-NEXT: vpbroadcastw %xmm0, %zmm2 # encoding: [0x62,0xf2,0x7d,0x48,0x79,0xd0]
301 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
302 ; X64-NEXT: vpbroadcastw %xmm0, %zmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x49,0x79,0xc8]
303 ; X64-NEXT: vpbroadcastw %xmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xc9,0x79,0xc0]
304 ; X64-NEXT: vpaddw %zmm0, %zmm1, %zmm0 # encoding: [0x62,0xf1,0x75,0x48,0xfd,0xc0]
305 ; X64-NEXT: vpaddw %zmm0, %zmm2, %zmm0 # encoding: [0x62,0xf1,0x6d,0x48,0xfd,0xc0]
306 ; X64-NEXT: retq # encoding: [0xc3]
307 %res = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 -1)
308 %res1 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> %x1, i32 %mask)
309 %res2 = call <32 x i16> @llvm.x86.avx512.pbroadcastw.512(<8 x i16> %x0, <32 x i16> zeroinitializer, i32 %mask)
310 %res3 = add <32 x i16> %res, %res1
311 %res4 = add <32 x i16> %res2, %res3
312 ret <32 x i16> %res4
313 }
315 declare void @llvm.x86.avx512.mask.storeu.b.128(i8*, <16 x i8>, i16)
317 define void@test_int_x86_avx512_mask_storeu_b_128(i8* %ptr1, i8* %ptr2, <16 x i8> %x1, i16 %x2) {
318 ; X86-LABEL: test_int_x86_avx512_mask_storeu_b_128:
319 ; X86: # %bb.0:
320 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
321 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
322 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
323 ; X86-NEXT: vmovdqu8 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x01]
324 ; X86-NEXT: vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
325 ; X86-NEXT: retl # encoding: [0xc3]
326 ;
327 ; X64-LABEL: test_int_x86_avx512_mask_storeu_b_128:
328 ; X64: # %bb.0:
329 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
330 ; X64-NEXT: vmovdqu8 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x7f,0x07]
331 ; X64-NEXT: vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
332 ; X64-NEXT: retq # encoding: [0xc3]
333 call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr1, <16 x i8> %x1, i16 %x2)
334 call void @llvm.x86.avx512.mask.storeu.b.128(i8* %ptr2, <16 x i8> %x1, i16 -1)
335 ret void
336 }
338 declare void @llvm.x86.avx512.mask.storeu.b.256(i8*, <32 x i8>, i32)
340 define void@test_int_x86_avx512_mask_storeu_b_256(i8* %ptr1, i8* %ptr2, <32 x i8> %x1, i32 %x2) {
341 ; X86-LABEL: test_int_x86_avx512_mask_storeu_b_256:
342 ; X86: # %bb.0:
343 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
344 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
345 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
346 ; X86-NEXT: vmovdqu8 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x01]
347 ; X86-NEXT: vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
348 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
349 ; X86-NEXT: retl # encoding: [0xc3]
350 ;
351 ; X64-LABEL: test_int_x86_avx512_mask_storeu_b_256:
352 ; X64: # %bb.0:
353 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
354 ; X64-NEXT: vmovdqu8 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x7f,0x07]
355 ; X64-NEXT: vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
356 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
357 ; X64-NEXT: retq # encoding: [0xc3]
358 call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr1, <32 x i8> %x1, i32 %x2)
359 call void @llvm.x86.avx512.mask.storeu.b.256(i8* %ptr2, <32 x i8> %x1, i32 -1)
360 ret void
361 }
363 declare void @llvm.x86.avx512.mask.storeu.w.128(i8*, <8 x i16>, i8)
365 define void@test_int_x86_avx512_mask_storeu_w_128(i8* %ptr1, i8* %ptr2, <8 x i16> %x1, i8 %x2) {
366 ; X86-LABEL: test_int_x86_avx512_mask_storeu_w_128:
367 ; X86: # %bb.0:
368 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
369 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
370 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
371 ; X86-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
372 ; X86-NEXT: vmovdqu16 %xmm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x01]
373 ; X86-NEXT: vmovdqu %xmm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x00]
374 ; X86-NEXT: retl # encoding: [0xc3]
375 ;
376 ; X64-LABEL: test_int_x86_avx512_mask_storeu_w_128:
377 ; X64: # %bb.0:
378 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
379 ; X64-NEXT: vmovdqu16 %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x7f,0x07]
380 ; X64-NEXT: vmovdqu %xmm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x7f,0x06]
381 ; X64-NEXT: retq # encoding: [0xc3]
382 call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr1, <8 x i16> %x1, i8 %x2)
383 call void @llvm.x86.avx512.mask.storeu.w.128(i8* %ptr2, <8 x i16> %x1, i8 -1)
384 ret void
385 }
387 declare void @llvm.x86.avx512.mask.storeu.w.256(i8*, <16 x i16>, i16)
389 define void@test_int_x86_avx512_mask_storeu_w_256(i8* %ptr1, i8* %ptr2, <16 x i16> %x1, i16 %x2) {
390 ; X86-LABEL: test_int_x86_avx512_mask_storeu_w_256:
391 ; X86: # %bb.0:
392 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
393 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
394 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
395 ; X86-NEXT: vmovdqu16 %ymm0, (%ecx) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x01]
396 ; X86-NEXT: vmovdqu %ymm0, (%eax) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x00]
397 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
398 ; X86-NEXT: retl # encoding: [0xc3]
399 ;
400 ; X64-LABEL: test_int_x86_avx512_mask_storeu_w_256:
401 ; X64: # %bb.0:
402 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
403 ; X64-NEXT: vmovdqu16 %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x7f,0x07]
404 ; X64-NEXT: vmovdqu %ymm0, (%rsi) # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x7f,0x06]
405 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
406 ; X64-NEXT: retq # encoding: [0xc3]
407 call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr1, <16 x i16> %x1, i16 %x2)
408 call void @llvm.x86.avx512.mask.storeu.w.256(i8* %ptr2, <16 x i16> %x1, i16 -1)
409 ret void
410 }
412 declare <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8*, <8 x i16>, i8)
414 define <8 x i16>@test_int_x86_avx512_mask_loadu_w_128(i8* %ptr, i8* %ptr2, <8 x i16> %x1, i8 %mask) {
415 ; X86-LABEL: test_int_x86_avx512_mask_loadu_w_128:
416 ; X86: # %bb.0:
417 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
418 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
419 ; X86-NEXT: vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
420 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %edx # encoding: [0x0f,0xb6,0x54,0x24,0x0c]
421 ; X86-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
422 ; X86-NEXT: vmovdqu16 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x00]
423 ; X86-NEXT: vmovdqu16 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x09]
424 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
425 ; X86-NEXT: retl # encoding: [0xc3]
426 ;
427 ; X64-LABEL: test_int_x86_avx512_mask_loadu_w_128:
428 ; X64: # %bb.0:
429 ; X64-NEXT: vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
430 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
431 ; X64-NEXT: vmovdqu16 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0xff,0x09,0x6f,0x06]
432 ; X64-NEXT: vmovdqu16 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0x89,0x6f,0x0f]
433 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
434 ; X64-NEXT: retq # encoding: [0xc3]
435 %res0 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> %x1, i8 -1)
436 %res = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr2, <8 x i16> %res0, i8 %mask)
437 %res1 = call <8 x i16> @llvm.x86.avx512.mask.loadu.w.128(i8* %ptr, <8 x i16> zeroinitializer, i8 %mask)
438 %res2 = add <8 x i16> %res, %res1
439 ret <8 x i16> %res2
440 }
442 declare <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8*, <16 x i16>, i16)
444 define <16 x i16>@test_int_x86_avx512_mask_loadu_w_256(i8* %ptr, i8* %ptr2, <16 x i16> %x1, i16 %mask) {
445 ; X86-LABEL: test_int_x86_avx512_mask_loadu_w_256:
446 ; X86: # %bb.0:
447 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
448 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
449 ; X86-NEXT: vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
450 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
451 ; X86-NEXT: vmovdqu16 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x00]
452 ; X86-NEXT: vmovdqu16 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x09]
453 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
454 ; X86-NEXT: retl # encoding: [0xc3]
455 ;
456 ; X64-LABEL: test_int_x86_avx512_mask_loadu_w_256:
457 ; X64: # %bb.0:
458 ; X64-NEXT: vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
459 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
460 ; X64-NEXT: vmovdqu16 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0xff,0x29,0x6f,0x06]
461 ; X64-NEXT: vmovdqu16 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0xff,0xa9,0x6f,0x0f]
462 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
463 ; X64-NEXT: retq # encoding: [0xc3]
464 %res0 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> %x1, i16 -1)
465 %res = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr2, <16 x i16> %res0, i16 %mask)
466 %res1 = call <16 x i16> @llvm.x86.avx512.mask.loadu.w.256(i8* %ptr, <16 x i16> zeroinitializer, i16 %mask)
467 %res2 = add <16 x i16> %res, %res1
468 ret <16 x i16> %res2
469 }
471 declare <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8*, <16 x i8>, i16)
473 define <16 x i8>@test_int_x86_avx512_mask_loadu_b_128(i8* %ptr, i8* %ptr2, <16 x i8> %x1, i16 %mask) {
474 ; X86-LABEL: test_int_x86_avx512_mask_loadu_b_128:
475 ; X86: # %bb.0:
476 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
477 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
478 ; X86-NEXT: vmovdqu (%ecx), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x01]
479 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x0c]
480 ; X86-NEXT: vmovdqu8 (%eax), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x00]
481 ; X86-NEXT: vmovdqu8 (%ecx), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x09]
482 ; X86-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
483 ; X86-NEXT: retl # encoding: [0xc3]
484 ;
485 ; X64-LABEL: test_int_x86_avx512_mask_loadu_b_128:
486 ; X64: # %bb.0:
487 ; X64-NEXT: vmovdqu (%rdi), %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x6f,0x07]
488 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
489 ; X64-NEXT: vmovdqu8 (%rsi), %xmm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x6f,0x06]
490 ; X64-NEXT: vmovdqu8 (%rdi), %xmm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x6f,0x0f]
491 ; X64-NEXT: vpaddb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc1]
492 ; X64-NEXT: retq # encoding: [0xc3]
493 %res0 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> %x1, i16 -1)
494 %res = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr2, <16 x i8> %res0, i16 %mask)
495 %res1 = call <16 x i8> @llvm.x86.avx512.mask.loadu.b.128(i8* %ptr, <16 x i8> zeroinitializer, i16 %mask)
496 %res2 = add <16 x i8> %res, %res1
497 ret <16 x i8> %res2
498 }
500 declare <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8*, <32 x i8>, i32)
502 define <32 x i8>@test_int_x86_avx512_mask_loadu_b_256(i8* %ptr, i8* %ptr2, <32 x i8> %x1, i32 %mask) {
503 ; X86-LABEL: test_int_x86_avx512_mask_loadu_b_256:
504 ; X86: # %bb.0:
505 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x08]
506 ; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x04]
507 ; X86-NEXT: vmovdqu (%ecx), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x01]
508 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x0c]
509 ; X86-NEXT: vmovdqu8 (%eax), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x00]
510 ; X86-NEXT: vmovdqu8 (%ecx), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x09]
511 ; X86-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
512 ; X86-NEXT: retl # encoding: [0xc3]
513 ;
514 ; X64-LABEL: test_int_x86_avx512_mask_loadu_b_256:
515 ; X64: # %bb.0:
516 ; X64-NEXT: vmovdqu (%rdi), %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x6f,0x07]
517 ; X64-NEXT: kmovd %edx, %k1 # encoding: [0xc5,0xfb,0x92,0xca]
518 ; X64-NEXT: vmovdqu8 (%rsi), %ymm0 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x6f,0x06]
519 ; X64-NEXT: vmovdqu8 (%rdi), %ymm1 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x6f,0x0f]
520 ; X64-NEXT: vpaddb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc1]
521 ; X64-NEXT: retq # encoding: [0xc3]
522 %res0 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> %x1, i32 -1)
523 %res = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr2, <32 x i8> %res0, i32 %mask)
524 %res1 = call <32 x i8> @llvm.x86.avx512.mask.loadu.b.256(i8* %ptr, <32 x i8> zeroinitializer, i32 %mask)
525 %res2 = add <32 x i8> %res, %res1
526 ret <32 x i8> %res2
527 }
529 declare <16 x i8> @llvm.x86.avx512.mask.palignr.128(<16 x i8>, <16 x i8>, i32, <16 x i8>, i16)
531 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) {
532 ; X86-LABEL: test_int_x86_avx512_mask_palignr_128:
533 ; X86: # %bb.0:
534 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
535 ; X86-NEXT: # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
536 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
537 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
538 ; X86-NEXT: # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
539 ; X86-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
540 ; X86-NEXT: # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
541 ; X86-NEXT: vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
542 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
543 ; X86-NEXT: retl # encoding: [0xc3]
544 ;
545 ; X64-LABEL: test_int_x86_avx512_mask_palignr_128:
546 ; X64: # %bb.0:
547 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x79,0x0f,0xd9,0x02]
548 ; X64-NEXT: # xmm3 = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
549 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
550 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x0f,0xd1,0x02]
551 ; X64-NEXT: # xmm2 {%k1} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
552 ; X64-NEXT: vpalignr $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x0f,0xc1,0x02]
553 ; X64-NEXT: # xmm0 {%k1} {z} = xmm1[2,3,4,5,6,7,8,9,10,11,12,13,14,15],xmm0[0,1]
554 ; X64-NEXT: vpaddb %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfc,0xc3]
555 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
556 ; X64-NEXT: retq # encoding: [0xc3]
557 %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)
558 %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)
559 %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)
560 %res3 = add <16 x i8> %res, %res1
561 %res4 = add <16 x i8> %res3, %res2
562 ret <16 x i8> %res4
563 }
565 declare <32 x i8> @llvm.x86.avx512.mask.palignr.256(<32 x i8>, <32 x i8>, i32, <32 x i8>, i32)
567 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) {
568 ; X86-LABEL: test_int_x86_avx512_mask_palignr_256:
569 ; X86: # %bb.0:
570 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
571 ; 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]
572 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
573 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
574 ; 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]
575 ; X86-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
576 ; 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]
577 ; X86-NEXT: vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
578 ; X86-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
579 ; X86-NEXT: retl # encoding: [0xc3]
580 ;
581 ; X64-LABEL: test_int_x86_avx512_mask_palignr_256:
582 ; X64: # %bb.0:
583 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x7d,0x0f,0xd9,0x02]
584 ; 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]
585 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
586 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x0f,0xd1,0x02]
587 ; 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]
588 ; X64-NEXT: vpalignr $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x0f,0xc1,0x02]
589 ; 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]
590 ; X64-NEXT: vpaddb %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfc,0xc3]
591 ; X64-NEXT: vpaddb %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc0]
592 ; X64-NEXT: retq # encoding: [0xc3]
593 %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)
594 %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)
595 %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)
596 %res3 = add <32 x i8> %res, %res1
597 %res4 = add <32 x i8> %res3, %res2
598 ret <32 x i8> %res4
599 }
601 declare <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16>, i32, <8 x i16>, i8)
603 define <8 x i16>@test_int_x86_avx512_mask_pshufh_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
604 ; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
605 ; X86: # %bb.0:
606 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
607 ; X86-NEXT: # xmm2 = xmm0[0,1,2,3,7,4,4,4]
608 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
609 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
610 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
611 ; X86-NEXT: # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
612 ; X86-NEXT: vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
613 ; X86-NEXT: # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
614 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
615 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
616 ; X86-NEXT: retl # encoding: [0xc3]
617 ;
618 ; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_128:
619 ; X64: # %bb.0:
620 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfa,0x70,0xd0,0x03]
621 ; X64-NEXT: # xmm2 = xmm0[0,1,2,3,7,4,4,4]
622 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
623 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x09,0x70,0xc8,0x03]
624 ; X64-NEXT: # xmm1 {%k1} = xmm0[0,1,2,3,7,4,4,4]
625 ; X64-NEXT: vpshufhw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0x89,0x70,0xc0,0x03]
626 ; X64-NEXT: # xmm0 {%k1} {z} = xmm0[0,1,2,3,7,4,4,4]
627 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
628 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
629 ; X64-NEXT: retq # encoding: [0xc3]
630 %res = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
631 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
632 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufh.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
633 %res3 = add <8 x i16> %res, %res1
634 %res4 = add <8 x i16> %res3, %res2
635 ret <8 x i16> %res4
636 }
638 declare <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16>, i32, <16 x i16>, i16)
640 define <16 x i16>@test_int_x86_avx512_mask_pshufh_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
641 ; X86-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
642 ; X86: # %bb.0:
643 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
644 ; X86-NEXT: # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
645 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
646 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
647 ; X86-NEXT: # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
648 ; X86-NEXT: vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
649 ; X86-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
650 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
651 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
652 ; X86-NEXT: retl # encoding: [0xc3]
653 ;
654 ; X64-LABEL: test_int_x86_avx512_mask_pshufh_w_256:
655 ; X64: # %bb.0:
656 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xfe,0x70,0xd0,0x03]
657 ; X64-NEXT: # ymm2 = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
658 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
659 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7e,0x29,0x70,0xc8,0x03]
660 ; X64-NEXT: # ymm1 {%k1} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
661 ; X64-NEXT: vpshufhw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7e,0xa9,0x70,0xc0,0x03]
662 ; X64-NEXT: # ymm0 {%k1} {z} = ymm0[0,1,2,3,7,4,4,4,8,9,10,11,15,12,12,12]
663 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
664 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
665 ; X64-NEXT: retq # encoding: [0xc3]
666 %res = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
667 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
668 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufh.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
669 %res3 = add <16 x i16> %res, %res1
670 %res4 = add <16 x i16> %res3, %res2
671 ret <16 x i16> %res4
672 }
674 declare <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16>, i32, <8 x i16>, i8)
676 define <8 x i16>@test_int_x86_avx512_mask_pshufl_w_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
677 ; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
678 ; X86: # %bb.0:
679 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
680 ; X86-NEXT: # xmm2 = xmm0[3,0,0,0,4,5,6,7]
681 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
682 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
683 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
684 ; X86-NEXT: # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
685 ; X86-NEXT: vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
686 ; X86-NEXT: # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
687 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
688 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
689 ; X86-NEXT: retl # encoding: [0xc3]
690 ;
691 ; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_128:
692 ; X64: # %bb.0:
693 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xfb,0x70,0xd0,0x03]
694 ; X64-NEXT: # xmm2 = xmm0[3,0,0,0,4,5,6,7]
695 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
696 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x09,0x70,0xc8,0x03]
697 ; X64-NEXT: # xmm1 {%k1} = xmm0[3,0,0,0,4,5,6,7]
698 ; X64-NEXT: vpshuflw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0x89,0x70,0xc0,0x03]
699 ; X64-NEXT: # xmm0 {%k1} {z} = xmm0[3,0,0,0,4,5,6,7]
700 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
701 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
702 ; X64-NEXT: retq # encoding: [0xc3]
703 %res = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
704 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
705 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pshufl.w.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
706 %res3 = add <8 x i16> %res, %res1
707 %res4 = add <8 x i16> %res3, %res2
708 ret <8 x i16> %res4
709 }
711 declare <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16>, i32, <16 x i16>, i16)
713 define <16 x i16>@test_int_x86_avx512_mask_pshufl_w_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
714 ; X86-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
715 ; X86: # %bb.0:
716 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
717 ; X86-NEXT: # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
718 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
719 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
720 ; X86-NEXT: # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
721 ; X86-NEXT: vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
722 ; X86-NEXT: # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
723 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
724 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
725 ; X86-NEXT: retl # encoding: [0xc3]
726 ;
727 ; X64-LABEL: test_int_x86_avx512_mask_pshufl_w_256:
728 ; X64: # %bb.0:
729 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xff,0x70,0xd0,0x03]
730 ; X64-NEXT: # ymm2 = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
731 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
732 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7f,0x29,0x70,0xc8,0x03]
733 ; X64-NEXT: # ymm1 {%k1} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
734 ; X64-NEXT: vpshuflw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7f,0xa9,0x70,0xc0,0x03]
735 ; X64-NEXT: # ymm0 {%k1} {z} = ymm0[3,0,0,0,4,5,6,7,11,8,8,8,12,13,14,15]
736 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
737 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
738 ; X64-NEXT: retq # encoding: [0xc3]
739 %res = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
740 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
741 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pshufl.w.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
742 %res3 = add <16 x i16> %res, %res1
743 %res4 = add <16 x i16> %res3, %res2
744 ret <16 x i16> %res4
745 }
747 define i32 @test_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b) {
748 ; CHECK-LABEL: test_pcmpeq_b_256:
749 ; CHECK: # %bb.0:
750 ; CHECK-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
751 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
752 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
753 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
754 %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
755 ret i32 %res
756 }
758 define i32 @test_mask_pcmpeq_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
759 ; X86-LABEL: test_mask_pcmpeq_b_256:
760 ; X86: # %bb.0:
761 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
762 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
763 ; X86-NEXT: andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
764 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
765 ; X86-NEXT: retl # encoding: [0xc3]
766 ;
767 ; X64-LABEL: test_mask_pcmpeq_b_256:
768 ; X64: # %bb.0:
769 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
770 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
771 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
772 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
773 ; X64-NEXT: retq # encoding: [0xc3]
774 %res = call i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
775 ret i32 %res
776 }
778 declare i32 @llvm.x86.avx512.mask.pcmpeq.b.256(<32 x i8>, <32 x i8>, i32)
780 define i16 @test_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b) {
781 ; CHECK-LABEL: test_pcmpeq_w_256:
782 ; CHECK: # %bb.0:
783 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
784 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
785 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
786 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
787 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
788 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
789 ret i16 %res
790 }
792 define i16 @test_mask_pcmpeq_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
793 ; X86-LABEL: test_mask_pcmpeq_w_256:
794 ; X86: # %bb.0:
795 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
796 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
797 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
798 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
799 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
800 ; X86-NEXT: retl # encoding: [0xc3]
801 ;
802 ; X64-LABEL: test_mask_pcmpeq_w_256:
803 ; X64: # %bb.0:
804 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
805 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
806 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
807 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
808 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
809 ; X64-NEXT: retq # encoding: [0xc3]
810 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
811 ret i16 %res
812 }
814 declare i16 @llvm.x86.avx512.mask.pcmpeq.w.256(<16 x i16>, <16 x i16>, i16)
816 define i32 @test_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b) {
817 ; CHECK-LABEL: test_pcmpgt_b_256:
818 ; CHECK: # %bb.0:
819 ; CHECK-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
820 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
821 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
822 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
823 %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 -1)
824 ret i32 %res
825 }
827 define i32 @test_mask_pcmpgt_b_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
828 ; X86-LABEL: test_mask_pcmpgt_b_256:
829 ; X86: # %bb.0:
830 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
831 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
832 ; X86-NEXT: andl {{[0-9]+}}(%esp), %eax # encoding: [0x23,0x44,0x24,0x04]
833 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
834 ; X86-NEXT: retl # encoding: [0xc3]
835 ;
836 ; X64-LABEL: test_mask_pcmpgt_b_256:
837 ; X64: # %bb.0:
838 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
839 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
840 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
841 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
842 ; X64-NEXT: retq # encoding: [0xc3]
843 %res = call i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8> %a, <32 x i8> %b, i32 %mask)
844 ret i32 %res
845 }
847 declare i32 @llvm.x86.avx512.mask.pcmpgt.b.256(<32 x i8>, <32 x i8>, i32)
849 define i16 @test_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b) {
850 ; CHECK-LABEL: test_pcmpgt_w_256:
851 ; CHECK: # %bb.0:
852 ; CHECK-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
853 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
854 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
855 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
856 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
857 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 -1)
858 ret i16 %res
859 }
861 define i16 @test_mask_pcmpgt_w_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
862 ; X86-LABEL: test_mask_pcmpgt_w_256:
863 ; X86: # %bb.0:
864 ; X86-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
865 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
866 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
867 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
868 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
869 ; X86-NEXT: retl # encoding: [0xc3]
870 ;
871 ; X64-LABEL: test_mask_pcmpgt_w_256:
872 ; X64: # %bb.0:
873 ; X64-NEXT: vpcmpgtw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xc1]
874 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
875 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
876 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
877 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
878 ; X64-NEXT: retq # encoding: [0xc3]
879 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16> %a, <16 x i16> %b, i16 %mask)
880 ret i16 %res
881 }
883 declare i16 @llvm.x86.avx512.mask.pcmpgt.w.256(<16 x i16>, <16 x i16>, i16)
885 define i16 @test_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b) {
886 ; CHECK-LABEL: test_pcmpeq_b_128:
887 ; CHECK: # %bb.0:
888 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
889 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
890 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
891 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
892 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
893 ret i16 %res
894 }
896 define i16 @test_mask_pcmpeq_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
897 ; X86-LABEL: test_mask_pcmpeq_b_128:
898 ; X86: # %bb.0:
899 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
900 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
901 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
902 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
903 ; X86-NEXT: retl # encoding: [0xc3]
904 ;
905 ; X64-LABEL: test_mask_pcmpeq_b_128:
906 ; X64: # %bb.0:
907 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
908 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
909 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
910 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
911 ; X64-NEXT: retq # encoding: [0xc3]
912 %res = call i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
913 ret i16 %res
914 }
916 declare i16 @llvm.x86.avx512.mask.pcmpeq.b.128(<16 x i8>, <16 x i8>, i16)
918 define i8 @test_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b) {
919 ; CHECK-LABEL: test_pcmpeq_w_128:
920 ; CHECK: # %bb.0:
921 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
922 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
923 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
924 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
925 %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
926 ret i8 %res
927 }
929 define i8 @test_mask_pcmpeq_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
930 ; X86-LABEL: test_mask_pcmpeq_w_128:
931 ; X86: # %bb.0:
932 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
933 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
934 ; X86-NEXT: andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
935 ; X86-NEXT: # kill: def $al killed $al killed $eax
936 ; X86-NEXT: retl # encoding: [0xc3]
937 ;
938 ; X64-LABEL: test_mask_pcmpeq_w_128:
939 ; X64: # %bb.0:
940 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
941 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
942 ; X64-NEXT: andb %dil, %al # encoding: [0x40,0x20,0xf8]
943 ; X64-NEXT: # kill: def $al killed $al killed $eax
944 ; X64-NEXT: retq # encoding: [0xc3]
945 %res = call i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
946 ret i8 %res
947 }
949 declare i8 @llvm.x86.avx512.mask.pcmpeq.w.128(<8 x i16>, <8 x i16>, i8)
951 define i16 @test_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b) {
952 ; CHECK-LABEL: test_pcmpgt_b_128:
953 ; CHECK: # %bb.0:
954 ; CHECK-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
955 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
956 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
957 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
958 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 -1)
959 ret i16 %res
960 }
962 define i16 @test_mask_pcmpgt_b_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
963 ; X86-LABEL: test_mask_pcmpgt_b_128:
964 ; X86: # %bb.0:
965 ; X86-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
966 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
967 ; X86-NEXT: andw {{[0-9]+}}(%esp), %ax # encoding: [0x66,0x23,0x44,0x24,0x04]
968 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
969 ; X86-NEXT: retl # encoding: [0xc3]
970 ;
971 ; X64-LABEL: test_mask_pcmpgt_b_128:
972 ; X64: # %bb.0:
973 ; X64-NEXT: vpcmpgtb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xc1]
974 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
975 ; X64-NEXT: andl %edi, %eax # encoding: [0x21,0xf8]
976 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
977 ; X64-NEXT: retq # encoding: [0xc3]
978 %res = call i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8> %a, <16 x i8> %b, i16 %mask)
979 ret i16 %res
980 }
982 declare i16 @llvm.x86.avx512.mask.pcmpgt.b.128(<16 x i8>, <16 x i8>, i16)
984 define i8 @test_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b) {
985 ; CHECK-LABEL: test_pcmpgt_w_128:
986 ; CHECK: # %bb.0:
987 ; CHECK-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
988 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
989 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
990 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
991 %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 -1)
992 ret i8 %res
993 }
995 define i8 @test_mask_pcmpgt_w_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
996 ; X86-LABEL: test_mask_pcmpgt_w_128:
997 ; X86: # %bb.0:
998 ; X86-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
999 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
1000 ; X86-NEXT: andb {{[0-9]+}}(%esp), %al # encoding: [0x22,0x44,0x24,0x04]
1001 ; X86-NEXT: # kill: def $al killed $al killed $eax
1002 ; X86-NEXT: retl # encoding: [0xc3]
1003 ;
1004 ; X64-LABEL: test_mask_pcmpgt_w_128:
1005 ; X64: # %bb.0:
1006 ; X64-NEXT: vpcmpgtw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xc1]
1007 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
1008 ; X64-NEXT: andb %dil, %al # encoding: [0x40,0x20,0xf8]
1009 ; X64-NEXT: # kill: def $al killed $al killed $eax
1010 ; X64-NEXT: retq # encoding: [0xc3]
1011 %res = call i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16> %a, <8 x i16> %b, i8 %mask)
1012 ret i8 %res
1013 }
1015 declare i8 @llvm.x86.avx512.mask.pcmpgt.w.128(<8 x i16>, <8 x i16>, i8)
1017 declare <16 x i8> @llvm.x86.avx512.mask.punpckhb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
1019 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) {
1020 ; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
1021 ; X86: # %bb.0:
1022 ; X86-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
1023 ; 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]
1024 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1025 ; X86-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
1026 ; 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]
1027 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1028 ; X86-NEXT: retl # encoding: [0xc3]
1029 ;
1030 ; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_128:
1031 ; X64: # %bb.0:
1032 ; X64-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x68,0xd9]
1033 ; 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]
1034 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1035 ; X64-NEXT: vpunpckhbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x68,0xd1]
1036 ; 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]
1037 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1038 ; X64-NEXT: retq # encoding: [0xc3]
1039 %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)
1040 %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)
1041 %res2 = add <16 x i8> %res, %res1
1042 ret <16 x i8> %res2
1043 }
1045 declare <16 x i8> @llvm.x86.avx512.mask.punpcklb.w.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
1047 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) {
1048 ; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
1049 ; X86: # %bb.0:
1050 ; X86-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
1051 ; 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]
1052 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1053 ; X86-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
1054 ; 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]
1055 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1056 ; X86-NEXT: retl # encoding: [0xc3]
1057 ;
1058 ; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_128:
1059 ; X64: # %bb.0:
1060 ; X64-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x60,0xd9]
1061 ; 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]
1062 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1063 ; X64-NEXT: vpunpcklbw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x60,0xd1]
1064 ; 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]
1065 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
1066 ; X64-NEXT: retq # encoding: [0xc3]
1067 %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)
1068 %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)
1069 %res2 = add <16 x i8> %res, %res1
1070 ret <16 x i8> %res2
1071 }
1073 declare <32 x i8> @llvm.x86.avx512.mask.punpckhb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
1075 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) {
1076 ; X86-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
1077 ; X86: # %bb.0:
1078 ; X86-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
1079 ; 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]
1080 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1081 ; X86-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
1082 ; 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]
1083 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1084 ; X86-NEXT: retl # encoding: [0xc3]
1085 ;
1086 ; X64-LABEL: test_int_x86_avx512_mask_punpckhb_w_256:
1087 ; X64: # %bb.0:
1088 ; X64-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x68,0xd9]
1089 ; 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]
1090 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1091 ; X64-NEXT: vpunpckhbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x68,0xd1]
1092 ; 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]
1093 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1094 ; X64-NEXT: retq # encoding: [0xc3]
1095 %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)
1096 %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)
1097 %res2 = add <32 x i8> %res, %res1
1098 ret <32 x i8> %res2
1099 }
1101 declare <32 x i8> @llvm.x86.avx512.mask.punpcklb.w.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
1103 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) {
1104 ; X86-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
1105 ; X86: # %bb.0:
1106 ; X86-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
1107 ; 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]
1108 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1109 ; X86-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
1110 ; 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]
1111 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1112 ; X86-NEXT: retl # encoding: [0xc3]
1113 ;
1114 ; X64-LABEL: test_int_x86_avx512_mask_punpcklb_w_256:
1115 ; X64: # %bb.0:
1116 ; X64-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x60,0xd9]
1117 ; 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]
1118 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1119 ; X64-NEXT: vpunpcklbw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x60,0xd1]
1120 ; 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]
1121 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
1122 ; X64-NEXT: retq # encoding: [0xc3]
1123 %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)
1124 %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)
1125 %res2 = add <32 x i8> %res, %res1
1126 ret <32 x i8> %res2
1127 }
1129 declare <8 x i16> @llvm.x86.avx512.mask.punpcklw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1131 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) {
1132 ; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
1133 ; X86: # %bb.0:
1134 ; X86-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
1135 ; X86-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1136 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1137 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1138 ; X86-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
1139 ; X86-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1140 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1141 ; X86-NEXT: retl # encoding: [0xc3]
1142 ;
1143 ; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_128:
1144 ; X64: # %bb.0:
1145 ; X64-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x61,0xd9]
1146 ; X64-NEXT: # xmm3 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1147 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1148 ; X64-NEXT: vpunpcklwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x61,0xd1]
1149 ; X64-NEXT: # xmm2 {%k1} = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
1150 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1151 ; X64-NEXT: retq # encoding: [0xc3]
1152 %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)
1153 %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)
1154 %res2 = add <8 x i16> %res, %res1
1155 ret <8 x i16> %res2
1156 }
1158 declare <8 x i16> @llvm.x86.avx512.mask.punpckhw.d.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1160 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) {
1161 ; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
1162 ; X86: # %bb.0:
1163 ; X86-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
1164 ; X86-NEXT: # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1165 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1166 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1167 ; X86-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
1168 ; X86-NEXT: # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1169 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1170 ; X86-NEXT: retl # encoding: [0xc3]
1171 ;
1172 ; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_128:
1173 ; X64: # %bb.0:
1174 ; X64-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x69,0xd9]
1175 ; X64-NEXT: # xmm3 = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1176 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1177 ; X64-NEXT: vpunpckhwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x69,0xd1]
1178 ; X64-NEXT: # xmm2 {%k1} = xmm0[4],xmm1[4],xmm0[5],xmm1[5],xmm0[6],xmm1[6],xmm0[7],xmm1[7]
1179 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
1180 ; X64-NEXT: retq # encoding: [0xc3]
1181 %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)
1182 %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)
1183 %res2 = add <8 x i16> %res, %res1
1184 ret <8 x i16> %res2
1185 }
1187 declare <16 x i16> @llvm.x86.avx512.mask.punpcklw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1189 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) {
1190 ; X86-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
1191 ; X86: # %bb.0:
1192 ; X86-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
1193 ; 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]
1194 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1195 ; X86-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
1196 ; 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]
1197 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1198 ; X86-NEXT: retl # encoding: [0xc3]
1199 ;
1200 ; X64-LABEL: test_int_x86_avx512_mask_punpcklw_d_256:
1201 ; X64: # %bb.0:
1202 ; X64-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x61,0xd9]
1203 ; 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]
1204 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1205 ; X64-NEXT: vpunpcklwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x61,0xd1]
1206 ; 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]
1207 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1208 ; X64-NEXT: retq # encoding: [0xc3]
1209 %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)
1210 %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)
1211 %res2 = add <16 x i16> %res, %res1
1212 ret <16 x i16> %res2
1213 }
1215 declare <16 x i16> @llvm.x86.avx512.mask.punpckhw.d.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1217 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) {
1218 ; X86-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
1219 ; X86: # %bb.0:
1220 ; X86-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
1221 ; 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]
1222 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1223 ; X86-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
1224 ; 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]
1225 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1226 ; X86-NEXT: retl # encoding: [0xc3]
1227 ;
1228 ; X64-LABEL: test_int_x86_avx512_mask_punpckhw_d_256:
1229 ; X64: # %bb.0:
1230 ; X64-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x69,0xd9]
1231 ; 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]
1232 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1233 ; X64-NEXT: vpunpckhwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x69,0xd1]
1234 ; 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]
1235 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
1236 ; X64-NEXT: retq # encoding: [0xc3]
1237 %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)
1238 %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)
1239 %res2 = add <16 x i16> %res, %res1
1240 ret <16 x i16> %res2
1241 }
1243 define <8 x i16> @test_mask_add_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1244 ; CHECK-LABEL: test_mask_add_epi16_rr_128:
1245 ; CHECK: # %bb.0:
1246 ; CHECK-NEXT: vpaddw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc1]
1247 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1248 %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)
1249 ret <8 x i16> %res
1250 }
1252 define <8 x i16> @test_mask_add_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1253 ; X86-LABEL: test_mask_add_epi16_rrk_128:
1254 ; X86: # %bb.0:
1255 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1256 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1257 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
1258 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1259 ; X86-NEXT: retl # encoding: [0xc3]
1260 ;
1261 ; X64-LABEL: test_mask_add_epi16_rrk_128:
1262 ; X64: # %bb.0:
1263 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1264 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0xd1]
1265 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1266 ; X64-NEXT: retq # encoding: [0xc3]
1267 %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)
1268 ret <8 x i16> %res
1269 }
1271 define <8 x i16> @test_mask_add_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1272 ; X86-LABEL: test_mask_add_epi16_rrkz_128:
1273 ; X86: # %bb.0:
1274 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1275 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1276 ; X86-NEXT: vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
1277 ; X86-NEXT: retl # encoding: [0xc3]
1278 ;
1279 ; X64-LABEL: test_mask_add_epi16_rrkz_128:
1280 ; X64: # %bb.0:
1281 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1282 ; X64-NEXT: vpaddw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0xc1]
1283 ; X64-NEXT: retq # encoding: [0xc3]
1284 %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)
1285 ret <8 x i16> %res
1286 }
1288 define <8 x i16> @test_mask_add_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1289 ; X86-LABEL: test_mask_add_epi16_rm_128:
1290 ; X86: # %bb.0:
1291 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1292 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x00]
1293 ; X86-NEXT: retl # encoding: [0xc3]
1294 ;
1295 ; X64-LABEL: test_mask_add_epi16_rm_128:
1296 ; X64: # %bb.0:
1297 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0x07]
1298 ; X64-NEXT: retq # encoding: [0xc3]
1299 %b = load <8 x i16>, <8 x i16>* %ptr_b
1300 %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)
1301 ret <8 x i16> %res
1302 }
1304 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) {
1305 ; X86-LABEL: test_mask_add_epi16_rmk_128:
1306 ; X86: # %bb.0:
1307 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1308 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1309 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1310 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x08]
