| blob | d723fc6c05a291b420b9427e29e3961df9b3a730 |
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 define <8 x i16> @test_mask_packs_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
6 ; CHECK-LABEL: test_mask_packs_epi32_rr_128:
7 ; CHECK: # %bb.0:
8 ; CHECK-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0xc1]
9 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
10 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
11 ret <8 x i16> %1
12 }
14 define <8 x i16> @test_mask_packs_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
15 ; X86-LABEL: test_mask_packs_epi32_rrk_128:
16 ; X86: # %bb.0:
17 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
18 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
19 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
20 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
21 ; X86-NEXT: retl # encoding: [0xc3]
22 ;
23 ; X64-LABEL: test_mask_packs_epi32_rrk_128:
24 ; X64: # %bb.0:
25 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
26 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0xd1]
27 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
28 ; X64-NEXT: retq # encoding: [0xc3]
29 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
30 %2 = bitcast i8 %mask to <8 x i1>
31 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
32 ret <8 x i16> %3
33 }
35 define <8 x i16> @test_mask_packs_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
36 ; X86-LABEL: test_mask_packs_epi32_rrkz_128:
37 ; X86: # %bb.0:
38 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
39 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
40 ; X86-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
41 ; X86-NEXT: retl # encoding: [0xc3]
42 ;
43 ; X64-LABEL: test_mask_packs_epi32_rrkz_128:
44 ; X64: # %bb.0:
45 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
46 ; X64-NEXT: vpackssdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0xc1]
47 ; X64-NEXT: retq # encoding: [0xc3]
48 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
49 %2 = bitcast i8 %mask to <8 x i1>
50 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
51 ret <8 x i16> %3
52 }
54 define <8 x i16> @test_mask_packs_epi32_rm_128(<4 x i32> %a, ptr %ptr_b) {
55 ; X86-LABEL: test_mask_packs_epi32_rm_128:
56 ; X86: # %bb.0:
57 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
58 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x00]
59 ; X86-NEXT: retl # encoding: [0xc3]
60 ;
61 ; X64-LABEL: test_mask_packs_epi32_rm_128:
62 ; X64: # %bb.0:
63 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6b,0x07]
64 ; X64-NEXT: retq # encoding: [0xc3]
65 %b = load <4 x i32>, ptr %ptr_b
66 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
67 ret <8 x i16> %1
68 }
70 define <8 x i16> @test_mask_packs_epi32_rmk_128(<4 x i32> %a, ptr %ptr_b, <8 x i16> %passThru, i8 %mask) {
71 ; X86-LABEL: test_mask_packs_epi32_rmk_128:
72 ; X86: # %bb.0:
73 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
74 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
75 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
76 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x08]
77 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
78 ; X86-NEXT: retl # encoding: [0xc3]
79 ;
80 ; X64-LABEL: test_mask_packs_epi32_rmk_128:
81 ; X64: # %bb.0:
82 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
83 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x6b,0x0f]
84 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
85 ; X64-NEXT: retq # encoding: [0xc3]
86 %b = load <4 x i32>, ptr %ptr_b
87 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
88 %2 = bitcast i8 %mask to <8 x i1>
89 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
90 ret <8 x i16> %3
91 }
93 define <8 x i16> @test_mask_packs_epi32_rmkz_128(<4 x i32> %a, ptr %ptr_b, i8 %mask) {
94 ; X86-LABEL: test_mask_packs_epi32_rmkz_128:
95 ; X86: # %bb.0:
96 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
97 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
98 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
99 ; X86-NEXT: vpackssdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x00]
100 ; X86-NEXT: retl # encoding: [0xc3]
101 ;
102 ; X64-LABEL: test_mask_packs_epi32_rmkz_128:
103 ; X64: # %bb.0:
104 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
105 ; X64-NEXT: vpackssdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x6b,0x07]
106 ; X64-NEXT: retq # encoding: [0xc3]
107 %b = load <4 x i32>, ptr %ptr_b
108 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
109 %2 = bitcast i8 %mask to <8 x i1>
110 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
111 ret <8 x i16> %3
112 }
114 define <8 x i16> @test_mask_packs_epi32_rmb_128(<4 x i32> %a, ptr %ptr_b) {
115 ; X86-LABEL: test_mask_packs_epi32_rmb_128:
116 ; X86: # %bb.0:
117 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
118 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x00]
119 ; X86-NEXT: retl # encoding: [0xc3]
120 ;
121 ; X64-LABEL: test_mask_packs_epi32_rmb_128:
122 ; X64: # %bb.0:
123 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf1,0x7d,0x18,0x6b,0x07]
124 ; X64-NEXT: retq # encoding: [0xc3]
125 %q = load i32, ptr %ptr_b
126 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
127 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
128 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
129 ret <8 x i16> %1
130 }
132 define <8 x i16> @test_mask_packs_epi32_rmbk_128(<4 x i32> %a, ptr %ptr_b, <8 x i16> %passThru, i8 %mask) {
133 ; X86-LABEL: test_mask_packs_epi32_rmbk_128:
134 ; X86: # %bb.0:
135 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
136 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
137 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
138 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x08]
139 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
140 ; X86-NEXT: retl # encoding: [0xc3]
141 ;
142 ; X64-LABEL: test_mask_packs_epi32_rmbk_128:
143 ; X64: # %bb.0:
144 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
145 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x19,0x6b,0x0f]
146 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
147 ; X64-NEXT: retq # encoding: [0xc3]
148 %q = load i32, ptr %ptr_b
149 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
150 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
151 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
152 %2 = bitcast i8 %mask to <8 x i1>
153 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
154 ret <8 x i16> %3
155 }
157 define <8 x i16> @test_mask_packs_epi32_rmbkz_128(<4 x i32> %a, ptr %ptr_b, i8 %mask) {
158 ; X86-LABEL: test_mask_packs_epi32_rmbkz_128:
159 ; X86: # %bb.0:
160 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
161 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
162 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
163 ; X86-NEXT: vpackssdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x00]
164 ; X86-NEXT: retl # encoding: [0xc3]
165 ;
166 ; X64-LABEL: test_mask_packs_epi32_rmbkz_128:
167 ; X64: # %bb.0:
168 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
169 ; X64-NEXT: vpackssdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x99,0x6b,0x07]
170 ; X64-NEXT: retq # encoding: [0xc3]
171 %q = load i32, ptr %ptr_b
172 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
173 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
174 %1 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %a, <4 x i32> %b)
175 %2 = bitcast i8 %mask to <8 x i1>
176 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
177 ret <8 x i16> %3
178 }
180 declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>)
182 define <16 x i16> @test_mask_packs_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
183 ; CHECK-LABEL: test_mask_packs_epi32_rr_256:
184 ; CHECK: # %bb.0:
185 ; CHECK-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0xc1]
186 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
187 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
188 ret <16 x i16> %1
189 }
191 define <16 x i16> @test_mask_packs_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
192 ; X86-LABEL: test_mask_packs_epi32_rrk_256:
193 ; X86: # %bb.0:
194 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
195 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
196 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
197 ; X86-NEXT: retl # encoding: [0xc3]
198 ;
199 ; X64-LABEL: test_mask_packs_epi32_rrk_256:
200 ; X64: # %bb.0:
201 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
202 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0xd1]
203 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
204 ; X64-NEXT: retq # encoding: [0xc3]
205 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
206 %2 = bitcast i16 %mask to <16 x i1>
207 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
208 ret <16 x i16> %3
209 }
211 define <16 x i16> @test_mask_packs_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
212 ; X86-LABEL: test_mask_packs_epi32_rrkz_256:
213 ; X86: # %bb.0:
214 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
215 ; X86-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
216 ; X86-NEXT: retl # encoding: [0xc3]
217 ;
218 ; X64-LABEL: test_mask_packs_epi32_rrkz_256:
219 ; X64: # %bb.0:
220 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
221 ; X64-NEXT: vpackssdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0xc1]
222 ; X64-NEXT: retq # encoding: [0xc3]
223 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
224 %2 = bitcast i16 %mask to <16 x i1>
225 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
226 ret <16 x i16> %3
227 }
229 define <16 x i16> @test_mask_packs_epi32_rm_256(<8 x i32> %a, ptr %ptr_b) {
230 ; X86-LABEL: test_mask_packs_epi32_rm_256:
231 ; X86: # %bb.0:
232 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
233 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x00]
234 ; X86-NEXT: retl # encoding: [0xc3]
235 ;
236 ; X64-LABEL: test_mask_packs_epi32_rm_256:
237 ; X64: # %bb.0:
238 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6b,0x07]
239 ; X64-NEXT: retq # encoding: [0xc3]
240 %b = load <8 x i32>, ptr %ptr_b
241 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
242 ret <16 x i16> %1
243 }
245 define <16 x i16> @test_mask_packs_epi32_rmk_256(<8 x i32> %a, ptr %ptr_b, <16 x i16> %passThru, i16 %mask) {
246 ; X86-LABEL: test_mask_packs_epi32_rmk_256:
247 ; X86: # %bb.0:
248 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
249 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
250 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x08]
251 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
252 ; X86-NEXT: retl # encoding: [0xc3]
253 ;
254 ; X64-LABEL: test_mask_packs_epi32_rmk_256:
255 ; X64: # %bb.0:
256 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
257 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x6b,0x0f]
258 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
259 ; X64-NEXT: retq # encoding: [0xc3]
260 %b = load <8 x i32>, ptr %ptr_b
261 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
262 %2 = bitcast i16 %mask to <16 x i1>
263 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
264 ret <16 x i16> %3
265 }
267 define <16 x i16> @test_mask_packs_epi32_rmkz_256(<8 x i32> %a, ptr %ptr_b, i16 %mask) {
268 ; X86-LABEL: test_mask_packs_epi32_rmkz_256:
269 ; X86: # %bb.0:
270 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
271 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
272 ; X86-NEXT: vpackssdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x00]
273 ; X86-NEXT: retl # encoding: [0xc3]
274 ;
275 ; X64-LABEL: test_mask_packs_epi32_rmkz_256:
276 ; X64: # %bb.0:
277 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
278 ; X64-NEXT: vpackssdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x6b,0x07]
279 ; X64-NEXT: retq # encoding: [0xc3]
280 %b = load <8 x i32>, ptr %ptr_b
281 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
282 %2 = bitcast i16 %mask to <16 x i1>
283 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
284 ret <16 x i16> %3
285 }
287 define <16 x i16> @test_mask_packs_epi32_rmb_256(<8 x i32> %a, ptr %ptr_b) {
288 ; X86-LABEL: test_mask_packs_epi32_rmb_256:
289 ; X86: # %bb.0:
290 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
291 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x00]
292 ; X86-NEXT: retl # encoding: [0xc3]
293 ;
294 ; X64-LABEL: test_mask_packs_epi32_rmb_256:
295 ; X64: # %bb.0:
296 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf1,0x7d,0x38,0x6b,0x07]
297 ; X64-NEXT: retq # encoding: [0xc3]
298 %q = load i32, ptr %ptr_b
299 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
300 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
301 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
302 ret <16 x i16> %1
303 }
305 define <16 x i16> @test_mask_packs_epi32_rmbk_256(<8 x i32> %a, ptr %ptr_b, <16 x i16> %passThru, i16 %mask) {
306 ; X86-LABEL: test_mask_packs_epi32_rmbk_256:
307 ; X86: # %bb.0:
308 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
309 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
310 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x08]