1311 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1312 ; X86-NEXT: retl # encoding: [0xc3]
1313 ;
1314 ; X64-LABEL: test_mask_add_epi16_rmk_128:
1315 ; X64: # %bb.0:
1316 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1317 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xfd,0x0f]
1318 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1319 ; X64-NEXT: retq # encoding: [0xc3]
1320 %b = load <8 x i16>, <8 x i16>* %ptr_b
1321 %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)
1322 ret <8 x i16> %res
1323 }
1325 define <8 x i16> @test_mask_add_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
1326 ; X86-LABEL: test_mask_add_epi16_rmkz_128:
1327 ; X86: # %bb.0:
1328 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1329 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1330 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1331 ; X86-NEXT: vpaddw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x00]
1332 ; X86-NEXT: retl # encoding: [0xc3]
1333 ;
1334 ; X64-LABEL: test_mask_add_epi16_rmkz_128:
1335 ; X64: # %bb.0:
1336 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1337 ; X64-NEXT: vpaddw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xfd,0x07]
1338 ; X64-NEXT: retq # encoding: [0xc3]
1339 %b = load <8 x i16>, <8 x i16>* %ptr_b
1340 %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)
1341 ret <8 x i16> %res
1342 }
1344 declare <8 x i16> @llvm.x86.avx512.mask.padd.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1346 define <16 x i16> @test_mask_add_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
1347 ; CHECK-LABEL: test_mask_add_epi16_rr_256:
1348 ; CHECK: # %bb.0:
1349 ; CHECK-NEXT: vpaddw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc1]
1350 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1351 %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)
1352 ret <16 x i16> %res
1353 }
1355 define <16 x i16> @test_mask_add_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
1356 ; X86-LABEL: test_mask_add_epi16_rrk_256:
1357 ; X86: # %bb.0:
1358 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1359 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
1360 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1361 ; X86-NEXT: retl # encoding: [0xc3]
1362 ;
1363 ; X64-LABEL: test_mask_add_epi16_rrk_256:
1364 ; X64: # %bb.0:
1365 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1366 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0xd1]
1367 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1368 ; X64-NEXT: retq # encoding: [0xc3]
1369 %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)
1370 ret <16 x i16> %res
1371 }
1373 define <16 x i16> @test_mask_add_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
1374 ; X86-LABEL: test_mask_add_epi16_rrkz_256:
1375 ; X86: # %bb.0:
1376 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1377 ; X86-NEXT: vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
1378 ; X86-NEXT: retl # encoding: [0xc3]
1379 ;
1380 ; X64-LABEL: test_mask_add_epi16_rrkz_256:
1381 ; X64: # %bb.0:
1382 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1383 ; X64-NEXT: vpaddw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0xc1]
1384 ; X64-NEXT: retq # encoding: [0xc3]
1385 %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)
1386 ret <16 x i16> %res
1387 }
1389 define <16 x i16> @test_mask_add_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
1390 ; X86-LABEL: test_mask_add_epi16_rm_256:
1391 ; X86: # %bb.0:
1392 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1393 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x00]
1394 ; X86-NEXT: retl # encoding: [0xc3]
1395 ;
1396 ; X64-LABEL: test_mask_add_epi16_rm_256:
1397 ; X64: # %bb.0:
1398 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0x07]
1399 ; X64-NEXT: retq # encoding: [0xc3]
1400 %b = load <16 x i16>, <16 x i16>* %ptr_b
1401 %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)
1402 ret <16 x i16> %res
1403 }
1405 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) {
1406 ; X86-LABEL: test_mask_add_epi16_rmk_256:
1407 ; X86: # %bb.0:
1408 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1409 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1410 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x08]
1411 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1412 ; X86-NEXT: retl # encoding: [0xc3]
1413 ;
1414 ; X64-LABEL: test_mask_add_epi16_rmk_256:
1415 ; X64: # %bb.0:
1416 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1417 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xfd,0x0f]
1418 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1419 ; X64-NEXT: retq # encoding: [0xc3]
1420 %b = load <16 x i16>, <16 x i16>* %ptr_b
1421 %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)
1422 ret <16 x i16> %res
1423 }
1425 define <16 x i16> @test_mask_add_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
1426 ; X86-LABEL: test_mask_add_epi16_rmkz_256:
1427 ; X86: # %bb.0:
1428 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1429 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1430 ; X86-NEXT: vpaddw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x00]
1431 ; X86-NEXT: retl # encoding: [0xc3]
1432 ;
1433 ; X64-LABEL: test_mask_add_epi16_rmkz_256:
1434 ; X64: # %bb.0:
1435 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1436 ; X64-NEXT: vpaddw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xfd,0x07]
1437 ; X64-NEXT: retq # encoding: [0xc3]
1438 %b = load <16 x i16>, <16 x i16>* %ptr_b
1439 %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)
1440 ret <16 x i16> %res
1441 }
1443 declare <16 x i16> @llvm.x86.avx512.mask.padd.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1445 define <8 x i16> @test_mask_sub_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1446 ; CHECK-LABEL: test_mask_sub_epi16_rr_128:
1447 ; CHECK: # %bb.0:
1448 ; CHECK-NEXT: vpsubw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0xc1]
1449 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1450 %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)
1451 ret <8 x i16> %res
1452 }
1454 define <8 x i16> @test_mask_sub_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1455 ; X86-LABEL: test_mask_sub_epi16_rrk_128:
1456 ; X86: # %bb.0:
1457 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1458 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1459 ; X86-NEXT: vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
1460 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1461 ; X86-NEXT: retl # encoding: [0xc3]
1462 ;
1463 ; X64-LABEL: test_mask_sub_epi16_rrk_128:
1464 ; X64: # %bb.0:
1465 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1466 ; X64-NEXT: vpsubw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0xd1]
1467 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1468 ; X64-NEXT: retq # encoding: [0xc3]
1469 %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)
1470 ret <8 x i16> %res
1471 }
1473 define <8 x i16> @test_mask_sub_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1474 ; X86-LABEL: test_mask_sub_epi16_rrkz_128:
1475 ; X86: # %bb.0:
1476 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1477 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1478 ; X86-NEXT: vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
1479 ; X86-NEXT: retl # encoding: [0xc3]
1480 ;
1481 ; X64-LABEL: test_mask_sub_epi16_rrkz_128:
1482 ; X64: # %bb.0:
1483 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1484 ; X64-NEXT: vpsubw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0xc1]
1485 ; X64-NEXT: retq # encoding: [0xc3]
1486 %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)
1487 ret <8 x i16> %res
1488 }
1490 define <8 x i16> @test_mask_sub_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1491 ; X86-LABEL: test_mask_sub_epi16_rm_128:
1492 ; X86: # %bb.0:
1493 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1494 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x00]
1495 ; X86-NEXT: retl # encoding: [0xc3]
1496 ;
1497 ; X64-LABEL: test_mask_sub_epi16_rm_128:
1498 ; X64: # %bb.0:
1499 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf9,0x07]
1500 ; X64-NEXT: retq # encoding: [0xc3]
1501 %b = load <8 x i16>, <8 x i16>* %ptr_b
1502 %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)
1503 ret <8 x i16> %res
1504 }
1506 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) {
1507 ; X86-LABEL: test_mask_sub_epi16_rmk_128:
1508 ; X86: # %bb.0:
1509 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1510 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1511 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1512 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x08]
1513 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1514 ; X86-NEXT: retl # encoding: [0xc3]
1515 ;
1516 ; X64-LABEL: test_mask_sub_epi16_rmk_128:
1517 ; X64: # %bb.0:
1518 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1519 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf9,0x0f]
1520 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1521 ; X64-NEXT: retq # encoding: [0xc3]
1522 %b = load <8 x i16>, <8 x i16>* %ptr_b
1523 %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)
1524 ret <8 x i16> %res
1525 }
1527 define <8 x i16> @test_mask_sub_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
1528 ; X86-LABEL: test_mask_sub_epi16_rmkz_128:
1529 ; X86: # %bb.0:
1530 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1531 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
1532 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
1533 ; X86-NEXT: vpsubw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x00]
1534 ; X86-NEXT: retl # encoding: [0xc3]
1535 ;
1536 ; X64-LABEL: test_mask_sub_epi16_rmkz_128:
1537 ; X64: # %bb.0:
1538 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1539 ; X64-NEXT: vpsubw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf9,0x07]
1540 ; X64-NEXT: retq # encoding: [0xc3]
1541 %b = load <8 x i16>, <8 x i16>* %ptr_b
1542 %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)
1543 ret <8 x i16> %res
1544 }
1546 declare <8 x i16> @llvm.x86.avx512.mask.psub.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
1548 define <16 x i16> @test_mask_sub_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
1549 ; CHECK-LABEL: test_mask_sub_epi16_rr_256:
1550 ; CHECK: # %bb.0:
1551 ; CHECK-NEXT: vpsubw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0xc1]
1552 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1553 %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)
1554 ret <16 x i16> %res
1555 }
1557 define <16 x i16> @test_mask_sub_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
1558 ; X86-LABEL: test_mask_sub_epi16_rrk_256:
1559 ; X86: # %bb.0:
1560 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1561 ; X86-NEXT: vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
1562 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1563 ; X86-NEXT: retl # encoding: [0xc3]
1564 ;
1565 ; X64-LABEL: test_mask_sub_epi16_rrk_256:
1566 ; X64: # %bb.0:
1567 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1568 ; X64-NEXT: vpsubw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0xd1]
1569 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1570 ; X64-NEXT: retq # encoding: [0xc3]
1571 %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)
1572 ret <16 x i16> %res
1573 }
1575 define <16 x i16> @test_mask_sub_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
1576 ; X86-LABEL: test_mask_sub_epi16_rrkz_256:
1577 ; X86: # %bb.0:
1578 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1579 ; X86-NEXT: vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
1580 ; X86-NEXT: retl # encoding: [0xc3]
1581 ;
1582 ; X64-LABEL: test_mask_sub_epi16_rrkz_256:
1583 ; X64: # %bb.0:
1584 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1585 ; X64-NEXT: vpsubw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0xc1]
1586 ; X64-NEXT: retq # encoding: [0xc3]
1587 %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)
1588 ret <16 x i16> %res
1589 }
1591 define <16 x i16> @test_mask_sub_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
1592 ; X86-LABEL: test_mask_sub_epi16_rm_256:
1593 ; X86: # %bb.0:
1594 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1595 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x00]
1596 ; X86-NEXT: retl # encoding: [0xc3]
1597 ;
1598 ; X64-LABEL: test_mask_sub_epi16_rm_256:
1599 ; X64: # %bb.0:
1600 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf9,0x07]
1601 ; X64-NEXT: retq # encoding: [0xc3]
1602 %b = load <16 x i16>, <16 x i16>* %ptr_b
1603 %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)
1604 ret <16 x i16> %res
1605 }
1607 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) {
1608 ; X86-LABEL: test_mask_sub_epi16_rmk_256:
1609 ; X86: # %bb.0:
1610 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1611 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1612 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x08]
1613 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1614 ; X86-NEXT: retl # encoding: [0xc3]
1615 ;
1616 ; X64-LABEL: test_mask_sub_epi16_rmk_256:
1617 ; X64: # %bb.0:
1618 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1619 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf9,0x0f]
1620 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1621 ; X64-NEXT: retq # encoding: [0xc3]
1622 %b = load <16 x i16>, <16 x i16>* %ptr_b
1623 %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)
1624 ret <16 x i16> %res
1625 }
1627 define <16 x i16> @test_mask_sub_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
1628 ; X86-LABEL: test_mask_sub_epi16_rmkz_256:
1629 ; X86: # %bb.0:
1630 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1631 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1632 ; X86-NEXT: vpsubw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x00]
1633 ; X86-NEXT: retl # encoding: [0xc3]
1634 ;
1635 ; X64-LABEL: test_mask_sub_epi16_rmkz_256:
1636 ; X64: # %bb.0:
1637 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1638 ; X64-NEXT: vpsubw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf9,0x07]
1639 ; X64-NEXT: retq # encoding: [0xc3]
1640 %b = load <16 x i16>, <16 x i16>* %ptr_b
1641 %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)
1642 ret <16 x i16> %res
1643 }
1645 declare <16 x i16> @llvm.x86.avx512.mask.psub.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
1647 define <32 x i16> @test_mask_add_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1648 ; CHECK-LABEL: test_mask_add_epi16_rr_512:
1649 ; CHECK: # %bb.0:
1650 ; CHECK-NEXT: vpaddw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0xc1]
1651 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1652 %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)
1653 ret <32 x i16> %res
1654 }
1656 define <32 x i16> @test_mask_add_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1657 ; X86-LABEL: test_mask_add_epi16_rrk_512:
1658 ; X86: # %bb.0:
1659 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1660 ; X86-NEXT: vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
1661 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1662 ; X86-NEXT: retl # encoding: [0xc3]
1663 ;
1664 ; X64-LABEL: test_mask_add_epi16_rrk_512:
1665 ; X64: # %bb.0:
1666 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1667 ; X64-NEXT: vpaddw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0xd1]
1668 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1669 ; X64-NEXT: retq # encoding: [0xc3]
1670 %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)
1671 ret <32 x i16> %res
1672 }
1674 define <32 x i16> @test_mask_add_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1675 ; X86-LABEL: test_mask_add_epi16_rrkz_512:
1676 ; X86: # %bb.0:
1677 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1678 ; X86-NEXT: vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
1679 ; X86-NEXT: retl # encoding: [0xc3]
1680 ;
1681 ; X64-LABEL: test_mask_add_epi16_rrkz_512:
1682 ; X64: # %bb.0:
1683 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1684 ; X64-NEXT: vpaddw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0xc1]
1685 ; X64-NEXT: retq # encoding: [0xc3]
1686 %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)
1687 ret <32 x i16> %res
1688 }
1690 define <32 x i16> @test_mask_add_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1691 ; X86-LABEL: test_mask_add_epi16_rm_512:
1692 ; X86: # %bb.0:
1693 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1694 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x00]
1695 ; X86-NEXT: retl # encoding: [0xc3]
1696 ;
1697 ; X64-LABEL: test_mask_add_epi16_rm_512:
1698 ; X64: # %bb.0:
1699 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xfd,0x07]
1700 ; X64-NEXT: retq # encoding: [0xc3]
1701 %b = load <32 x i16>, <32 x i16>* %ptr_b
1702 %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)
1703 ret <32 x i16> %res
1704 }
1706 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) {
1707 ; X86-LABEL: test_mask_add_epi16_rmk_512:
1708 ; X86: # %bb.0:
1709 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1710 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1711 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x08]
1712 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1713 ; X86-NEXT: retl # encoding: [0xc3]
1714 ;
1715 ; X64-LABEL: test_mask_add_epi16_rmk_512:
1716 ; X64: # %bb.0:
1717 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1718 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xfd,0x0f]
1719 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1720 ; X64-NEXT: retq # encoding: [0xc3]
1721 %b = load <32 x i16>, <32 x i16>* %ptr_b
1722 %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)
1723 ret <32 x i16> %res
1724 }
1726 define <32 x i16> @test_mask_add_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1727 ; X86-LABEL: test_mask_add_epi16_rmkz_512:
1728 ; X86: # %bb.0:
1729 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1730 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1731 ; X86-NEXT: vpaddw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x00]
1732 ; X86-NEXT: retl # encoding: [0xc3]
1733 ;
1734 ; X64-LABEL: test_mask_add_epi16_rmkz_512:
1735 ; X64: # %bb.0:
1736 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1737 ; X64-NEXT: vpaddw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xfd,0x07]
1738 ; X64-NEXT: retq # encoding: [0xc3]
1739 %b = load <32 x i16>, <32 x i16>* %ptr_b
1740 %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)
1741 ret <32 x i16> %res
1742 }
1744 declare <32 x i16> @llvm.x86.avx512.mask.padd.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1746 define <32 x i16> @test_mask_sub_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1747 ; CHECK-LABEL: test_mask_sub_epi16_rr_512:
1748 ; CHECK: # %bb.0:
1749 ; CHECK-NEXT: vpsubw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0xc1]
1750 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1751 %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)
1752 ret <32 x i16> %res
1753 }
1755 define <32 x i16> @test_mask_sub_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1756 ; X86-LABEL: test_mask_sub_epi16_rrk_512:
1757 ; X86: # %bb.0:
1758 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1759 ; X86-NEXT: vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
1760 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1761 ; X86-NEXT: retl # encoding: [0xc3]
1762 ;
1763 ; X64-LABEL: test_mask_sub_epi16_rrk_512:
1764 ; X64: # %bb.0:
1765 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1766 ; X64-NEXT: vpsubw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0xd1]
1767 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1768 ; X64-NEXT: retq # encoding: [0xc3]
1769 %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)
1770 ret <32 x i16> %res
1771 }
1773 define <32 x i16> @test_mask_sub_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1774 ; X86-LABEL: test_mask_sub_epi16_rrkz_512:
1775 ; X86: # %bb.0:
1776 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1777 ; X86-NEXT: vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
1778 ; X86-NEXT: retl # encoding: [0xc3]
1779 ;
1780 ; X64-LABEL: test_mask_sub_epi16_rrkz_512:
1781 ; X64: # %bb.0:
1782 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1783 ; X64-NEXT: vpsubw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0xc1]
1784 ; X64-NEXT: retq # encoding: [0xc3]
1785 %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)
1786 ret <32 x i16> %res
1787 }
1789 define <32 x i16> @test_mask_sub_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1790 ; X86-LABEL: test_mask_sub_epi16_rm_512:
1791 ; X86: # %bb.0:
1792 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1793 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x00]
1794 ; X86-NEXT: retl # encoding: [0xc3]
1795 ;
1796 ; X64-LABEL: test_mask_sub_epi16_rm_512:
1797 ; X64: # %bb.0:
1798 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xf9,0x07]
1799 ; X64-NEXT: retq # encoding: [0xc3]
1800 %b = load <32 x i16>, <32 x i16>* %ptr_b
1801 %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)
1802 ret <32 x i16> %res
1803 }
1805 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) {
1806 ; X86-LABEL: test_mask_sub_epi16_rmk_512:
1807 ; X86: # %bb.0:
1808 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1809 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1810 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x08]
1811 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1812 ; X86-NEXT: retl # encoding: [0xc3]
1813 ;
1814 ; X64-LABEL: test_mask_sub_epi16_rmk_512:
1815 ; X64: # %bb.0:
1816 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1817 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xf9,0x0f]
1818 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1819 ; X64-NEXT: retq # encoding: [0xc3]
1820 %b = load <32 x i16>, <32 x i16>* %ptr_b
1821 %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)
1822 ret <32 x i16> %res
1823 }
1825 define <32 x i16> @test_mask_sub_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1826 ; X86-LABEL: test_mask_sub_epi16_rmkz_512:
1827 ; X86: # %bb.0:
1828 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1829 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1830 ; X86-NEXT: vpsubw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x00]
1831 ; X86-NEXT: retl # encoding: [0xc3]
1832 ;
1833 ; X64-LABEL: test_mask_sub_epi16_rmkz_512:
1834 ; X64: # %bb.0:
1835 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1836 ; X64-NEXT: vpsubw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xf9,0x07]
1837 ; X64-NEXT: retq # encoding: [0xc3]
1838 %b = load <32 x i16>, <32 x i16>* %ptr_b
1839 %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)
1840 ret <32 x i16> %res
1841 }
1843 declare <32 x i16> @llvm.x86.avx512.mask.psub.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1845 define <32 x i16> @test_mask_mullo_epi16_rr_512(<32 x i16> %a, <32 x i16> %b) {
1846 ; CHECK-LABEL: test_mask_mullo_epi16_rr_512:
1847 ; CHECK: # %bb.0:
1848 ; CHECK-NEXT: vpmullw %zmm1, %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0xc1]
1849 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1850 %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)
1851 ret <32 x i16> %res
1852 }
1854 define <32 x i16> @test_mask_mullo_epi16_rrk_512(<32 x i16> %a, <32 x i16> %b, <32 x i16> %passThru, i32 %mask) {
1855 ; X86-LABEL: test_mask_mullo_epi16_rrk_512:
1856 ; X86: # %bb.0:
1857 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1858 ; X86-NEXT: vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
1859 ; X86-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1860 ; X86-NEXT: retl # encoding: [0xc3]
1861 ;
1862 ; X64-LABEL: test_mask_mullo_epi16_rrk_512:
1863 ; X64: # %bb.0:
1864 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1865 ; X64-NEXT: vpmullw %zmm1, %zmm0, %zmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0xd1]
1866 ; X64-NEXT: vmovdqa64 %zmm2, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc2]
1867 ; X64-NEXT: retq # encoding: [0xc3]
1868 %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)
1869 ret <32 x i16> %res
1870 }
1872 define <32 x i16> @test_mask_mullo_epi16_rrkz_512(<32 x i16> %a, <32 x i16> %b, i32 %mask) {
1873 ; X86-LABEL: test_mask_mullo_epi16_rrkz_512:
1874 ; X86: # %bb.0:
1875 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1876 ; X86-NEXT: vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
1877 ; X86-NEXT: retl # encoding: [0xc3]
1878 ;
1879 ; X64-LABEL: test_mask_mullo_epi16_rrkz_512:
1880 ; X64: # %bb.0:
1881 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1882 ; X64-NEXT: vpmullw %zmm1, %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0xc1]
1883 ; X64-NEXT: retq # encoding: [0xc3]
1884 %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)
1885 ret <32 x i16> %res
1886 }
1888 define <32 x i16> @test_mask_mullo_epi16_rm_512(<32 x i16> %a, <32 x i16>* %ptr_b) {
1889 ; X86-LABEL: test_mask_mullo_epi16_rm_512:
1890 ; X86: # %bb.0:
1891 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1892 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x00]
1893 ; X86-NEXT: retl # encoding: [0xc3]
1894 ;
1895 ; X64-LABEL: test_mask_mullo_epi16_rm_512:
1896 ; X64: # %bb.0:
1897 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm0 # encoding: [0x62,0xf1,0x7d,0x48,0xd5,0x07]
1898 ; X64-NEXT: retq # encoding: [0xc3]
1899 %b = load <32 x i16>, <32 x i16>* %ptr_b
1900 %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)
1901 ret <32 x i16> %res
1902 }
1904 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) {
1905 ; X86-LABEL: test_mask_mullo_epi16_rmk_512:
1906 ; X86: # %bb.0:
1907 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1908 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1909 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x08]
1910 ; X86-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1911 ; X86-NEXT: retl # encoding: [0xc3]
1912 ;
1913 ; X64-LABEL: test_mask_mullo_epi16_rmk_512:
1914 ; X64: # %bb.0:
1915 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1916 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x49,0xd5,0x0f]
1917 ; X64-NEXT: vmovdqa64 %zmm1, %zmm0 # encoding: [0x62,0xf1,0xfd,0x48,0x6f,0xc1]
1918 ; X64-NEXT: retq # encoding: [0xc3]
1919 %b = load <32 x i16>, <32 x i16>* %ptr_b
1920 %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)
1921 ret <32 x i16> %res
1922 }
1924 define <32 x i16> @test_mask_mullo_epi16_rmkz_512(<32 x i16> %a, <32 x i16>* %ptr_b, i32 %mask) {
1925 ; X86-LABEL: test_mask_mullo_epi16_rmkz_512:
1926 ; X86: # %bb.0:
1927 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1928 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1929 ; X86-NEXT: vpmullw (%eax), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x00]
1930 ; X86-NEXT: retl # encoding: [0xc3]
1931 ;
1932 ; X64-LABEL: test_mask_mullo_epi16_rmkz_512:
1933 ; X64: # %bb.0:
1934 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1935 ; X64-NEXT: vpmullw (%rdi), %zmm0, %zmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xc9,0xd5,0x07]
1936 ; X64-NEXT: retq # encoding: [0xc3]
1937 %b = load <32 x i16>, <32 x i16>* %ptr_b
1938 %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)
1939 ret <32 x i16> %res
1940 }
1942 declare <32 x i16> @llvm.x86.avx512.mask.pmull.w.512(<32 x i16>, <32 x i16>, <32 x i16>, i32)
1944 define <8 x i16> @test_mask_mullo_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
1945 ; CHECK-LABEL: test_mask_mullo_epi16_rr_128:
1946 ; CHECK: # %bb.0:
1947 ; CHECK-NEXT: vpmullw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0xc1]
1948 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1949 %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)
1950 ret <8 x i16> %res
1951 }
1953 define <8 x i16> @test_mask_mullo_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
1954 ; X86-LABEL: test_mask_mullo_epi16_rrk_128:
1955 ; X86: # %bb.0:
1956 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1957 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1958 ; X86-NEXT: vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
1959 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1960 ; X86-NEXT: retl # encoding: [0xc3]
1961 ;
1962 ; X64-LABEL: test_mask_mullo_epi16_rrk_128:
1963 ; X64: # %bb.0:
1964 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1965 ; X64-NEXT: vpmullw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0xd1]
1966 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1967 ; X64-NEXT: retq # encoding: [0xc3]
1968 %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)
1969 ret <8 x i16> %res
1970 }
1972 define <8 x i16> @test_mask_mullo_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
1973 ; X86-LABEL: test_mask_mullo_epi16_rrkz_128:
1974 ; X86: # %bb.0:
1975 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1976 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1977 ; X86-NEXT: vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
1978 ; X86-NEXT: retl # encoding: [0xc3]
1979 ;
1980 ; X64-LABEL: test_mask_mullo_epi16_rrkz_128:
1981 ; X64: # %bb.0:
1982 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1983 ; X64-NEXT: vpmullw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0xc1]
1984 ; X64-NEXT: retq # encoding: [0xc3]
1985 %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)
1986 ret <8 x i16> %res
1987 }
1989 define <8 x i16> @test_mask_mullo_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
1990 ; X86-LABEL: test_mask_mullo_epi16_rm_128:
1991 ; X86: # %bb.0:
1992 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1993 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x00]
1994 ; X86-NEXT: retl # encoding: [0xc3]
1995 ;
1996 ; X64-LABEL: test_mask_mullo_epi16_rm_128:
1997 ; X64: # %bb.0:
1998 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd5,0x07]
1999 ; X64-NEXT: retq # encoding: [0xc3]
2000 %b = load <8 x i16>, <8 x i16>* %ptr_b
2001 %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)
2002 ret <8 x i16> %res
2003 }
2005 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) {
2006 ; X86-LABEL: test_mask_mullo_epi16_rmk_128:
2007 ; X86: # %bb.0:
2008 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2009 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
2010 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
2011 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x08]
2012 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
2013 ; X86-NEXT: retl # encoding: [0xc3]
2014 ;
2015 ; X64-LABEL: test_mask_mullo_epi16_rmk_128:
2016 ; X64: # %bb.0:
2017 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2018 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd5,0x0f]
2019 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
2020 ; X64-NEXT: retq # encoding: [0xc3]
2021 %b = load <8 x i16>, <8 x i16>* %ptr_b
2022 %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)
2023 ret <8 x i16> %res
2024 }
2026 define <8 x i16> @test_mask_mullo_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
2027 ; X86-LABEL: test_mask_mullo_epi16_rmkz_128:
2028 ; X86: # %bb.0:
2029 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2030 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
2031 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
2032 ; X86-NEXT: vpmullw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x00]
2033 ; X86-NEXT: retl # encoding: [0xc3]
2034 ;
2035 ; X64-LABEL: test_mask_mullo_epi16_rmkz_128:
2036 ; X64: # %bb.0:
2037 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2038 ; X64-NEXT: vpmullw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd5,0x07]
2039 ; X64-NEXT: retq # encoding: [0xc3]
2040 %b = load <8 x i16>, <8 x i16>* %ptr_b
2041 %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)
2042 ret <8 x i16> %res
2043 }
2045 declare <8 x i16> @llvm.x86.avx512.mask.pmull.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2047 define <16 x i16> @test_mask_mullo_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
2048 ; CHECK-LABEL: test_mask_mullo_epi16_rr_256:
2049 ; CHECK: # %bb.0:
2050 ; CHECK-NEXT: vpmullw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0xc1]
2051 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2052 %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)
2053 ret <16 x i16> %res
2054 }
2056 define <16 x i16> @test_mask_mullo_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
2057 ; X86-LABEL: test_mask_mullo_epi16_rrk_256:
2058 ; X86: # %bb.0:
2059 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2060 ; X86-NEXT: vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
2061 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2062 ; X86-NEXT: retl # encoding: [0xc3]
2063 ;
2064 ; X64-LABEL: test_mask_mullo_epi16_rrk_256:
2065 ; X64: # %bb.0:
2066 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2067 ; X64-NEXT: vpmullw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0xd1]
2068 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2069 ; X64-NEXT: retq # encoding: [0xc3]
2070 %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)
2071 ret <16 x i16> %res
2072 }
2074 define <16 x i16> @test_mask_mullo_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
2075 ; X86-LABEL: test_mask_mullo_epi16_rrkz_256:
2076 ; X86: # %bb.0:
2077 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2078 ; X86-NEXT: vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
2079 ; X86-NEXT: retl # encoding: [0xc3]
2080 ;
2081 ; X64-LABEL: test_mask_mullo_epi16_rrkz_256:
2082 ; X64: # %bb.0:
2083 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2084 ; X64-NEXT: vpmullw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0xc1]
2085 ; X64-NEXT: retq # encoding: [0xc3]
2086 %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)
2087 ret <16 x i16> %res
2088 }
2090 define <16 x i16> @test_mask_mullo_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
2091 ; X86-LABEL: test_mask_mullo_epi16_rm_256:
2092 ; X86: # %bb.0:
2093 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2094 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x00]
2095 ; X86-NEXT: retl # encoding: [0xc3]
2096 ;
2097 ; X64-LABEL: test_mask_mullo_epi16_rm_256:
2098 ; X64: # %bb.0:
2099 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd5,0x07]
2100 ; X64-NEXT: retq # encoding: [0xc3]
2101 %b = load <16 x i16>, <16 x i16>* %ptr_b
2102 %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)
2103 ret <16 x i16> %res
2104 }
2106 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) {
2107 ; X86-LABEL: test_mask_mullo_epi16_rmk_256:
2108 ; X86: # %bb.0:
2109 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2110 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2111 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x08]
2112 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
2113 ; X86-NEXT: retl # encoding: [0xc3]
2114 ;
2115 ; X64-LABEL: test_mask_mullo_epi16_rmk_256:
2116 ; X64: # %bb.0:
2117 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2118 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd5,0x0f]
2119 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
2120 ; X64-NEXT: retq # encoding: [0xc3]
2121 %b = load <16 x i16>, <16 x i16>* %ptr_b
2122 %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)
2123 ret <16 x i16> %res
2124 }
2126 define <16 x i16> @test_mask_mullo_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
2127 ; X86-LABEL: test_mask_mullo_epi16_rmkz_256:
2128 ; X86: # %bb.0:
2129 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
2130 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2131 ; X86-NEXT: vpmullw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x00]
2132 ; X86-NEXT: retl # encoding: [0xc3]
2133 ;
2134 ; X64-LABEL: test_mask_mullo_epi16_rmkz_256:
2135 ; X64: # %bb.0:
2136 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2137 ; X64-NEXT: vpmullw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd5,0x07]
2138 ; X64-NEXT: retq # encoding: [0xc3]
2139 %b = load <16 x i16>, <16 x i16>* %ptr_b
2140 %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)
2141 ret <16 x i16> %res
2142 }
2144 declare <16 x i16> @llvm.x86.avx512.mask.pmull.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2146 declare <16 x i8> @llvm.x86.avx512.mask.pmaxs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2148 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) {
2149 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
2150 ; X86: # %bb.0:
2151 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2152 ; X86-NEXT: vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
2153 ; X86-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
2154 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2155 ; X86-NEXT: retl # encoding: [0xc3]
2156 ;
2157 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_128:
2158 ; X64: # %bb.0:
2159 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2160 ; X64-NEXT: vpmaxsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3c,0xd1]
2161 ; X64-NEXT: vpmaxsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x3c,0xc1]
2162 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2163 ; X64-NEXT: retq # encoding: [0xc3]
2164 %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)
2165 %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)
2166 %res2 = add <16 x i8> %res, %res1
2167 ret <16 x i8> %res2
2168 }
2170 declare <32 x i8> @llvm.x86.avx512.mask.pmaxs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2172 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) {
2173 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
2174 ; X86: # %bb.0:
2175 ; X86-NEXT: vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
2176 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2177 ; X86-NEXT: vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
2178 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2179 ; X86-NEXT: retl # encoding: [0xc3]
2180 ;
2181 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_b_256:
2182 ; X64: # %bb.0:
2183 ; X64-NEXT: vpmaxsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x3c,0xd9]
2184 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2185 ; X64-NEXT: vpmaxsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3c,0xd1]
2186 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2187 ; X64-NEXT: retq # encoding: [0xc3]
2188 %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)
2189 %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)
2190 %res2 = add <32 x i8> %res, %res1
2191 ret <32 x i8> %res2
2192 }
2194 declare <8 x i16> @llvm.x86.avx512.mask.pmaxs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2196 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) {
2197 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
2198 ; X86: # %bb.0:
2199 ; X86-NEXT: vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
2200 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2201 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2202 ; X86-NEXT: vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
2203 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2204 ; X86-NEXT: retl # encoding: [0xc3]
2205 ;
2206 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_128:
2207 ; X64: # %bb.0:
2208 ; X64-NEXT: vpmaxsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xee,0xd9]
2209 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2210 ; X64-NEXT: vpmaxsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xee,0xd1]
2211 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2212 ; X64-NEXT: retq # encoding: [0xc3]
2213 %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)
2214 %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)
2215 %res2 = add <8 x i16> %res, %res1
2216 ret <8 x i16> %res2
2217 }
2219 declare <16 x i16> @llvm.x86.avx512.mask.pmaxs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2221 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) {
2222 ; X86-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
2223 ; X86: # %bb.0:
2224 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2225 ; X86-NEXT: vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
2226 ; X86-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
2227 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2228 ; X86-NEXT: retl # encoding: [0xc3]
2229 ;
2230 ; X64-LABEL: test_int_x86_avx512_mask_pmaxs_w_256:
2231 ; X64: # %bb.0:
2232 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2233 ; X64-NEXT: vpmaxsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xee,0xd1]
2234 ; X64-NEXT: vpmaxsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xee,0xc1]
2235 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2236 ; X64-NEXT: retq # encoding: [0xc3]
2237 %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)
2238 %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)
2239 %res2 = add <16 x i16> %res, %res1
2240 ret <16 x i16> %res2
2241 }
2243 declare <16 x i8> @llvm.x86.avx512.mask.pmaxu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2245 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) {
2246 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
2247 ; X86: # %bb.0:
2248 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2249 ; X86-NEXT: vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
2250 ; X86-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
2251 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2252 ; X86-NEXT: retl # encoding: [0xc3]
2253 ;
2254 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_128:
2255 ; X64: # %bb.0:
2256 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2257 ; X64-NEXT: vpmaxub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xde,0xd1]
2258 ; X64-NEXT: vpmaxub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xde,0xc1]
2259 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2260 ; X64-NEXT: retq # encoding: [0xc3]
2261 %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)
2262 %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)
2263 %res2 = add <16 x i8> %res, %res1
2264 ret <16 x i8> %res2
2265 }
2267 declare <32 x i8> @llvm.x86.avx512.mask.pmaxu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2269 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) {
2270 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
2271 ; X86: # %bb.0:
2272 ; X86-NEXT: vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
2273 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2274 ; X86-NEXT: vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
2275 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2276 ; X86-NEXT: retl # encoding: [0xc3]
2277 ;
2278 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_b_256:
2279 ; X64: # %bb.0:
2280 ; X64-NEXT: vpmaxub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xde,0xd9]
2281 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2282 ; X64-NEXT: vpmaxub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xde,0xd1]
2283 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2284 ; X64-NEXT: retq # encoding: [0xc3]
2285 %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)
2286 %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)
2287 %res2 = add <32 x i8> %res, %res1
2288 ret <32 x i8> %res2
2289 }
2291 declare <8 x i16> @llvm.x86.avx512.mask.pmaxu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2293 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) {
2294 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
2295 ; X86: # %bb.0:
2296 ; X86-NEXT: vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
2297 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2298 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2299 ; X86-NEXT: vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
2300 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2301 ; X86-NEXT: retl # encoding: [0xc3]
2302 ;
2303 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_128:
2304 ; X64: # %bb.0:
2305 ; X64-NEXT: vpmaxuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3e,0xd9]
2306 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2307 ; X64-NEXT: vpmaxuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3e,0xd1]
2308 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2309 ; X64-NEXT: retq # encoding: [0xc3]
2310 %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)
2311 %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)
2312 %res2 = add <8 x i16> %res, %res1
2313 ret <8 x i16> %res2
2314 }
2316 declare <16 x i16> @llvm.x86.avx512.mask.pmaxu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2318 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) {
2319 ; X86-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
2320 ; X86: # %bb.0:
2321 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2322 ; X86-NEXT: vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
2323 ; X86-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
2324 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2325 ; X86-NEXT: retl # encoding: [0xc3]
2326 ;
2327 ; X64-LABEL: test_int_x86_avx512_mask_pmaxu_w_256:
2328 ; X64: # %bb.0:
2329 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2330 ; X64-NEXT: vpmaxuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3e,0xd1]
2331 ; X64-NEXT: vpmaxuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3e,0xc1]
2332 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2333 ; X64-NEXT: retq # encoding: [0xc3]
2334 %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)
2335 %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)
2336 %res2 = add <16 x i16> %res, %res1
2337 ret <16 x i16> %res2
2338 }
2340 declare <16 x i8> @llvm.x86.avx512.mask.pmins.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2342 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) {
2343 ; X86-LABEL: test_int_x86_avx512_mask_pmins_b_128:
2344 ; X86: # %bb.0:
2345 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2346 ; X86-NEXT: vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
2347 ; X86-NEXT: vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
2348 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2349 ; X86-NEXT: retl # encoding: [0xc3]
2350 ;
2351 ; X64-LABEL: test_int_x86_avx512_mask_pmins_b_128:
2352 ; X64: # %bb.0:
2353 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2354 ; X64-NEXT: vpminsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x38,0xd1]
2355 ; X64-NEXT: vpminsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x38,0xc1]
2356 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2357 ; X64-NEXT: retq # encoding: [0xc3]
2358 %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)
2359 %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)
2360 %res2 = add <16 x i8> %res, %res1
2361 ret <16 x i8> %res2
2362 }
2364 declare <32 x i8> @llvm.x86.avx512.mask.pmins.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2366 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) {
2367 ; X86-LABEL: test_int_x86_avx512_mask_pmins_b_256:
2368 ; X86: # %bb.0:
2369 ; X86-NEXT: vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
2370 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2371 ; X86-NEXT: vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
2372 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2373 ; X86-NEXT: retl # encoding: [0xc3]
2374 ;
2375 ; X64-LABEL: test_int_x86_avx512_mask_pmins_b_256:
2376 ; X64: # %bb.0:
2377 ; X64-NEXT: vpminsb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x38,0xd9]
2378 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2379 ; X64-NEXT: vpminsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x38,0xd1]
2380 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2381 ; X64-NEXT: retq # encoding: [0xc3]
2382 %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)
2383 %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)
2384 %res2 = add <32 x i8> %res, %res1
2385 ret <32 x i8> %res2
2386 }
2388 declare <8 x i16> @llvm.x86.avx512.mask.pmins.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2390 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) {
2391 ; X86-LABEL: test_int_x86_avx512_mask_pmins_w_128:
2392 ; X86: # %bb.0:
2393 ; X86-NEXT: vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
2394 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2395 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2396 ; X86-NEXT: vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
2397 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2398 ; X86-NEXT: retl # encoding: [0xc3]
2399 ;
2400 ; X64-LABEL: test_int_x86_avx512_mask_pmins_w_128:
2401 ; X64: # %bb.0:
2402 ; X64-NEXT: vpminsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xea,0xd9]
2403 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2404 ; X64-NEXT: vpminsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xea,0xd1]
2405 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2406 ; X64-NEXT: retq # encoding: [0xc3]
2407 %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)
2408 %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)
2409 %res2 = add <8 x i16> %res, %res1
2410 ret <8 x i16> %res2
2411 }
2413 declare <16 x i16> @llvm.x86.avx512.mask.pmins.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2415 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) {
2416 ; X86-LABEL: test_int_x86_avx512_mask_pmins_w_256:
2417 ; X86: # %bb.0:
2418 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2419 ; X86-NEXT: vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
2420 ; X86-NEXT: vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
2421 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2422 ; X86-NEXT: retl # encoding: [0xc3]
2423 ;
2424 ; X64-LABEL: test_int_x86_avx512_mask_pmins_w_256:
2425 ; X64: # %bb.0:
2426 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2427 ; X64-NEXT: vpminsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xea,0xd1]
2428 ; X64-NEXT: vpminsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xea,0xc1]
2429 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2430 ; X64-NEXT: retq # encoding: [0xc3]
2431 %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)
2432 %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)
2433 %res2 = add <16 x i16> %res, %res1
2434 ret <16 x i16> %res2
2435 }
2437 declare <16 x i8> @llvm.x86.avx512.mask.pminu.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2439 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) {
2440 ; X86-LABEL: test_int_x86_avx512_mask_pminu_b_128:
2441 ; X86: # %bb.0:
2442 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2443 ; X86-NEXT: vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
2444 ; X86-NEXT: vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
2445 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2446 ; X86-NEXT: retl # encoding: [0xc3]
2447 ;
2448 ; X64-LABEL: test_int_x86_avx512_mask_pminu_b_128:
2449 ; X64: # %bb.0:
2450 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2451 ; X64-NEXT: vpminub %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xda,0xd1]
2452 ; X64-NEXT: vpminub %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xda,0xc1]
2453 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
2454 ; X64-NEXT: retq # encoding: [0xc3]
2455 %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)
2456 %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)
2457 %res2 = add <16 x i8> %res, %res1
2458 ret <16 x i8> %res2
2459 }
2461 declare <32 x i8> @llvm.x86.avx512.mask.pminu.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2463 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) {
2464 ; X86-LABEL: test_int_x86_avx512_mask_pminu_b_256:
2465 ; X86: # %bb.0:
2466 ; X86-NEXT: vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
2467 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2468 ; X86-NEXT: vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
2469 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2470 ; X86-NEXT: retl # encoding: [0xc3]
2471 ;
2472 ; X64-LABEL: test_int_x86_avx512_mask_pminu_b_256:
2473 ; X64: # %bb.0:
2474 ; X64-NEXT: vpminub %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xda,0xd9]
2475 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2476 ; X64-NEXT: vpminub %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xda,0xd1]
2477 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2478 ; X64-NEXT: retq # encoding: [0xc3]
2479 %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)
2480 %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)
2481 %res2 = add <32 x i8> %res, %res1
2482 ret <32 x i8> %res2
2483 }
2485 declare <8 x i16> @llvm.x86.avx512.mask.pminu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2487 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) {
2488 ; X86-LABEL: test_int_x86_avx512_mask_pminu_w_128:
2489 ; X86: # %bb.0:
2490 ; X86-NEXT: vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
2491 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2492 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2493 ; X86-NEXT: vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
2494 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2495 ; X86-NEXT: retl # encoding: [0xc3]
2496 ;
2497 ; X64-LABEL: test_int_x86_avx512_mask_pminu_w_128:
2498 ; X64: # %bb.0:
2499 ; X64-NEXT: vpminuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x3a,0xd9]
2500 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2501 ; X64-NEXT: vpminuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x3a,0xd1]
2502 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
2503 ; X64-NEXT: retq # encoding: [0xc3]
2504 %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)
2505 %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)
2506 %res2 = add <8 x i16> %res, %res1
2507 ret <8 x i16> %res2
2508 }
2510 declare <16 x i16> @llvm.x86.avx512.mask.pminu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2512 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) {
2513 ; X86-LABEL: test_int_x86_avx512_mask_pminu_w_256:
2514 ; X86: # %bb.0:
2515 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2516 ; X86-NEXT: vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
2517 ; X86-NEXT: vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
2518 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2519 ; X86-NEXT: retl # encoding: [0xc3]
2520 ;
2521 ; X64-LABEL: test_int_x86_avx512_mask_pminu_w_256:
2522 ; X64: # %bb.0:
2523 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2524 ; X64-NEXT: vpminuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x3a,0xd1]
2525 ; X64-NEXT: vpminuw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x3a,0xc1]
2526 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2527 ; X64-NEXT: retq # encoding: [0xc3]
2528 %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)
2529 %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)
2530 %res2 = add <16 x i16> %res, %res1
2531 ret <16 x i16> %res2
2532 }
2534 declare <8 x i16> @llvm.x86.avx512.mask.psrl.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2536 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) {
2537 ; X86-LABEL: test_int_x86_avx512_mask_psrl_w_128:
2538 ; X86: # %bb.0:
2539 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
2540 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2541 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2542 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
2543 ; X86-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
2544 ; X86-NEXT: vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
2545 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2546 ; X86-NEXT: retl # encoding: [0xc3]
2547 ;
2548 ; X64-LABEL: test_int_x86_avx512_mask_psrl_w_128:
2549 ; X64: # %bb.0:
2550 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd1,0xd9]
2551 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2552 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd1,0xd1]
2553 ; X64-NEXT: vpsrlw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd1,0xc1]
2554 ; X64-NEXT: vpaddw %xmm0, %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe1,0xfd,0xc0]
2555 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2556 ; X64-NEXT: retq # encoding: [0xc3]
2557 %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)
2558 %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)
2559 %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)
2560 %res3 = add <8 x i16> %res, %res1
2561 %res4 = add <8 x i16> %res2, %res3
2562 ret <8 x i16> %res4
2563 }
2565 declare <16 x i16> @llvm.x86.avx512.mask.psrl.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2567 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) {
2568 ; X86-LABEL: test_int_x86_avx512_mask_psrl_w_256:
2569 ; X86: # %bb.0:
2570 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
2571 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2572 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
2573 ; X86-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
2574 ; X86-NEXT: vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
2575 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2576 ; X86-NEXT: retl # encoding: [0xc3]
2577 ;
2578 ; X64-LABEL: test_int_x86_avx512_mask_psrl_w_256:
2579 ; X64: # %bb.0:
2580 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd1,0xd9]
2581 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2582 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd1,0xd1]
2583 ; X64-NEXT: vpsrlw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd1,0xc1]
2584 ; X64-NEXT: vpaddw %ymm0, %ymm3, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xe5,0xfd,0xc0]
2585 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2586 ; X64-NEXT: retq # encoding: [0xc3]
2587 %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)
2588 %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)
2589 %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)
2590 %res3 = add <16 x i16> %res, %res1
2591 %res4 = add <16 x i16> %res3, %res2
2592 ret <16 x i16> %res4
2593 }
2595 declare <8 x i16> @llvm.x86.avx512.mask.psra.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2597 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) {
2598 ; X86-LABEL: test_int_x86_avx512_mask_psra_w_128:
2599 ; X86: # %bb.0:
2600 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
2601 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2602 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2603 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
2604 ; X86-NEXT: vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
2605 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2606 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2607 ; X86-NEXT: retl # encoding: [0xc3]
2608 ;
2609 ; X64-LABEL: test_int_x86_avx512_mask_psra_w_128:
2610 ; X64: # %bb.0:
2611 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe1,0xd9]
2612 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2613 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe1,0xd1]
2614 ; X64-NEXT: vpsraw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe1,0xc1]
2615 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2616 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2617 ; X64-NEXT: retq # encoding: [0xc3]
2618 %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)
2619 %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)
2620 %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)
2621 %res3 = add <8 x i16> %res, %res1
2622 %res4 = add <8 x i16> %res3, %res2
2623 ret <8 x i16> %res4
2624 }
2626 declare <16 x i16> @llvm.x86.avx512.mask.psra.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2628 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) {
2629 ; X86-LABEL: test_int_x86_avx512_mask_psra_w_256:
2630 ; X86: # %bb.0:
2631 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
2632 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2633 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
2634 ; X86-NEXT: vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
2635 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2636 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2637 ; X86-NEXT: retl # encoding: [0xc3]
2638 ;
2639 ; X64-LABEL: test_int_x86_avx512_mask_psra_w_256:
2640 ; X64: # %bb.0:
2641 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe1,0xd9]
2642 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2643 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe1,0xd1]
2644 ; X64-NEXT: vpsraw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe1,0xc1]
2645 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2646 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2647 ; X64-NEXT: retq # encoding: [0xc3]
2648 %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)
2649 %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)
2650 %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)
2651 %res3 = add <16 x i16> %res, %res1
2652 %res4 = add <16 x i16> %res3, %res2
2653 ret <16 x i16> %res4
2654 }
2656 declare <8 x i16> @llvm.x86.avx512.mask.psll.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2658 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) {
2659 ; X86-LABEL: test_int_x86_avx512_mask_psll_w_128:
2660 ; X86: # %bb.0:
2661 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
2662 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2663 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2664 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
2665 ; X86-NEXT: vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
2666 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2667 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2668 ; X86-NEXT: retl # encoding: [0xc3]
2669 ;
2670 ; X64-LABEL: test_int_x86_avx512_mask_psll_w_128:
2671 ; X64: # %bb.0:
2672 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf1,0xd9]
2673 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2674 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf1,0xd1]
2675 ; X64-NEXT: vpsllw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xf1,0xc1]
2676 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
2677 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2678 ; X64-NEXT: retq # encoding: [0xc3]
2679 %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)
2680 %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)
2681 %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)
2682 %res3 = add <8 x i16> %res, %res1
2683 %res4 = add <8 x i16> %res3, %res2
2684 ret <8 x i16> %res4
2685 }
2687 declare <16 x i16> @llvm.x86.avx512.mask.psll.w.256(<16 x i16>, <8 x i16>, <16 x i16>, i16)
2689 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) {
2690 ; X86-LABEL: test_int_x86_avx512_mask_psll_w_256:
2691 ; X86: # %bb.0:
2692 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
2693 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2694 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
2695 ; X86-NEXT: vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
2696 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2697 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2698 ; X86-NEXT: retl # encoding: [0xc3]
2699 ;
2700 ; X64-LABEL: test_int_x86_avx512_mask_psll_w_256:
2701 ; X64: # %bb.0:
2702 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf1,0xd9]
2703 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2704 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf1,0xd1]
2705 ; X64-NEXT: vpsllw %xmm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xf1,0xc1]
2706 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
2707 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2708 ; X64-NEXT: retq # encoding: [0xc3]
2709 %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)
2710 %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)
2711 %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)
2712 %res3 = add <16 x i16> %res, %res1
2713 %res4 = add <16 x i16> %res3, %res2
2714 ret <16 x i16> %res4
2715 }
2717 declare <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2719 define <8 x i16>@test_int_x86_avx512_mask_psrl_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2720 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
2721 ; X86: # %bb.0:
2722 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
2723 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2724 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2725 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
2726 ; X86-NEXT: vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
2727 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2728 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2729 ; X86-NEXT: retl # encoding: [0xc3]
2730 ;
2731 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_128:
2732 ; X64: # %bb.0:
2733 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xd0,0x03]
2734 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2735 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xd0,0x03]
2736 ; X64-NEXT: vpsrlw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xd0,0x03]
2737 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
2738 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2739 ; X64-NEXT: retq # encoding: [0xc3]
2740 %res = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2741 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2742 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psrl.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2743 %res3 = add <8 x i16> %res, %res1
2744 %res4 = add <8 x i16> %res2, %res3
2745 ret <8 x i16> %res4
2746 }
2748 declare <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2750 define <16 x i16>@test_int_x86_avx512_mask_psrl_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2751 ; X86-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
2752 ; X86: # %bb.0:
2753 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
2754 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2755 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
2756 ; X86-NEXT: vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
2757 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2758 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2759 ; X86-NEXT: retl # encoding: [0xc3]
2760 ;
2761 ; X64-LABEL: test_int_x86_avx512_mask_psrl_wi_256:
2762 ; X64: # %bb.0:
2763 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xd0,0x03]
2764 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2765 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xd0,0x03]
2766 ; X64-NEXT: vpsrlw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xd0,0x03]
2767 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
2768 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2769 ; X64-NEXT: retq # encoding: [0xc3]
2770 %res = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2771 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2772 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psrl.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2773 %res3 = add <16 x i16> %res, %res1
2774 %res4 = add <16 x i16> %res3, %res2
2775 ret <16 x i16> %res4
2776 }
2778 declare <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2780 define <8 x i16>@test_int_x86_avx512_mask_psra_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2781 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_128:
2782 ; X86: # %bb.0:
2783 ; X86-NEXT: vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
2784 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2785 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2786 ; X86-NEXT: vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
2787 ; X86-NEXT: vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
2788 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2789 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2790 ; X86-NEXT: retl # encoding: [0xc3]
2791 ;
2792 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_128:
2793 ; X64: # %bb.0:
2794 ; X64-NEXT: vpsraw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xe0,0x03]
2795 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2796 ; X64-NEXT: vpsraw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xe0,0x03]
2797 ; X64-NEXT: vpsraw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xe0,0x03]
2798 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2799 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2800 ; X64-NEXT: retq # encoding: [0xc3]
2801 %res = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2802 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2803 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psra.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2804 %res3 = add <8 x i16> %res, %res1
2805 %res4 = add <8 x i16> %res3, %res2
2806 ret <8 x i16> %res4
2807 }
2809 declare <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2811 define <16 x i16>@test_int_x86_avx512_mask_psra_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2812 ; X86-LABEL: test_int_x86_avx512_mask_psra_wi_256:
2813 ; X86: # %bb.0:
2814 ; X86-NEXT: vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
2815 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2816 ; X86-NEXT: vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
2817 ; X86-NEXT: vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
2818 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2819 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2820 ; X86-NEXT: retl # encoding: [0xc3]
2821 ;
2822 ; X64-LABEL: test_int_x86_avx512_mask_psra_wi_256:
2823 ; X64: # %bb.0:
2824 ; X64-NEXT: vpsraw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xe0,0x03]
2825 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2826 ; X64-NEXT: vpsraw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xe0,0x03]
2827 ; X64-NEXT: vpsraw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xe0,0x03]
2828 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2829 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2830 ; X64-NEXT: retq # encoding: [0xc3]
2831 %res = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2832 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2833 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psra.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2834 %res3 = add <16 x i16> %res, %res1
2835 %res4 = add <16 x i16> %res3, %res2
2836 ret <16 x i16> %res4
2837 }
2839 declare <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16>, i32, <8 x i16>, i8)
2841 define <8 x i16>@test_int_x86_avx512_mask_psll_wi_128(<8 x i16> %x0, i32 %x1, <8 x i16> %x2, i8 %x3) {
2842 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_128:
2843 ; X86: # %bb.0:
2844 ; X86-NEXT: vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
2845 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
2846 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2847 ; X86-NEXT: vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
2848 ; X86-NEXT: vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
2849 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2850 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2851 ; X86-NEXT: retl # encoding: [0xc3]
2852 ;
2853 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_128:
2854 ; X64: # %bb.0:
2855 ; X64-NEXT: vpsllw $3, %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x71,0xf0,0x03]
2856 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2857 ; X64-NEXT: vpsllw $3, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x71,0xf0,0x03]
2858 ; X64-NEXT: vpsllw $3, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x71,0xf0,0x03]
2859 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2860 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2861 ; X64-NEXT: retq # encoding: [0xc3]
2862 %res = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 %x3)
2863 %res1 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> zeroinitializer, i8 %x3)
2864 %res2 = call <8 x i16> @llvm.x86.avx512.mask.psll.wi.128(<8 x i16> %x0, i32 3, <8 x i16> %x2, i8 -1)
2865 %res3 = add <8 x i16> %res, %res1
2866 %res4 = add <8 x i16> %res3, %res2
2867 ret <8 x i16> %res4
2868 }
2870 declare <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16>, i32, <16 x i16>, i16)
2872 define <16 x i16>@test_int_x86_avx512_mask_psll_wi_256(<16 x i16> %x0, i32 %x1, <16 x i16> %x2, i16 %x3) {
2873 ; X86-LABEL: test_int_x86_avx512_mask_psll_wi_256:
2874 ; X86: # %bb.0:
2875 ; X86-NEXT: vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
2876 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
2877 ; X86-NEXT: vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
2878 ; X86-NEXT: vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
2879 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2880 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2881 ; X86-NEXT: retl # encoding: [0xc3]
2882 ;
2883 ; X64-LABEL: test_int_x86_avx512_mask_psll_wi_256:
2884 ; X64: # %bb.0:
2885 ; X64-NEXT: vpsllw $3, %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc5,0xed,0x71,0xf0,0x03]
2886 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
2887 ; X64-NEXT: vpsllw $3, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x71,0xf0,0x03]
2888 ; X64-NEXT: vpsllw $3, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x71,0xf0,0x03]
2889 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2890 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2891 ; X64-NEXT: retq # encoding: [0xc3]
2892 %res = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 %x3)
2893 %res1 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> zeroinitializer, i16 %x3)
2894 %res2 = call <16 x i16> @llvm.x86.avx512.mask.psll.wi.256(<16 x i16> %x0, i32 3, <16 x i16> %x2, i16 -1)
2895 %res3 = add <16 x i16> %res, %res1
2896 %res4 = add <16 x i16> %res3, %res2
2897 ret <16 x i16> %res4
2898 }
2900 declare <16 x i8> @llvm.x86.avx512.mask.pshuf.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
2902 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) {
2903 ; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
2904 ; X86: # %bb.0:
2905 ; X86-NEXT: vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
2906 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2907 ; X86-NEXT: vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
2908 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
2909 ; X86-NEXT: retl # encoding: [0xc3]
2910 ;
2911 ; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_128:
2912 ; X64: # %bb.0:
2913 ; X64-NEXT: vpshufb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x00,0xd9]
2914 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2915 ; X64-NEXT: vpshufb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x00,0xd1]
2916 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
2917 ; X64-NEXT: retq # encoding: [0xc3]
2918 %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)
2919 %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)
2920 %res2 = add <16 x i8> %res, %res1
2921 ret <16 x i8> %res2
2922 }
2924 declare <32 x i8> @llvm.x86.avx512.mask.pshuf.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
2926 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) {
2927 ; X86-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
2928 ; X86: # %bb.0:
2929 ; X86-NEXT: vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
2930 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
2931 ; X86-NEXT: vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
2932 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2933 ; X86-NEXT: retl # encoding: [0xc3]
2934 ;
2935 ; X64-LABEL: test_int_x86_avx512_mask_pshuf_b_256:
2936 ; X64: # %bb.0:
2937 ; X64-NEXT: vpshufb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x00,0xd9]
2938 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2939 ; X64-NEXT: vpshufb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x00,0xd1]
2940 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
2941 ; X64-NEXT: retq # encoding: [0xc3]
2942 %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)
2943 %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)
2944 %res2 = add <32 x i8> %res, %res1
2945 ret <32 x i8> %res2
2946 }
2948 declare <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8>, <8 x i16>, i8)
2950 define <8 x i16>@test_int_x86_avx512_mask_pmovzxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
2951 ; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
2952 ; X86: # %bb.0:
2953 ; X86-NEXT: vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
2954 ; 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
2955 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2956 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2957 ; X86-NEXT: vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
2958 ; 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
2959 ; X86-NEXT: vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
2960 ; 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
2961 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2962 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2963 ; X86-NEXT: retl # encoding: [0xc3]
2964 ;
2965 ; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_128:
2966 ; X64: # %bb.0:
2967 ; X64-NEXT: vpmovzxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x30,0xd0]
2968 ; 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
2969 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2970 ; X64-NEXT: vpmovzxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x30,0xc8]
2971 ; 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
2972 ; X64-NEXT: vpmovzxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x30,0xc0]
2973 ; 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
2974 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
2975 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
2976 ; X64-NEXT: retq # encoding: [0xc3]
2977 %res = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
2978 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
2979 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovzxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
2980 %res3 = add <8 x i16> %res, %res1
2981 %res4 = add <8 x i16> %res3, %res2
2982 ret <8 x i16> %res4
2983 }
2985 declare <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8>, <16 x i16>, i16)
2987 define <16 x i16>@test_int_x86_avx512_mask_pmovzxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
2988 ; X86-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
2989 ; X86: # %bb.0:
2990 ; X86-NEXT: vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
2991 ; 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
2992 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2993 ; X86-NEXT: vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
2994 ; 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
2995 ; X86-NEXT: vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
2996 ; 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
2997 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
2998 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
2999 ; X86-NEXT: retl # encoding: [0xc3]
3000 ;
3001 ; X64-LABEL: test_int_x86_avx512_mask_pmovzxb_w_256:
3002 ; X64: # %bb.0:
3003 ; X64-NEXT: vpmovzxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x30,0xd0]
3004 ; 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
3005 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3006 ; X64-NEXT: vpmovzxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x30,0xc8]
3007 ; 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
3008 ; X64-NEXT: vpmovzxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x30,0xc0]
3009 ; 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
3010 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3011 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3012 ; X64-NEXT: retq # encoding: [0xc3]
3013 %res = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
3014 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
3015 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovzxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
3016 %res3 = add <16 x i16> %res, %res1
3017 %res4 = add <16 x i16> %res3, %res2
3018 ret <16 x i16> %res4
3019 }
3022 declare <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8>, <8 x i16>, i8)
3024 define <8 x i16>@test_int_x86_avx512_mask_pmovsxb_w_128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2) {
3025 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
3026 ; X86: # %bb.0:
3027 ; X86-NEXT: vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
3028 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3029 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3030 ; X86-NEXT: vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
3031 ; X86-NEXT: vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
3032 ; X86-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
3033 ; X86-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
3034 ; X86-NEXT: retl # encoding: [0xc3]
3035 ;
3036 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_128:
3037 ; X64: # %bb.0:
3038 ; X64-NEXT: vpmovsxbw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x20,0xd0]
3039 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3040 ; X64-NEXT: vpmovsxbw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x20,0xc8]
3041 ; X64-NEXT: vpmovsxbw %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x20,0xc0]
3042 ; X64-NEXT: vpaddw %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc2]
3043 ; X64-NEXT: vpaddw %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc0]
3044 ; X64-NEXT: retq # encoding: [0xc3]
3045 %res = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 %x2)
3046 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> zeroinitializer, i8 %x2)
3047 %res2 = call <8 x i16> @llvm.x86.avx512.mask.pmovsxb.w.128(<16 x i8> %x0, <8 x i16> %x1, i8 -1)
3048 %res3 = add <8 x i16> %res, %res1
3049 %res4 = add <8 x i16> %res3, %res2
3050 ret <8 x i16> %res4
3051 }
3053 declare <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8>, <16 x i16>, i16)
3055 define <16 x i16>@test_int_x86_avx512_mask_pmovsxb_w_256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2) {
3056 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
3057 ; X86: # %bb.0:
3058 ; X86-NEXT: vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
3059 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3060 ; X86-NEXT: vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
3061 ; X86-NEXT: vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
3062 ; X86-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3063 ; X86-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3064 ; X86-NEXT: retl # encoding: [0xc3]
3065 ;
3066 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxb_w_256:
3067 ; X64: # %bb.0:
3068 ; X64-NEXT: vpmovsxbw %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x20,0xd0]
3069 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3070 ; X64-NEXT: vpmovsxbw %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x20,0xc8]
3071 ; X64-NEXT: vpmovsxbw %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x20,0xc0]
3072 ; X64-NEXT: vpaddw %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc2]
3073 ; X64-NEXT: vpaddw %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc0]
3074 ; X64-NEXT: retq # encoding: [0xc3]
3075 %res = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 %x2)
3076 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> zeroinitializer, i16 %x2)
3077 %res2 = call <16 x i16> @llvm.x86.avx512.mask.pmovsxb.w.256(<16 x i8> %x0, <16 x i16> %x1, i16 -1)
3078 %res3 = add <16 x i16> %res, %res1
3079 %res4 = add <16 x i16> %res3, %res2
3080 ret <16 x i16> %res4
3081 }
3083 declare <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32>, <2 x i64>, i8)
3085 define <2 x i64>@test_int_x86_avx512_mask_pmovsxd_q_128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2) {
3086 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
3087 ; X86: # %bb.0:
3088 ; X86-NEXT: vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
3089 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3090 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3091 ; X86-NEXT: vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
3092 ; X86-NEXT: vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
3093 ; X86-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
3094 ; X86-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
3095 ; X86-NEXT: retl # encoding: [0xc3]
3096 ;
3097 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_128:
3098 ; X64: # %bb.0:
3099 ; X64-NEXT: vpmovsxdq %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x25,0xd0]
3100 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3101 ; X64-NEXT: vpmovsxdq %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x25,0xc8]
3102 ; X64-NEXT: vpmovsxdq %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x25,0xc0]
3103 ; X64-NEXT: vpaddq %xmm2, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd4,0xc2]
3104 ; X64-NEXT: vpaddq %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xd4,0xc0]
3105 ; X64-NEXT: retq # encoding: [0xc3]
3106 %res = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 %x2)
3107 %res1 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> zeroinitializer, i8 %x2)
3108 %res2 = call <2 x i64> @llvm.x86.avx512.mask.pmovsxd.q.128(<4 x i32> %x0, <2 x i64> %x1, i8 -1)
3109 %res3 = add <2 x i64> %res, %res1
3110 %res4 = add <2 x i64> %res3, %res2
3111 ret <2 x i64> %res4
3112 }
3114 declare <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32>, <4 x i64>, i8)
3116 define <4 x i64>@test_int_x86_avx512_mask_pmovsxd_q_256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2) {
3117 ; X86-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
3118 ; X86: # %bb.0:
3119 ; X86-NEXT: vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
3120 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3121 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3122 ; X86-NEXT: vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
3123 ; X86-NEXT: vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
3124 ; X86-NEXT: vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
3125 ; X86-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
3126 ; X86-NEXT: retl # encoding: [0xc3]
3127 ;
3128 ; X64-LABEL: test_int_x86_avx512_mask_pmovsxd_q_256:
3129 ; X64: # %bb.0:
3130 ; X64-NEXT: vpmovsxdq %xmm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x25,0xd0]
3131 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3132 ; X64-NEXT: vpmovsxdq %xmm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x25,0xc8]
3133 ; X64-NEXT: vpmovsxdq %xmm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x25,0xc0]
3134 ; X64-NEXT: vpaddq %ymm2, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd4,0xc2]
3135 ; X64-NEXT: vpaddq %ymm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xd4,0xc0]
3136 ; X64-NEXT: retq # encoding: [0xc3]
3137 %res = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 %x2)
3138 %res1 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> zeroinitializer, i8 %x2)
3139 %res2 = call <4 x i64> @llvm.x86.avx512.mask.pmovsxd.q.256(<4 x i32> %x0, <4 x i64> %x1, i8 -1)
3140 %res3 = add <4 x i64> %res, %res1
3141 %res4 = add <4 x i64> %res3, %res2
3142 ret <4 x i64> %res4
3143 }
3146 declare <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16)
3148 define <16 x i8>@test_int_x86_avx512_cvtmask2b_128(i16 %x0) {
3149 ; X86-LABEL: test_int_x86_avx512_cvtmask2b_128:
3150 ; X86: # %bb.0:
3151 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
3152 ; X86-NEXT: vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
3153 ; X86-NEXT: retl # encoding: [0xc3]
3154 ;
3155 ; X64-LABEL: test_int_x86_avx512_cvtmask2b_128:
3156 ; X64: # %bb.0:
3157 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3158 ; X64-NEXT: vpmovm2b %k0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x08,0x28,0xc0]
3159 ; X64-NEXT: retq # encoding: [0xc3]
3160 %res = call <16 x i8> @llvm.x86.avx512.cvtmask2b.128(i16 %x0)
3161 ret <16 x i8> %res
3162 }
3164 declare <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32)
3166 define <32 x i8>@test_int_x86_avx512_cvtmask2b_256(i32 %x0) {
3167 ; X86-LABEL: test_int_x86_avx512_cvtmask2b_256:
3168 ; X86: # %bb.0:
3169 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k0 # encoding: [0xc4,0xe1,0xf9,0x90,0x44,0x24,0x04]
3170 ; X86-NEXT: vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
3171 ; X86-NEXT: retl # encoding: [0xc3]
3172 ;
3173 ; X64-LABEL: test_int_x86_avx512_cvtmask2b_256:
3174 ; X64: # %bb.0:
3175 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3176 ; X64-NEXT: vpmovm2b %k0, %ymm0 # encoding: [0x62,0xf2,0x7e,0x28,0x28,0xc0]
3177 ; X64-NEXT: retq # encoding: [0xc3]
3178 %res = call <32 x i8> @llvm.x86.avx512.cvtmask2b.256(i32 %x0)
3179 ret <32 x i8> %res
3180 }
3182 declare <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8)
3184 define <8 x i16>@test_int_x86_avx512_cvtmask2w_128(i8 %x0) {
3185 ; X86-LABEL: test_int_x86_avx512_cvtmask2w_128:
3186 ; X86: # %bb.0:
3187 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3188 ; X86-NEXT: kmovd %eax, %k0 # encoding: [0xc5,0xfb,0x92,0xc0]
3189 ; X86-NEXT: vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
3190 ; X86-NEXT: retl # encoding: [0xc3]
3191 ;
3192 ; X64-LABEL: test_int_x86_avx512_cvtmask2w_128:
3193 ; X64: # %bb.0:
3194 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3195 ; X64-NEXT: vpmovm2w %k0, %xmm0 # encoding: [0x62,0xf2,0xfe,0x08,0x28,0xc0]
3196 ; X64-NEXT: retq # encoding: [0xc3]
3197 %res = call <8 x i16> @llvm.x86.avx512.cvtmask2w.128(i8 %x0)
3198 ret <8 x i16> %res
3199 }
3201 declare <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16)
3203 define <16 x i16>@test_int_x86_avx512_cvtmask2w_256(i16 %x0) {
3204 ; X86-LABEL: test_int_x86_avx512_cvtmask2w_256:
3205 ; X86: # %bb.0:
3206 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k0 # encoding: [0xc5,0xf8,0x90,0x44,0x24,0x04]
3207 ; X86-NEXT: vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
3208 ; X86-NEXT: retl # encoding: [0xc3]
3209 ;
3210 ; X64-LABEL: test_int_x86_avx512_cvtmask2w_256:
3211 ; X64: # %bb.0:
3212 ; X64-NEXT: kmovd %edi, %k0 # encoding: [0xc5,0xfb,0x92,0xc7]
3213 ; X64-NEXT: vpmovm2w %k0, %ymm0 # encoding: [0x62,0xf2,0xfe,0x28,0x28,0xc0]
3214 ; X64-NEXT: retq # encoding: [0xc3]
3215 %res = call <16 x i16> @llvm.x86.avx512.cvtmask2w.256(i16 %x0)
3216 ret <16 x i16> %res
3217 }
3218 define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
3219 ; CHECK-LABEL: test_mask_packs_epi32_rr_128:
3220 ; CHECK: # %bb.0:
3221 ; CHECK-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
3222 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3223 %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)
3224 ret <8 x i16> %res
3225 }
3227 define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
3228 ; X86-LABEL: test_mask_packs_epi32_rrk_128:
3229 ; X86: # %bb.0:
3230 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3231 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3232 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
3233 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3234 ; X86-NEXT: retl # encoding: [0xc3]
3235 ;
3236 ; X64-LABEL: test_mask_packs_epi32_rrk_128:
3237 ; X64: # %bb.0:
3238 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3239 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
3240 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3241 ; X64-NEXT: retq # encoding: [0xc3]
3242 %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)
3243 ret <8 x i16> %res
3244 }
3246 define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
3247 ; X86-LABEL: test_mask_packs_epi32_rrkz_128:
3248 ; X86: # %bb.0:
3249 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3250 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3251 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
3252 ; X86-NEXT: retl # encoding: [0xc3]
3253 ;
3254 ; X64-LABEL: test_mask_packs_epi32_rrkz_128:
3255 ; X64: # %bb.0:
3256 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3257 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
3258 ; X64-NEXT: retq # encoding: [0xc3]
3259 %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)
3260 ret <8 x i16> %res
3261 }
3263 define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
3264 ; X86-LABEL: test_mask_packs_epi32_rm_128:
3265 ; X86: # %bb.0:
3266 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3267 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x00]
3268 ; X86-NEXT: retl # encoding: [0xc3]
3269 ;
3270 ; X64-LABEL: test_mask_packs_epi32_rm_128:
3271 ; X64: # %bb.0:
3272 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
3273 ; X64-NEXT: retq # encoding: [0xc3]
3274 %b = load <4 x i32>, <4 x i32>* %ptr_b
3275 %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)
3276 ret <8 x i16> %res
3277 }
3279 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) {
3280 ; X86-LABEL: test_mask_packs_epi32_rmk_128:
3281 ; X86: # %bb.0:
3282 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3283 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3284 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3285 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x08]
3286 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3287 ; X86-NEXT: retl # encoding: [0xc3]
3288 ;
3289 ; X64-LABEL: test_mask_packs_epi32_rmk_128:
3290 ; X64: # %bb.0:
3291 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3292 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
3293 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3294 ; X64-NEXT: retq # encoding: [0xc3]
3295 %b = load <4 x i32>, <4 x i32>* %ptr_b
3296 %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)
3297 ret <8 x i16> %res
3298 }
3300 define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
3301 ; X86-LABEL: test_mask_packs_epi32_rmkz_128:
3302 ; X86: # %bb.0:
3303 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3304 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3305 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3306 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x00]
3307 ; X86-NEXT: retl # encoding: [0xc3]
3308 ;
3309 ; X64-LABEL: test_mask_packs_epi32_rmkz_128:
3310 ; X64: # %bb.0:
3311 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3312 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
3313 ; X64-NEXT: retq # encoding: [0xc3]
3314 %b = load <4 x i32>, <4 x i32>* %ptr_b
3315 %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)
3316 ret <8 x i16> %res
3317 }
3319 define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
3320 ; X86-LABEL: test_mask_packs_epi32_rmb_128:
3321 ; X86: # %bb.0:
3322 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3323 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x00]
3324 ; X86-NEXT: retl # encoding: [0xc3]
3325 ;
3326 ; X64-LABEL: test_mask_packs_epi32_rmb_128:
3327 ; X64: # %bb.0:
3328 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
3329 ; X64-NEXT: retq # encoding: [0xc3]
3330 %q = load i32, i32* %ptr_b
3331 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3332 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3333 %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)
3334 ret <8 x i16> %res
3335 }
3337 define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3338 ; X86-LABEL: test_mask_packs_epi32_rmbk_128:
3339 ; X86: # %bb.0:
3340 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3341 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3342 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3343 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x08]
3344 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3345 ; X86-NEXT: retl # encoding: [0xc3]
3346 ;
3347 ; X64-LABEL: test_mask_packs_epi32_rmbk_128:
3348 ; X64: # %bb.0:
3349 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3350 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
3351 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3352 ; X64-NEXT: retq # encoding: [0xc3]
3353 %q = load i32, i32* %ptr_b
3354 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3355 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3356 %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)
3357 ret <8 x i16> %res
3358 }
3360 define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
3361 ; X86-LABEL: test_mask_packs_epi32_rmbkz_128:
3362 ; X86: # %bb.0:
3363 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3364 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3365 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3366 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x00]
3367 ; X86-NEXT: retl # encoding: [0xc3]
3368 ;
3369 ; X64-LABEL: test_mask_packs_epi32_rmbkz_128:
3370 ; X64: # %bb.0:
3371 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3372 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
3373 ; X64-NEXT: retq # encoding: [0xc3]
3374 %q = load i32, i32* %ptr_b
3375 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3376 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3377 %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)
3378 ret <8 x i16> %res
3379 }
3381 declare <8 x i16> @llvm.x86.avx512.mask.packssdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
3383 define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
3384 ; CHECK-LABEL: test_mask_packs_epi32_rr_256:
3385 ; CHECK: # %bb.0:
3386 ; CHECK-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
3387 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3388 %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)
3389 ret <16 x i16> %res
3390 }
3392 define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
3393 ; X86-LABEL: test_mask_packs_epi32_rrk_256:
3394 ; X86: # %bb.0:
3395 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3396 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
3397 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3398 ; X86-NEXT: retl # encoding: [0xc3]
3399 ;
3400 ; X64-LABEL: test_mask_packs_epi32_rrk_256:
3401 ; X64: # %bb.0:
3402 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3403 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
3404 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3405 ; X64-NEXT: retq # encoding: [0xc3]
3406 %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)
3407 ret <16 x i16> %res
3408 }
3410 define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
3411 ; X86-LABEL: test_mask_packs_epi32_rrkz_256:
3412 ; X86: # %bb.0:
3413 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3414 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
3415 ; X86-NEXT: retl # encoding: [0xc3]
3416 ;
3417 ; X64-LABEL: test_mask_packs_epi32_rrkz_256:
3418 ; X64: # %bb.0:
3419 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3420 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
3421 ; X64-NEXT: retq # encoding: [0xc3]
3422 %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)
3423 ret <16 x i16> %res
3424 }
3426 define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
3427 ; X86-LABEL: test_mask_packs_epi32_rm_256:
3428 ; X86: # %bb.0:
3429 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3430 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x00]
3431 ; X86-NEXT: retl # encoding: [0xc3]
3432 ;
3433 ; X64-LABEL: test_mask_packs_epi32_rm_256:
3434 ; X64: # %bb.0:
3435 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
3436 ; X64-NEXT: retq # encoding: [0xc3]
3437 %b = load <8 x i32>, <8 x i32>* %ptr_b
3438 %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)
3439 ret <16 x i16> %res
3440 }
3442 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) {
3443 ; X86-LABEL: test_mask_packs_epi32_rmk_256:
3444 ; X86: # %bb.0:
3445 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3446 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3447 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x08]
3448 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3449 ; X86-NEXT: retl # encoding: [0xc3]
3450 ;
3451 ; X64-LABEL: test_mask_packs_epi32_rmk_256:
3452 ; X64: # %bb.0:
3453 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3454 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
3455 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3456 ; X64-NEXT: retq # encoding: [0xc3]
3457 %b = load <8 x i32>, <8 x i32>* %ptr_b
3458 %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)
3459 ret <16 x i16> %res
3460 }
3462 define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
3463 ; X86-LABEL: test_mask_packs_epi32_rmkz_256:
3464 ; X86: # %bb.0:
3465 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3466 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3467 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x00]
3468 ; X86-NEXT: retl # encoding: [0xc3]
3469 ;
3470 ; X64-LABEL: test_mask_packs_epi32_rmkz_256:
3471 ; X64: # %bb.0:
3472 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3473 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
3474 ; X64-NEXT: retq # encoding: [0xc3]
3475 %b = load <8 x i32>, <8 x i32>* %ptr_b
3476 %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)
3477 ret <16 x i16> %res