311 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
312 ; X86-NEXT: retl # encoding: [0xc3]
313 ;
314 ; X64-LABEL: test_mask_packs_epi32_rmbk_256:
315 ; X64: # %bb.0:
316 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
317 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x39,0x6b,0x0f]
318 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
319 ; X64-NEXT: retq # encoding: [0xc3]
320 %q = load i32, ptr %ptr_b
321 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
322 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
323 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
324 %2 = bitcast i16 %mask to <16 x i1>
325 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
326 ret <16 x i16> %3
327 }
329 define <16 x i16> @test_mask_packs_epi32_rmbkz_256(<8 x i32> %a, ptr %ptr_b, i16 %mask) {
330 ; X86-LABEL: test_mask_packs_epi32_rmbkz_256:
331 ; X86: # %bb.0:
332 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
333 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
334 ; X86-NEXT: vpackssdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x00]
335 ; X86-NEXT: retl # encoding: [0xc3]
336 ;
337 ; X64-LABEL: test_mask_packs_epi32_rmbkz_256:
338 ; X64: # %bb.0:
339 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
340 ; X64-NEXT: vpackssdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xb9,0x6b,0x07]
341 ; X64-NEXT: retq # encoding: [0xc3]
342 %q = load i32, ptr %ptr_b
343 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
344 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
345 %1 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %a, <8 x i32> %b)
346 %2 = bitcast i16 %mask to <16 x i1>
347 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
348 ret <16 x i16> %3
349 }
351 declare <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32>, <8 x i32>)
353 define <16 x i8> @test_mask_packs_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
354 ; CHECK-LABEL: test_mask_packs_epi16_rr_128:
355 ; CHECK: # %bb.0:
356 ; CHECK-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0xc1]
357 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
358 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
359 ret <16 x i8> %1
360 }
362 define <16 x i8> @test_mask_packs_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
363 ; X86-LABEL: test_mask_packs_epi16_rrk_128:
364 ; X86: # %bb.0:
365 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
366 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
367 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
368 ; X86-NEXT: retl # encoding: [0xc3]
369 ;
370 ; X64-LABEL: test_mask_packs_epi16_rrk_128:
371 ; X64: # %bb.0:
372 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
373 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0xd1]
374 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
375 ; X64-NEXT: retq # encoding: [0xc3]
376 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
377 %2 = bitcast i16 %mask to <16 x i1>
378 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
379 ret <16 x i8> %3
380 }
382 define <16 x i8> @test_mask_packs_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
383 ; X86-LABEL: test_mask_packs_epi16_rrkz_128:
384 ; X86: # %bb.0:
385 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
386 ; X86-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
387 ; X86-NEXT: retl # encoding: [0xc3]
388 ;
389 ; X64-LABEL: test_mask_packs_epi16_rrkz_128:
390 ; X64: # %bb.0:
391 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
392 ; X64-NEXT: vpacksswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0xc1]
393 ; X64-NEXT: retq # encoding: [0xc3]
394 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
395 %2 = bitcast i16 %mask to <16 x i1>
396 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
397 ret <16 x i8> %3
398 }
400 define <16 x i8> @test_mask_packs_epi16_rm_128(<8 x i16> %a, ptr %ptr_b) {
401 ; X86-LABEL: test_mask_packs_epi16_rm_128:
402 ; X86: # %bb.0:
403 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
404 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x00]
405 ; X86-NEXT: retl # encoding: [0xc3]
406 ;
407 ; X64-LABEL: test_mask_packs_epi16_rm_128:
408 ; X64: # %bb.0:
409 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x63,0x07]
410 ; X64-NEXT: retq # encoding: [0xc3]
411 %b = load <8 x i16>, ptr %ptr_b
412 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
413 ret <16 x i8> %1
414 }
416 define <16 x i8> @test_mask_packs_epi16_rmk_128(<8 x i16> %a, ptr %ptr_b, <16 x i8> %passThru, i16 %mask) {
417 ; X86-LABEL: test_mask_packs_epi16_rmk_128:
418 ; X86: # %bb.0:
419 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
420 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
421 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x08]
422 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
423 ; X86-NEXT: retl # encoding: [0xc3]
424 ;
425 ; X64-LABEL: test_mask_packs_epi16_rmk_128:
426 ; X64: # %bb.0:
427 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
428 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x63,0x0f]
429 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
430 ; X64-NEXT: retq # encoding: [0xc3]
431 %b = load <8 x i16>, ptr %ptr_b
432 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
433 %2 = bitcast i16 %mask to <16 x i1>
434 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
435 ret <16 x i8> %3
436 }
438 define <16 x i8> @test_mask_packs_epi16_rmkz_128(<8 x i16> %a, ptr %ptr_b, i16 %mask) {
439 ; X86-LABEL: test_mask_packs_epi16_rmkz_128:
440 ; X86: # %bb.0:
441 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
442 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
443 ; X86-NEXT: vpacksswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x00]
444 ; X86-NEXT: retl # encoding: [0xc3]
445 ;
446 ; X64-LABEL: test_mask_packs_epi16_rmkz_128:
447 ; X64: # %bb.0:
448 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
449 ; X64-NEXT: vpacksswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x63,0x07]
450 ; X64-NEXT: retq # encoding: [0xc3]
451 %b = load <8 x i16>, ptr %ptr_b
452 %1 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %a, <8 x i16> %b)
453 %2 = bitcast i16 %mask to <16 x i1>
454 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
455 ret <16 x i8> %3
456 }
458 declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>)
460 define <32 x i8> @test_mask_packs_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
461 ; CHECK-LABEL: test_mask_packs_epi16_rr_256:
462 ; CHECK: # %bb.0:
463 ; CHECK-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0xc1]
464 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
465 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
466 ret <32 x i8> %1
467 }
469 define <32 x i8> @test_mask_packs_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
470 ; X86-LABEL: test_mask_packs_epi16_rrk_256:
471 ; X86: # %bb.0:
472 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
473 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
474 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
475 ; X86-NEXT: retl # encoding: [0xc3]
476 ;
477 ; X64-LABEL: test_mask_packs_epi16_rrk_256:
478 ; X64: # %bb.0:
479 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
480 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0xd1]
481 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
482 ; X64-NEXT: retq # encoding: [0xc3]
483 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
484 %2 = bitcast i32 %mask to <32 x i1>
485 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
486 ret <32 x i8> %3
487 }
489 define <32 x i8> @test_mask_packs_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
490 ; X86-LABEL: test_mask_packs_epi16_rrkz_256:
491 ; X86: # %bb.0:
492 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
493 ; X86-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
494 ; X86-NEXT: retl # encoding: [0xc3]
495 ;
496 ; X64-LABEL: test_mask_packs_epi16_rrkz_256:
497 ; X64: # %bb.0:
498 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
499 ; X64-NEXT: vpacksswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0xc1]
500 ; X64-NEXT: retq # encoding: [0xc3]
501 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
502 %2 = bitcast i32 %mask to <32 x i1>
503 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
504 ret <32 x i8> %3
505 }
507 define <32 x i8> @test_mask_packs_epi16_rm_256(<16 x i16> %a, ptr %ptr_b) {
508 ; X86-LABEL: test_mask_packs_epi16_rm_256:
509 ; X86: # %bb.0:
510 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
511 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x00]
512 ; X86-NEXT: retl # encoding: [0xc3]
513 ;
514 ; X64-LABEL: test_mask_packs_epi16_rm_256:
515 ; X64: # %bb.0:
516 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x63,0x07]
517 ; X64-NEXT: retq # encoding: [0xc3]
518 %b = load <16 x i16>, ptr %ptr_b
519 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
520 ret <32 x i8> %1
521 }
523 define <32 x i8> @test_mask_packs_epi16_rmk_256(<16 x i16> %a, ptr %ptr_b, <32 x i8> %passThru, i32 %mask) {
524 ; X86-LABEL: test_mask_packs_epi16_rmk_256:
525 ; X86: # %bb.0:
526 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
527 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
528 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x08]
529 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
530 ; X86-NEXT: retl # encoding: [0xc3]
531 ;
532 ; X64-LABEL: test_mask_packs_epi16_rmk_256:
533 ; X64: # %bb.0:
534 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
535 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x63,0x0f]
536 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
537 ; X64-NEXT: retq # encoding: [0xc3]
538 %b = load <16 x i16>, ptr %ptr_b
539 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
540 %2 = bitcast i32 %mask to <32 x i1>
541 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
542 ret <32 x i8> %3
543 }
545 define <32 x i8> @test_mask_packs_epi16_rmkz_256(<16 x i16> %a, ptr %ptr_b, i32 %mask) {
546 ; X86-LABEL: test_mask_packs_epi16_rmkz_256:
547 ; X86: # %bb.0:
548 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
549 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
550 ; X86-NEXT: vpacksswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x00]
551 ; X86-NEXT: retl # encoding: [0xc3]
552 ;
553 ; X64-LABEL: test_mask_packs_epi16_rmkz_256:
554 ; X64: # %bb.0:
555 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
556 ; X64-NEXT: vpacksswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x63,0x07]
557 ; X64-NEXT: retq # encoding: [0xc3]
558 %b = load <16 x i16>, ptr %ptr_b
559 %1 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %a, <16 x i16> %b)
560 %2 = bitcast i32 %mask to <32 x i1>
561 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
562 ret <32 x i8> %3
563 }
565 declare <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16>, <16 x i16>)
568 define <8 x i16> @test_mask_packus_epi32_rr_128(<4 x i32> %a, <4 x i32> %b) {
569 ; CHECK-LABEL: test_mask_packus_epi32_rr_128:
570 ; CHECK: # %bb.0:
571 ; CHECK-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0xc1]
572 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
573 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
574 ret <8 x i16> %1
575 }
577 define <8 x i16> @test_mask_packus_epi32_rrk_128(<4 x i32> %a, <4 x i32> %b, <8 x i16> %passThru, i8 %mask) {
578 ; X86-LABEL: test_mask_packus_epi32_rrk_128:
579 ; X86: # %bb.0:
580 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
581 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
582 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
583 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
584 ; X86-NEXT: retl # encoding: [0xc3]
585 ;
586 ; X64-LABEL: test_mask_packus_epi32_rrk_128:
587 ; X64: # %bb.0:
588 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
589 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0xd1]
590 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
591 ; X64-NEXT: retq # encoding: [0xc3]
592 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
593 %2 = bitcast i8 %mask to <8 x i1>
594 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
595 ret <8 x i16> %3
596 }
598 define <8 x i16> @test_mask_packus_epi32_rrkz_128(<4 x i32> %a, <4 x i32> %b, i8 %mask) {
599 ; X86-LABEL: test_mask_packus_epi32_rrkz_128:
600 ; X86: # %bb.0:
601 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
602 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
603 ; X86-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
604 ; X86-NEXT: retl # encoding: [0xc3]
605 ;
606 ; X64-LABEL: test_mask_packus_epi32_rrkz_128:
607 ; X64: # %bb.0:
608 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
609 ; X64-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0xc1]
610 ; X64-NEXT: retq # encoding: [0xc3]
611 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
612 %2 = bitcast i8 %mask to <8 x i1>
613 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
614 ret <8 x i16> %3
615 }
617 define <8 x i16> @test_mask_packus_epi32_rm_128(<4 x i32> %a, ptr %ptr_b) {
618 ; X86-LABEL: test_mask_packus_epi32_rm_128:
619 ; X86: # %bb.0:
620 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