3478 }
3480 define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
3481 ; X86-LABEL: test_mask_packs_epi32_rmb_256:
3482 ; X86: # %bb.0:
3483 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3484 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x00]
3485 ; X86-NEXT: retl # encoding: [0xc3]
3486 ;
3487 ; X64-LABEL: test_mask_packs_epi32_rmb_256:
3488 ; X64: # %bb.0:
3489 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
3490 ; X64-NEXT: retq # encoding: [0xc3]
3491 %q = load i32, i32* %ptr_b
3492 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3493 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3494 %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)
3495 ret <16 x i16> %res
3496 }
3498 define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
3499 ; X86-LABEL: test_mask_packs_epi32_rmbk_256:
3500 ; X86: # %bb.0:
3501 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3502 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3503 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x08]
3504 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3505 ; X86-NEXT: retl # encoding: [0xc3]
3506 ;
3507 ; X64-LABEL: test_mask_packs_epi32_rmbk_256:
3508 ; X64: # %bb.0:
3509 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3510 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
3511 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3512 ; X64-NEXT: retq # encoding: [0xc3]
3513 %q = load i32, i32* %ptr_b
3514 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3515 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3516 %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)
3517 ret <16 x i16> %res
3518 }
3520 define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
3521 ; X86-LABEL: test_mask_packs_epi32_rmbkz_256:
3522 ; X86: # %bb.0:
3523 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3524 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3525 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x00]
3526 ; X86-NEXT: retl # encoding: [0xc3]
3527 ;
3528 ; X64-LABEL: test_mask_packs_epi32_rmbkz_256:
3529 ; X64: # %bb.0:
3530 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3531 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
3532 ; X64-NEXT: retq # encoding: [0xc3]
3533 %q = load i32, i32* %ptr_b
3534 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
3535 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
3536 %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)
3537 ret <16 x i16> %res
3538 }
3540 declare <16 x i16> @llvm.x86.avx512.mask.packssdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
3542 define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
3543 ; CHECK-LABEL: test_mask_packs_epi16_rr_128:
3544 ; CHECK: # %bb.0:
3545 ; CHECK-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
3546 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3547 %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)
3548 ret <16 x i8> %res
3549 }
3551 define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
3552 ; X86-LABEL: test_mask_packs_epi16_rrk_128:
3553 ; X86: # %bb.0:
3554 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3555 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
3556 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3557 ; X86-NEXT: retl # encoding: [0xc3]
3558 ;
3559 ; X64-LABEL: test_mask_packs_epi16_rrk_128:
3560 ; X64: # %bb.0:
3561 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3562 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
3563 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3564 ; X64-NEXT: retq # encoding: [0xc3]
3565 %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)
3566 ret <16 x i8> %res
3567 }
3569 define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
3570 ; X86-LABEL: test_mask_packs_epi16_rrkz_128:
3571 ; X86: # %bb.0:
3572 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3573 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
3574 ; X86-NEXT: retl # encoding: [0xc3]
3575 ;
3576 ; X64-LABEL: test_mask_packs_epi16_rrkz_128:
3577 ; X64: # %bb.0:
3578 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3579 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
3580 ; X64-NEXT: retq # encoding: [0xc3]
3581 %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)
3582 ret <16 x i8> %res
3583 }
3585 define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
3586 ; X86-LABEL: test_mask_packs_epi16_rm_128:
3587 ; X86: # %bb.0:
3588 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3589 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x00]
3590 ; X86-NEXT: retl # encoding: [0xc3]
3591 ;
3592 ; X64-LABEL: test_mask_packs_epi16_rm_128:
3593 ; X64: # %bb.0:
3594 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
3595 ; X64-NEXT: retq # encoding: [0xc3]
3596 %b = load <8 x i16>, <8 x i16>* %ptr_b
3597 %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)
3598 ret <16 x i8> %res
3599 }
3601 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) {
3602 ; X86-LABEL: test_mask_packs_epi16_rmk_128:
3603 ; X86: # %bb.0:
3604 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3605 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3606 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x08]
3607 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3608 ; X86-NEXT: retl # encoding: [0xc3]
3609 ;
3610 ; X64-LABEL: test_mask_packs_epi16_rmk_128:
3611 ; X64: # %bb.0:
3612 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3613 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
3614 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3615 ; X64-NEXT: retq # encoding: [0xc3]
3616 %b = load <8 x i16>, <8 x i16>* %ptr_b
3617 %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)
3618 ret <16 x i8> %res
3619 }
3621 define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
3622 ; X86-LABEL: test_mask_packs_epi16_rmkz_128:
3623 ; X86: # %bb.0:
3624 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3625 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3626 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x00]
3627 ; X86-NEXT: retl # encoding: [0xc3]
3628 ;
3629 ; X64-LABEL: test_mask_packs_epi16_rmkz_128:
3630 ; X64: # %bb.0:
3631 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3632 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
3633 ; X64-NEXT: retq # encoding: [0xc3]
3634 %b = load <8 x i16>, <8 x i16>* %ptr_b
3635 %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)
3636 ret <16 x i8> %res
3637 }
3639 declare <16 x i8> @llvm.x86.avx512.mask.packsswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
3641 define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
3642 ; CHECK-LABEL: test_mask_packs_epi16_rr_256:
3643 ; CHECK: # %bb.0:
3644 ; CHECK-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
3645 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3646 %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)
3647 ret <32 x i8> %res
3648 }
3650 define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
3651 ; X86-LABEL: test_mask_packs_epi16_rrk_256:
3652 ; X86: # %bb.0:
3653 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
3654 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
3655 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3656 ; X86-NEXT: retl # encoding: [0xc3]
3657 ;
3658 ; X64-LABEL: test_mask_packs_epi16_rrk_256:
3659 ; X64: # %bb.0:
3660 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3661 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
3662 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3663 ; X64-NEXT: retq # encoding: [0xc3]
3664 %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)
3665 ret <32 x i8> %res
3666 }
3668 define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
3669 ; X86-LABEL: test_mask_packs_epi16_rrkz_256:
3670 ; X86: # %bb.0:
3671 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
3672 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
3673 ; X86-NEXT: retl # encoding: [0xc3]
3674 ;
3675 ; X64-LABEL: test_mask_packs_epi16_rrkz_256:
3676 ; X64: # %bb.0:
3677 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3678 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
3679 ; X64-NEXT: retq # encoding: [0xc3]
3680 %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)
3681 ret <32 x i8> %res
3682 }
3684 define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
3685 ; X86-LABEL: test_mask_packs_epi16_rm_256:
3686 ; X86: # %bb.0:
3687 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3688 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x00]
3689 ; X86-NEXT: retl # encoding: [0xc3]
3690 ;
3691 ; X64-LABEL: test_mask_packs_epi16_rm_256:
3692 ; X64: # %bb.0:
3693 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
3694 ; X64-NEXT: retq # encoding: [0xc3]
3695 %b = load <16 x i16>, <16 x i16>* %ptr_b
3696 %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)
3697 ret <32 x i8> %res
3698 }
3700 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) {
3701 ; X86-LABEL: test_mask_packs_epi16_rmk_256:
3702 ; X86: # %bb.0:
3703 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3704 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
3705 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x08]
3706 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3707 ; X86-NEXT: retl # encoding: [0xc3]
3708 ;
3709 ; X64-LABEL: test_mask_packs_epi16_rmk_256:
3710 ; X64: # %bb.0:
3711 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3712 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
3713 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3714 ; X64-NEXT: retq # encoding: [0xc3]
3715 %b = load <16 x i16>, <16 x i16>* %ptr_b
3716 %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)
3717 ret <32 x i8> %res
3718 }
3720 define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
3721 ; X86-LABEL: test_mask_packs_epi16_rmkz_256:
3722 ; X86: # %bb.0:
3723 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3724 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
3725 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x00]
3726 ; X86-NEXT: retl # encoding: [0xc3]
3727 ;
3728 ; X64-LABEL: test_mask_packs_epi16_rmkz_256:
3729 ; X64: # %bb.0:
3730 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3731 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
3732 ; X64-NEXT: retq # encoding: [0xc3]
3733 %b = load <16 x i16>, <16 x i16>* %ptr_b
3734 %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)
3735 ret <32 x i8> %res
3736 }
3738 declare <32 x i8> @llvm.x86.avx512.mask.packsswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
3741 define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
3742 ; CHECK-LABEL: test_mask_packus_epi32_rr_128:
3743 ; CHECK: # %bb.0:
3744 ; CHECK-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
3745 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3746 %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)
3747 ret <8 x i16> %res
3748 }
3750 define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
3751 ; X86-LABEL: test_mask_packus_epi32_rrk_128:
3752 ; X86: # %bb.0:
3753 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3754 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3755 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
3756 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3757 ; X86-NEXT: retl # encoding: [0xc3]
3758 ;
3759 ; X64-LABEL: test_mask_packus_epi32_rrk_128:
3760 ; X64: # %bb.0:
3761 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3762 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
3763 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
3764 ; X64-NEXT: retq # encoding: [0xc3]
3765 %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)
3766 ret <8 x i16> %res
3767 }
3769 define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
3770 ; X86-LABEL: test_mask_packus_epi32_rrkz_128:
3771 ; X86: # %bb.0:
3772 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
3773 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
3774 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
3775 ; X86-NEXT: retl # encoding: [0xc3]
3776 ;
3777 ; X64-LABEL: test_mask_packus_epi32_rrkz_128:
3778 ; X64: # %bb.0:
3779 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3780 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
3781 ; X64-NEXT: retq # encoding: [0xc3]
3782 %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)
3783 ret <8 x i16> %res
3784 }
3786 define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, <4 x i32>* %ptr_b) {
3787 ; X86-LABEL: test_mask_packus_epi32_rm_128:
3788 ; X86: # %bb.0:
3789 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3790 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x00]
3791 ; X86-NEXT: retl # encoding: [0xc3]
3792 ;
3793 ; X64-LABEL: test_mask_packus_epi32_rm_128:
3794 ; X64: # %bb.0:
3795 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
3796 ; X64-NEXT: retq # encoding: [0xc3]
3797 %b = load <4 x i32>, <4 x i32>* %ptr_b
3798 %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)
3799 ret <8 x i16> %res
3800 }
3802 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) {
3803 ; X86-LABEL: test_mask_packus_epi32_rmk_128:
3804 ; X86: # %bb.0:
3805 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3806 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3807 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3808 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x08]
3809 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3810 ; X86-NEXT: retl # encoding: [0xc3]
3811 ;
3812 ; X64-LABEL: test_mask_packus_epi32_rmk_128:
3813 ; X64: # %bb.0:
3814 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3815 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
3816 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3817 ; X64-NEXT: retq # encoding: [0xc3]
3818 %b = load <4 x i32>, <4 x i32>* %ptr_b
3819 %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)
3820 ret <8 x i16> %res
3821 }
3823 define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, <4 x i32>* %ptr_b, i8 %mask) {
3824 ; X86-LABEL: test_mask_packus_epi32_rmkz_128:
3825 ; X86: # %bb.0:
3826 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3827 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3828 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3829 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x00]
3830 ; X86-NEXT: retl # encoding: [0xc3]
3831 ;
3832 ; X64-LABEL: test_mask_packus_epi32_rmkz_128:
3833 ; X64: # %bb.0:
3834 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3835 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
3836 ; X64-NEXT: retq # encoding: [0xc3]
3837 %b = load <4 x i32>, <4 x i32>* %ptr_b
3838 %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)
3839 ret <8 x i16> %res
3840 }
3842 define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, i32* %ptr_b) {
3843 ; X86-LABEL: test_mask_packus_epi32_rmb_128:
3844 ; X86: # %bb.0:
3845 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3846 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x00]
3847 ; X86-NEXT: retl # encoding: [0xc3]
3848 ;
3849 ; X64-LABEL: test_mask_packus_epi32_rmb_128:
3850 ; X64: # %bb.0:
3851 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
3852 ; X64-NEXT: retq # encoding: [0xc3]
3853 %q = load i32, i32* %ptr_b
3854 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3855 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3856 %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)
3857 ret <8 x i16> %res
3858 }
3860 define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, i32* %ptr_b, <8 x i16> %passThru, i8 %mask) {
3861 ; X86-LABEL: test_mask_packus_epi32_rmbk_128:
3862 ; X86: # %bb.0:
3863 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3864 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3865 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3866 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x08]
3867 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3868 ; X86-NEXT: retl # encoding: [0xc3]
3869 ;
3870 ; X64-LABEL: test_mask_packus_epi32_rmbk_128:
3871 ; X64: # %bb.0:
3872 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3873 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
3874 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
3875 ; X64-NEXT: retq # encoding: [0xc3]
3876 %q = load i32, i32* %ptr_b
3877 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3878 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3879 %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)
3880 ret <8 x i16> %res
3881 }
3883 define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, i32* %ptr_b, i8 %mask) {
3884 ; X86-LABEL: test_mask_packus_epi32_rmbkz_128:
3885 ; X86: # %bb.0:
3886 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3887 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
3888 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
3889 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x00]
3890 ; X86-NEXT: retl # encoding: [0xc3]
3891 ;
3892 ; X64-LABEL: test_mask_packus_epi32_rmbkz_128:
3893 ; X64: # %bb.0:
3894 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3895 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
3896 ; X64-NEXT: retq # encoding: [0xc3]
3897 %q = load i32, i32* %ptr_b
3898 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
3899 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
3900 %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)
3901 ret <8 x i16> %res
3902 }
3904 declare <8 x i16> @llvm.x86.avx512.mask.packusdw.128(<4 x i32>, <4 x i32>, <8 x i16>, i8)
3906 define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
3907 ; CHECK-LABEL: test_mask_packus_epi32_rr_256:
3908 ; CHECK: # %bb.0:
3909 ; CHECK-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
3910 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
3911 %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)
3912 ret <16 x i16> %res
3913 }
3915 define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
3916 ; X86-LABEL: test_mask_packus_epi32_rrk_256:
3917 ; X86: # %bb.0:
3918 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3919 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
3920 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3921 ; X86-NEXT: retl # encoding: [0xc3]
3922 ;
3923 ; X64-LABEL: test_mask_packus_epi32_rrk_256:
3924 ; X64: # %bb.0:
3925 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3926 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
3927 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
3928 ; X64-NEXT: retq # encoding: [0xc3]
3929 %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)
3930 ret <16 x i16> %res
3931 }
3933 define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
3934 ; X86-LABEL: test_mask_packus_epi32_rrkz_256:
3935 ; X86: # %bb.0:
3936 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
3937 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
3938 ; X86-NEXT: retl # encoding: [0xc3]
3939 ;
3940 ; X64-LABEL: test_mask_packus_epi32_rrkz_256:
3941 ; X64: # %bb.0:
3942 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
3943 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
3944 ; X64-NEXT: retq # encoding: [0xc3]
3945 %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)
3946 ret <16 x i16> %res
3947 }
3949 define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, <8 x i32>* %ptr_b) {
3950 ; X86-LABEL: test_mask_packus_epi32_rm_256:
3951 ; X86: # %bb.0:
3952 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3953 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x00]
3954 ; X86-NEXT: retl # encoding: [0xc3]
3955 ;
3956 ; X64-LABEL: test_mask_packus_epi32_rm_256:
3957 ; X64: # %bb.0:
3958 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
3959 ; X64-NEXT: retq # encoding: [0xc3]
3960 %b = load <8 x i32>, <8 x i32>* %ptr_b
3961 %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)
3962 ret <16 x i16> %res
3963 }
3965 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) {
3966 ; X86-LABEL: test_mask_packus_epi32_rmk_256:
3967 ; X86: # %bb.0:
3968 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3969 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3970 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x08]
3971 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3972 ; X86-NEXT: retl # encoding: [0xc3]
3973 ;
3974 ; X64-LABEL: test_mask_packus_epi32_rmk_256:
3975 ; X64: # %bb.0:
3976 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3977 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
3978 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
3979 ; X64-NEXT: retq # encoding: [0xc3]
3980 %b = load <8 x i32>, <8 x i32>* %ptr_b
3981 %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)
3982 ret <16 x i16> %res
3983 }
3985 define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, <8 x i32>* %ptr_b, i16 %mask) {
3986 ; X86-LABEL: test_mask_packus_epi32_rmkz_256:
3987 ; X86: # %bb.0:
3988 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
3989 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
3990 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x00]
3991 ; X86-NEXT: retl # encoding: [0xc3]
3992 ;
3993 ; X64-LABEL: test_mask_packus_epi32_rmkz_256:
3994 ; X64: # %bb.0:
3995 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
3996 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
3997 ; X64-NEXT: retq # encoding: [0xc3]
3998 %b = load <8 x i32>, <8 x i32>* %ptr_b
3999 %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)
4000 ret <16 x i16> %res
4001 }
4003 define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, i32* %ptr_b) {
4004 ; X86-LABEL: test_mask_packus_epi32_rmb_256:
4005 ; X86: # %bb.0:
4006 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4007 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x00]
4008 ; X86-NEXT: retl # encoding: [0xc3]
4009 ;
4010 ; X64-LABEL: test_mask_packus_epi32_rmb_256:
4011 ; X64: # %bb.0:
4012 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
4013 ; X64-NEXT: retq # encoding: [0xc3]
4014 %q = load i32, i32* %ptr_b
4015 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4016 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4017 %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)
4018 ret <16 x i16> %res
4019 }
4021 define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, i32* %ptr_b, <16 x i16> %passThru, i16 %mask) {
4022 ; X86-LABEL: test_mask_packus_epi32_rmbk_256:
4023 ; X86: # %bb.0:
4024 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4025 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4026 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x08]
4027 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4028 ; X86-NEXT: retl # encoding: [0xc3]
4029 ;
4030 ; X64-LABEL: test_mask_packus_epi32_rmbk_256:
4031 ; X64: # %bb.0:
4032 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4033 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
4034 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4035 ; X64-NEXT: retq # encoding: [0xc3]
4036 %q = load i32, i32* %ptr_b
4037 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4038 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4039 %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)
4040 ret <16 x i16> %res
4041 }
4043 define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, i32* %ptr_b, i16 %mask) {
4044 ; X86-LABEL: test_mask_packus_epi32_rmbkz_256:
4045 ; X86: # %bb.0:
4046 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4047 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4048 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x00]
4049 ; X86-NEXT: retl # encoding: [0xc3]
4050 ;
4051 ; X64-LABEL: test_mask_packus_epi32_rmbkz_256:
4052 ; X64: # %bb.0:
4053 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4054 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
4055 ; X64-NEXT: retq # encoding: [0xc3]
4056 %q = load i32, i32* %ptr_b
4057 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
4058 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
4059 %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)
4060 ret <16 x i16> %res
4061 }
4063 declare <16 x i16> @llvm.x86.avx512.mask.packusdw.256(<8 x i32>, <8 x i32>, <16 x i16>, i16)
4065 define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
4066 ; CHECK-LABEL: test_mask_packus_epi16_rr_128:
4067 ; CHECK: # %bb.0:
4068 ; CHECK-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
4069 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4070 %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)
4071 ret <16 x i8> %res
4072 }
4074 define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
4075 ; X86-LABEL: test_mask_packus_epi16_rrk_128:
4076 ; X86: # %bb.0:
4077 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
4078 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
4079 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
4080 ; X86-NEXT: retl # encoding: [0xc3]
4081 ;
4082 ; X64-LABEL: test_mask_packus_epi16_rrk_128:
4083 ; X64: # %bb.0:
4084 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4085 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
4086 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
4087 ; X64-NEXT: retq # encoding: [0xc3]
4088 %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)
4089 ret <16 x i8> %res
4090 }
4092 define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
4093 ; X86-LABEL: test_mask_packus_epi16_rrkz_128:
4094 ; X86: # %bb.0:
4095 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
4096 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
4097 ; X86-NEXT: retl # encoding: [0xc3]
4098 ;
4099 ; X64-LABEL: test_mask_packus_epi16_rrkz_128:
4100 ; X64: # %bb.0:
4101 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4102 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
4103 ; X64-NEXT: retq # encoding: [0xc3]
4104 %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)
4105 ret <16 x i8> %res
4106 }
4108 define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
4109 ; X86-LABEL: test_mask_packus_epi16_rm_128:
4110 ; X86: # %bb.0:
4111 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4112 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x00]
4113 ; X86-NEXT: retl # encoding: [0xc3]
4114 ;
4115 ; X64-LABEL: test_mask_packus_epi16_rm_128:
4116 ; X64: # %bb.0:
4117 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
4118 ; X64-NEXT: retq # encoding: [0xc3]
4119 %b = load <8 x i16>, <8 x i16>* %ptr_b
4120 %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)
4121 ret <16 x i8> %res
4122 }
4124 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) {
4125 ; X86-LABEL: test_mask_packus_epi16_rmk_128:
4126 ; X86: # %bb.0:
4127 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4128 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4129 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x08]
4130 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
4131 ; X86-NEXT: retl # encoding: [0xc3]
4132 ;
4133 ; X64-LABEL: test_mask_packus_epi16_rmk_128:
4134 ; X64: # %bb.0:
4135 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4136 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
4137 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
4138 ; X64-NEXT: retq # encoding: [0xc3]
4139 %b = load <8 x i16>, <8 x i16>* %ptr_b
4140 %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)
4141 ret <16 x i8> %res
4142 }
4144 define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i16 %mask) {
4145 ; X86-LABEL: test_mask_packus_epi16_rmkz_128:
4146 ; X86: # %bb.0:
4147 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4148 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
4149 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x00]
4150 ; X86-NEXT: retl # encoding: [0xc3]
4151 ;
4152 ; X64-LABEL: test_mask_packus_epi16_rmkz_128:
4153 ; X64: # %bb.0:
4154 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4155 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
4156 ; X64-NEXT: retq # encoding: [0xc3]
4157 %b = load <8 x i16>, <8 x i16>* %ptr_b
4158 %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)
4159 ret <16 x i8> %res
4160 }
4162 declare <16 x i8> @llvm.x86.avx512.mask.packuswb.128(<8 x i16>, <8 x i16>, <16 x i8>, i16)
4164 define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
4165 ; CHECK-LABEL: test_mask_packus_epi16_rr_256:
4166 ; CHECK: # %bb.0:
4167 ; CHECK-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
4168 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4169 %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)
4170 ret <32 x i8> %res
4171 }
4173 define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
4174 ; X86-LABEL: test_mask_packus_epi16_rrk_256:
4175 ; X86: # %bb.0:
4176 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
4177 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
4178 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
4179 ; X86-NEXT: retl # encoding: [0xc3]
4180 ;
4181 ; X64-LABEL: test_mask_packus_epi16_rrk_256:
4182 ; X64: # %bb.0:
4183 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4184 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
4185 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
4186 ; X64-NEXT: retq # encoding: [0xc3]
4187 %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)
4188 ret <32 x i8> %res
4189 }
4191 define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
4192 ; X86-LABEL: test_mask_packus_epi16_rrkz_256:
4193 ; X86: # %bb.0:
4194 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
4195 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
4196 ; X86-NEXT: retl # encoding: [0xc3]
4197 ;
4198 ; X64-LABEL: test_mask_packus_epi16_rrkz_256:
4199 ; X64: # %bb.0:
4200 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4201 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
4202 ; X64-NEXT: retq # encoding: [0xc3]
4203 %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)
4204 ret <32 x i8> %res
4205 }
4207 define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
4208 ; X86-LABEL: test_mask_packus_epi16_rm_256:
4209 ; X86: # %bb.0:
4210 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4211 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x00]
4212 ; X86-NEXT: retl # encoding: [0xc3]
4213 ;
4214 ; X64-LABEL: test_mask_packus_epi16_rm_256:
4215 ; X64: # %bb.0:
4216 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
4217 ; X64-NEXT: retq # encoding: [0xc3]
4218 %b = load <16 x i16>, <16 x i16>* %ptr_b
4219 %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)
4220 ret <32 x i8> %res
4221 }
4223 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) {
4224 ; X86-LABEL: test_mask_packus_epi16_rmk_256:
4225 ; X86: # %bb.0:
4226 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4227 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
4228 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x08]
4229 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4230 ; X86-NEXT: retl # encoding: [0xc3]
4231 ;
4232 ; X64-LABEL: test_mask_packus_epi16_rmk_256:
4233 ; X64: # %bb.0:
4234 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4235 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
4236 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
4237 ; X64-NEXT: retq # encoding: [0xc3]
4238 %b = load <16 x i16>, <16 x i16>* %ptr_b
4239 %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)
4240 ret <32 x i8> %res
4241 }
4243 define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i32 %mask) {
4244 ; X86-LABEL: test_mask_packus_epi16_rmkz_256:
4245 ; X86: # %bb.0:
4246 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4247 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
4248 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x00]
4249 ; X86-NEXT: retl # encoding: [0xc3]
4250 ;
4251 ; X64-LABEL: test_mask_packus_epi16_rmkz_256:
4252 ; X64: # %bb.0:
4253 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
4254 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
4255 ; X64-NEXT: retq # encoding: [0xc3]
4256 %b = load <16 x i16>, <16 x i16>* %ptr_b
4257 %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)
4258 ret <32 x i8> %res
4259 }
4261 declare <32 x i8> @llvm.x86.avx512.mask.packuswb.256(<16 x i16>, <16 x i16>, <32 x i8>, i32)
4263 define <8 x i32> @test_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
4264 ; X86-LABEL: test_cmp_b_256:
4265 ; X86: # %bb.0:
4266 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4267 ; X86-NEXT: .cfi_def_cfa_offset 8
4268 ; X86-NEXT: pushl %edi # encoding: [0x57]
4269 ; X86-NEXT: .cfi_def_cfa_offset 12
4270 ; X86-NEXT: pushl %esi # encoding: [0x56]
4271 ; X86-NEXT: .cfi_def_cfa_offset 16
4272 ; X86-NEXT: .cfi_offset %esi, -16
4273 ; X86-NEXT: .cfi_offset %edi, -12
4274 ; X86-NEXT: .cfi_offset %ebx, -8
4275 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4276 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4277 ; X86-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
4278 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4279 ; X86-NEXT: vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
4280 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4281 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4282 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4283 ; X86-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
4284 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4285 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
4286 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4287 ; X86-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4288 ; X86-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4289 ; X86-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
4290 ; X86-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4291 ; X86-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4292 ; X86-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4293 ; X86-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4294 ; X86-NEXT: vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
4295 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4296 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4297 ; X86-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4298 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4299 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4300 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4301 ; X86-NEXT: popl %esi # encoding: [0x5e]
4302 ; X86-NEXT: .cfi_def_cfa_offset 12
4303 ; X86-NEXT: popl %edi # encoding: [0x5f]
4304 ; X86-NEXT: .cfi_def_cfa_offset 8
4305 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4306 ; X86-NEXT: .cfi_def_cfa_offset 4
4307 ; X86-NEXT: retl # encoding: [0xc3]
4308 ;
4309 ; X64-LABEL: test_cmp_b_256:
4310 ; X64: # %bb.0:
4311 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4312 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4313 ; X64-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 # encoding: [0x62,0xf1,0x75,0x28,0x64,0xc0]
4314 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4315 ; X64-NEXT: vpcmpleb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x02]
4316 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4317 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4318 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4319 ; X64-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x05]
4320 ; X64-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4321 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x64,0xc1]
4322 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4323 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4324 ; X64-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4325 ; X64-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
4326 ; X64-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4327 ; X64-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4328 ; X64-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4329 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4330 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4331 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4332 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4333 ; X64-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4334 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4335 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4336 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4337 ; X64-NEXT: retq # encoding: [0xc3]
4338 %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
4339 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4340 %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
4341 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4342 %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
4343 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4344 %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
4345 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4346 %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
4347 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4348 %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
4349 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4350 %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
4351 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4352 %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
4353 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4354 ret <8 x i32> %vec7
4355 }
4357 define <8 x i32> @test_mask_cmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
4358 ; X86-LABEL: test_mask_cmp_b_256:
4359 ; X86: # %bb.0:
4360 ; X86-NEXT: pushl %ebp # encoding: [0x55]
4361 ; X86-NEXT: .cfi_def_cfa_offset 8
4362 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4363 ; X86-NEXT: .cfi_def_cfa_offset 12
4364 ; X86-NEXT: pushl %edi # encoding: [0x57]
4365 ; X86-NEXT: .cfi_def_cfa_offset 16
4366 ; X86-NEXT: pushl %esi # encoding: [0x56]
4367 ; X86-NEXT: .cfi_def_cfa_offset 20
4368 ; X86-NEXT: .cfi_offset %esi, -20
4369 ; X86-NEXT: .cfi_offset %edi, -16
4370 ; X86-NEXT: .cfi_offset %ebx, -12
4371 ; X86-NEXT: .cfi_offset %ebp, -8
4372 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
4373 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4374 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4375 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4376 ; X86-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
4377 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4378 ; X86-NEXT: vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
4379 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4380 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4381 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4382 ; X86-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
4383 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4384 ; X86-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
4385 ; X86-NEXT: kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
4386 ; X86-NEXT: vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
4387 ; X86-NEXT: vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
4388 ; X86-NEXT: vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
4389 ; X86-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
4390 ; X86-NEXT: vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
4391 ; X86-NEXT: vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
4392 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4393 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4394 ; X86-NEXT: vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
4395 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4396 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4397 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4398 ; X86-NEXT: popl %esi # encoding: [0x5e]
4399 ; X86-NEXT: .cfi_def_cfa_offset 16
4400 ; X86-NEXT: popl %edi # encoding: [0x5f]
4401 ; X86-NEXT: .cfi_def_cfa_offset 12
4402 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4403 ; X86-NEXT: .cfi_def_cfa_offset 8
4404 ; X86-NEXT: popl %ebp # encoding: [0x5d]
4405 ; X86-NEXT: .cfi_def_cfa_offset 4
4406 ; X86-NEXT: retl # encoding: [0xc3]
4407 ;
4408 ; X64-LABEL: test_mask_cmp_b_256:
4409 ; X64: # %bb.0:
4410 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4411 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4412 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4413 ; X64-NEXT: vpcmpgtb %ymm0, %ymm1, %k0 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x64,0xc0]
4414 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4415 ; X64-NEXT: vpcmpleb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x02]
4416 ; X64-NEXT: kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
4417 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4418 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4419 ; X64-NEXT: vpcmpnltb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x05]
4420 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4421 ; X64-NEXT: vpcmpgtb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x64,0xc1]
4422 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4423 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4424 ; X64-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
4425 ; X64-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
4426 ; X64-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
4427 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4428 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4429 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4430 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4431 ; X64-NEXT: vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
4432 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4433 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4434 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4435 ; X64-NEXT: retq # encoding: [0xc3]
4436 %res0 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
4437 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4438 %res1 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
4439 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4440 %res2 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
4441 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4442 %res3 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
4443 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4444 %res4 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
4445 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4446 %res5 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
4447 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4448 %res6 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
4449 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4450 %res7 = call i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
4451 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4452 ret <8 x i32> %vec7
4453 }
4455 declare i32 @llvm.x86.avx512.mask.cmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
4457 define <8 x i32> @test_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1) {
4458 ; X86-LABEL: test_ucmp_b_256:
4459 ; X86: # %bb.0:
4460 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4461 ; X86-NEXT: .cfi_def_cfa_offset 8
4462 ; X86-NEXT: pushl %edi # encoding: [0x57]
4463 ; X86-NEXT: .cfi_def_cfa_offset 12
4464 ; X86-NEXT: pushl %esi # encoding: [0x56]
4465 ; X86-NEXT: .cfi_def_cfa_offset 16
4466 ; X86-NEXT: .cfi_offset %esi, -16
4467 ; X86-NEXT: .cfi_offset %edi, -12
4468 ; X86-NEXT: .cfi_offset %ebx, -8
4469 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4470 ; X86-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4471 ; X86-NEXT: vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
4472 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4473 ; X86-NEXT: vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
4474 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4475 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4476 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4477 ; X86-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
4478 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4479 ; X86-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
4480 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4481 ; X86-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4482 ; X86-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4483 ; X86-NEXT: vpinsrd $2, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x02]
4484 ; X86-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4485 ; X86-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4486 ; X86-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4487 ; X86-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4488 ; X86-NEXT: vmovd %eax, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd0]
4489 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4490 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4491 ; X86-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4492 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4493 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4494 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4495 ; X86-NEXT: popl %esi # encoding: [0x5e]
4496 ; X86-NEXT: .cfi_def_cfa_offset 12
4497 ; X86-NEXT: popl %edi # encoding: [0x5f]
4498 ; X86-NEXT: .cfi_def_cfa_offset 8
4499 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4500 ; X86-NEXT: .cfi_def_cfa_offset 4
4501 ; X86-NEXT: retl # encoding: [0xc3]
4502 ;
4503 ; X64-LABEL: test_ucmp_b_256:
4504 ; X64: # %bb.0:
4505 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x74,0xc1]
4506 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4507 ; X64-NEXT: vpcmpltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x01]
4508 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4509 ; X64-NEXT: vpcmpleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x02]
4510 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4511 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3f,0xc1,0x04]
4512 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4513 ; X64-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x05]
4514 ; X64-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4515 ; X64-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 # encoding: [0x62,0xf3,0x7d,0x28,0x3e,0xc1,0x06]
4516 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4517 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4518 ; X64-NEXT: vpinsrd $1, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x01]
4519 ; X64-NEXT: vpinsrd $2, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x02]
4520 ; X64-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4521 ; X64-NEXT: vpblendd $8, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x02,0xc1,0x08]
4522 ; X64-NEXT: # xmm0 = xmm0[0,1,2],xmm1[3]
4523 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4524 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4525 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4526 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4527 ; X64-NEXT: vmovd %edx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd2]
4528 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4529 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4530 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4531 ; X64-NEXT: retq # encoding: [0xc3]
4532 %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 -1)
4533 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4534 %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 -1)
4535 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4536 %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 -1)
4537 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4538 %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 -1)
4539 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4540 %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 -1)
4541 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4542 %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 -1)
4543 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4544 %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 -1)
4545 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4546 %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 -1)
4547 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4548 ret <8 x i32> %vec7
4549 }
4551 define <8 x i32> @test_mask_ucmp_b_256(<32 x i8> %a0, <32 x i8> %a1, i32 %mask) {
4552 ; X86-LABEL: test_mask_ucmp_b_256:
4553 ; X86: # %bb.0:
4554 ; X86-NEXT: pushl %ebp # encoding: [0x55]
4555 ; X86-NEXT: .cfi_def_cfa_offset 8
4556 ; X86-NEXT: pushl %ebx # encoding: [0x53]
4557 ; X86-NEXT: .cfi_def_cfa_offset 12
4558 ; X86-NEXT: pushl %edi # encoding: [0x57]
4559 ; X86-NEXT: .cfi_def_cfa_offset 16
4560 ; X86-NEXT: pushl %esi # encoding: [0x56]
4561 ; X86-NEXT: .cfi_def_cfa_offset 20
4562 ; X86-NEXT: .cfi_offset %esi, -20
4563 ; X86-NEXT: .cfi_offset %edi, -16
4564 ; X86-NEXT: .cfi_offset %ebx, -12
4565 ; X86-NEXT: .cfi_offset %ebp, -8
4566 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x14]
4567 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4568 ; X86-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4569 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4570 ; X86-NEXT: vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
4571 ; X86-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4572 ; X86-NEXT: vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
4573 ; X86-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4574 ; X86-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4575 ; X86-NEXT: kmovd %k0, %edi # encoding: [0xc5,0xfb,0x93,0xf8]
4576 ; X86-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
4577 ; X86-NEXT: kmovd %k0, %ebx # encoding: [0xc5,0xfb,0x93,0xd8]
4578 ; X86-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
4579 ; X86-NEXT: kmovd %k0, %ebp # encoding: [0xc5,0xfb,0x93,0xe8]
4580 ; X86-NEXT: vmovd %edi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc7]
4581 ; X86-NEXT: vpinsrd $1, %ebx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc3,0x01]
4582 ; X86-NEXT: vpinsrd $2, %ebp, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc5,0x02]
4583 ; X86-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x03]
4584 ; X86-NEXT: vmovd %edx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xca]
4585 ; X86-NEXT: vmovd %ecx, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd1]
4586 ; X86-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4587 ; X86-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4588 ; X86-NEXT: vmovd %esi, %xmm2 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xd6]