621 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x00]
622 ; X86-NEXT: retl # encoding: [0xc3]
623 ;
624 ; X64-LABEL: test_mask_packus_epi32_rm_128:
625 ; X64: # %bb.0:
626 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x2b,0x07]
627 ; X64-NEXT: retq # encoding: [0xc3]
628 %b = load <4 x i32>, ptr %ptr_b
629 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
630 ret <8 x i16> %1
631 }
633 define <8 x i16> @test_mask_packus_epi32_rmk_128(<4 x i32> %a, ptr %ptr_b, <8 x i16> %passThru, i8 %mask) {
634 ; X86-LABEL: test_mask_packus_epi32_rmk_128:
635 ; X86: # %bb.0:
636 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
637 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
638 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
639 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x08]
640 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
641 ; X86-NEXT: retl # encoding: [0xc3]
642 ;
643 ; X64-LABEL: test_mask_packus_epi32_rmk_128:
644 ; X64: # %bb.0:
645 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
646 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x2b,0x0f]
647 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
648 ; X64-NEXT: retq # encoding: [0xc3]
649 %b = load <4 x i32>, ptr %ptr_b
650 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
651 %2 = bitcast i8 %mask to <8 x i1>
652 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
653 ret <8 x i16> %3
654 }
656 define <8 x i16> @test_mask_packus_epi32_rmkz_128(<4 x i32> %a, ptr %ptr_b, i8 %mask) {
657 ; X86-LABEL: test_mask_packus_epi32_rmkz_128:
658 ; X86: # %bb.0:
659 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
660 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
661 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
662 ; X86-NEXT: vpackusdw (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x00]
663 ; X86-NEXT: retl # encoding: [0xc3]
664 ;
665 ; X64-LABEL: test_mask_packus_epi32_rmkz_128:
666 ; X64: # %bb.0:
667 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
668 ; X64-NEXT: vpackusdw (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x89,0x2b,0x07]
669 ; X64-NEXT: retq # encoding: [0xc3]
670 %b = load <4 x i32>, ptr %ptr_b
671 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
672 %2 = bitcast i8 %mask to <8 x i1>
673 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
674 ret <8 x i16> %3
675 }
677 define <8 x i16> @test_mask_packus_epi32_rmb_128(<4 x i32> %a, ptr %ptr_b) {
678 ; X86-LABEL: test_mask_packus_epi32_rmb_128:
679 ; X86: # %bb.0:
680 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
681 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x00]
682 ; X86-NEXT: retl # encoding: [0xc3]
683 ;
684 ; X64-LABEL: test_mask_packus_epi32_rmb_128:
685 ; X64: # %bb.0:
686 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0x7d,0x18,0x2b,0x07]
687 ; X64-NEXT: retq # encoding: [0xc3]
688 %q = load i32, ptr %ptr_b
689 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
690 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
691 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
692 ret <8 x i16> %1
693 }
695 define <8 x i16> @test_mask_packus_epi32_rmbk_128(<4 x i32> %a, ptr %ptr_b, <8 x i16> %passThru, i8 %mask) {
696 ; X86-LABEL: test_mask_packus_epi32_rmbk_128:
697 ; X86: # %bb.0:
698 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
699 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
700 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
701 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x08]
702 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
703 ; X86-NEXT: retl # encoding: [0xc3]
704 ;
705 ; X64-LABEL: test_mask_packus_epi32_rmbk_128:
706 ; X64: # %bb.0:
707 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
708 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x19,0x2b,0x0f]
709 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
710 ; X64-NEXT: retq # encoding: [0xc3]
711 %q = load i32, ptr %ptr_b
712 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
713 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
714 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
715 %2 = bitcast i8 %mask to <8 x i1>
716 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %passThru
717 ret <8 x i16> %3
718 }
720 define <8 x i16> @test_mask_packus_epi32_rmbkz_128(<4 x i32> %a, ptr %ptr_b, i8 %mask) {
721 ; X86-LABEL: test_mask_packus_epi32_rmbkz_128:
722 ; X86: # %bb.0:
723 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
724 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
725 ; X86-NEXT: kmovd %ecx, %k1 # encoding: [0xc5,0xfb,0x92,0xc9]
726 ; X86-NEXT: vpackusdw (%eax){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x00]
727 ; X86-NEXT: retl # encoding: [0xc3]
728 ;
729 ; X64-LABEL: test_mask_packus_epi32_rmbkz_128:
730 ; X64: # %bb.0:
731 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
732 ; X64-NEXT: vpackusdw (%rdi){1to4}, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0x99,0x2b,0x07]
733 ; X64-NEXT: retq # encoding: [0xc3]
734 %q = load i32, ptr %ptr_b
735 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
736 %b = shufflevector <4 x i32> %vecinit.i, <4 x i32> undef, <4 x i32> zeroinitializer
737 %1 = call <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32> %a, <4 x i32> %b)
738 %2 = bitcast i8 %mask to <8 x i1>
739 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
740 ret <8 x i16> %3
741 }
743 declare <8 x i16> @llvm.x86.sse41.packusdw(<4 x i32>, <4 x i32>)
745 define <16 x i16> @test_mask_packus_epi32_rr_256(<8 x i32> %a, <8 x i32> %b) {
746 ; CHECK-LABEL: test_mask_packus_epi32_rr_256:
747 ; CHECK: # %bb.0:
748 ; CHECK-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0xc1]
749 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
750 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
751 ret <16 x i16> %1
752 }
754 define <16 x i16> @test_mask_packus_epi32_rrk_256(<8 x i32> %a, <8 x i32> %b, <16 x i16> %passThru, i16 %mask) {
755 ; X86-LABEL: test_mask_packus_epi32_rrk_256:
756 ; X86: # %bb.0:
757 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
758 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
759 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
760 ; X86-NEXT: retl # encoding: [0xc3]
761 ;
762 ; X64-LABEL: test_mask_packus_epi32_rrk_256:
763 ; X64: # %bb.0:
764 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
765 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0xd1]
766 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
767 ; X64-NEXT: retq # encoding: [0xc3]
768 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
769 %2 = bitcast i16 %mask to <16 x i1>
770 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
771 ret <16 x i16> %3
772 }
774 define <16 x i16> @test_mask_packus_epi32_rrkz_256(<8 x i32> %a, <8 x i32> %b, i16 %mask) {
775 ; X86-LABEL: test_mask_packus_epi32_rrkz_256:
776 ; X86: # %bb.0:
777 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
778 ; X86-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
779 ; X86-NEXT: retl # encoding: [0xc3]
780 ;
781 ; X64-LABEL: test_mask_packus_epi32_rrkz_256:
782 ; X64: # %bb.0:
783 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
784 ; X64-NEXT: vpackusdw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0xc1]
785 ; X64-NEXT: retq # encoding: [0xc3]
786 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
787 %2 = bitcast i16 %mask to <16 x i1>
788 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
789 ret <16 x i16> %3
790 }
792 define <16 x i16> @test_mask_packus_epi32_rm_256(<8 x i32> %a, ptr %ptr_b) {
793 ; X86-LABEL: test_mask_packus_epi32_rm_256:
794 ; X86: # %bb.0:
795 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
796 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x00]
797 ; X86-NEXT: retl # encoding: [0xc3]
798 ;
799 ; X64-LABEL: test_mask_packus_epi32_rm_256:
800 ; X64: # %bb.0:
801 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7d,0x2b,0x07]
802 ; X64-NEXT: retq # encoding: [0xc3]
803 %b = load <8 x i32>, ptr %ptr_b
804 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
805 ret <16 x i16> %1
806 }
808 define <16 x i16> @test_mask_packus_epi32_rmk_256(<8 x i32> %a, ptr %ptr_b, <16 x i16> %passThru, i16 %mask) {
809 ; X86-LABEL: test_mask_packus_epi32_rmk_256:
810 ; X86: # %bb.0:
811 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
812 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
813 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x08]
814 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
815 ; X86-NEXT: retl # encoding: [0xc3]
816 ;
817 ; X64-LABEL: test_mask_packus_epi32_rmk_256:
818 ; X64: # %bb.0:
819 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
820 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x2b,0x0f]
821 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
822 ; X64-NEXT: retq # encoding: [0xc3]
823 %b = load <8 x i32>, ptr %ptr_b
824 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
825 %2 = bitcast i16 %mask to <16 x i1>
826 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
827 ret <16 x i16> %3
828 }
830 define <16 x i16> @test_mask_packus_epi32_rmkz_256(<8 x i32> %a, ptr %ptr_b, i16 %mask) {
831 ; X86-LABEL: test_mask_packus_epi32_rmkz_256:
832 ; X86: # %bb.0:
833 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
834 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
835 ; X86-NEXT: vpackusdw (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x00]
836 ; X86-NEXT: retl # encoding: [0xc3]
837 ;
838 ; X64-LABEL: test_mask_packus_epi32_rmkz_256:
839 ; X64: # %bb.0:
840 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
841 ; X64-NEXT: vpackusdw (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xa9,0x2b,0x07]
842 ; X64-NEXT: retq # encoding: [0xc3]
843 %b = load <8 x i32>, ptr %ptr_b
844 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
845 %2 = bitcast i16 %mask to <16 x i1>
846 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
847 ret <16 x i16> %3
848 }
850 define <16 x i16> @test_mask_packus_epi32_rmb_256(<8 x i32> %a, ptr %ptr_b) {
851 ; X86-LABEL: test_mask_packus_epi32_rmb_256:
852 ; X86: # %bb.0:
853 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
854 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x00]
855 ; X86-NEXT: retl # encoding: [0xc3]
856 ;
857 ; X64-LABEL: test_mask_packus_epi32_rmb_256:
858 ; X64: # %bb.0:
859 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0x7d,0x38,0x2b,0x07]
860 ; X64-NEXT: retq # encoding: [0xc3]
861 %q = load i32, ptr %ptr_b
862 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
863 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
864 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
865 ret <16 x i16> %1
866 }
868 define <16 x i16> @test_mask_packus_epi32_rmbk_256(<8 x i32> %a, ptr %ptr_b, <16 x i16> %passThru, i16 %mask) {
869 ; X86-LABEL: test_mask_packus_epi32_rmbk_256:
870 ; X86: # %bb.0:
871 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
872 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
873 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x08]
874 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
875 ; X86-NEXT: retl # encoding: [0xc3]
876 ;
877 ; X64-LABEL: test_mask_packus_epi32_rmbk_256:
878 ; X64: # %bb.0:
879 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
880 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0x7d,0x39,0x2b,0x0f]
881 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
882 ; X64-NEXT: retq # encoding: [0xc3]
883 %q = load i32, ptr %ptr_b
884 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
885 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
886 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
887 %2 = bitcast i16 %mask to <16 x i1>
888 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %passThru
889 ret <16 x i16> %3
890 }
892 define <16 x i16> @test_mask_packus_epi32_rmbkz_256(<8 x i32> %a, ptr %ptr_b, i16 %mask) {
893 ; X86-LABEL: test_mask_packus_epi32_rmbkz_256:
894 ; X86: # %bb.0:
895 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
896 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
897 ; X86-NEXT: vpackusdw (%eax){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x00]
898 ; X86-NEXT: retl # encoding: [0xc3]
899 ;
900 ; X64-LABEL: test_mask_packus_epi32_rmbkz_256:
901 ; X64: # %bb.0:
902 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
903 ; X64-NEXT: vpackusdw (%rdi){1to8}, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0x7d,0xb9,0x2b,0x07]
904 ; X64-NEXT: retq # encoding: [0xc3]
905 %q = load i32, ptr %ptr_b
906 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
907 %b = shufflevector <8 x i32> %vecinit.i, <8 x i32> undef, <8 x i32> zeroinitializer
908 %1 = call <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32> %a, <8 x i32> %b)
909 %2 = bitcast i16 %mask to <16 x i1>
910 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
911 ret <16 x i16> %3
912 }
914 declare <16 x i16> @llvm.x86.avx2.packusdw(<8 x i32>, <8 x i32>)
916 define <16 x i8> @test_mask_packus_epi16_rr_128(<8 x i16> %a, <8 x i16> %b) {
917 ; CHECK-LABEL: test_mask_packus_epi16_rr_128:
918 ; CHECK: # %bb.0:
919 ; CHECK-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0xc1]
920 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