4589 ; X86-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4590 ; X86-NEXT: # xmm1 = xmm1[0],xmm2[0]
4591 ; X86-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4592 ; X86-NEXT: popl %esi # encoding: [0x5e]
4593 ; X86-NEXT: .cfi_def_cfa_offset 16
4594 ; X86-NEXT: popl %edi # encoding: [0x5f]
4595 ; X86-NEXT: .cfi_def_cfa_offset 12
4596 ; X86-NEXT: popl %ebx # encoding: [0x5b]
4597 ; X86-NEXT: .cfi_def_cfa_offset 8
4598 ; X86-NEXT: popl %ebp # encoding: [0x5d]
4599 ; X86-NEXT: .cfi_def_cfa_offset 4
4600 ; X86-NEXT: retl # encoding: [0xc3]
4601 ;
4602 ; X64-LABEL: test_mask_ucmp_b_256:
4603 ; X64: # %bb.0:
4604 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4605 ; X64-NEXT: vpcmpeqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x74,0xc1]
4606 ; X64-NEXT: kmovd %k0, %r8d # encoding: [0xc5,0x7b,0x93,0xc0]
4607 ; X64-NEXT: vpcmpltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x01]
4608 ; X64-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4609 ; X64-NEXT: vpcmpleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x02]
4610 ; X64-NEXT: kmovd %k0, %r9d # encoding: [0xc5,0x7b,0x93,0xc8]
4611 ; X64-NEXT: vpcmpneqb %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3f,0xc1,0x04]
4612 ; X64-NEXT: kmovd %k0, %esi # encoding: [0xc5,0xfb,0x93,0xf0]
4613 ; X64-NEXT: vpcmpnltub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x05]
4614 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4615 ; X64-NEXT: vpcmpnleub %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x3e,0xc1,0x06]
4616 ; X64-NEXT: kmovd %k0, %edx # encoding: [0xc5,0xfb,0x93,0xd0]
4617 ; X64-NEXT: vmovd %esi, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc6]
4618 ; X64-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc0,0x01]
4619 ; X64-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc2,0x02]
4620 ; X64-NEXT: vpinsrd $3, %edi, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x22,0xc7,0x03]
4621 ; X64-NEXT: vmovd %ecx, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6e,0xc9]
4622 ; X64-NEXT: vmovd %r8d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd0]
4623 ; X64-NEXT: vpunpckldq %xmm1, %xmm2, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0x62,0xc9]
4624 ; X64-NEXT: # xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
4625 ; X64-NEXT: vmovd %r9d, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xc1,0x79,0x6e,0xd1]
4626 ; X64-NEXT: vpunpcklqdq %xmm2, %xmm1, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0x6c,0xca]
4627 ; X64-NEXT: # xmm1 = xmm1[0],xmm2[0]
4628 ; X64-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe3,0x75,0x38,0xc0,0x01]
4629 ; X64-NEXT: retq # encoding: [0xc3]
4630 %res0 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 0, i32 %mask)
4631 %vec0 = insertelement <8 x i32> undef, i32 %res0, i32 0
4632 %res1 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 1, i32 %mask)
4633 %vec1 = insertelement <8 x i32> %vec0, i32 %res1, i32 1
4634 %res2 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 2, i32 %mask)
4635 %vec2 = insertelement <8 x i32> %vec1, i32 %res2, i32 2
4636 %res3 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 3, i32 %mask)
4637 %vec3 = insertelement <8 x i32> %vec2, i32 %res3, i32 3
4638 %res4 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 4, i32 %mask)
4639 %vec4 = insertelement <8 x i32> %vec3, i32 %res4, i32 4
4640 %res5 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 5, i32 %mask)
4641 %vec5 = insertelement <8 x i32> %vec4, i32 %res5, i32 5
4642 %res6 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 6, i32 %mask)
4643 %vec6 = insertelement <8 x i32> %vec5, i32 %res6, i32 6
4644 %res7 = call i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8> %a0, <32 x i8> %a1, i32 7, i32 %mask)
4645 %vec7 = insertelement <8 x i32> %vec6, i32 %res7, i32 7
4646 ret <8 x i32> %vec7
4647 }
4649 declare i32 @llvm.x86.avx512.mask.ucmp.b.256(<32 x i8>, <32 x i8>, i32, i32) nounwind readnone
4651 define <8 x i16> @test_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
4652 ; CHECK-LABEL: test_cmp_w_256:
4653 ; CHECK: # %bb.0:
4654 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
4655 ; CHECK-NEXT: vpcmpgtw %ymm0, %ymm1, %k1 # encoding: [0x62,0xf1,0x75,0x28,0x65,0xc8]
4656 ; CHECK-NEXT: vpcmplew %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd1,0x02]
4657 ; CHECK-NEXT: vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
4658 ; CHECK-NEXT: vpcmpnltw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xe1,0x05]
4659 ; CHECK-NEXT: vpcmpgtw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf1,0x7d,0x28,0x65,0xe9]
4660 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4661 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4662 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4663 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4664 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4665 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4666 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4667 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4668 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4669 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4670 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4671 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4672 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4673 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4674 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4675 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4676 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4677 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4678 %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
4679 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4680 %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
4681 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4682 %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
4683 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4684 %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
4685 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4686 %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
4687 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4688 %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
4689 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4690 %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
4691 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4692 %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
4693 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4694 ret <8 x i16> %vec7
4695 }
4697 define <8 x i16> @test_mask_cmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
4698 ; X86-LABEL: test_mask_cmp_w_256:
4699 ; X86: # %bb.0:
4700 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4701 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4702 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4703 ; X86-NEXT: vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
4704 ; X86-NEXT: vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
4705 ; X86-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4706 ; X86-NEXT: vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
4707 ; X86-NEXT: vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
4708 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4709 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4710 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4711 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4712 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4713 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4714 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4715 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4716 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4717 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4718 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4719 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4720 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4721 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4722 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4723 ; X86-NEXT: retl # encoding: [0xc3]
4724 ;
4725 ; X64-LABEL: test_mask_cmp_w_256:
4726 ; X64: # %bb.0:
4727 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4728 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4729 ; X64-NEXT: vpcmpgtw %ymm0, %ymm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x29,0x65,0xd0]
4730 ; X64-NEXT: vpcmplew %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xd9,0x02]
4731 ; X64-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4732 ; X64-NEXT: vpcmpnltw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe9,0x05]
4733 ; X64-NEXT: vpcmpgtw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x65,0xc9]
4734 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4735 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4736 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4737 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4738 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4739 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4740 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4741 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4742 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4743 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4744 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4745 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4746 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4747 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4748 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4749 ; X64-NEXT: retq # encoding: [0xc3]
4750 %res0 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
4751 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4752 %res1 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
4753 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4754 %res2 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
4755 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4756 %res3 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
4757 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4758 %res4 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
4759 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4760 %res5 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
4761 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4762 %res6 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
4763 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4764 %res7 = call i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
4765 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4766 ret <8 x i16> %vec7
4767 }
4769 declare i16 @llvm.x86.avx512.mask.cmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
4771 define <8 x i16> @test_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1) {
4772 ; CHECK-LABEL: test_ucmp_w_256:
4773 ; CHECK: # %bb.0:
4774 ; CHECK-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf1,0x7d,0x28,0x75,0xc1]
4775 ; CHECK-NEXT: vpcmpltuw %ymm1, %ymm0, %k1 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xc9,0x01]
4776 ; CHECK-NEXT: vpcmpleuw %ymm1, %ymm0, %k2 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xd1,0x02]
4777 ; CHECK-NEXT: vpcmpneqw %ymm1, %ymm0, %k3 # encoding: [0x62,0xf3,0xfd,0x28,0x3f,0xd9,0x04]
4778 ; CHECK-NEXT: vpcmpnltuw %ymm1, %ymm0, %k4 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe1,0x05]
4779 ; CHECK-NEXT: vpcmpnleuw %ymm1, %ymm0, %k5 # encoding: [0x62,0xf3,0xfd,0x28,0x3e,0xe9,0x06]
4780 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4781 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4782 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4783 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4784 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4785 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4786 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4787 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4788 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4789 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4790 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4791 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4792 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4793 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4794 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4795 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4796 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4797 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4798 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 -1)
4799 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4800 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 -1)
4801 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4802 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 -1)
4803 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4804 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 -1)
4805 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4806 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 -1)
4807 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4808 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 -1)
4809 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4810 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 -1)
4811 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4812 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 -1)
4813 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4814 ret <8 x i16> %vec7
4815 }
4817 define <8 x i16> @test_mask_ucmp_w_256(<16 x i16> %a0, <16 x i16> %a1, i16 %mask) {
4818 ; X86-LABEL: test_mask_ucmp_w_256:
4819 ; X86: # %bb.0:
4820 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4821 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4822 ; X86-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4823 ; X86-NEXT: vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
4824 ; X86-NEXT: vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
4825 ; X86-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4826 ; X86-NEXT: vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
4827 ; X86-NEXT: vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
4828 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4829 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4830 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4831 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4832 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4833 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4834 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4835 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4836 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4837 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4838 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4839 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4840 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4841 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4842 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4843 ; X86-NEXT: retl # encoding: [0xc3]
4844 ;
4845 ; X64-LABEL: test_mask_ucmp_w_256:
4846 ; X64: # %bb.0:
4847 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4848 ; X64-NEXT: vpcmpeqw %ymm1, %ymm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x75,0xc1]
4849 ; X64-NEXT: vpcmpltuw %ymm1, %ymm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd1,0x01]
4850 ; X64-NEXT: vpcmpleuw %ymm1, %ymm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xd9,0x02]
4851 ; X64-NEXT: vpcmpneqw %ymm1, %ymm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3f,0xe1,0x04]
4852 ; X64-NEXT: vpcmpnltuw %ymm1, %ymm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xe9,0x05]
4853 ; X64-NEXT: vpcmpnleuw %ymm1, %ymm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x29,0x3e,0xc9,0x06]
4854 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4855 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4856 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4857 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4858 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4859 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4860 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4861 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4862 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4863 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4864 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4865 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4866 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4867 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4868 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
4869 ; X64-NEXT: retq # encoding: [0xc3]
4870 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 0, i16 %mask)
4871 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4872 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 1, i16 %mask)
4873 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4874 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 2, i16 %mask)
4875 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4876 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 3, i16 %mask)
4877 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4878 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 4, i16 %mask)
4879 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4880 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 5, i16 %mask)
4881 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4882 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 6, i16 %mask)
4883 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4884 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16> %a0, <16 x i16> %a1, i32 7, i16 %mask)
4885 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4886 ret <8 x i16> %vec7
4887 }
4889 declare i16 @llvm.x86.avx512.mask.ucmp.w.256(<16 x i16>, <16 x i16>, i32, i16) nounwind readnone
4891 define <8 x i16> @test_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
4892 ; CHECK-LABEL: test_cmp_b_128:
4893 ; CHECK: # %bb.0:
4894 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
4895 ; CHECK-NEXT: vpcmpgtb %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x64,0xc8]
4896 ; CHECK-NEXT: vpcmpleb %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd1,0x02]
4897 ; CHECK-NEXT: vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
4898 ; CHECK-NEXT: vpcmpnltb %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xe1,0x05]
4899 ; CHECK-NEXT: vpcmpgtb %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x64,0xe9]
4900 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4901 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4902 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4903 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4904 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4905 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4906 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4907 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4908 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4909 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4910 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4911 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4912 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4913 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
4914 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
4915 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
4916 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
4917 %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
4918 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4919 %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
4920 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4921 %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
4922 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4923 %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
4924 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4925 %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
4926 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4927 %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
4928 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4929 %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
4930 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
4931 %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
4932 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
4933 ret <8 x i16> %vec7
4934 }
4936 define <8 x i16> @test_mask_cmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
4937 ; X86-LABEL: test_mask_cmp_b_128:
4938 ; X86: # %bb.0:
4939 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
4940 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
4941 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
4942 ; X86-NEXT: vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
4943 ; X86-NEXT: vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
4944 ; X86-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
4945 ; X86-NEXT: vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
4946 ; X86-NEXT: vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
4947 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
4948 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4949 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
4950 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
4951 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
4952 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
4953 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
4954 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
4955 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
4956 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
4957 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
4958 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
4959 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
4960 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
4961 ; X86-NEXT: retl # encoding: [0xc3]
4962 ;
4963 ; X64-LABEL: test_mask_cmp_b_128:
4964 ; X64: # %bb.0:
4965 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
4966 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
4967 ; X64-NEXT: vpcmpgtb %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x64,0xd0]
4968 ; X64-NEXT: vpcmpleb %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xd9,0x02]
4969 ; X64-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
4970 ; X64-NEXT: vpcmpnltb %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe9,0x05]
4971 ; X64-NEXT: vpcmpgtb %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x64,0xc9]
4972 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
4973 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
4974 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
4975 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
4976 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
4977 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
4978 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
4979 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
4980 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
4981 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
4982 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
4983 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
4984 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
4985 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
4986 ; X64-NEXT: retq # encoding: [0xc3]
4987 %res0 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
4988 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
4989 %res1 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
4990 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
4991 %res2 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
4992 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
4993 %res3 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
4994 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
4995 %res4 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
4996 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
4997 %res5 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
4998 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
4999 %res6 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
5000 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
5001 %res7 = call i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
5002 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
5003 ret <8 x i16> %vec7
5004 }
5006 declare i16 @llvm.x86.avx512.mask.cmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
5008 define <8 x i16> @test_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1) {
5009 ; CHECK-LABEL: test_ucmp_b_128:
5010 ; CHECK: # %bb.0:
5011 ; CHECK-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x74,0xc1]
5012 ; CHECK-NEXT: vpcmpltub %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xc9,0x01]
5013 ; CHECK-NEXT: vpcmpleub %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xd1,0x02]
5014 ; CHECK-NEXT: vpcmpneqb %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0x7d,0x08,0x3f,0xd9,0x04]
5015 ; CHECK-NEXT: vpcmpnltub %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe1,0x05]
5016 ; CHECK-NEXT: vpcmpnleub %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0x7d,0x08,0x3e,0xe9,0x06]
5017 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5018 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5019 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5020 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5021 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5022 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5023 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5024 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5025 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5026 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5027 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5028 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5029 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5030 ; CHECK-NEXT: vpcmpeqd %xmm1, %xmm1, %xmm1 # encoding: [0xc5,0xf1,0x76,0xc9]
5031 ; CHECK-NEXT: vpblendw $128, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0x79,0x0e,0xc1,0x80]
5032 ; CHECK-NEXT: # xmm0 = xmm0[0,1,2,3,4,5,6],xmm1[7]
5033 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5034 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 -1)
5035 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
5036 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 -1)
5037 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
5038 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 -1)
5039 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
5040 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 -1)
5041 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
5042 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 -1)
5043 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
5044 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 -1)
5045 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
5046 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 -1)
5047 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
5048 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 -1)
5049 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
5050 ret <8 x i16> %vec7
5051 }
5053 define <8 x i16> @test_mask_ucmp_b_128(<16 x i8> %a0, <16 x i8> %a1, i16 %mask) {
5054 ; X86-LABEL: test_mask_ucmp_b_128:
5055 ; X86: # %bb.0:
5056 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5057 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5058 ; X86-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
5059 ; X86-NEXT: vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
5060 ; X86-NEXT: vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
5061 ; X86-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
5062 ; X86-NEXT: vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
5063 ; X86-NEXT: vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
5064 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5065 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5066 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
5067 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5068 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
5069 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5070 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
5071 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5072 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
5073 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5074 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
5075 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5076 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
5077 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5078 ; X86-NEXT: retl # encoding: [0xc3]
5079 ;
5080 ; X64-LABEL: test_mask_ucmp_b_128:
5081 ; X64: # %bb.0:
5082 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5083 ; X64-NEXT: vpcmpeqb %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x74,0xc1]
5084 ; X64-NEXT: vpcmpltub %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd1,0x01]
5085 ; X64-NEXT: vpcmpleub %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xd9,0x02]
5086 ; X64-NEXT: vpcmpneqb %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3f,0xe1,0x04]
5087 ; X64-NEXT: vpcmpnltub %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xe9,0x05]
5088 ; X64-NEXT: vpcmpnleub %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x3e,0xc9,0x06]
5089 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5090 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5091 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5092 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5093 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5094 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5095 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5096 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5097 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5098 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5099 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5100 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5101 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5102 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
5103 ; X64-NEXT: retq # encoding: [0xc3]
5104 %res0 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 0, i16 %mask)
5105 %vec0 = insertelement <8 x i16> undef, i16 %res0, i32 0
5106 %res1 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 1, i16 %mask)
5107 %vec1 = insertelement <8 x i16> %vec0, i16 %res1, i32 1
5108 %res2 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 2, i16 %mask)
5109 %vec2 = insertelement <8 x i16> %vec1, i16 %res2, i32 2
5110 %res3 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 3, i16 %mask)
5111 %vec3 = insertelement <8 x i16> %vec2, i16 %res3, i32 3
5112 %res4 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 4, i16 %mask)
5113 %vec4 = insertelement <8 x i16> %vec3, i16 %res4, i32 4
5114 %res5 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 5, i16 %mask)
5115 %vec5 = insertelement <8 x i16> %vec4, i16 %res5, i32 5
5116 %res6 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 6, i16 %mask)
5117 %vec6 = insertelement <8 x i16> %vec5, i16 %res6, i32 6
5118 %res7 = call i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8> %a0, <16 x i8> %a1, i32 7, i16 %mask)
5119 %vec7 = insertelement <8 x i16> %vec6, i16 %res7, i32 7
5120 ret <8 x i16> %vec7
5121 }
5123 declare i16 @llvm.x86.avx512.mask.ucmp.b.128(<16 x i8>, <16 x i8>, i32, i16) nounwind readnone
5125 define <8 x i8> @test_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
5126 ; CHECK-LABEL: test_cmp_w_128:
5127 ; CHECK: # %bb.0:
5128 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
5129 ; CHECK-NEXT: vpcmpgtw %xmm0, %xmm1, %k1 # encoding: [0x62,0xf1,0x75,0x08,0x65,0xc8]
5130 ; CHECK-NEXT: vpcmplew %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd1,0x02]
5131 ; CHECK-NEXT: vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
5132 ; CHECK-NEXT: vpcmpnltw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xe1,0x05]
5133 ; CHECK-NEXT: vpcmpgtw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf1,0x7d,0x08,0x65,0xe9]
5134 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5135 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5136 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5137 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5138 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5139 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5140 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5141 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5142 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5143 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5144 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5145 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5146 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5147 ; CHECK-NEXT: movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
5148 ; CHECK-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5149 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5150 %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
5151 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5152 %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
5153 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5154 %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
5155 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5156 %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
5157 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5158 %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
5159 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5160 %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
5161 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5162 %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
5163 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5164 %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
5165 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5166 ret <8 x i8> %vec7
5167 }
5169 define <8 x i8> @test_mask_cmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
5170 ; X86-LABEL: test_mask_cmp_w_128:
5171 ; X86: # %bb.0:
5172 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5173 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5174 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5175 ; X86-NEXT: vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
5176 ; X86-NEXT: vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
5177 ; X86-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5178 ; X86-NEXT: vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
5179 ; X86-NEXT: vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
5180 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5181 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5182 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
5183 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5184 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
5185 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5186 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
5187 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5188 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
5189 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5190 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
5191 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5192 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
5193 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5194 ; X86-NEXT: retl # encoding: [0xc3]
5195 ;
5196 ; X64-LABEL: test_mask_cmp_w_128:
5197 ; X64: # %bb.0:
5198 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5199 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5200 ; X64-NEXT: vpcmpgtw %xmm0, %xmm1, %k2 {%k1} # encoding: [0x62,0xf1,0x75,0x09,0x65,0xd0]
5201 ; X64-NEXT: vpcmplew %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xd9,0x02]
5202 ; X64-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5203 ; X64-NEXT: vpcmpnltw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe9,0x05]
5204 ; X64-NEXT: vpcmpgtw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x65,0xc9]
5205 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5206 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5207 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5208 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5209 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5210 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5211 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5212 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5213 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5214 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5215 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5216 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5217 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5218 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
5219 ; X64-NEXT: retq # encoding: [0xc3]
5220 %res0 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
5221 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5222 %res1 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
5223 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5224 %res2 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
5225 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5226 %res3 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
5227 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5228 %res4 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
5229 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5230 %res5 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
5231 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5232 %res6 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
5233 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5234 %res7 = call i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
5235 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5236 ret <8 x i8> %vec7
5237 }
5239 declare i8 @llvm.x86.avx512.mask.cmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
5241 define <8 x i8> @test_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1) {
5242 ; CHECK-LABEL: test_ucmp_w_128:
5243 ; CHECK: # %bb.0:
5244 ; CHECK-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf1,0x7d,0x08,0x75,0xc1]
5245 ; CHECK-NEXT: vpcmpltuw %xmm1, %xmm0, %k1 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xc9,0x01]
5246 ; CHECK-NEXT: vpcmpleuw %xmm1, %xmm0, %k2 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xd1,0x02]
5247 ; CHECK-NEXT: vpcmpneqw %xmm1, %xmm0, %k3 # encoding: [0x62,0xf3,0xfd,0x08,0x3f,0xd9,0x04]
5248 ; CHECK-NEXT: vpcmpnltuw %xmm1, %xmm0, %k4 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe1,0x05]
5249 ; CHECK-NEXT: vpcmpnleuw %xmm1, %xmm0, %k5 # encoding: [0x62,0xf3,0xfd,0x08,0x3e,0xe9,0x06]
5250 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5251 ; CHECK-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5252 ; CHECK-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5253 ; CHECK-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5254 ; CHECK-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5255 ; CHECK-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5256 ; CHECK-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5257 ; CHECK-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5258 ; CHECK-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5259 ; CHECK-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5260 ; CHECK-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5261 ; CHECK-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5262 ; CHECK-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5263 ; CHECK-NEXT: movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
5264 ; CHECK-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5265 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5266 %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 -1)
5267 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5268 %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 -1)
5269 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5270 %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 -1)
5271 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5272 %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 -1)
5273 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5274 %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 -1)
5275 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5276 %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 -1)
5277 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5278 %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 -1)
5279 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5280 %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 -1)
5281 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5282 ret <8 x i8> %vec7
5283 }
5285 define <8 x i8> @test_mask_ucmp_w_128(<8 x i16> %a0, <8 x i16> %a1, i8 %mask) {
5286 ; X86-LABEL: test_mask_ucmp_w_128:
5287 ; X86: # %bb.0:
5288 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5289 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5290 ; X86-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5291 ; X86-NEXT: vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
5292 ; X86-NEXT: vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
5293 ; X86-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5294 ; X86-NEXT: vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
5295 ; X86-NEXT: vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
5296 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5297 ; X86-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5298 ; X86-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x00]
5299 ; X86-NEXT: kmovd %k2, %ecx # encoding: [0xc5,0xfb,0x93,0xca]
5300 ; X86-NEXT: vpinsrw $1, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x01]
5301 ; X86-NEXT: kmovd %k3, %ecx # encoding: [0xc5,0xfb,0x93,0xcb]
5302 ; X86-NEXT: vpinsrw $2, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x02]
5303 ; X86-NEXT: kmovd %k4, %ecx # encoding: [0xc5,0xfb,0x93,0xcc]
5304 ; X86-NEXT: vpinsrw $4, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x04]
5305 ; X86-NEXT: kmovd %k5, %ecx # encoding: [0xc5,0xfb,0x93,0xcd]
5306 ; X86-NEXT: vpinsrw $5, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x05]
5307 ; X86-NEXT: kmovd %k1, %ecx # encoding: [0xc5,0xfb,0x93,0xc9]
5308 ; X86-NEXT: vpinsrw $6, %ecx, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc1,0x06]
5309 ; X86-NEXT: vpinsrw $7, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x07]
5310 ; X86-NEXT: retl # encoding: [0xc3]
5311 ;
5312 ; X64-LABEL: test_mask_ucmp_w_128:
5313 ; X64: # %bb.0:
5314 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5315 ; X64-NEXT: vpcmpeqw %xmm1, %xmm0, %k0 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x75,0xc1]
5316 ; X64-NEXT: vpcmpltuw %xmm1, %xmm0, %k2 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd1,0x01]
5317 ; X64-NEXT: vpcmpleuw %xmm1, %xmm0, %k3 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xd9,0x02]
5318 ; X64-NEXT: vpcmpneqw %xmm1, %xmm0, %k4 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3f,0xe1,0x04]
5319 ; X64-NEXT: vpcmpnltuw %xmm1, %xmm0, %k5 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xe9,0x05]
5320 ; X64-NEXT: vpcmpnleuw %xmm1, %xmm0, %k1 {%k1} # encoding: [0x62,0xf3,0xfd,0x09,0x3e,0xc9,0x06]
5321 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5322 ; X64-NEXT: vpxor %xmm0, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xef,0xc0]
5323 ; X64-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x00]
5324 ; X64-NEXT: kmovd %k2, %eax # encoding: [0xc5,0xfb,0x93,0xc2]
5325 ; X64-NEXT: vpinsrw $1, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x01]
5326 ; X64-NEXT: kmovd %k3, %eax # encoding: [0xc5,0xfb,0x93,0xc3]
5327 ; X64-NEXT: vpinsrw $2, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x02]
5328 ; X64-NEXT: kmovd %k4, %eax # encoding: [0xc5,0xfb,0x93,0xc4]
5329 ; X64-NEXT: vpinsrw $4, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x04]
5330 ; X64-NEXT: kmovd %k5, %eax # encoding: [0xc5,0xfb,0x93,0xc5]
5331 ; X64-NEXT: vpinsrw $5, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x05]
5332 ; X64-NEXT: kmovd %k1, %eax # encoding: [0xc5,0xfb,0x93,0xc1]
5333 ; X64-NEXT: vpinsrw $6, %eax, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc0,0x06]
5334 ; X64-NEXT: vpinsrw $7, %edi, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xc4,0xc7,0x07]
5335 ; X64-NEXT: retq # encoding: [0xc3]
5336 %res0 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 0, i8 %mask)
5337 %vec0 = insertelement <8 x i8> undef, i8 %res0, i32 0
5338 %res1 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 1, i8 %mask)
5339 %vec1 = insertelement <8 x i8> %vec0, i8 %res1, i32 1
5340 %res2 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 2, i8 %mask)
5341 %vec2 = insertelement <8 x i8> %vec1, i8 %res2, i32 2
5342 %res3 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 3, i8 %mask)
5343 %vec3 = insertelement <8 x i8> %vec2, i8 %res3, i32 3
5344 %res4 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 4, i8 %mask)
5345 %vec4 = insertelement <8 x i8> %vec3, i8 %res4, i32 4
5346 %res5 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 5, i8 %mask)
5347 %vec5 = insertelement <8 x i8> %vec4, i8 %res5, i32 5
5348 %res6 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 6, i8 %mask)
5349 %vec6 = insertelement <8 x i8> %vec5, i8 %res6, i32 6
5350 %res7 = call i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16> %a0, <8 x i16> %a1, i32 7, i8 %mask)
5351 %vec7 = insertelement <8 x i8> %vec6, i8 %res7, i32 7
5352 ret <8 x i8> %vec7
5353 }
5355 declare i8 @llvm.x86.avx512.mask.ucmp.w.128(<8 x i16>, <8 x i16>, i32, i8) nounwind readnone
5357 define <16 x i8>@mm_mask_avg_epu8(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
5358 ; X86-LABEL: mm_mask_avg_epu8:
5359 ; X86: # %bb.0:
5360 ; X86-NEXT: vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
5361 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5362 ; X86-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
5363 ; X86-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
5364 ; X86-NEXT: retl # encoding: [0xc3]
5365 ;
5366 ; X64-LABEL: mm_mask_avg_epu8:
5367 ; X64: # %bb.0:
5368 ; X64-NEXT: vpavgb %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe0,0xd9]
5369 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5370 ; X64-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
5371 ; X64-NEXT: vpaddb %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc3]
5372 ; X64-NEXT: retq # encoding: [0xc3]
5373 %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)
5374 %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)
5375 %res2 = add <16 x i8> %res, %res1
5376 ret <16 x i8> %res2
5377 }
5379 declare <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8>, <16 x i8>, i16)
5381 define <16 x i8>@test_int_x86_avx512_mask_pabs_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5382 ; X86-LABEL: test_int_x86_avx512_mask_pabs_b_128:
5383 ; X86: # %bb.0:
5384 ; X86-NEXT: vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
5385 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5386 ; X86-NEXT: vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
5387 ; X86-NEXT: vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
5388 ; X86-NEXT: retl # encoding: [0xc3]
5389 ;
5390 ; X64-LABEL: test_int_x86_avx512_mask_pabs_b_128:
5391 ; X64: # %bb.0:
5392 ; X64-NEXT: vpabsb %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1c,0xd0]
5393 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5394 ; X64-NEXT: vpabsb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1c,0xc8]
5395 ; X64-NEXT: vpaddb %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc2]
5396 ; X64-NEXT: retq # encoding: [0xc3]
5397 %res = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5398 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pabs.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 -1)
5399 %res2 = add <16 x i8> %res, %res1
5400 ret <16 x i8> %res2
5401 }
5403 declare <16 x i8> @llvm.x86.avx512.mask.pavg.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
5405 define <32 x i8>@mm256_mask_avg_epu8(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
5406 ; X86-LABEL: mm256_mask_avg_epu8:
5407 ; X86: # %bb.0:
5408 ; X86-NEXT: vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
5409 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
5410 ; X86-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
5411 ; X86-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
5412 ; X86-NEXT: retl # encoding: [0xc3]
5413 ;
5414 ; X64-LABEL: mm256_mask_avg_epu8:
5415 ; X64: # %bb.0:
5416 ; X64-NEXT: vpavgb %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe0,0xd9]
5417 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5418 ; X64-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
5419 ; X64-NEXT: vpaddb %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfc,0xc3]
5420 ; X64-NEXT: retq # encoding: [0xc3]
5421 %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)
5422 %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)
5423 %res2 = add <32 x i8> %res, %res1
5424 ret <32 x i8> %res2
5425 }
5427 declare <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8>, <32 x i8>, i32)
5429 define <32 x i8>@test_int_x86_avx512_mask_pabs_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5430 ; X86-LABEL: test_int_x86_avx512_mask_pabs_b_256:
5431 ; X86: # %bb.0:
5432 ; X86-NEXT: vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
5433 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
5434 ; X86-NEXT: vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
5435 ; X86-NEXT: vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
5436 ; X86-NEXT: retl # encoding: [0xc3]
5437 ;
5438 ; X64-LABEL: test_int_x86_avx512_mask_pabs_b_256:
5439 ; X64: # %bb.0:
5440 ; X64-NEXT: vpabsb %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1c,0xd0]