921 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
922 ret <16 x i8> %1
923 }
925 define <16 x i8> @test_mask_packus_epi16_rrk_128(<8 x i16> %a, <8 x i16> %b, <16 x i8> %passThru, i16 %mask) {
926 ; X86-LABEL: test_mask_packus_epi16_rrk_128:
927 ; X86: # %bb.0:
928 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
929 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
930 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
931 ; X86-NEXT: retl # encoding: [0xc3]
932 ;
933 ; X64-LABEL: test_mask_packus_epi16_rrk_128:
934 ; X64: # %bb.0:
935 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
936 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0xd1]
937 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
938 ; X64-NEXT: retq # encoding: [0xc3]
939 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
940 %2 = bitcast i16 %mask to <16 x i1>
941 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
942 ret <16 x i8> %3
943 }
945 define <16 x i8> @test_mask_packus_epi16_rrkz_128(<8 x i16> %a, <8 x i16> %b, i16 %mask) {
946 ; X86-LABEL: test_mask_packus_epi16_rrkz_128:
947 ; X86: # %bb.0:
948 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
949 ; X86-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
950 ; X86-NEXT: retl # encoding: [0xc3]
951 ;
952 ; X64-LABEL: test_mask_packus_epi16_rrkz_128:
953 ; X64: # %bb.0:
954 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
955 ; X64-NEXT: vpackuswb %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0xc1]
956 ; X64-NEXT: retq # encoding: [0xc3]
957 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
958 %2 = bitcast i16 %mask to <16 x i1>
959 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
960 ret <16 x i8> %3
961 }
963 define <16 x i8> @test_mask_packus_epi16_rm_128(<8 x i16> %a, ptr %ptr_b) {
964 ; X86-LABEL: test_mask_packus_epi16_rm_128:
965 ; X86: # %bb.0:
966 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
967 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x00]
968 ; X86-NEXT: retl # encoding: [0xc3]
969 ;
970 ; X64-LABEL: test_mask_packus_epi16_rm_128:
971 ; X64: # %bb.0:
972 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x67,0x07]
973 ; X64-NEXT: retq # encoding: [0xc3]
974 %b = load <8 x i16>, ptr %ptr_b
975 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
976 ret <16 x i8> %1
977 }
979 define <16 x i8> @test_mask_packus_epi16_rmk_128(<8 x i16> %a, ptr %ptr_b, <16 x i8> %passThru, i16 %mask) {
980 ; X86-LABEL: test_mask_packus_epi16_rmk_128:
981 ; X86: # %bb.0:
982 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
983 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
984 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x08]
985 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
986 ; X86-NEXT: retl # encoding: [0xc3]
987 ;
988 ; X64-LABEL: test_mask_packus_epi16_rmk_128:
989 ; X64: # %bb.0:
990 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
991 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0x67,0x0f]
992 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
993 ; X64-NEXT: retq # encoding: [0xc3]
994 %b = load <8 x i16>, ptr %ptr_b
995 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
996 %2 = bitcast i16 %mask to <16 x i1>
997 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %passThru
998 ret <16 x i8> %3
999 }
1001 define <16 x i8> @test_mask_packus_epi16_rmkz_128(<8 x i16> %a, ptr %ptr_b, i16 %mask) {
1002 ; X86-LABEL: test_mask_packus_epi16_rmkz_128:
1003 ; X86: # %bb.0:
1004 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1005 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1006 ; X86-NEXT: vpackuswb (%eax), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x00]
1007 ; X86-NEXT: retl # encoding: [0xc3]
1008 ;
1009 ; X64-LABEL: test_mask_packus_epi16_rmkz_128:
1010 ; X64: # %bb.0:
1011 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1012 ; X64-NEXT: vpackuswb (%rdi), %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0x89,0x67,0x07]
1013 ; X64-NEXT: retq # encoding: [0xc3]
1014 %b = load <8 x i16>, ptr %ptr_b
1015 %1 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %a, <8 x i16> %b)
1016 %2 = bitcast i16 %mask to <16 x i1>
1017 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
1018 ret <16 x i8> %3
1019 }
1021 declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>)
1023 define <32 x i8> @test_mask_packus_epi16_rr_256(<16 x i16> %a, <16 x i16> %b) {
1024 ; CHECK-LABEL: test_mask_packus_epi16_rr_256:
1025 ; CHECK: # %bb.0:
1026 ; CHECK-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0xc1]
1027 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1028 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1029 ret <32 x i8> %1
1030 }
1032 define <32 x i8> @test_mask_packus_epi16_rrk_256(<16 x i16> %a, <16 x i16> %b, <32 x i8> %passThru, i32 %mask) {
1033 ; X86-LABEL: test_mask_packus_epi16_rrk_256:
1034 ; X86: # %bb.0:
1035 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1036 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
1037 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1038 ; X86-NEXT: retl # encoding: [0xc3]
1039 ;
1040 ; X64-LABEL: test_mask_packus_epi16_rrk_256:
1041 ; X64: # %bb.0:
1042 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1043 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0xd1]
1044 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1045 ; X64-NEXT: retq # encoding: [0xc3]
1046 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1047 %2 = bitcast i32 %mask to <32 x i1>
1048 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
1049 ret <32 x i8> %3
1050 }
1052 define <32 x i8> @test_mask_packus_epi16_rrkz_256(<16 x i16> %a, <16 x i16> %b, i32 %mask) {
1053 ; X86-LABEL: test_mask_packus_epi16_rrkz_256:
1054 ; X86: # %bb.0:
1055 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1056 ; X86-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
1057 ; X86-NEXT: retl # encoding: [0xc3]
1058 ;
1059 ; X64-LABEL: test_mask_packus_epi16_rrkz_256:
1060 ; X64: # %bb.0:
1061 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1062 ; X64-NEXT: vpackuswb %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0xc1]
1063 ; X64-NEXT: retq # encoding: [0xc3]
1064 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1065 %2 = bitcast i32 %mask to <32 x i1>
1066 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
1067 ret <32 x i8> %3
1068 }
1070 define <32 x i8> @test_mask_packus_epi16_rm_256(<16 x i16> %a, ptr %ptr_b) {
1071 ; X86-LABEL: test_mask_packus_epi16_rm_256:
1072 ; X86: # %bb.0:
1073 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1074 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x00]