5441 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5442 ; X64-NEXT: vpabsb %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1c,0xc8]
5443 ; X64-NEXT: vpaddb %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfc,0xc2]
5444 ; X64-NEXT: retq # encoding: [0xc3]
5445 %res = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5446 %res1 = call <32 x i8> @llvm.x86.avx512.mask.pabs.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 -1)
5447 %res2 = add <32 x i8> %res, %res1
5448 ret <32 x i8> %res2
5449 }
5451 declare <32 x i8> @llvm.x86.avx512.mask.pavg.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
5453 define <8 x i16>@mm_mask_avg_epu16(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
5454 ; X86-LABEL: mm_mask_avg_epu16:
5455 ; X86: # %bb.0:
5456 ; X86-NEXT: vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
5457 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5458 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5459 ; X86-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
5460 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5461 ; X86-NEXT: retl # encoding: [0xc3]
5462 ;
5463 ; X64-LABEL: mm_mask_avg_epu16:
5464 ; X64: # %bb.0:
5465 ; X64-NEXT: vpavgw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe3,0xd9]
5466 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5467 ; X64-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
5468 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5469 ; X64-NEXT: retq # encoding: [0xc3]
5470 %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)
5471 %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)
5472 %res2 = add <8 x i16> %res, %res1
5473 ret <8 x i16> %res2
5474 }
5476 declare <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16>, <8 x i16>, i8)
5478 define <8 x i16>@test_int_x86_avx512_mask_pabs_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5479 ; X86-LABEL: test_int_x86_avx512_mask_pabs_w_128:
5480 ; X86: # %bb.0:
5481 ; X86-NEXT: vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
5482 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5483 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5484 ; X86-NEXT: vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
5485 ; X86-NEXT: vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
5486 ; X86-NEXT: retl # encoding: [0xc3]
5487 ;
5488 ; X64-LABEL: test_int_x86_avx512_mask_pabs_w_128:
5489 ; X64: # %bb.0:
5490 ; X64-NEXT: vpabsw %xmm0, %xmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x1d,0xd0]
5491 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5492 ; X64-NEXT: vpabsw %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x1d,0xc8]
5493 ; X64-NEXT: vpaddw %xmm2, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc2]
5494 ; X64-NEXT: retq # encoding: [0xc3]
5495 %res = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5496 %res1 = call <8 x i16> @llvm.x86.avx512.mask.pabs.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 -1)
5497 %res2 = add <8 x i16> %res, %res1
5498 ret <8 x i16> %res2
5499 }
5501 declare <8 x i16> @llvm.x86.avx512.mask.pavg.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5503 define <16 x i16>@mm256_mask_avg_epu16(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
5504 ; X86-LABEL: mm256_mask_avg_epu16:
5505 ; X86: # %bb.0:
5506 ; X86-NEXT: vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
5507 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5508 ; X86-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
5509 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5510 ; X86-NEXT: retl # encoding: [0xc3]
5511 ;
5512 ; X64-LABEL: mm256_mask_avg_epu16:
5513 ; X64: # %bb.0:
5514 ; X64-NEXT: vpavgw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe3,0xd9]
5515 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5516 ; X64-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
5517 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5518 ; X64-NEXT: retq # encoding: [0xc3]
5519 %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)
5520 %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)
5521 %res2 = add <16 x i16> %res, %res1
5522 ret <16 x i16> %res2
5523 }
5525 declare <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16>, <16 x i16>, i16)
5527 define <16 x i16>@test_int_x86_avx512_mask_pabs_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5528 ; X86-LABEL: test_int_x86_avx512_mask_pabs_w_256:
5529 ; X86: # %bb.0:
5530 ; X86-NEXT: vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
5531 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5532 ; X86-NEXT: vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
5533 ; X86-NEXT: vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
5534 ; X86-NEXT: retl # encoding: [0xc3]
5535 ;
5536 ; X64-LABEL: test_int_x86_avx512_mask_pabs_w_256:
5537 ; X64: # %bb.0:
5538 ; X64-NEXT: vpabsw %ymm0, %ymm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x1d,0xd0]
5539 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5540 ; X64-NEXT: vpabsw %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x1d,0xc8]
5541 ; X64-NEXT: vpaddw %ymm2, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc2]
5542 ; X64-NEXT: retq # encoding: [0xc3]
5543 %res = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5544 %res1 = call <16 x i16> @llvm.x86.avx512.mask.pabs.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 -1)
5545 %res2 = add <16 x i16> %res, %res1
5546 ret <16 x i16> %res2
5547 }
5549 declare <16 x i16> @llvm.x86.avx512.mask.pavg.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5551 declare i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8>, <16 x i8>, i16)
5553 define i16@test_int_x86_avx512_ptestm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5554 ; X86-LABEL: test_int_x86_avx512_ptestm_b_128:
5555 ; X86: # %bb.0:
5556 ; X86-NEXT: vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
5557 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5558 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5559 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5560 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5561 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5562 ; X86-NEXT: retl # encoding: [0xc3]
5563 ;
5564 ; X64-LABEL: test_int_x86_avx512_ptestm_b_128:
5565 ; X64: # %bb.0:
5566 ; X64-NEXT: vptestmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7d,0x08,0x26,0xc1]
5567 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5568 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5569 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5570 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5571 ; X64-NEXT: retq # encoding: [0xc3]
5572 %res = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5573 %res1 = call i16 @llvm.x86.avx512.ptestm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
5574 %res2 = add i16 %res, %res1
5575 ret i16 %res2
5576 }
5578 declare i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8>, <32 x i8>, i32)
5580 define i32@test_int_x86_avx512_ptestm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5581 ; X86-LABEL: test_int_x86_avx512_ptestm_b_256:
5582 ; X86: # %bb.0:
5583 ; X86-NEXT: vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
5584 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5585 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5586 ; X86-NEXT: andl %ecx, %eax # encoding: [0x21,0xc8]
5587 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5588 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5589 ; X86-NEXT: retl # encoding: [0xc3]
5590 ;
5591 ; X64-LABEL: test_int_x86_avx512_ptestm_b_256:
5592 ; X64: # %bb.0:
5593 ; X64-NEXT: vptestmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7d,0x28,0x26,0xc1]
5594 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5595 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5596 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5597 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5598 ; X64-NEXT: retq # encoding: [0xc3]
5599 %res = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5600 %res1 = call i32 @llvm.x86.avx512.ptestm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
5601 %res2 = add i32 %res, %res1
5602 ret i32 %res2
5603 }
5605 declare i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16>, <8 x i16>, i8)
5607 define i8@test_int_x86_avx512_ptestm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5608 ; X86-LABEL: test_int_x86_avx512_ptestm_w_128:
5609 ; X86: # %bb.0:
5610 ; X86-NEXT: vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
5611 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5612 ; X86-NEXT: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
5613 ; X86-NEXT: andb %cl, %al # encoding: [0x20,0xc8]
5614 ; X86-NEXT: addb %cl, %al # encoding: [0x00,0xc8]
5615 ; X86-NEXT: retl # encoding: [0xc3]
5616 ;
5617 ; X64-LABEL: test_int_x86_avx512_ptestm_w_128:
5618 ; X64: # %bb.0:
5619 ; X64-NEXT: vptestmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfd,0x08,0x26,0xc1]
5620 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5621 ; X64-NEXT: andb %al, %dil # encoding: [0x40,0x20,0xc7]
5622 ; X64-NEXT: addb %dil, %al # encoding: [0x40,0x00,0xf8]
5623 ; X64-NEXT: # kill: def $al killed $al killed $eax
5624 ; X64-NEXT: retq # encoding: [0xc3]
5625 %res = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5626 %res1 = call i8 @llvm.x86.avx512.ptestm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
5627 %res2 = add i8 %res, %res1
5628 ret i8 %res2
5629 }
5631 declare i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16>, <16 x i16>, i16)
5633 define i16@test_int_x86_avx512_ptestm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5634 ; X86-LABEL: test_int_x86_avx512_ptestm_w_256:
5635 ; X86: # %bb.0:
5636 ; X86-NEXT: vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
5637 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5638 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5639 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5640 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5641 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5642 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5643 ; X86-NEXT: retl # encoding: [0xc3]
5644 ;
5645 ; X64-LABEL: test_int_x86_avx512_ptestm_w_256:
5646 ; X64: # %bb.0:
5647 ; X64-NEXT: vptestmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfd,0x28,0x26,0xc1]
5648 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5649 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5650 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5651 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5652 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5653 ; X64-NEXT: retq # encoding: [0xc3]
5654 %res = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5655 %res1 = call i16 @llvm.x86.avx512.ptestm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
5656 %res2 = add i16 %res, %res1
5657 ret i16 %res2
5658 }
5660 declare i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8>, <16 x i8>, i16)
5662 define i16@test_int_x86_avx512_ptestnm_b_128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2) {
5663 ; X86-LABEL: test_int_x86_avx512_ptestnm_b_128:
5664 ; X86: # %bb.0:
5665 ; X86-NEXT: vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
5666 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5667 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5668 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5669 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5670 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5671 ; X86-NEXT: retl # encoding: [0xc3]
5672 ;
5673 ; X64-LABEL: test_int_x86_avx512_ptestnm_b_128:
5674 ; X64: # %bb.0:
5675 ; X64-NEXT: vptestnmb %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x26,0xc1]
5676 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5677 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5678 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5679 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5680 ; X64-NEXT: retq # encoding: [0xc3]
5681 %res = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16 %x2)
5682 %res1 = call i16 @llvm.x86.avx512.ptestnm.b.128(<16 x i8> %x0, <16 x i8> %x1, i16-1)
5683 %res2 = add i16 %res, %res1
5684 ret i16 %res2
5685 }
5687 declare i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8>, <32 x i8>, i32)
5689 define i32@test_int_x86_avx512_ptestnm_b_256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2) {
5690 ; X86-LABEL: test_int_x86_avx512_ptestnm_b_256:
5691 ; X86: # %bb.0:
5692 ; X86-NEXT: vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
5693 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5694 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
5695 ; X86-NEXT: andl %ecx, %eax # encoding: [0x21,0xc8]
5696 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5697 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5698 ; X86-NEXT: retl # encoding: [0xc3]
5699 ;
5700 ; X64-LABEL: test_int_x86_avx512_ptestnm_b_256:
5701 ; X64: # %bb.0:
5702 ; X64-NEXT: vptestnmb %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x26,0xc1]
5703 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5704 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5705 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5706 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5707 ; X64-NEXT: retq # encoding: [0xc3]
5708 %res = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32 %x2)
5709 %res1 = call i32 @llvm.x86.avx512.ptestnm.b.256(<32 x i8> %x0, <32 x i8> %x1, i32-1)
5710 %res2 = add i32 %res, %res1
5711 ret i32 %res2
5712 }
5714 declare i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16>, <8 x i16>, i8 %x2)
5716 define i8@test_int_x86_avx512_ptestnm_w_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2) {
5717 ; X86-LABEL: test_int_x86_avx512_ptestnm_w_128:
5718 ; X86: # %bb.0:
5719 ; X86-NEXT: vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
5720 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5721 ; X86-NEXT: movb {{[0-9]+}}(%esp), %al # encoding: [0x8a,0x44,0x24,0x04]
5722 ; X86-NEXT: andb %cl, %al # encoding: [0x20,0xc8]
5723 ; X86-NEXT: addb %cl, %al # encoding: [0x00,0xc8]
5724 ; X86-NEXT: retl # encoding: [0xc3]
5725 ;
5726 ; X64-LABEL: test_int_x86_avx512_ptestnm_w_128:
5727 ; X64: # %bb.0:
5728 ; X64-NEXT: vptestnmw %xmm1, %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x26,0xc1]
5729 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5730 ; X64-NEXT: andb %al, %dil # encoding: [0x40,0x20,0xc7]
5731 ; X64-NEXT: addb %dil, %al # encoding: [0x40,0x00,0xf8]
5732 ; X64-NEXT: # kill: def $al killed $al killed $eax
5733 ; X64-NEXT: retq # encoding: [0xc3]
5734 %res = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8 %x2)
5735 %res1 = call i8 @llvm.x86.avx512.ptestnm.w.128(<8 x i16> %x0, <8 x i16> %x1, i8-1)
5736 %res2 = add i8 %res, %res1
5737 ret i8 %res2
5738 }
5740 declare i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16>, <16 x i16>, i16 %x2)
5742 define i16@test_int_x86_avx512_ptestnm_w_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2) {
5743 ; X86-LABEL: test_int_x86_avx512_ptestnm_w_256:
5744 ; X86: # %bb.0:
5745 ; X86-NEXT: vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
5746 ; X86-NEXT: kmovd %k0, %ecx # encoding: [0xc5,0xfb,0x93,0xc8]
5747 ; X86-NEXT: movzwl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb7,0x44,0x24,0x04]
5748 ; X86-NEXT: andw %cx, %ax # encoding: [0x66,0x21,0xc8]
5749 ; X86-NEXT: addl %ecx, %eax # encoding: [0x01,0xc8]
5750 ; X86-NEXT: # kill: def $ax killed $ax killed $eax
5751 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5752 ; X86-NEXT: retl # encoding: [0xc3]
5753 ;
5754 ; X64-LABEL: test_int_x86_avx512_ptestnm_w_256:
5755 ; X64: # %bb.0:
5756 ; X64-NEXT: vptestnmw %ymm1, %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x26,0xc1]
5757 ; X64-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5758 ; X64-NEXT: andl %eax, %edi # encoding: [0x21,0xc7]
5759 ; X64-NEXT: addl %edi, %eax # encoding: [0x01,0xf8]
5760 ; X64-NEXT: # kill: def $ax killed $ax killed $eax
5761 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5762 ; X64-NEXT: retq # encoding: [0xc3]
5763 %res = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16 %x2)
5764 %res1 = call i16 @llvm.x86.avx512.ptestnm.w.256(<16 x i16> %x0, <16 x i16> %x1, i16-1)
5765 %res2 = add i16 %res, %res1
5766 ret i16 %res2
5767 }
5769 declare i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8>)
5771 define i16@test_int_x86_avx512_cvtb2mask_128(<16 x i8> %x0) {
5772 ; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_128:
5773 ; CHECK: # %bb.0:
5774 ; CHECK-NEXT: vpmovb2m %xmm0, %k0 # encoding: [0x62,0xf2,0x7e,0x08,0x29,0xc0]
5775 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5776 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
5777 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5778 %res = call i16 @llvm.x86.avx512.cvtb2mask.128(<16 x i8> %x0)
5779 ret i16 %res
5780 }
5782 declare i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8>)
5784 define i32@test_int_x86_avx512_cvtb2mask_256(<32 x i8> %x0) {
5785 ; CHECK-LABEL: test_int_x86_avx512_cvtb2mask_256:
5786 ; CHECK: # %bb.0:
5787 ; CHECK-NEXT: vpmovb2m %ymm0, %k0 # encoding: [0x62,0xf2,0x7e,0x28,0x29,0xc0]
5788 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5789 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5790 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5791 %res = call i32 @llvm.x86.avx512.cvtb2mask.256(<32 x i8> %x0)
5792 ret i32 %res
5793 }
5795 declare i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16>)
5797 define i8@test_int_x86_avx512_cvtw2mask_128(<8 x i16> %x0) {
5798 ; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_128:
5799 ; CHECK: # %bb.0:
5800 ; CHECK-NEXT: vpmovw2m %xmm0, %k0 # encoding: [0x62,0xf2,0xfe,0x08,0x29,0xc0]
5801 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5802 ; CHECK-NEXT: # kill: def $al killed $al killed $eax
5803 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5804 %res = call i8 @llvm.x86.avx512.cvtw2mask.128(<8 x i16> %x0)
5805 ret i8 %res
5806 }
5808 declare i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16>)
5810 define i16@test_int_x86_avx512_cvtw2mask_256(<16 x i16> %x0) {
5811 ; CHECK-LABEL: test_int_x86_avx512_cvtw2mask_256:
5812 ; CHECK: # %bb.0:
5813 ; CHECK-NEXT: vpmovw2m %ymm0, %k0 # encoding: [0x62,0xf2,0xfe,0x28,0x29,0xc0]
5814 ; CHECK-NEXT: kmovd %k0, %eax # encoding: [0xc5,0xfb,0x93,0xc0]
5815 ; CHECK-NEXT: # kill: def $ax killed $ax killed $eax
5816 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
5817 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
5818 %res = call i16 @llvm.x86.avx512.cvtw2mask.256(<16 x i16> %x0)
5819 ret i16 %res
5820 }
5822 declare <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5824 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) {
5825 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
5826 ; X86: # %bb.0:
5827 ; X86-NEXT: vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
5828 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5829 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5830 ; X86-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
5831 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5832 ; X86-NEXT: retl # encoding: [0xc3]
5833 ;
5834 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
5835 ; X64: # %bb.0:
5836 ; X64-NEXT: vpmulhuw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe4,0xd9]
5837 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5838 ; X64-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
5839 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5840 ; X64-NEXT: retq # encoding: [0xc3]
5841 %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)
5842 %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)
5843 %res2 = add <8 x i16> %res, %res1
5844 ret <8 x i16> %res2
5845 }
5847 declare <16 x i16> @llvm.x86.avx512.mask.pmulhu.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5849 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) {
5850 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
5851 ; X86: # %bb.0:
5852 ; X86-NEXT: vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
5853 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5854 ; X86-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
5855 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5856 ; X86-NEXT: retl # encoding: [0xc3]
5857 ;
5858 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
5859 ; X64: # %bb.0:
5860 ; X64-NEXT: vpmulhuw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe4,0xd9]
5861 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5862 ; X64-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
5863 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5864 ; X64-NEXT: retq # encoding: [0xc3]
5865 %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)
5866 %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)
5867 %res2 = add <16 x i16> %res, %res1
5868 ret <16 x i16> %res2
5869 }
5871 declare <8 x i16> @llvm.x86.avx512.mask.pmulh.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5873 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) {
5874 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
5875 ; X86: # %bb.0:
5876 ; X86-NEXT: vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
5877 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5878 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5879 ; X86-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
5880 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5881 ; X86-NEXT: retl # encoding: [0xc3]
5882 ;
5883 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
5884 ; X64: # %bb.0:
5885 ; X64-NEXT: vpmulhw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe5,0xd9]
5886 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5887 ; X64-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
5888 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5889 ; X64-NEXT: retq # encoding: [0xc3]
5890 %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)
5891 %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)
5892 %res2 = add <8 x i16> %res, %res1
5893 ret <8 x i16> %res2
5894 }
5896 declare <16 x i16> @llvm.x86.avx512.mask.pmulh.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5898 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) {
5899 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
5900 ; X86: # %bb.0:
5901 ; X86-NEXT: vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
5902 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5903 ; X86-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
5904 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5905 ; X86-NEXT: retl # encoding: [0xc3]
5906 ;
5907 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
5908 ; X64: # %bb.0:
5909 ; X64-NEXT: vpmulhw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe5,0xd9]
5910 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5911 ; X64-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
5912 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5913 ; X64-NEXT: retq # encoding: [0xc3]
5914 %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)
5915 %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)
5916 %res2 = add <16 x i16> %res, %res1
5917 ret <16 x i16> %res2
5918 }
5920 declare <8 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
5922 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) {
5923 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
5924 ; X86: # %bb.0:
5925 ; X86-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
5926 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5927 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5928 ; X86-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
5929 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5930 ; X86-NEXT: retl # encoding: [0xc3]
5931 ;
5932 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
5933 ; X64: # %bb.0:
5934 ; X64-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x0b,0xd9]
5935 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5936 ; X64-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
5937 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5938 ; X64-NEXT: retq # encoding: [0xc3]
5939 %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)
5940 %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)
5941 %res2 = add <8 x i16> %res, %res1
5942 ret <8 x i16> %res2
5943 }
5945 declare <16 x i16> @llvm.x86.avx512.mask.pmul.hr.sw.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
5947 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) {
5948 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
5949 ; X86: # %bb.0:
5950 ; X86-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
5951 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
5952 ; X86-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
5953 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5954 ; X86-NEXT: retl # encoding: [0xc3]
5955 ;
5956 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
5957 ; X64: # %bb.0:
5958 ; X64-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x0b,0xd9]
5959 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5960 ; X64-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
5961 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
5962 ; X64-NEXT: retq # encoding: [0xc3]
5963 %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)
5964 %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)
5965 %res2 = add <16 x i16> %res, %res1
5966 ret <16 x i16> %res2
5967 }
5969 declare <8 x i16> @llvm.x86.avx512.mask.pmaddubs.w.128(<16 x i8>, <16 x i8>, <8 x i16>, i8)
5971 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) {
5972 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
5973 ; X86: # %bb.0:
5974 ; X86-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
5975 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
5976 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
5977 ; X86-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
5978 ; X86-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5979 ; X86-NEXT: retl # encoding: [0xc3]
5980 ;
5981 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
5982 ; X64: # %bb.0:
5983 ; X64-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x04,0xd9]
5984 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
5985 ; X64-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
5986 ; X64-NEXT: vpaddw %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc3]
5987 ; X64-NEXT: retq # encoding: [0xc3]
5988 %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)
5989 %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)
5990 %res2 = add <8 x i16> %res, %res1
5991 ret <8 x i16> %res2
5992 }
5994 declare <16 x i16> @llvm.x86.avx512.mask.pmaddubs.w.256(<32 x i8>, <32 x i8>, <16 x i16>, i16)
5996 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) {
5997 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
5998 ; X86: # %bb.0:
5999 ; X86-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
6000 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6001 ; X86-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
6002 ; X86-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
6003 ; X86-NEXT: retl # encoding: [0xc3]
6004 ;
6005 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
6006 ; X64: # %bb.0:
6007 ; X64-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x04,0xd9]
6008 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6009 ; X64-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
6010 ; X64-NEXT: vpaddw %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc3]
6011 ; X64-NEXT: retq # encoding: [0xc3]
6012 %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)
6013 %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)
6014 %res2 = add <16 x i16> %res, %res1
6015 ret <16 x i16> %res2
6016 }
6018 declare <4 x i32> @llvm.x86.avx512.mask.pmaddw.d.128(<8 x i16>, <8 x i16>, <4 x i32>, i8)
6020 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) {
6021 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
6022 ; X86: # %bb.0:
6023 ; X86-NEXT: vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
6024 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6025 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6026 ; X86-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
6027 ; X86-NEXT: vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
6028 ; X86-NEXT: retl # encoding: [0xc3]
6029 ;
6030 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
6031 ; X64: # %bb.0:
6032 ; X64-NEXT: vpmaddwd %xmm1, %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xf5,0xd9]
6033 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6034 ; X64-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
6035 ; X64-NEXT: vpaddd %xmm3, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfe,0xc3]
6036 ; X64-NEXT: retq # encoding: [0xc3]
6037 %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)
6038 %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)
6039 %res2 = add <4 x i32> %res, %res1
6040 ret <4 x i32> %res2
6041 }
6043 declare <8 x i32> @llvm.x86.avx512.mask.pmaddw.d.256(<16 x i16>, <16 x i16>, <8 x i32>, i8)
6045 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) {
6046 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
6047 ; X86: # %bb.0:
6048 ; X86-NEXT: vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
6049 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6050 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6051 ; X86-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
6052 ; X86-NEXT: vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
6053 ; X86-NEXT: retl # encoding: [0xc3]
6054 ;
6055 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
6056 ; X64: # %bb.0:
6057 ; X64-NEXT: vpmaddwd %ymm1, %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xf5,0xd9]
6058 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6059 ; X64-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
6060 ; X64-NEXT: vpaddd %ymm3, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfe,0xc3]
6061 ; X64-NEXT: retq # encoding: [0xc3]
6062 %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)
6063 %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)
6064 %res2 = add <8 x i32> %res, %res1
6065 ret <8 x i32> %res2
6066 }
6068 declare <8 x i16> @llvm.x86.avx512.mask.permvar.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6070 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) {
6071 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
6072 ; X86: # %bb.0:
6073 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
6074 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6075 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6076 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
6077 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
6078 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6079 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6080 ; X86-NEXT: retl # encoding: [0xc3]
6081 ;
6082 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
6083 ; X64: # %bb.0:
6084 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xd8]
6085 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6086 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
6087 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
6088 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6089 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6090 ; X64-NEXT: retq # encoding: [0xc3]
6091 %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)
6092 %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)
6093 %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)
6094 %res3 = add <8 x i16> %res, %res1
6095 %res4 = add <8 x i16> %res3, %res2
6096 ret <8 x i16> %res4
6097 }
6099 declare <16 x i16> @llvm.x86.avx512.mask.permvar.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6101 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) {
6102 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
6103 ; X86: # %bb.0:
6104 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
6105 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6106 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
6107 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
6108 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6109 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6110 ; X86-NEXT: retl # encoding: [0xc3]
6111 ;
6112 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
6113 ; X64: # %bb.0:
6114 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xd8]
6115 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6116 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
6117 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
6118 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6119 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6120 ; X64-NEXT: retq # encoding: [0xc3]
6121 %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)
6122 %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)
6123 %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)
6124 %res3 = add <16 x i16> %res, %res1
6125 %res4 = add <16 x i16> %res3, %res2
6126 ret <16 x i16> %res4
6127 }
6129 declare <8 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6131 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) {
6132 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
6133 ; X86: # %bb.0:
6134 ; X86-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6135 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6136 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6137 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6138 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
6139 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6140 ; X86-NEXT: retl # encoding: [0xc3]
6141 ;
6142 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
6143 ; X64: # %bb.0:
6144 ; X64-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6145 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6146 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6147 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
6148 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6149 ; X64-NEXT: retq # encoding: [0xc3]
6150 %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)
6151 %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)
6152 %res2 = add <8 x i16> %res, %res1
6153 ret <8 x i16> %res2
6154 }
6156 declare <8 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6158 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) {
6159 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
6160 ; X86: # %bb.0:
6161 ; X86-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6162 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6163 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6164 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6165 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
6166 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6167 ; X86-NEXT: retl # encoding: [0xc3]
6168 ;
6169 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
6170 ; X64: # %bb.0:
6171 ; X64-NEXT: vmovdqa %xmm1, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd9]
6172 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x7d,0xda]
6173 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6174 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x7d,0xca]
6175 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6176 ; X64-NEXT: retq # encoding: [0xc3]
6177 %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)
6178 %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)
6179 %res2 = add <8 x i16> %res, %res1
6180 ret <8 x i16> %res2
6181 }
6183 declare <16 x i16> @llvm.x86.avx512.mask.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6185 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) {
6186 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
6187 ; X86: # %bb.0:
6188 ; X86-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6189 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6190 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6191 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
6192 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6193 ; X86-NEXT: retl # encoding: [0xc3]
6194 ;
6195 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
6196 ; X64: # %bb.0:
6197 ; X64-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6198 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6199 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6200 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
6201 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6202 ; X64-NEXT: retq # encoding: [0xc3]
6203 %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)
6204 %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)
6205 %res2 = add <16 x i16> %res, %res1
6206 ret <16 x i16> %res2
6207 }
6209 declare <16 x i16> @llvm.x86.avx512.maskz.vpermt2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6211 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) {
6212 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
6213 ; X86: # %bb.0:
6214 ; X86-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6215 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6216 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6217 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
6218 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6219 ; X86-NEXT: retl # encoding: [0xc3]
6220 ;
6221 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
6222 ; X64: # %bb.0:
6223 ; X64-NEXT: vmovdqa %ymm1, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd9]
6224 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x7d,0xda]
6225 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6226 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x7d,0xca]
6227 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6228 ; X64-NEXT: retq # encoding: [0xc3]
6229 %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)
6230 %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)
6231 %res2 = add <16 x i16> %res, %res1
6232 ret <16 x i16> %res2
6233 }
6235 declare <8 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6237 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) {
6238 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
6239 ; X86: # %bb.0:
6240 ; X86-NEXT: vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
6241 ; X86-NEXT: vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
6242 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6243 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6244 ; X86-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
6245 ; X86-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6246 ; X86-NEXT: retl # encoding: [0xc3]
6247 ;
6248 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
6249 ; X64: # %bb.0:
6250 ; X64-NEXT: vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
6251 ; X64-NEXT: vpermt2w %xmm2, %xmm1, %xmm3 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xda]
6252 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6253 ; X64-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
6254 ; X64-NEXT: vpaddw %xmm3, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfd,0xc3]
6255 ; X64-NEXT: retq # encoding: [0xc3]
6256 %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)
6257 %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)
6258 %res2 = add <8 x i16> %res, %res1
6259 ret <8 x i16> %res2
6260 }
6262 declare <16 x i16> @llvm.x86.avx512.mask.vpermi2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6264 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) {
6265 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
6266 ; X86: # %bb.0:
6267 ; X86-NEXT: vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
6268 ; X86-NEXT: vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
6269 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6270 ; X86-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
6271 ; X86-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6272 ; X86-NEXT: retl # encoding: [0xc3]
6273 ;
6274 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
6275 ; X64: # %bb.0:
6276 ; X64-NEXT: vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
6277 ; X64-NEXT: vpermt2w %ymm2, %ymm1, %ymm3 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xda]
6278 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6279 ; X64-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
6280 ; X64-NEXT: vpaddw %ymm3, %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xf5,0xfd,0xc3]
6281 ; X64-NEXT: retq # encoding: [0xc3]
6282 %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)
6283 %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)
6284 %res2 = add <16 x i16> %res, %res1
6285 ret <16 x i16> %res2
6286 }
6288 declare <8 x i16> @llvm.x86.avx512.mask.dbpsadbw.128(<16 x i8>, <16 x i8>, i32, <8 x i16>, i8)
6290 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) {
6291 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
6292 ; X86: # %bb.0:
6293 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
6294 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6295 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6296 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
6297 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
6298 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6299 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6300 ; X86-NEXT: retl # encoding: [0xc3]
6301 ;
6302 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
6303 ; X64: # %bb.0:
6304 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd9,0x02]
6305 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6306 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xd1,0x02]
6307 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xc1,0x02]
6308 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
6309 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
6310 ; X64-NEXT: retq # encoding: [0xc3]
6311 %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)
6312 %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)
6313 %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)
6314 %res3 = add <8 x i16> %res, %res1
6315 %res4 = add <8 x i16> %res2, %res3
6316 ret <8 x i16> %res4
6317 }
6319 declare <16 x i16> @llvm.x86.avx512.mask.dbpsadbw.256(<32 x i8>, <32 x i8>, i32, <16 x i16>, i16)
6321 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) {
6322 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
6323 ; X86: # %bb.0:
6324 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
6325 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6326 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
6327 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
6328 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6329 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6330 ; X86-NEXT: retl # encoding: [0xc3]
6331 ;
6332 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
6333 ; X64: # %bb.0:
6334 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd9,0x02]
6335 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6336 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xd1,0x02]
6337 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xc1,0x02]
6338 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
6339 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
6340 ; X64-NEXT: retq # encoding: [0xc3]
6341 %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)
6342 %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)
6343 %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)
6344 %res3 = add <16 x i16> %res, %res1
6345 %res4 = add <16 x i16> %res3, %res2
6346 ret <16 x i16> %res4
6347 }
6349 define <8 x i16> @test_mask_adds_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
6350 ; CHECK-LABEL: test_mask_adds_epu16_rr_128:
6351 ; CHECK: # %bb.0:
6352 ; CHECK-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0xc1]
6353 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6354 %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)
6355 ret <8 x i16> %res
6356 }
6358 define <8 x i16> @test_mask_adds_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
6359 ; X86-LABEL: test_mask_adds_epu16_rrk_128:
6360 ; X86: # %bb.0:
6361 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6362 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6363 ; X86-NEXT: vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
6364 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6365 ; X86-NEXT: retl # encoding: [0xc3]
6366 ;
6367 ; X64-LABEL: test_mask_adds_epu16_rrk_128:
6368 ; X64: # %bb.0:
6369 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6370 ; X64-NEXT: vpaddusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0xd1]
6371 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6372 ; X64-NEXT: retq # encoding: [0xc3]
6373 %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)
6374 ret <8 x i16> %res
6375 }
6377 define <8 x i16> @test_mask_adds_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
6378 ; X86-LABEL: test_mask_adds_epu16_rrkz_128:
6379 ; X86: # %bb.0:
6380 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6381 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6382 ; X86-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
6383 ; X86-NEXT: retl # encoding: [0xc3]
6384 ;
6385 ; X64-LABEL: test_mask_adds_epu16_rrkz_128:
6386 ; X64: # %bb.0:
6387 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6388 ; X64-NEXT: vpaddusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0xc1]
6389 ; X64-NEXT: retq # encoding: [0xc3]
6390 %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)
6391 ret <8 x i16> %res
6392 }
6394 define <8 x i16> @test_mask_adds_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
6395 ; X86-LABEL: test_mask_adds_epu16_rm_128:
6396 ; X86: # %bb.0:
6397 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6398 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x00]
6399 ; X86-NEXT: retl # encoding: [0xc3]
6400 ;
6401 ; X64-LABEL: test_mask_adds_epu16_rm_128:
6402 ; X64: # %bb.0:
6403 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdd,0x07]
6404 ; X64-NEXT: retq # encoding: [0xc3]
6405 %b = load <8 x i16>, <8 x i16>* %ptr_b
6406 %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)
6407 ret <8 x i16> %res
6408 }
6410 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) {
6411 ; X86-LABEL: test_mask_adds_epu16_rmk_128:
6412 ; X86: # %bb.0:
6413 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6414 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6415 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6416 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x08]
6417 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6418 ; X86-NEXT: retl # encoding: [0xc3]
6419 ;
6420 ; X64-LABEL: test_mask_adds_epu16_rmk_128:
6421 ; X64: # %bb.0:
6422 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6423 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdd,0x0f]
6424 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6425 ; X64-NEXT: retq # encoding: [0xc3]
6426 %b = load <8 x i16>, <8 x i16>* %ptr_b
6427 %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)
6428 ret <8 x i16> %res
6429 }
6431 define <8 x i16> @test_mask_adds_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
6432 ; X86-LABEL: test_mask_adds_epu16_rmkz_128:
6433 ; X86: # %bb.0:
6434 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6435 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6436 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6437 ; X86-NEXT: vpaddusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x00]
6438 ; X86-NEXT: retl # encoding: [0xc3]
6439 ;
6440 ; X64-LABEL: test_mask_adds_epu16_rmkz_128:
6441 ; X64: # %bb.0:
6442 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6443 ; X64-NEXT: vpaddusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdd,0x07]
6444 ; X64-NEXT: retq # encoding: [0xc3]
6445 %b = load <8 x i16>, <8 x i16>* %ptr_b
6446 %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)
6447 ret <8 x i16> %res
6448 }
6450 declare <8 x i16> @llvm.x86.avx512.mask.paddus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6452 define <16 x i16> @test_mask_adds_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
6453 ; CHECK-LABEL: test_mask_adds_epu16_rr_256:
6454 ; CHECK: # %bb.0:
6455 ; CHECK-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0xc1]
6456 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6457 %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)
6458 ret <16 x i16> %res
6459 }
6461 define <16 x i16> @test_mask_adds_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
6462 ; X86-LABEL: test_mask_adds_epu16_rrk_256:
6463 ; X86: # %bb.0:
6464 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6465 ; X86-NEXT: vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
6466 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6467 ; X86-NEXT: retl # encoding: [0xc3]
6468 ;
6469 ; X64-LABEL: test_mask_adds_epu16_rrk_256:
6470 ; X64: # %bb.0:
6471 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6472 ; X64-NEXT: vpaddusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0xd1]
6473 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6474 ; X64-NEXT: retq # encoding: [0xc3]
6475 %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)
6476 ret <16 x i16> %res
6477 }
6479 define <16 x i16> @test_mask_adds_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
6480 ; X86-LABEL: test_mask_adds_epu16_rrkz_256:
6481 ; X86: # %bb.0:
6482 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6483 ; X86-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
6484 ; X86-NEXT: retl # encoding: [0xc3]
6485 ;
6486 ; X64-LABEL: test_mask_adds_epu16_rrkz_256:
6487 ; X64: # %bb.0:
6488 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6489 ; X64-NEXT: vpaddusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0xc1]
6490 ; X64-NEXT: retq # encoding: [0xc3]
6491 %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)
6492 ret <16 x i16> %res
6493 }
6495 define <16 x i16> @test_mask_adds_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
6496 ; X86-LABEL: test_mask_adds_epu16_rm_256:
6497 ; X86: # %bb.0:
6498 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6499 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x00]
6500 ; X86-NEXT: retl # encoding: [0xc3]
6501 ;
6502 ; X64-LABEL: test_mask_adds_epu16_rm_256:
6503 ; X64: # %bb.0:
6504 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdd,0x07]
6505 ; X64-NEXT: retq # encoding: [0xc3]
6506 %b = load <16 x i16>, <16 x i16>* %ptr_b
6507 %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)
6508 ret <16 x i16> %res
6509 }
6511 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) {
6512 ; X86-LABEL: test_mask_adds_epu16_rmk_256:
6513 ; X86: # %bb.0:
6514 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6515 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6516 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x08]
6517 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6518 ; X86-NEXT: retl # encoding: [0xc3]
6519 ;
6520 ; X64-LABEL: test_mask_adds_epu16_rmk_256:
6521 ; X64: # %bb.0:
6522 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6523 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdd,0x0f]
6524 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6525 ; X64-NEXT: retq # encoding: [0xc3]
6526 %b = load <16 x i16>, <16 x i16>* %ptr_b
6527 %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)
6528 ret <16 x i16> %res
6529 }
6531 define <16 x i16> @test_mask_adds_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
6532 ; X86-LABEL: test_mask_adds_epu16_rmkz_256:
6533 ; X86: # %bb.0:
6534 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6535 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6536 ; X86-NEXT: vpaddusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x00]
6537 ; X86-NEXT: retl # encoding: [0xc3]
6538 ;
6539 ; X64-LABEL: test_mask_adds_epu16_rmkz_256:
6540 ; X64: # %bb.0:
6541 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6542 ; X64-NEXT: vpaddusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdd,0x07]
6543 ; X64-NEXT: retq # encoding: [0xc3]
6544 %b = load <16 x i16>, <16 x i16>* %ptr_b
6545 %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)
6546 ret <16 x i16> %res
6547 }
6549 declare <16 x i16> @llvm.x86.avx512.mask.paddus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6551 define <8 x i16> @test_mask_subs_epu16_rr_128(<8 x i16> %a, <8 x i16> %b) {
6552 ; CHECK-LABEL: test_mask_subs_epu16_rr_128:
6553 ; CHECK: # %bb.0:
6554 ; CHECK-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0xc1]
6555 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6556 %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)
6557 ret <8 x i16> %res
6558 }
6560 define <8 x i16> @test_mask_subs_epu16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
6561 ; X86-LABEL: test_mask_subs_epu16_rrk_128:
6562 ; X86: # %bb.0:
6563 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6564 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6565 ; X86-NEXT: vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
6566 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6567 ; X86-NEXT: retl # encoding: [0xc3]
6568 ;
6569 ; X64-LABEL: test_mask_subs_epu16_rrk_128:
6570 ; X64: # %bb.0:
6571 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6572 ; X64-NEXT: vpsubusw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0xd1]
6573 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6574 ; X64-NEXT: retq # encoding: [0xc3]
6575 %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)
6576 ret <8 x i16> %res
6577 }
6579 define <8 x i16> @test_mask_subs_epu16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
6580 ; X86-LABEL: test_mask_subs_epu16_rrkz_128:
6581 ; X86: # %bb.0:
6582 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
6583 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
6584 ; X86-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
6585 ; X86-NEXT: retl # encoding: [0xc3]
6586 ;
6587 ; X64-LABEL: test_mask_subs_epu16_rrkz_128:
6588 ; X64: # %bb.0:
6589 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6590 ; X64-NEXT: vpsubusw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0xc1]
6591 ; X64-NEXT: retq # encoding: [0xc3]
6592 %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)
6593 ret <8 x i16> %res
6594 }
6596 define <8 x i16> @test_mask_subs_epu16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
6597 ; X86-LABEL: test_mask_subs_epu16_rm_128:
6598 ; X86: # %bb.0:
6599 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6600 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x00]
6601 ; X86-NEXT: retl # encoding: [0xc3]
6602 ;
6603 ; X64-LABEL: test_mask_subs_epu16_rm_128:
6604 ; X64: # %bb.0:
6605 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd9,0x07]
6606 ; X64-NEXT: retq # encoding: [0xc3]
6607 %b = load <8 x i16>, <8 x i16>* %ptr_b
6608 %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)
6609 ret <8 x i16> %res
6610 }
6612 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) {
6613 ; X86-LABEL: test_mask_subs_epu16_rmk_128:
6614 ; X86: # %bb.0:
6615 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6616 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6617 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6618 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x08]
6619 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6620 ; X86-NEXT: retl # encoding: [0xc3]
6621 ;
6622 ; X64-LABEL: test_mask_subs_epu16_rmk_128:
6623 ; X64: # %bb.0:
6624 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6625 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd9,0x0f]
6626 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6627 ; X64-NEXT: retq # encoding: [0xc3]
6628 %b = load <8 x i16>, <8 x i16>* %ptr_b
6629 %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)
6630 ret <8 x i16> %res
6631 }
6633 define <8 x i16> @test_mask_subs_epu16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
6634 ; X86-LABEL: test_mask_subs_epu16_rmkz_128:
6635 ; X86: # %bb.0:
6636 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6637 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
6638 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
6639 ; X86-NEXT: vpsubusw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x00]
6640 ; X86-NEXT: retl # encoding: [0xc3]
6641 ;
6642 ; X64-LABEL: test_mask_subs_epu16_rmkz_128:
6643 ; X64: # %bb.0:
6644 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6645 ; X64-NEXT: vpsubusw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd9,0x07]
6646 ; X64-NEXT: retq # encoding: [0xc3]
6647 %b = load <8 x i16>, <8 x i16>* %ptr_b
6648 %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)
6649 ret <8 x i16> %res
6650 }
6652 declare <8 x i16> @llvm.x86.avx512.mask.psubus.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
6654 define <16 x i16> @test_mask_subs_epu16_rr_256(<16 x i16> %a, <16 x i16> %b) {
6655 ; CHECK-LABEL: test_mask_subs_epu16_rr_256:
6656 ; CHECK: # %bb.0:
6657 ; CHECK-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0xc1]
6658 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6659 %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)
6660 ret <16 x i16> %res
6661 }
6663 define <16 x i16> @test_mask_subs_epu16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
6664 ; X86-LABEL: test_mask_subs_epu16_rrk_256:
6665 ; X86: # %bb.0:
6666 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6667 ; X86-NEXT: vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
6668 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6669 ; X86-NEXT: retl # encoding: [0xc3]
6670 ;
6671 ; X64-LABEL: test_mask_subs_epu16_rrk_256:
6672 ; X64: # %bb.0:
6673 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6674 ; X64-NEXT: vpsubusw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0xd1]
6675 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6676 ; X64-NEXT: retq # encoding: [0xc3]
6677 %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)
6678 ret <16 x i16> %res
6679 }
6681 define <16 x i16> @test_mask_subs_epu16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
6682 ; X86-LABEL: test_mask_subs_epu16_rrkz_256:
6683 ; X86: # %bb.0:
6684 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6685 ; X86-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
6686 ; X86-NEXT: retl # encoding: [0xc3]
6687 ;
6688 ; X64-LABEL: test_mask_subs_epu16_rrkz_256:
6689 ; X64: # %bb.0:
6690 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6691 ; X64-NEXT: vpsubusw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0xc1]
6692 ; X64-NEXT: retq # encoding: [0xc3]
6693 %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)
6694 ret <16 x i16> %res
6695 }
6697 define <16 x i16> @test_mask_subs_epu16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
6698 ; X86-LABEL: test_mask_subs_epu16_rm_256:
6699 ; X86: # %bb.0:
6700 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6701 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x00]
6702 ; X86-NEXT: retl # encoding: [0xc3]
6703 ;
6704 ; X64-LABEL: test_mask_subs_epu16_rm_256:
6705 ; X64: # %bb.0:
6706 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd9,0x07]
6707 ; X64-NEXT: retq # encoding: [0xc3]
6708 %b = load <16 x i16>, <16 x i16>* %ptr_b
6709 %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)
6710 ret <16 x i16> %res
6711 }
6713 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) {
6714 ; X86-LABEL: test_mask_subs_epu16_rmk_256:
6715 ; X86: # %bb.0:
6716 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6717 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6718 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x08]
6719 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6720 ; X86-NEXT: retl # encoding: [0xc3]
6721 ;
6722 ; X64-LABEL: test_mask_subs_epu16_rmk_256:
6723 ; X64: # %bb.0:
6724 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6725 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd9,0x0f]
6726 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6727 ; X64-NEXT: retq # encoding: [0xc3]
6728 %b = load <16 x i16>, <16 x i16>* %ptr_b
6729 %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)
6730 ret <16 x i16> %res
6731 }
6733 define <16 x i16> @test_mask_subs_epu16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
6734 ; X86-LABEL: test_mask_subs_epu16_rmkz_256:
6735 ; X86: # %bb.0:
6736 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6737 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6738 ; X86-NEXT: vpsubusw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x00]
6739 ; X86-NEXT: retl # encoding: [0xc3]
6740 ;
6741 ; X64-LABEL: test_mask_subs_epu16_rmkz_256:
6742 ; X64: # %bb.0:
6743 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6744 ; X64-NEXT: vpsubusw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd9,0x07]
6745 ; X64-NEXT: retq # encoding: [0xc3]
6746 %b = load <16 x i16>, <16 x i16>* %ptr_b
6747 %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)
6748 ret <16 x i16> %res
6749 }
6751 declare <16 x i16> @llvm.x86.avx512.mask.psubus.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
6753 define <16 x i8> @test_mask_adds_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
6754 ; CHECK-LABEL: test_mask_adds_epu8_rr_128:
6755 ; CHECK: # %bb.0:
6756 ; CHECK-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0xc1]
6757 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6758 %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)
6759 ret <16 x i8> %res
6760 }
6762 define <16 x i8> @test_mask_adds_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
6763 ; X86-LABEL: test_mask_adds_epu8_rrk_128:
6764 ; X86: # %bb.0:
6765 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6766 ; X86-NEXT: vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
6767 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6768 ; X86-NEXT: retl # encoding: [0xc3]
6769 ;
6770 ; X64-LABEL: test_mask_adds_epu8_rrk_128:
6771 ; X64: # %bb.0:
6772 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6773 ; X64-NEXT: vpaddusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0xd1]
6774 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6775 ; X64-NEXT: retq # encoding: [0xc3]
6776 %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)
6777 ret <16 x i8> %res
6778 }
6780 define <16 x i8> @test_mask_adds_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
6781 ; X86-LABEL: test_mask_adds_epu8_rrkz_128:
6782 ; X86: # %bb.0:
6783 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6784 ; X86-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
6785 ; X86-NEXT: retl # encoding: [0xc3]
6786 ;
6787 ; X64-LABEL: test_mask_adds_epu8_rrkz_128:
6788 ; X64: # %bb.0:
6789 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6790 ; X64-NEXT: vpaddusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0xc1]
6791 ; X64-NEXT: retq # encoding: [0xc3]
6792 %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)
6793 ret <16 x i8> %res
6794 }
6796 define <16 x i8> @test_mask_adds_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
6797 ; X86-LABEL: test_mask_adds_epu8_rm_128:
6798 ; X86: # %bb.0:
6799 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6800 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x00]
6801 ; X86-NEXT: retl # encoding: [0xc3]
6802 ;
6803 ; X64-LABEL: test_mask_adds_epu8_rm_128:
6804 ; X64: # %bb.0:
6805 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xdc,0x07]
6806 ; X64-NEXT: retq # encoding: [0xc3]
6807 %b = load <16 x i8>, <16 x i8>* %ptr_b
6808 %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)
6809 ret <16 x i8> %res
6810 }
6812 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) {
6813 ; X86-LABEL: test_mask_adds_epu8_rmk_128:
6814 ; X86: # %bb.0:
6815 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6816 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6817 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x08]
6818 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6819 ; X86-NEXT: retl # encoding: [0xc3]
6820 ;
6821 ; X64-LABEL: test_mask_adds_epu8_rmk_128:
6822 ; X64: # %bb.0:
6823 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6824 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xdc,0x0f]
6825 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
6826 ; X64-NEXT: retq # encoding: [0xc3]
6827 %b = load <16 x i8>, <16 x i8>* %ptr_b
6828 %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)
6829 ret <16 x i8> %res
6830 }
6832 define <16 x i8> @test_mask_adds_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
6833 ; X86-LABEL: test_mask_adds_epu8_rmkz_128:
6834 ; X86: # %bb.0:
6835 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6836 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
6837 ; X86-NEXT: vpaddusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x00]
6838 ; X86-NEXT: retl # encoding: [0xc3]
6839 ;
6840 ; X64-LABEL: test_mask_adds_epu8_rmkz_128:
6841 ; X64: # %bb.0:
6842 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6843 ; X64-NEXT: vpaddusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xdc,0x07]
6844 ; X64-NEXT: retq # encoding: [0xc3]
6845 %b = load <16 x i8>, <16 x i8>* %ptr_b
6846 %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)
6847 ret <16 x i8> %res
6848 }
6850 declare <16 x i8> @llvm.x86.avx512.mask.paddus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
6852 define <32 x i8> @test_mask_adds_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
6853 ; CHECK-LABEL: test_mask_adds_epu8_rr_256:
6854 ; CHECK: # %bb.0:
6855 ; CHECK-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0xc1]
6856 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6857 %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)
6858 ret <32 x i8> %res
6859 }
6861 define <32 x i8> @test_mask_adds_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
6862 ; X86-LABEL: test_mask_adds_epu8_rrk_256:
6863 ; X86: # %bb.0:
6864 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
6865 ; X86-NEXT: vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
6866 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6867 ; X86-NEXT: retl # encoding: [0xc3]
6868 ;
6869 ; X64-LABEL: test_mask_adds_epu8_rrk_256:
6870 ; X64: # %bb.0:
6871 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6872 ; X64-NEXT: vpaddusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0xd1]
6873 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
6874 ; X64-NEXT: retq # encoding: [0xc3]
6875 %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)
6876 ret <32 x i8> %res
6877 }
6879 define <32 x i8> @test_mask_adds_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
6880 ; X86-LABEL: test_mask_adds_epu8_rrkz_256:
6881 ; X86: # %bb.0:
6882 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
6883 ; X86-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
6884 ; X86-NEXT: retl # encoding: [0xc3]
6885 ;
6886 ; X64-LABEL: test_mask_adds_epu8_rrkz_256:
6887 ; X64: # %bb.0:
6888 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6889 ; X64-NEXT: vpaddusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0xc1]
6890 ; X64-NEXT: retq # encoding: [0xc3]
6891 %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)
6892 ret <32 x i8> %res
6893 }
6895 define <32 x i8> @test_mask_adds_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
6896 ; X86-LABEL: test_mask_adds_epu8_rm_256:
6897 ; X86: # %bb.0:
6898 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6899 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x00]
6900 ; X86-NEXT: retl # encoding: [0xc3]
6901 ;
6902 ; X64-LABEL: test_mask_adds_epu8_rm_256:
6903 ; X64: # %bb.0:
6904 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xdc,0x07]
6905 ; X64-NEXT: retq # encoding: [0xc3]
6906 %b = load <32 x i8>, <32 x i8>* %ptr_b
6907 %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)
6908 ret <32 x i8> %res
6909 }
6911 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) {
6912 ; X86-LABEL: test_mask_adds_epu8_rmk_256:
6913 ; X86: # %bb.0:
6914 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6915 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
6916 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x08]
6917 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6918 ; X86-NEXT: retl # encoding: [0xc3]
6919 ;
6920 ; X64-LABEL: test_mask_adds_epu8_rmk_256:
6921 ; X64: # %bb.0:
6922 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6923 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xdc,0x0f]
6924 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
6925 ; X64-NEXT: retq # encoding: [0xc3]
6926 %b = load <32 x i8>, <32 x i8>* %ptr_b
6927 %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)
6928 ret <32 x i8> %res
6929 }
6931 define <32 x i8> @test_mask_adds_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
6932 ; X86-LABEL: test_mask_adds_epu8_rmkz_256:
6933 ; X86: # %bb.0:
6934 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6935 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
6936 ; X86-NEXT: vpaddusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x00]
6937 ; X86-NEXT: retl # encoding: [0xc3]
6938 ;
6939 ; X64-LABEL: test_mask_adds_epu8_rmkz_256:
6940 ; X64: # %bb.0:
6941 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
6942 ; X64-NEXT: vpaddusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xdc,0x07]
6943 ; X64-NEXT: retq # encoding: [0xc3]
6944 %b = load <32 x i8>, <32 x i8>* %ptr_b
6945 %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)
6946 ret <32 x i8> %res
6947 }
6949 declare <32 x i8> @llvm.x86.avx512.mask.paddus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
6951 define <16 x i8> @test_mask_subs_epu8_rr_128(<16 x i8> %a, <16 x i8> %b) {
6952 ; CHECK-LABEL: test_mask_subs_epu8_rr_128:
6953 ; CHECK: # %bb.0:
6954 ; CHECK-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0xc1]
6955 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
6956 %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)
6957 ret <16 x i8> %res
6958 }
6960 define <16 x i8> @test_mask_subs_epu8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
6961 ; X86-LABEL: test_mask_subs_epu8_rrk_128:
6962 ; X86: # %bb.0:
6963 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6964 ; X86-NEXT: vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
6965 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6966 ; X86-NEXT: retl # encoding: [0xc3]
6967 ;
6968 ; X64-LABEL: test_mask_subs_epu8_rrk_128:
6969 ; X64: # %bb.0:
6970 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6971 ; X64-NEXT: vpsubusb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0xd1]
6972 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
6973 ; X64-NEXT: retq # encoding: [0xc3]
6974 %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)
6975 ret <16 x i8> %res
6976 }
6978 define <16 x i8> @test_mask_subs_epu8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
6979 ; X86-LABEL: test_mask_subs_epu8_rrkz_128:
6980 ; X86: # %bb.0:
6981 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
6982 ; X86-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
6983 ; X86-NEXT: retl # encoding: [0xc3]
6984 ;
6985 ; X64-LABEL: test_mask_subs_epu8_rrkz_128:
6986 ; X64: # %bb.0:
6987 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
6988 ; X64-NEXT: vpsubusb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0xc1]
6989 ; X64-NEXT: retq # encoding: [0xc3]
6990 %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)
6991 ret <16 x i8> %res
6992 }
6994 define <16 x i8> @test_mask_subs_epu8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
6995 ; X86-LABEL: test_mask_subs_epu8_rm_128:
6996 ; X86: # %bb.0:
6997 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
6998 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x00]
6999 ; X86-NEXT: retl # encoding: [0xc3]
7000 ;
7001 ; X64-LABEL: test_mask_subs_epu8_rm_128:
7002 ; X64: # %bb.0:
7003 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xd8,0x07]
7004 ; X64-NEXT: retq # encoding: [0xc3]
7005 %b = load <16 x i8>, <16 x i8>* %ptr_b
7006 %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)
7007 ret <16 x i8> %res
7008 }
7010 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) {
7011 ; X86-LABEL: test_mask_subs_epu8_rmk_128:
7012 ; X86: # %bb.0:
7013 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7014 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7015 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x08]
7016 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7017 ; X86-NEXT: retl # encoding: [0xc3]
7018 ;
7019 ; X64-LABEL: test_mask_subs_epu8_rmk_128:
7020 ; X64: # %bb.0:
7021 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7022 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xd8,0x0f]
7023 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7024 ; X64-NEXT: retq # encoding: [0xc3]
7025 %b = load <16 x i8>, <16 x i8>* %ptr_b
7026 %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)
7027 ret <16 x i8> %res
7028 }
7030 define <16 x i8> @test_mask_subs_epu8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
7031 ; X86-LABEL: test_mask_subs_epu8_rmkz_128:
7032 ; X86: # %bb.0:
7033 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7034 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7035 ; X86-NEXT: vpsubusb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x00]
7036 ; X86-NEXT: retl # encoding: [0xc3]
7037 ;
7038 ; X64-LABEL: test_mask_subs_epu8_rmkz_128:
7039 ; X64: # %bb.0:
7040 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7041 ; X64-NEXT: vpsubusb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xd8,0x07]
7042 ; X64-NEXT: retq # encoding: [0xc3]
7043 %b = load <16 x i8>, <16 x i8>* %ptr_b
7044 %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)
7045 ret <16 x i8> %res
7046 }
7048 declare <16 x i8> @llvm.x86.avx512.mask.psubus.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
7050 define <32 x i8> @test_mask_subs_epu8_rr_256(<32 x i8> %a, <32 x i8> %b) {
7051 ; CHECK-LABEL: test_mask_subs_epu8_rr_256:
7052 ; CHECK: # %bb.0:
7053 ; CHECK-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0xc1]
7054 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7055 %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)
7056 ret <32 x i8> %res
7057 }
7059 define <32 x i8> @test_mask_subs_epu8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
7060 ; X86-LABEL: test_mask_subs_epu8_rrk_256:
7061 ; X86: # %bb.0:
7062 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7063 ; X86-NEXT: vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
7064 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7065 ; X86-NEXT: retl # encoding: [0xc3]
7066 ;
7067 ; X64-LABEL: test_mask_subs_epu8_rrk_256:
7068 ; X64: # %bb.0:
7069 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7070 ; X64-NEXT: vpsubusb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0xd1]
7071 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7072 ; X64-NEXT: retq # encoding: [0xc3]
7073 %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)
7074 ret <32 x i8> %res
7075 }
7077 define <32 x i8> @test_mask_subs_epu8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
7078 ; X86-LABEL: test_mask_subs_epu8_rrkz_256:
7079 ; X86: # %bb.0:
7080 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7081 ; X86-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
7082 ; X86-NEXT: retl # encoding: [0xc3]
7083 ;
7084 ; X64-LABEL: test_mask_subs_epu8_rrkz_256:
7085 ; X64: # %bb.0:
7086 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7087 ; X64-NEXT: vpsubusb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0xc1]
7088 ; X64-NEXT: retq # encoding: [0xc3]
7089 %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)
7090 ret <32 x i8> %res
7091 }
7093 define <32 x i8> @test_mask_subs_epu8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
7094 ; X86-LABEL: test_mask_subs_epu8_rm_256:
7095 ; X86: # %bb.0:
7096 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7097 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x00]
7098 ; X86-NEXT: retl # encoding: [0xc3]
7099 ;
7100 ; X64-LABEL: test_mask_subs_epu8_rm_256:
7101 ; X64: # %bb.0:
7102 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xd8,0x07]
7103 ; X64-NEXT: retq # encoding: [0xc3]
7104 %b = load <32 x i8>, <32 x i8>* %ptr_b
7105 %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)
7106 ret <32 x i8> %res
7107 }
7109 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) {
7110 ; X86-LABEL: test_mask_subs_epu8_rmk_256:
7111 ; X86: # %bb.0:
7112 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7113 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7114 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x08]
7115 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7116 ; X86-NEXT: retl # encoding: [0xc3]
7117 ;
7118 ; X64-LABEL: test_mask_subs_epu8_rmk_256:
7119 ; X64: # %bb.0:
7120 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7121 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xd8,0x0f]
7122 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7123 ; X64-NEXT: retq # encoding: [0xc3]
7124 %b = load <32 x i8>, <32 x i8>* %ptr_b
7125 %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)
7126 ret <32 x i8> %res
7127 }
7129 define <32 x i8> @test_mask_subs_epu8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
7130 ; X86-LABEL: test_mask_subs_epu8_rmkz_256:
7131 ; X86: # %bb.0:
7132 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7133 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7134 ; X86-NEXT: vpsubusb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x00]
7135 ; X86-NEXT: retl # encoding: [0xc3]
7136 ;
7137 ; X64-LABEL: test_mask_subs_epu8_rmkz_256:
7138 ; X64: # %bb.0:
7139 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7140 ; X64-NEXT: vpsubusb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xd8,0x07]
7141 ; X64-NEXT: retq # encoding: [0xc3]
7142 %b = load <32 x i8>, <32 x i8>* %ptr_b
7143 %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)
7144 ret <32 x i8> %res
7145 }
7147 declare <32 x i8> @llvm.x86.avx512.mask.psubus.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
7149 define <8 x i16> @test_mask_adds_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
7150 ; CHECK-LABEL: test_mask_adds_epi16_rr_128:
7151 ; CHECK: # %bb.0:
7152 ; CHECK-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0xc1]
7153 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7154 %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)
7155 ret <8 x i16> %res
7156 }
7158 define <8 x i16> @test_mask_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7159 ; X86-LABEL: test_mask_adds_epi16_rrk_128:
7160 ; X86: # %bb.0:
7161 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7162 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7163 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7164 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7165 ; X86-NEXT: retl # encoding: [0xc3]
7166 ;
7167 ; X64-LABEL: test_mask_adds_epi16_rrk_128:
7168 ; X64: # %bb.0:
7169 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7170 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7171 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7172 ; X64-NEXT: retq # encoding: [0xc3]
7173 %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)
7174 ret <8 x i16> %res
7175 }
7177 define <8 x i16> @test_mask_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7178 ; X86-LABEL: test_mask_adds_epi16_rrkz_128:
7179 ; X86: # %bb.0:
7180 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7181 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7182 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7183 ; X86-NEXT: retl # encoding: [0xc3]
7184 ;
7185 ; X64-LABEL: test_mask_adds_epi16_rrkz_128:
7186 ; X64: # %bb.0:
7187 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7188 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7189 ; X64-NEXT: retq # encoding: [0xc3]
7190 %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)
7191 ret <8 x i16> %res
7192 }
7194 define <8 x i16> @test_mask_adds_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
7195 ; X86-LABEL: test_mask_adds_epi16_rm_128:
7196 ; X86: # %bb.0:
7197 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7198 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x00]
7199 ; X86-NEXT: retl # encoding: [0xc3]
7200 ;
7201 ; X64-LABEL: test_mask_adds_epi16_rm_128:
7202 ; X64: # %bb.0:
7203 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xed,0x07]
7204 ; X64-NEXT: retq # encoding: [0xc3]
7205 %b = load <8 x i16>, <8 x i16>* %ptr_b
7206 %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)
7207 ret <8 x i16> %res
7208 }
7210 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) {
7211 ; X86-LABEL: test_mask_adds_epi16_rmk_128:
7212 ; X86: # %bb.0:
7213 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7214 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7215 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7216 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
7217 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7218 ; X86-NEXT: retl # encoding: [0xc3]
7219 ;
7220 ; X64-LABEL: test_mask_adds_epi16_rmk_128:
7221 ; X64: # %bb.0:
7222 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7223 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
7224 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7225 ; X64-NEXT: retq # encoding: [0xc3]
7226 %b = load <8 x i16>, <8 x i16>* %ptr_b
7227 %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)
7228 ret <8 x i16> %res
7229 }
7231 define <8 x i16> @test_mask_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7232 ; X86-LABEL: test_mask_adds_epi16_rmkz_128:
7233 ; X86: # %bb.0:
7234 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7235 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7236 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7237 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
7238 ; X86-NEXT: retl # encoding: [0xc3]
7239 ;
7240 ; X64-LABEL: test_mask_adds_epi16_rmkz_128:
7241 ; X64: # %bb.0:
7242 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7243 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
7244 ; X64-NEXT: retq # encoding: [0xc3]
7245 %b = load <8 x i16>, <8 x i16>* %ptr_b
7246 %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)
7247 ret <8 x i16> %res
7248 }
7250 declare <8 x i16> @llvm.x86.avx512.mask.padds.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
7252 define <16 x i16> @test_mask_adds_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
7253 ; CHECK-LABEL: test_mask_adds_epi16_rr_256:
7254 ; CHECK: # %bb.0:
7255 ; CHECK-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0xc1]
7256 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7257 %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)
7258 ret <16 x i16> %res
7259 }
7261 define <16 x i16> @test_mask_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7262 ; X86-LABEL: test_mask_adds_epi16_rrk_256:
7263 ; X86: # %bb.0:
7264 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7265 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7266 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7267 ; X86-NEXT: retl # encoding: [0xc3]
7268 ;
7269 ; X64-LABEL: test_mask_adds_epi16_rrk_256:
7270 ; X64: # %bb.0:
7271 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7272 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7273 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7274 ; X64-NEXT: retq # encoding: [0xc3]
7275 %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)
7276 ret <16 x i16> %res
7277 }
7279 define <16 x i16> @test_mask_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7280 ; X86-LABEL: test_mask_adds_epi16_rrkz_256:
7281 ; X86: # %bb.0:
7282 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7283 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7284 ; X86-NEXT: retl # encoding: [0xc3]
7285 ;
7286 ; X64-LABEL: test_mask_adds_epi16_rrkz_256:
7287 ; X64: # %bb.0:
7288 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7289 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7290 ; X64-NEXT: retq # encoding: [0xc3]
7291 %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)
7292 ret <16 x i16> %res
7293 }
7295 define <16 x i16> @test_mask_adds_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
7296 ; X86-LABEL: test_mask_adds_epi16_rm_256:
7297 ; X86: # %bb.0:
7298 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7299 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x00]
7300 ; X86-NEXT: retl # encoding: [0xc3]
7301 ;
7302 ; X64-LABEL: test_mask_adds_epi16_rm_256:
7303 ; X64: # %bb.0:
7304 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xed,0x07]
7305 ; X64-NEXT: retq # encoding: [0xc3]
7306 %b = load <16 x i16>, <16 x i16>* %ptr_b
7307 %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)
7308 ret <16 x i16> %res
7309 }
7311 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) {
7312 ; X86-LABEL: test_mask_adds_epi16_rmk_256:
7313 ; X86: # %bb.0:
7314 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7315 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7316 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
7317 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7318 ; X86-NEXT: retl # encoding: [0xc3]
7319 ;
7320 ; X64-LABEL: test_mask_adds_epi16_rmk_256:
7321 ; X64: # %bb.0:
7322 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7323 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
7324 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7325 ; X64-NEXT: retq # encoding: [0xc3]
7326 %b = load <16 x i16>, <16 x i16>* %ptr_b
7327 %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)
7328 ret <16 x i16> %res
7329 }
7331 define <16 x i16> @test_mask_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7332 ; X86-LABEL: test_mask_adds_epi16_rmkz_256:
7333 ; X86: # %bb.0:
7334 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7335 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7336 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
7337 ; X86-NEXT: retl # encoding: [0xc3]
7338 ;
7339 ; X64-LABEL: test_mask_adds_epi16_rmkz_256:
7340 ; X64: # %bb.0:
7341 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7342 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
7343 ; X64-NEXT: retq # encoding: [0xc3]
7344 %b = load <16 x i16>, <16 x i16>* %ptr_b
7345 %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)
7346 ret <16 x i16> %res
7347 }
7349 declare <16 x i16> @llvm.x86.avx512.mask.padds.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
7351 declare <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16>, <8 x i16>) #0
7353 define <8 x i16> @test_test_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7354 ; X86-LABEL: test_test_subs_epi16_rrk_128:
7355 ; X86: # %bb.0:
7356 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7357 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7358 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7359 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7360 ; X86-NEXT: retl # encoding: [0xc3]
7361 ;
7362 ; X64-LABEL: test_test_subs_epi16_rrk_128:
7363 ; X64: # %bb.0:
7364 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7365 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7366 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7367 ; X64-NEXT: retq # encoding: [0xc3]
7368 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7369 %2 = bitcast i8 %mask to <8 x i1>
7370 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7371 ret <8 x i16> %3
7372 }
7374 define <8 x i16> @test_test_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7375 ; X86-LABEL: test_test_subs_epi16_rrkz_128:
7376 ; X86: # %bb.0:
7377 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7378 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7379 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7380 ; X86-NEXT: retl # encoding: [0xc3]
7381 ;
7382 ; X64-LABEL: test_test_subs_epi16_rrkz_128:
7383 ; X64: # %bb.0:
7384 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7385 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7386 ; X64-NEXT: retq # encoding: [0xc3]
7387 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7388 %2 = bitcast i8 %mask to <8 x i1>
7389 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7390 ret <8 x i16> %3
7391 }
7393 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) {
7394 ; X86-LABEL: test_test_subs_epi16_rmk_128:
7395 ; X86: # %bb.0:
7396 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7397 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7398 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7399 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
7400 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7401 ; X86-NEXT: retl # encoding: [0xc3]
7402 ;
7403 ; X64-LABEL: test_test_subs_epi16_rmk_128:
7404 ; X64: # %bb.0:
7405 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7406 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
7407 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7408 ; X64-NEXT: retq # encoding: [0xc3]
7409 %b = load <8 x i16>, <8 x i16>* %ptr_b
7410 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7411 %2 = bitcast i8 %mask to <8 x i1>
7412 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7413 ret <8 x i16> %3
7414 }
7416 define <8 x i16> @test_test_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7417 ; X86-LABEL: test_test_subs_epi16_rmkz_128:
7418 ; X86: # %bb.0:
7419 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7420 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7421 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7422 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
7423 ; X86-NEXT: retl # encoding: [0xc3]
7424 ;
7425 ; X64-LABEL: test_test_subs_epi16_rmkz_128:
7426 ; X64: # %bb.0:
7427 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7428 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
7429 ; X64-NEXT: retq # encoding: [0xc3]
7430 %b = load <8 x i16>, <8 x i16>* %ptr_b
7431 %1 = call <8 x i16> @llvm.x86.sse2.psubs.w(<8 x i16> %a, <8 x i16> %b)
7432 %2 = bitcast i8 %mask to <8 x i1>
7433 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7434 ret <8 x i16> %3
7435 }
7437 declare <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16>, <16 x i16>) #0
7439 define <16 x i16> @test_test_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7440 ; X86-LABEL: test_test_subs_epi16_rrk_256:
7441 ; X86: # %bb.0:
7442 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7443 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7444 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7445 ; X86-NEXT: retl # encoding: [0xc3]
7446 ;
7447 ; X64-LABEL: test_test_subs_epi16_rrk_256:
7448 ; X64: # %bb.0:
7449 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7450 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7451 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7452 ; X64-NEXT: retq # encoding: [0xc3]
7453 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7454 %2 = bitcast i16 %mask to <16 x i1>
7455 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7456 ret <16 x i16> %3
7457 }
7459 define <16 x i16> @test_test_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7460 ; X86-LABEL: test_test_subs_epi16_rrkz_256:
7461 ; X86: # %bb.0:
7462 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7463 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7464 ; X86-NEXT: retl # encoding: [0xc3]
7465 ;
7466 ; X64-LABEL: test_test_subs_epi16_rrkz_256:
7467 ; X64: # %bb.0:
7468 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7469 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7470 ; X64-NEXT: retq # encoding: [0xc3]
7471 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7472 %2 = bitcast i16 %mask to <16 x i1>
7473 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
7474 ret <16 x i16> %3
7475 }
7477 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) {
7478 ; X86-LABEL: test_test_subs_epi16_rmk_256:
7479 ; X86: # %bb.0:
7480 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7481 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7482 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
7483 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7484 ; X86-NEXT: retl # encoding: [0xc3]
7485 ;
7486 ; X64-LABEL: test_test_subs_epi16_rmk_256:
7487 ; X64: # %bb.0:
7488 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7489 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
7490 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7491 ; X64-NEXT: retq # encoding: [0xc3]
7492 %b = load <16 x i16>, <16 x i16>* %ptr_b
7493 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7494 %2 = bitcast i16 %mask to <16 x i1>
7495 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7496 ret <16 x i16> %3
7497 }
7499 define <16 x i16> @test_test_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7500 ; X86-LABEL: test_test_subs_epi16_rmkz_256:
7501 ; X86: # %bb.0:
7502 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7503 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7504 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
7505 ; X86-NEXT: retl # encoding: [0xc3]
7506 ;
7507 ; X64-LABEL: test_test_subs_epi16_rmkz_256:
7508 ; X64: # %bb.0:
7509 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7510 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
7511 ; X64-NEXT: retq # encoding: [0xc3]
7512 %b = load <16 x i16>, <16 x i16>* %ptr_b
7513 %1 = call <16 x i16> @llvm.x86.avx2.psubs.w(<16 x i16> %a, <16 x i16> %b)
7514 %2 = bitcast i16 %mask to <16 x i1>
7515 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
7516 ret <16 x i16> %3
7517 }
7519 declare <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8>, <16 x i8>) #0
7521 define <16 x i8> @test_test_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
7522 ; X86-LABEL: test_test_subs_epi8_rrk_128:
7523 ; X86: # %bb.0:
7524 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7525 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
7526 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7527 ; X86-NEXT: retl # encoding: [0xc3]
7528 ;
7529 ; X64-LABEL: test_test_subs_epi8_rrk_128:
7530 ; X64: # %bb.0:
7531 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7532 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
7533 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7534 ; X64-NEXT: retq # encoding: [0xc3]
7535 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7536 %2 = bitcast i16 %mask to <16 x i1>
7537 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
7538 ret <16 x i8> %3
7539 }
7541 define <16 x i8> @test_test_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
7542 ; X86-LABEL: test_test_subs_epi8_rrkz_128:
7543 ; X86: # %bb.0:
7544 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7545 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
7546 ; X86-NEXT: retl # encoding: [0xc3]
7547 ;
7548 ; X64-LABEL: test_test_subs_epi8_rrkz_128:
7549 ; X64: # %bb.0:
7550 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7551 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
7552 ; X64-NEXT: retq # encoding: [0xc3]
7553 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7554 %2 = bitcast i16 %mask to <16 x i1>
7555 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
7556 ret <16 x i8> %3
7557 }
7559 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) {
7560 ; X86-LABEL: test_test_subs_epi8_rmk_128:
7561 ; X86: # %bb.0:
7562 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7563 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7564 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
7565 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7566 ; X86-NEXT: retl # encoding: [0xc3]
7567 ;
7568 ; X64-LABEL: test_test_subs_epi8_rmk_128:
7569 ; X64: # %bb.0:
7570 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7571 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
7572 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7573 ; X64-NEXT: retq # encoding: [0xc3]
7574 %b = load <16 x i8>, <16 x i8>* %ptr_b
7575 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7576 %2 = bitcast i16 %mask to <16 x i1>
7577 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
7578 ret <16 x i8> %3
7579 }
7581 define <16 x i8> @test_test_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
7582 ; X86-LABEL: test_test_subs_epi8_rmkz_128:
7583 ; X86: # %bb.0:
7584 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7585 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7586 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
7587 ; X86-NEXT: retl # encoding: [0xc3]
7588 ;
7589 ; X64-LABEL: test_test_subs_epi8_rmkz_128:
7590 ; X64: # %bb.0:
7591 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7592 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
7593 ; X64-NEXT: retq # encoding: [0xc3]
7594 %b = load <16 x i8>, <16 x i8>* %ptr_b
7595 %1 = call <16 x i8> @llvm.x86.sse2.psubs.b(<16 x i8> %a, <16 x i8> %b)
7596 %2 = bitcast i16 %mask to <16 x i1>
7597 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
7598 ret <16 x i8> %3
7599 }
7601 declare <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8>, <32 x i8>) #0
7603 define <32 x i8> @test_test_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
7604 ; X86-LABEL: test_test_subs_epi8_rrk_256:
7605 ; X86: # %bb.0:
7606 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7607 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
7608 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7609 ; X86-NEXT: retl # encoding: [0xc3]
7610 ;
7611 ; X64-LABEL: test_test_subs_epi8_rrk_256:
7612 ; X64: # %bb.0:
7613 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7614 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
7615 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7616 ; X64-NEXT: retq # encoding: [0xc3]
7617 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7618 %2 = bitcast i32 %mask to <32 x i1>
7619 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
7620 ret <32 x i8> %3
7621 }
7623 define <32 x i8> @test_test_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
7624 ; X86-LABEL: test_test_subs_epi8_rrkz_256:
7625 ; X86: # %bb.0:
7626 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
7627 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
7628 ; X86-NEXT: retl # encoding: [0xc3]
7629 ;
7630 ; X64-LABEL: test_test_subs_epi8_rrkz_256:
7631 ; X64: # %bb.0:
7632 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7633 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
7634 ; X64-NEXT: retq # encoding: [0xc3]
7635 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7636 %2 = bitcast i32 %mask to <32 x i1>
7637 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
7638 ret <32 x i8> %3
7639 }
7641 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) {
7642 ; X86-LABEL: test_test_subs_epi8_rmk_256:
7643 ; X86: # %bb.0:
7644 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7645 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7646 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
7647 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7648 ; X86-NEXT: retl # encoding: [0xc3]
7649 ;
7650 ; X64-LABEL: test_test_subs_epi8_rmk_256:
7651 ; X64: # %bb.0:
7652 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7653 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
7654 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7655 ; X64-NEXT: retq # encoding: [0xc3]
7656 %b = load <32 x i8>, <32 x i8>* %ptr_b
7657 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7658 %2 = bitcast i32 %mask to <32 x i1>
7659 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
7660 ret <32 x i8> %3
7661 }
7663 define <32 x i8> @test_test_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
7664 ; X86-LABEL: test_test_subs_epi8_rmkz_256:
7665 ; X86: # %bb.0:
7666 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7667 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
7668 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
7669 ; X86-NEXT: retl # encoding: [0xc3]
7670 ;
7671 ; X64-LABEL: test_test_subs_epi8_rmkz_256:
7672 ; X64: # %bb.0:
7673 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7674 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
7675 ; X64-NEXT: retq # encoding: [0xc3]
7676 %b = load <32 x i8>, <32 x i8>* %ptr_b
7677 %1 = call <32 x i8> @llvm.x86.avx2.psubs.b(<32 x i8> %a, <32 x i8> %b)
7678 %2 = bitcast i32 %mask to <32 x i1>
7679 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
7680 ret <32 x i8> %3
7681 }
7683 define <8 x i16> @test_mask_subs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
7684 ; CHECK-LABEL: test_mask_subs_epi16_rr_128:
7685 ; CHECK: # %bb.0:
7686 ; CHECK-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0xc1]
7687 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7688 %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)
7689 ret <8 x i16> %res
7690 }
7692 define <8 x i16> @test_mask_subs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7693 ; X86-LABEL: test_mask_subs_epi16_rrk_128:
7694 ; X86: # %bb.0:
7695 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7696 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7697 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7698 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7699 ; X86-NEXT: retl # encoding: [0xc3]
7700 ;
7701 ; X64-LABEL: test_mask_subs_epi16_rrk_128:
7702 ; X64: # %bb.0:
7703 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7704 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0xd1]
7705 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7706 ; X64-NEXT: retq # encoding: [0xc3]
7707 %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)
7708 ret <8 x i16> %res
7709 }
7711 define <8 x i16> @test_mask_subs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7712 ; X86-LABEL: test_mask_subs_epi16_rrkz_128:
7713 ; X86: # %bb.0:
7714 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7715 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7716 ; X86-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7717 ; X86-NEXT: retl # encoding: [0xc3]
7718 ;
7719 ; X64-LABEL: test_mask_subs_epi16_rrkz_128:
7720 ; X64: # %bb.0:
7721 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7722 ; X64-NEXT: vpsubsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0xc1]
7723 ; X64-NEXT: retq # encoding: [0xc3]
7724 %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)
7725 ret <8 x i16> %res
7726 }
7728 define <8 x i16> @test_mask_subs_epi16_rm_128(<8 x i16> %a, <8 x i16>* %ptr_b) {
7729 ; X86-LABEL: test_mask_subs_epi16_rm_128:
7730 ; X86: # %bb.0:
7731 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7732 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x00]
7733 ; X86-NEXT: retl # encoding: [0xc3]
7734 ;
7735 ; X64-LABEL: test_mask_subs_epi16_rm_128:
7736 ; X64: # %bb.0:
7737 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe9,0x07]
7738 ; X64-NEXT: retq # encoding: [0xc3]
7739 %b = load <8 x i16>, <8 x i16>* %ptr_b
7740 %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)
7741 ret <8 x i16> %res
7742 }
7744 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) {
7745 ; X86-LABEL: test_mask_subs_epi16_rmk_128:
7746 ; X86: # %bb.0:
7747 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7748 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7749 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7750 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x08]
7751 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7752 ; X86-NEXT: retl # encoding: [0xc3]
7753 ;
7754 ; X64-LABEL: test_mask_subs_epi16_rmk_128:
7755 ; X64: # %bb.0:
7756 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7757 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe9,0x0f]
7758 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7759 ; X64-NEXT: retq # encoding: [0xc3]
7760 %b = load <8 x i16>, <8 x i16>* %ptr_b
7761 %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)
7762 ret <8 x i16> %res
7763 }
7765 define <8 x i16> @test_mask_subs_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7766 ; X86-LABEL: test_mask_subs_epi16_rmkz_128:
7767 ; X86: # %bb.0:
7768 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7769 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7770 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7771 ; X86-NEXT: vpsubsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x00]
7772 ; X86-NEXT: retl # encoding: [0xc3]
7773 ;
7774 ; X64-LABEL: test_mask_subs_epi16_rmkz_128:
7775 ; X64: # %bb.0:
7776 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7777 ; X64-NEXT: vpsubsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe9,0x07]
7778 ; X64-NEXT: retq # encoding: [0xc3]
7779 %b = load <8 x i16>, <8 x i16>* %ptr_b
7780 %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)
7781 ret <8 x i16> %res
7782 }
7784 declare <8 x i16> @llvm.x86.avx512.mask.psubs.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8)
7786 define <16 x i16> @test_mask_subs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
7787 ; CHECK-LABEL: test_mask_subs_epi16_rr_256:
7788 ; CHECK: # %bb.0:
7789 ; CHECK-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0xc1]
7790 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
7791 %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)
7792 ret <16 x i16> %res
7793 }
7795 define <16 x i16> @test_mask_subs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7796 ; X86-LABEL: test_mask_subs_epi16_rrk_256:
7797 ; X86: # %bb.0:
7798 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7799 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7800 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7801 ; X86-NEXT: retl # encoding: [0xc3]
7802 ;
7803 ; X64-LABEL: test_mask_subs_epi16_rrk_256:
7804 ; X64: # %bb.0:
7805 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7806 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0xd1]
7807 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7808 ; X64-NEXT: retq # encoding: [0xc3]
7809 %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)
7810 ret <16 x i16> %res
7811 }
7813 define <16 x i16> @test_mask_subs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7814 ; X86-LABEL: test_mask_subs_epi16_rrkz_256:
7815 ; X86: # %bb.0:
7816 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7817 ; X86-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7818 ; X86-NEXT: retl # encoding: [0xc3]
7819 ;
7820 ; X64-LABEL: test_mask_subs_epi16_rrkz_256:
7821 ; X64: # %bb.0:
7822 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7823 ; X64-NEXT: vpsubsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0xc1]
7824 ; X64-NEXT: retq # encoding: [0xc3]
7825 %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)
7826 ret <16 x i16> %res
7827 }
7829 define <16 x i16> @test_mask_subs_epi16_rm_256(<16 x i16> %a, <16 x i16>* %ptr_b) {
7830 ; X86-LABEL: test_mask_subs_epi16_rm_256:
7831 ; X86: # %bb.0:
7832 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7833 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x00]
7834 ; X86-NEXT: retl # encoding: [0xc3]
7835 ;
7836 ; X64-LABEL: test_mask_subs_epi16_rm_256:
7837 ; X64: # %bb.0:
7838 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe9,0x07]
7839 ; X64-NEXT: retq # encoding: [0xc3]
7840 %b = load <16 x i16>, <16 x i16>* %ptr_b
7841 %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)
7842 ret <16 x i16> %res
7843 }
7845 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) {
7846 ; X86-LABEL: test_mask_subs_epi16_rmk_256:
7847 ; X86: # %bb.0:
7848 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7849 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7850 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x08]
7851 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7852 ; X86-NEXT: retl # encoding: [0xc3]
7853 ;
7854 ; X64-LABEL: test_mask_subs_epi16_rmk_256:
7855 ; X64: # %bb.0:
7856 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7857 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe9,0x0f]
7858 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
7859 ; X64-NEXT: retq # encoding: [0xc3]
7860 %b = load <16 x i16>, <16 x i16>* %ptr_b
7861 %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)
7862 ret <16 x i16> %res
7863 }
7865 define <16 x i16> @test_mask_subs_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
7866 ; X86-LABEL: test_mask_subs_epi16_rmkz_256:
7867 ; X86: # %bb.0:
7868 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7869 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
7870 ; X86-NEXT: vpsubsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x00]
7871 ; X86-NEXT: retl # encoding: [0xc3]
7872 ;
7873 ; X64-LABEL: test_mask_subs_epi16_rmkz_256:
7874 ; X64: # %bb.0:
7875 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7876 ; X64-NEXT: vpsubsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe9,0x07]
7877 ; X64-NEXT: retq # encoding: [0xc3]
7878 %b = load <16 x i16>, <16 x i16>* %ptr_b
7879 %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)
7880 ret <16 x i16> %res
7881 }
7883 declare <16 x i16> @llvm.x86.avx512.mask.psubs.w.256(<16 x i16>, <16 x i16>, <16 x i16>, i16)
7885 declare <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16>, <8 x i16>) #0
7887 define <8 x i16> @test_adds_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <8 x i16> %passThru, i8 %mask) {
7888 ; X86-LABEL: test_adds_epi16_rrk_128:
7889 ; X86: # %bb.0:
7890 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7891 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7892 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7893 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7894 ; X86-NEXT: retl # encoding: [0xc3]
7895 ;
7896 ; X64-LABEL: test_adds_epi16_rrk_128:
7897 ; X64: # %bb.0:
7898 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7899 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0xd1]
7900 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
7901 ; X64-NEXT: retq # encoding: [0xc3]
7902 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7903 %2 = bitcast i8 %mask to <8 x i1>
7904 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7905 ret <8 x i16> %3
7906 }
7908 define <8 x i16> @test_adds_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i8 %mask) {
7909 ; X86-LABEL: test_adds_epi16_rrkz_128:
7910 ; X86: # %bb.0:
7911 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
7912 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
7913 ; X86-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7914 ; X86-NEXT: retl # encoding: [0xc3]
7915 ;
7916 ; X64-LABEL: test_adds_epi16_rrkz_128:
7917 ; X64: # %bb.0:
7918 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7919 ; X64-NEXT: vpaddsw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0xc1]
7920 ; X64-NEXT: retq # encoding: [0xc3]
7921 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7922 %2 = bitcast i8 %mask to <8 x i1>
7923 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7924 ret <8 x i16> %3
7925 }
7927 define <8 x i16> @test_adds_epi16_rmk_128(<8 x i16> %a, <8 x i16>* %ptr_b, <8 x i16> %passThru, i8 %mask) {
7928 ; X86-LABEL: test_adds_epi16_rmk_128:
7929 ; X86: # %bb.0:
7930 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7931 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7932 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7933 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x08]
7934 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7935 ; X86-NEXT: retl # encoding: [0xc3]
7936 ;
7937 ; X64-LABEL: test_adds_epi16_rmk_128:
7938 ; X64: # %bb.0:
7939 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7940 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xed,0x0f]
7941 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
7942 ; X64-NEXT: retq # encoding: [0xc3]
7943 %b = load <8 x i16>, <8 x i16>* %ptr_b
7944 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7945 %2 = bitcast i8 %mask to <8 x i1>
7946 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
7947 ret <8 x i16> %3
7948 }
7950 define <8 x i16> @test_adds_epi16_rmkz_128(<8 x i16> %a, <8 x i16>* %ptr_b, i8 %mask) {
7951 ; X86-LABEL: test_adds_epi16_rmkz_128:
7952 ; X86: # %bb.0:
7953 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
7954 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
7955 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
7956 ; X86-NEXT: vpaddsw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x00]
7957 ; X86-NEXT: retl # encoding: [0xc3]
7958 ;
7959 ; X64-LABEL: test_adds_epi16_rmkz_128:
7960 ; X64: # %bb.0:
7961 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
7962 ; X64-NEXT: vpaddsw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xed,0x07]
7963 ; X64-NEXT: retq # encoding: [0xc3]
7964 %b = load <8 x i16>, <8 x i16>* %ptr_b
7965 %1 = call <8 x i16> @llvm.x86.sse2.padds.w(<8 x i16> %a, <8 x i16> %b)
7966 %2 = bitcast i8 %mask to <8 x i1>
7967 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
7968 ret <8 x i16> %3
7969 }
7971 declare <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16>, <16 x i16>) #0
7973 define <16 x i16> @test_adds_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <16 x i16> %passThru, i16 %mask) {
7974 ; X86-LABEL: test_adds_epi16_rrk_256:
7975 ; X86: # %bb.0:
7976 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7977 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7978 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7979 ; X86-NEXT: retl # encoding: [0xc3]
7980 ;
7981 ; X64-LABEL: test_adds_epi16_rrk_256:
7982 ; X64: # %bb.0:
7983 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
7984 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0xd1]
7985 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
7986 ; X64-NEXT: retq # encoding: [0xc3]
7987 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
7988 %2 = bitcast i16 %mask to <16 x i1>
7989 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
7990 ret <16 x i16> %3
7991 }
7993 define <16 x i16> @test_adds_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i16 %mask) {
7994 ; X86-LABEL: test_adds_epi16_rrkz_256:
7995 ; X86: # %bb.0:
7996 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
7997 ; X86-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
7998 ; X86-NEXT: retl # encoding: [0xc3]
7999 ;
8000 ; X64-LABEL: test_adds_epi16_rrkz_256:
8001 ; X64: # %bb.0:
8002 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8003 ; X64-NEXT: vpaddsw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0xc1]
8004 ; X64-NEXT: retq # encoding: [0xc3]
8005 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8006 %2 = bitcast i16 %mask to <16 x i1>
8007 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
8008 ret <16 x i16> %3
8009 }
8011 define <16 x i16> @test_adds_epi16_rmk_256(<16 x i16> %a, <16 x i16>* %ptr_b, <16 x i16> %passThru, i16 %mask) {
8012 ; X86-LABEL: test_adds_epi16_rmk_256:
8013 ; X86: # %bb.0:
8014 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8015 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8016 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x08]
8017 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8018 ; X86-NEXT: retl # encoding: [0xc3]
8019 ;
8020 ; X64-LABEL: test_adds_epi16_rmk_256:
8021 ; X64: # %bb.0:
8022 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8023 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xed,0x0f]
8024 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8025 ; X64-NEXT: retq # encoding: [0xc3]
8026 %b = load <16 x i16>, <16 x i16>* %ptr_b
8027 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8028 %2 = bitcast i16 %mask to <16 x i1>
8029 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
8030 ret <16 x i16> %3
8031 }
8033 define <16 x i16> @test_adds_epi16_rmkz_256(<16 x i16> %a, <16 x i16>* %ptr_b, i16 %mask) {
8034 ; X86-LABEL: test_adds_epi16_rmkz_256:
8035 ; X86: # %bb.0:
8036 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8037 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8038 ; X86-NEXT: vpaddsw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x00]
8039 ; X86-NEXT: retl # encoding: [0xc3]
8040 ;
8041 ; X64-LABEL: test_adds_epi16_rmkz_256:
8042 ; X64: # %bb.0:
8043 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8044 ; X64-NEXT: vpaddsw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xed,0x07]
8045 ; X64-NEXT: retq # encoding: [0xc3]
8046 %b = load <16 x i16>, <16 x i16>* %ptr_b
8047 %1 = call <16 x i16> @llvm.x86.avx2.padds.w(<16 x i16> %a, <16 x i16> %b)
8048 %2 = bitcast i16 %mask to <16 x i1>
8049 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
8050 ret <16 x i16> %3
8051 }
8053 declare <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8>, <16 x i8>) #0
8055 define <16 x i8> @test_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8056 ; X86-LABEL: test_adds_epi8_rrk_128:
8057 ; X86: # %bb.0:
8058 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8059 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8060 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8061 ; X86-NEXT: retl # encoding: [0xc3]
8062 ;
8063 ; X64-LABEL: test_adds_epi8_rrk_128:
8064 ; X64: # %bb.0:
8065 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8066 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8067 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8068 ; X64-NEXT: retq # encoding: [0xc3]
8069 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8070 %2 = bitcast i16 %mask to <16 x i1>
8071 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
8072 ret <16 x i8> %3
8073 }
8075 define <16 x i8> @test_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8076 ; X86-LABEL: test_adds_epi8_rrkz_128:
8077 ; X86: # %bb.0:
8078 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8079 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8080 ; X86-NEXT: retl # encoding: [0xc3]
8081 ;
8082 ; X64-LABEL: test_adds_epi8_rrkz_128:
8083 ; X64: # %bb.0:
8084 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8085 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8086 ; X64-NEXT: retq # encoding: [0xc3]
8087 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8088 %2 = bitcast i16 %mask to <16 x i1>
8089 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
8090 ret <16 x i8> %3
8091 }
8093 define <16 x i8> @test_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8094 ; X86-LABEL: test_adds_epi8_rm_128:
8095 ; X86: # %bb.0:
8096 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8097 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
8098 ; X86-NEXT: retl # encoding: [0xc3]
8099 ;
8100 ; X64-LABEL: test_adds_epi8_rm_128:
8101 ; X64: # %bb.0:
8102 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
8103 ; X64-NEXT: retq # encoding: [0xc3]
8104 %b = load <16 x i8>, <16 x i8>* %ptr_b
8105 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8106 ret <16 x i8> %1
8107 }
8109 define <16 x i8> @test_adds_epi8_rmk_128(<16 x i8> %a, <16 x i8>* %ptr_b, <16 x i8> %passThru, i16 %mask) {
8110 ; X86-LABEL: test_adds_epi8_rmk_128:
8111 ; X86: # %bb.0:
8112 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8113 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8114 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
8115 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8116 ; X86-NEXT: retl # encoding: [0xc3]
8117 ;
8118 ; X64-LABEL: test_adds_epi8_rmk_128:
8119 ; X64: # %bb.0:
8120 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8121 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
8122 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8123 ; X64-NEXT: retq # encoding: [0xc3]
8124 %b = load <16 x i8>, <16 x i8>* %ptr_b
8125 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8126 %2 = bitcast i16 %mask to <16 x i1>
8127 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
8128 ret <16 x i8> %3
8129 }
8131 define <16 x i8> @test_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8132 ; X86-LABEL: test_adds_epi8_rmkz_128:
8133 ; X86: # %bb.0:
8134 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8135 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8136 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
8137 ; X86-NEXT: retl # encoding: [0xc3]
8138 ;
8139 ; X64-LABEL: test_adds_epi8_rmkz_128:
8140 ; X64: # %bb.0:
8141 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8142 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
8143 ; X64-NEXT: retq # encoding: [0xc3]
8144 %b = load <16 x i8>, <16 x i8>* %ptr_b
8145 %1 = call <16 x i8> @llvm.x86.sse2.padds.b(<16 x i8> %a, <16 x i8> %b)
8146 %2 = bitcast i16 %mask to <16 x i1>
8147 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
8148 ret <16 x i8> %3
8149 }
8151 declare <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8>, <32 x i8>) #0
8153 define <32 x i8> @test_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8154 ; X86-LABEL: test_adds_epi8_rrk_256:
8155 ; X86: # %bb.0:
8156 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8157 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8158 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8159 ; X86-NEXT: retl # encoding: [0xc3]
8160 ;
8161 ; X64-LABEL: test_adds_epi8_rrk_256:
8162 ; X64: # %bb.0:
8163 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8164 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8165 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8166 ; X64-NEXT: retq # encoding: [0xc3]
8167 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8168 %2 = bitcast i32 %mask to <32 x i1>
8169 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
8170 ret <32 x i8> %3
8171 }
8173 define <32 x i8> @test_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8174 ; X86-LABEL: test_adds_epi8_rrkz_256:
8175 ; X86: # %bb.0:
8176 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8177 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8178 ; X86-NEXT: retl # encoding: [0xc3]
8179 ;
8180 ; X64-LABEL: test_adds_epi8_rrkz_256:
8181 ; X64: # %bb.0:
8182 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8183 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8184 ; X64-NEXT: retq # encoding: [0xc3]
8185 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8186 %2 = bitcast i32 %mask to <32 x i1>
8187 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
8188 ret <32 x i8> %3
8189 }
8191 define <32 x i8> @test_adds_epi8_rmk_256(<32 x i8> %a, <32 x i8>* %ptr_b, <32 x i8> %passThru, i32 %mask) {
8192 ; X86-LABEL: test_adds_epi8_rmk_256:
8193 ; X86: # %bb.0:
8194 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8195 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8196 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
8197 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8198 ; X86-NEXT: retl # encoding: [0xc3]
8199 ;
8200 ; X64-LABEL: test_adds_epi8_rmk_256:
8201 ; X64: # %bb.0:
8202 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8203 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
8204 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8205 ; X64-NEXT: retq # encoding: [0xc3]
8206 %b = load <32 x i8>, <32 x i8>* %ptr_b
8207 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8208 %2 = bitcast i32 %mask to <32 x i1>
8209 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
8210 ret <32 x i8> %3
8211 }
8213 define <32 x i8> @test_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8214 ; X86-LABEL: test_adds_epi8_rmkz_256:
8215 ; X86: # %bb.0:
8216 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8217 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8218 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
8219 ; X86-NEXT: retl # encoding: [0xc3]
8220 ;
8221 ; X64-LABEL: test_adds_epi8_rmkz_256:
8222 ; X64: # %bb.0:
8223 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8224 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
8225 ; X64-NEXT: retq # encoding: [0xc3]
8226 %b = load <32 x i8>, <32 x i8>* %ptr_b
8227 %1 = call <32 x i8> @llvm.x86.avx2.padds.b(<32 x i8> %a, <32 x i8> %b)
8228 %2 = bitcast i32 %mask to <32 x i1>
8229 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
8230 ret <32 x i8> %3
8231 }
8233 define <16 x i8> @test_mask_adds_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
8234 ; CHECK-LABEL: test_mask_adds_epi8_rr_128:
8235 ; CHECK: # %bb.0:
8236 ; CHECK-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0xc1]
8237 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8238 %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)
8239 ret <16 x i8> %res
8240 }
8242 define <16 x i8> @test_mask_adds_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8243 ; X86-LABEL: test_mask_adds_epi8_rrk_128:
8244 ; X86: # %bb.0:
8245 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8246 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8247 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8248 ; X86-NEXT: retl # encoding: [0xc3]
8249 ;
8250 ; X64-LABEL: test_mask_adds_epi8_rrk_128:
8251 ; X64: # %bb.0:
8252 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8253 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0xd1]
8254 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8255 ; X64-NEXT: retq # encoding: [0xc3]
8256 %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)
8257 ret <16 x i8> %res
8258 }
8260 define <16 x i8> @test_mask_adds_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8261 ; X86-LABEL: test_mask_adds_epi8_rrkz_128:
8262 ; X86: # %bb.0:
8263 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8264 ; X86-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8265 ; X86-NEXT: retl # encoding: [0xc3]
8266 ;
8267 ; X64-LABEL: test_mask_adds_epi8_rrkz_128:
8268 ; X64: # %bb.0:
8269 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8270 ; X64-NEXT: vpaddsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0xc1]
8271 ; X64-NEXT: retq # encoding: [0xc3]
8272 %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)
8273 ret <16 x i8> %res
8274 }
8276 define <16 x i8> @test_mask_adds_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8277 ; X86-LABEL: test_mask_adds_epi8_rm_128:
8278 ; X86: # %bb.0:
8279 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8280 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x00]
8281 ; X86-NEXT: retl # encoding: [0xc3]
8282 ;
8283 ; X64-LABEL: test_mask_adds_epi8_rm_128:
8284 ; X64: # %bb.0:
8285 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xec,0x07]
8286 ; X64-NEXT: retq # encoding: [0xc3]
8287 %b = load <16 x i8>, <16 x i8>* %ptr_b
8288 %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)
8289 ret <16 x i8> %res
8290 }
8292 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) {
8293 ; X86-LABEL: test_mask_adds_epi8_rmk_128:
8294 ; X86: # %bb.0:
8295 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8296 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8297 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x08]
8298 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8299 ; X86-NEXT: retl # encoding: [0xc3]
8300 ;
8301 ; X64-LABEL: test_mask_adds_epi8_rmk_128:
8302 ; X64: # %bb.0:
8303 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8304 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xec,0x0f]
8305 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8306 ; X64-NEXT: retq # encoding: [0xc3]
8307 %b = load <16 x i8>, <16 x i8>* %ptr_b
8308 %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)
8309 ret <16 x i8> %res
8310 }
8312 define <16 x i8> @test_mask_adds_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8313 ; X86-LABEL: test_mask_adds_epi8_rmkz_128:
8314 ; X86: # %bb.0:
8315 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8316 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8317 ; X86-NEXT: vpaddsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x00]
8318 ; X86-NEXT: retl # encoding: [0xc3]
8319 ;
8320 ; X64-LABEL: test_mask_adds_epi8_rmkz_128:
8321 ; X64: # %bb.0:
8322 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8323 ; X64-NEXT: vpaddsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xec,0x07]
8324 ; X64-NEXT: retq # encoding: [0xc3]
8325 %b = load <16 x i8>, <16 x i8>* %ptr_b
8326 %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)
8327 ret <16 x i8> %res
8328 }
8330 declare <16 x i8> @llvm.x86.avx512.mask.padds.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
8332 define <32 x i8> @test_mask_adds_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
8333 ; CHECK-LABEL: test_mask_adds_epi8_rr_256:
8334 ; CHECK: # %bb.0:
8335 ; CHECK-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0xc1]
8336 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8337 %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)
8338 ret <32 x i8> %res
8339 }
8341 define <32 x i8> @test_mask_adds_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8342 ; X86-LABEL: test_mask_adds_epi8_rrk_256:
8343 ; X86: # %bb.0:
8344 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8345 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8346 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8347 ; X86-NEXT: retl # encoding: [0xc3]
8348 ;
8349 ; X64-LABEL: test_mask_adds_epi8_rrk_256:
8350 ; X64: # %bb.0:
8351 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8352 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0xd1]
8353 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8354 ; X64-NEXT: retq # encoding: [0xc3]
8355 %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)
8356 ret <32 x i8> %res
8357 }
8359 define <32 x i8> @test_mask_adds_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8360 ; X86-LABEL: test_mask_adds_epi8_rrkz_256:
8361 ; X86: # %bb.0:
8362 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8363 ; X86-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8364 ; X86-NEXT: retl # encoding: [0xc3]
8365 ;
8366 ; X64-LABEL: test_mask_adds_epi8_rrkz_256:
8367 ; X64: # %bb.0:
8368 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8369 ; X64-NEXT: vpaddsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0xc1]
8370 ; X64-NEXT: retq # encoding: [0xc3]
8371 %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)
8372 ret <32 x i8> %res
8373 }
8375 define <32 x i8> @test_mask_adds_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
8376 ; X86-LABEL: test_mask_adds_epi8_rm_256:
8377 ; X86: # %bb.0:
8378 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8379 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x00]
8380 ; X86-NEXT: retl # encoding: [0xc3]
8381 ;
8382 ; X64-LABEL: test_mask_adds_epi8_rm_256:
8383 ; X64: # %bb.0:
8384 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xec,0x07]
8385 ; X64-NEXT: retq # encoding: [0xc3]
8386 %b = load <32 x i8>, <32 x i8>* %ptr_b
8387 %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)
8388 ret <32 x i8> %res
8389 }
8391 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) {
8392 ; X86-LABEL: test_mask_adds_epi8_rmk_256:
8393 ; X86: # %bb.0:
8394 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8395 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8396 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x08]
8397 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8398 ; X86-NEXT: retl # encoding: [0xc3]
8399 ;
8400 ; X64-LABEL: test_mask_adds_epi8_rmk_256:
8401 ; X64: # %bb.0:
8402 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8403 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xec,0x0f]
8404 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8405 ; X64-NEXT: retq # encoding: [0xc3]
8406 %b = load <32 x i8>, <32 x i8>* %ptr_b
8407 %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)
8408 ret <32 x i8> %res
8409 }
8411 define <32 x i8> @test_mask_adds_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8412 ; X86-LABEL: test_mask_adds_epi8_rmkz_256:
8413 ; X86: # %bb.0:
8414 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8415 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8416 ; X86-NEXT: vpaddsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x00]
8417 ; X86-NEXT: retl # encoding: [0xc3]
8418 ;
8419 ; X64-LABEL: test_mask_adds_epi8_rmkz_256:
8420 ; X64: # %bb.0:
8421 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8422 ; X64-NEXT: vpaddsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xec,0x07]
8423 ; X64-NEXT: retq # encoding: [0xc3]
8424 %b = load <32 x i8>, <32 x i8>* %ptr_b
8425 %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)
8426 ret <32 x i8> %res
8427 }
8429 declare <32 x i8> @llvm.x86.avx512.mask.padds.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
8431 define <16 x i8> @test_mask_subs_epi8_rr_128(<16 x i8> %a, <16 x i8> %b) {
8432 ; CHECK-LABEL: test_mask_subs_epi8_rr_128:
8433 ; CHECK: # %bb.0:
8434 ; CHECK-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0xc1]
8435 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8436 %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)
8437 ret <16 x i8> %res
8438 }
8440 define <16 x i8> @test_mask_subs_epi8_rrk_128(<16 x i8> %a, <16 x i8> %b, <16 x i8> %passThru, i16 %mask) {
8441 ; X86-LABEL: test_mask_subs_epi8_rrk_128:
8442 ; X86: # %bb.0:
8443 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8444 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
8445 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8446 ; X86-NEXT: retl # encoding: [0xc3]
8447 ;
8448 ; X64-LABEL: test_mask_subs_epi8_rrk_128:
8449 ; X64: # %bb.0:
8450 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8451 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0xd1]
8452 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
8453 ; X64-NEXT: retq # encoding: [0xc3]
8454 %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)
8455 ret <16 x i8> %res
8456 }
8458 define <16 x i8> @test_mask_subs_epi8_rrkz_128(<16 x i8> %a, <16 x i8> %b, i16 %mask) {
8459 ; X86-LABEL: test_mask_subs_epi8_rrkz_128:
8460 ; X86: # %bb.0:
8461 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8462 ; X86-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
8463 ; X86-NEXT: retl # encoding: [0xc3]
8464 ;
8465 ; X64-LABEL: test_mask_subs_epi8_rrkz_128:
8466 ; X64: # %bb.0:
8467 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8468 ; X64-NEXT: vpsubsb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0xc1]
8469 ; X64-NEXT: retq # encoding: [0xc3]
8470 %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)
8471 ret <16 x i8> %res
8472 }
8474 define <16 x i8> @test_mask_subs_epi8_rm_128(<16 x i8> %a, <16 x i8>* %ptr_b) {
8475 ; X86-LABEL: test_mask_subs_epi8_rm_128:
8476 ; X86: # %bb.0:
8477 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8478 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x00]
8479 ; X86-NEXT: retl # encoding: [0xc3]
8480 ;
8481 ; X64-LABEL: test_mask_subs_epi8_rm_128:
8482 ; X64: # %bb.0:
8483 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xe8,0x07]
8484 ; X64-NEXT: retq # encoding: [0xc3]
8485 %b = load <16 x i8>, <16 x i8>* %ptr_b
8486 %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)
8487 ret <16 x i8> %res
8488 }
8490 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) {
8491 ; X86-LABEL: test_mask_subs_epi8_rmk_128:
8492 ; X86: # %bb.0:
8493 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8494 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8495 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x08]
8496 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8497 ; X86-NEXT: retl # encoding: [0xc3]
8498 ;
8499 ; X64-LABEL: test_mask_subs_epi8_rmk_128:
8500 ; X64: # %bb.0:
8501 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8502 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe8,0x0f]
8503 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
8504 ; X64-NEXT: retq # encoding: [0xc3]
8505 %b = load <16 x i8>, <16 x i8>* %ptr_b
8506 %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)
8507 ret <16 x i8> %res
8508 }
8510 define <16 x i8> @test_mask_subs_epi8_rmkz_128(<16 x i8> %a, <16 x i8>* %ptr_b, i16 %mask) {
8511 ; X86-LABEL: test_mask_subs_epi8_rmkz_128:
8512 ; X86: # %bb.0:
8513 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8514 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
8515 ; X86-NEXT: vpsubsb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x00]
8516 ; X86-NEXT: retl # encoding: [0xc3]
8517 ;
8518 ; X64-LABEL: test_mask_subs_epi8_rmkz_128:
8519 ; X64: # %bb.0:
8520 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8521 ; X64-NEXT: vpsubsb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0xe8,0x07]
8522 ; X64-NEXT: retq # encoding: [0xc3]
8523 %b = load <16 x i8>, <16 x i8>* %ptr_b
8524 %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)
8525 ret <16 x i8> %res
8526 }
8528 declare <16 x i8> @llvm.x86.avx512.mask.psubs.b.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
8530 define <32 x i8> @test_mask_subs_epi8_rr_256(<32 x i8> %a, <32 x i8> %b) {
8531 ; CHECK-LABEL: test_mask_subs_epi8_rr_256:
8532 ; CHECK: # %bb.0:
8533 ; CHECK-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0xc1]
8534 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
8535 %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)
8536 ret <32 x i8> %res
8537 }
8539 define <32 x i8> @test_mask_subs_epi8_rrk_256(<32 x i8> %a, <32 x i8> %b, <32 x i8> %passThru, i32 %mask) {
8540 ; X86-LABEL: test_mask_subs_epi8_rrk_256:
8541 ; X86: # %bb.0:
8542 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8543 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
8544 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8545 ; X86-NEXT: retl # encoding: [0xc3]
8546 ;
8547 ; X64-LABEL: test_mask_subs_epi8_rrk_256:
8548 ; X64: # %bb.0:
8549 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8550 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0xd1]
8551 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
8552 ; X64-NEXT: retq # encoding: [0xc3]
8553 %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)
8554 ret <32 x i8> %res
8555 }
8557 define <32 x i8> @test_mask_subs_epi8_rrkz_256(<32 x i8> %a, <32 x i8> %b, i32 %mask) {
8558 ; X86-LABEL: test_mask_subs_epi8_rrkz_256:
8559 ; X86: # %bb.0:
8560 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
8561 ; X86-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
8562 ; X86-NEXT: retl # encoding: [0xc3]
8563 ;
8564 ; X64-LABEL: test_mask_subs_epi8_rrkz_256:
8565 ; X64: # %bb.0:
8566 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8567 ; X64-NEXT: vpsubsb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0xc1]
8568 ; X64-NEXT: retq # encoding: [0xc3]
8569 %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)
8570 ret <32 x i8> %res
8571 }
8573 define <32 x i8> @test_mask_subs_epi8_rm_256(<32 x i8> %a, <32 x i8>* %ptr_b) {
8574 ; X86-LABEL: test_mask_subs_epi8_rm_256:
8575 ; X86: # %bb.0:
8576 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8577 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x00]
8578 ; X86-NEXT: retl # encoding: [0xc3]
8579 ;
8580 ; X64-LABEL: test_mask_subs_epi8_rm_256:
8581 ; X64: # %bb.0:
8582 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xe8,0x07]
8583 ; X64-NEXT: retq # encoding: [0xc3]
8584 %b = load <32 x i8>, <32 x i8>* %ptr_b
8585 %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)
8586 ret <32 x i8> %res
8587 }
8589 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) {
8590 ; X86-LABEL: test_mask_subs_epi8_rmk_256:
8591 ; X86: # %bb.0:
8592 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8593 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8594 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x08]
8595 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8596 ; X86-NEXT: retl # encoding: [0xc3]
8597 ;
8598 ; X64-LABEL: test_mask_subs_epi8_rmk_256:
8599 ; X64: # %bb.0:
8600 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8601 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe8,0x0f]
8602 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
8603 ; X64-NEXT: retq # encoding: [0xc3]
8604 %b = load <32 x i8>, <32 x i8>* %ptr_b
8605 %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)
8606 ret <32 x i8> %res
8607 }
8609 define <32 x i8> @test_mask_subs_epi8_rmkz_256(<32 x i8> %a, <32 x i8>* %ptr_b, i32 %mask) {
8610 ; X86-LABEL: test_mask_subs_epi8_rmkz_256:
8611 ; X86: # %bb.0:
8612 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
8613 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
8614 ; X86-NEXT: vpsubsb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x00]
8615 ; X86-NEXT: retl # encoding: [0xc3]
8616 ;
8617 ; X64-LABEL: test_mask_subs_epi8_rmkz_256:
8618 ; X64: # %bb.0:
8619 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
8620 ; X64-NEXT: vpsubsb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0xe8,0x07]
8621 ; X64-NEXT: retq # encoding: [0xc3]
8622 %b = load <32 x i8>, <32 x i8>* %ptr_b
8623 %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)
8624 ret <32 x i8> %res
8625 }
8627 declare <32 x i8> @llvm.x86.avx512.mask.psubs.b.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
8629 declare <16 x i16> @llvm.x86.avx512.mask.psrav16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8631 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) {
8632 ; X86-LABEL: test_int_x86_avx512_mask_psrav16_hi:
8633 ; X86: # %bb.0:
8634 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x11,0xd9]
8635 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8636 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
8637 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
8638 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8639 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8640 ; X86-NEXT: retl # encoding: [0xc3]
8641 ;
8642 ; X64-LABEL: test_int_x86_avx512_mask_psrav16_hi:
8643 ; X64: # %bb.0:
8644 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x11,0xd9]
8645 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8646 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
8647 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
8648 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8649 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8650 ; X64-NEXT: retq # encoding: [0xc3]
8651 %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)
8652 %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)
8653 %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)
8654 %res3 = add <16 x i16> %res, %res1
8655 %res4 = add <16 x i16> %res3, %res2
8656 ret <16 x i16> %res4
8657 }
8659 declare <8 x i16> @llvm.x86.avx512.mask.psrav8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8661 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) {
8662 ; X86-LABEL: test_int_x86_avx512_mask_psrav8_hi:
8663 ; X86: # %bb.0:
8664 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x11,0xd9]
8665 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8666 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8667 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
8668 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
8669 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8670 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8671 ; X86-NEXT: retl # encoding: [0xc3]
8672 ;
8673 ; X64-LABEL: test_int_x86_avx512_mask_psrav8_hi:
8674 ; X64: # %bb.0:
8675 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x11,0xd9]
8676 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8677 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
8678 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
8679 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8680 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8681 ; X64-NEXT: retq # encoding: [0xc3]
8682 %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)
8683 %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)
8684 %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)
8685 %res3 = add <8 x i16> %res, %res1
8686 %res4 = add <8 x i16> %res3, %res2
8687 ret <8 x i16> %res4
8688 }
8690 declare <16 x i16> @llvm.x86.avx512.mask.psllv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8692 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) {
8693 ; X86-LABEL: test_int_x86_avx512_mask_psllv16_hi:
8694 ; X86: # %bb.0:
8695 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0xd9]
8696 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8697 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
8698 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
8699 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8700 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8701 ; X86-NEXT: retl # encoding: [0xc3]
8702 ;
8703 ; X64-LABEL: test_int_x86_avx512_mask_psllv16_hi:
8704 ; X64: # %bb.0:
8705 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0xd9]
8706 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8707 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
8708 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
8709 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8710 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8711 ; X64-NEXT: retq # encoding: [0xc3]
8712 %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)
8713 %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)
8714 %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)
8715 %res3 = add <16 x i16> %res, %res1
8716 %res4 = add <16 x i16> %res3, %res2
8717 ret <16 x i16> %res4
8718 }
8720 declare <8 x i16> @llvm.x86.avx512.mask.psllv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8722 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) {
8723 ; X86-LABEL: test_int_x86_avx512_mask_psllv8_hi:
8724 ; X86: # %bb.0:
8725 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0xd9]
8726 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8727 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8728 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
8729 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
8730 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8731 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8732 ; X86-NEXT: retl # encoding: [0xc3]
8733 ;
8734 ; X64-LABEL: test_int_x86_avx512_mask_psllv8_hi:
8735 ; X64: # %bb.0:
8736 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0xd9]
8737 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8738 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
8739 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
8740 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8741 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8742 ; X64-NEXT: retq # encoding: [0xc3]
8743 %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)
8744 %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)
8745 %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)
8746 %res3 = add <8 x i16> %res, %res1
8747 %res4 = add <8 x i16> %res3, %res2
8748 ret <8 x i16> %res4
8749 }
8751 declare <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
8753 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) {
8754 ; X86-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
8755 ; X86: # %bb.0:
8756 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0xd9]
8757 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8758 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
8759 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
8760 ; X86-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8761 ; X86-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8762 ; X86-NEXT: retl # encoding: [0xc3]
8763 ;
8764 ; X64-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
8765 ; X64: # %bb.0:
8766 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm3 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0xd9]
8767 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8768 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
8769 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
8770 ; X64-NEXT: vpaddw %ymm3, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0xfd,0xc3]
8771 ; X64-NEXT: vpaddw %ymm0, %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xed,0xfd,0xc0]
8772 ; X64-NEXT: retq # encoding: [0xc3]
8773 %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)
8774 %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)
8775 %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)
8776 %res3 = add <16 x i16> %res, %res1
8777 %res4 = add <16 x i16> %res3, %res2
8778 ret <16 x i16> %res4
8779 }
8781 declare <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
8783 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) {
8784 ; X86-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
8785 ; X86: # %bb.0:
8786 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0xd9]
8787 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
8788 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
8789 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
8790 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
8791 ; X86-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8792 ; X86-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8793 ; X86-NEXT: retl # encoding: [0xc3]
8794 ;
8795 ; X64-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
8796 ; X64: # %bb.0:
8797 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm3 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0xd9]
8798 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8799 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
8800 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
8801 ; X64-NEXT: vpaddw %xmm3, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0xfd,0xc3]
8802 ; X64-NEXT: vpaddw %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfd,0xc0]
8803 ; X64-NEXT: retq # encoding: [0xc3]
8804 %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)
8805 %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)
8806 %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)
8807 %res3 = add <8 x i16> %res, %res1
8808 %res4 = add <8 x i16> %res3, %res2
8809 ret <8 x i16> %res4
8810 }
8812 declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16>, <16 x i8>, i16)
8814 define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
8815 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
8816 ; X86: # %bb.0:
8817 ; X86-NEXT: vpmovwb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc2]
8818 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
8819 ; X86-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
8820 ; X86-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
8821 ; X86-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
8822 ; X86-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
8823 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
8824 ; X86-NEXT: retl # encoding: [0xc3]
8825 ;
8826 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
8827 ; X64: # %bb.0:
8828 ; X64-NEXT: vpmovwb %ymm0, %xmm2 # encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc2]
8829 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
8830 ; X64-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
8831 ; X64-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
8832 ; X64-NEXT: vpaddb %xmm0, %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf1,0xfc,0xc0]
8833 ; X64-NEXT: vpaddb %xmm0, %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xe9,0xfc,0xc0]
8834 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
8835 ; X64-NEXT: retq # encoding: [0xc3]
8836 %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
8837 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
8838 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
8839 %res3 = add <16 x i8> %res0, %res1
8840 %res4 = add <16 x i8> %res3, %res2
8841 ret <16 x i8> %res4
8842 }