1075 ; X86-NEXT: retl # encoding: [0xc3]
1076 ;
1077 ; X64-LABEL: test_mask_packus_epi16_rm_256:
1078 ; X64: # %bb.0:
1079 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x67,0x07]
1080 ; X64-NEXT: retq # encoding: [0xc3]
1081 %b = load <16 x i16>, ptr %ptr_b
1082 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1083 ret <32 x i8> %1
1084 }
1086 define <32 x i8> @test_mask_packus_epi16_rmk_256(<16 x i16> %a, ptr %ptr_b, <32 x i8> %passThru, i32 %mask) {
1087 ; X86-LABEL: test_mask_packus_epi16_rmk_256:
1088 ; X86: # %bb.0:
1089 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1090 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1091 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x08]
1092 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1093 ; X86-NEXT: retl # encoding: [0xc3]
1094 ;
1095 ; X64-LABEL: test_mask_packus_epi16_rmk_256:
1096 ; X64: # %bb.0:
1097 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1098 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0x67,0x0f]
1099 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1100 ; X64-NEXT: retq # encoding: [0xc3]
1101 %b = load <16 x i16>, ptr %ptr_b
1102 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1103 %2 = bitcast i32 %mask to <32 x i1>
1104 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %passThru
1105 ret <32 x i8> %3
1106 }
1108 define <32 x i8> @test_mask_packus_epi16_rmkz_256(<16 x i16> %a, ptr %ptr_b, i32 %mask) {
1109 ; X86-LABEL: test_mask_packus_epi16_rmkz_256:
1110 ; X86: # %bb.0:
1111 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1112 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x08]
1113 ; X86-NEXT: vpackuswb (%eax), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x00]
1114 ; X86-NEXT: retl # encoding: [0xc3]
1115 ;
1116 ; X64-LABEL: test_mask_packus_epi16_rmkz_256:
1117 ; X64: # %bb.0:
1118 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1119 ; X64-NEXT: vpackuswb (%rdi), %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf1,0x7d,0xa9,0x67,0x07]
1120 ; X64-NEXT: retq # encoding: [0xc3]
1121 %b = load <16 x i16>, ptr %ptr_b
1122 %1 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %a, <16 x i16> %b)
1123 %2 = bitcast i32 %mask to <32 x i1>
1124 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> zeroinitializer
1125 ret <32 x i8> %3
1126 }
1128 declare <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16>, <16 x i16>)
1130 define <8 x i16>@test_int_x86_avx512_vpermt2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2) {
1131 ; CHECK-LABEL: test_int_x86_avx512_vpermt2var_hi_128:
1132 ; CHECK: # %bb.0:
1133 ; CHECK-NEXT: vpermi2w %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x75,0xc2]
1134 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1135 %1 = call <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16> %x1, <8 x i16> %x0, <8 x i16> %x2)
1136 ret <8 x i16> %1
1137 }
1139 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) {
1140 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
1141 ; X86: # %bb.0:
1142 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1143 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1144 ; X86-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
1145 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1146 ; X86-NEXT: retl # encoding: [0xc3]
1147 ;
1148 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_128:
1149 ; X64: # %bb.0:
1150 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1151 ; X64-NEXT: vpermt2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x7d,0xca]
1152 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1153 ; X64-NEXT: retq # encoding: [0xc3]
1154 %1 = call <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16> %x1, <8 x i16> %x0, <8 x i16> %x2)
1155 %2 = bitcast i8 %x3 to <8 x i1>
1156 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x1
1157 ret <8 x i16> %3
1158 }
1160 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) {
1161 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
1162 ; X86: # %bb.0:
1163 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1164 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1165 ; X86-NEXT: vpermi2w %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x75,0xc2]
1166 ; X86-NEXT: retl # encoding: [0xc3]
1167 ;
1168 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_128:
1169 ; X64: # %bb.0:
1170 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1171 ; X64-NEXT: vpermi2w %xmm2, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x75,0xc2]
1172 ; X64-NEXT: retq # encoding: [0xc3]
1173 %1 = call <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16> %x1, <8 x i16> %x0, <8 x i16> %x2)
1174 %2 = bitcast i8 %x3 to <8 x i1>
1175 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
1176 ret <8 x i16> %3
1177 }
1179 define <16 x i16>@test_int_x86_avx512_vpermt2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2) {
1180 ; CHECK-LABEL: test_int_x86_avx512_vpermt2var_hi_256:
1181 ; CHECK: # %bb.0:
1182 ; CHECK-NEXT: vpermi2w %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x75,0xc2]
1183 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1184 %1 = call <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16> %x1, <16 x i16> %x0, <16 x i16> %x2)
1185 ret <16 x i16> %1
1186 }
1188 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) {
1189 ; X86-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
1190 ; X86: # %bb.0:
1191 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1192 ; X86-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
1193 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1194 ; X86-NEXT: retl # encoding: [0xc3]
1195 ;
1196 ; X64-LABEL: test_int_x86_avx512_mask_vpermt2var_hi_256:
1197 ; X64: # %bb.0:
1198 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1199 ; X64-NEXT: vpermt2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x7d,0xca]
1200 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1201 ; X64-NEXT: retq # encoding: [0xc3]
1202 %1 = call <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16> %x1, <16 x i16> %x0, <16 x i16> %x2)
1203 %2 = bitcast i16 %x3 to <16 x i1>
1204 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x1
1205 ret <16 x i16> %3
1206 }
1208 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) {
1209 ; X86-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
1210 ; X86: # %bb.0:
1211 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1212 ; X86-NEXT: vpermi2w %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x75,0xc2]
1213 ; X86-NEXT: retl # encoding: [0xc3]
1214 ;
1215 ; X64-LABEL: test_int_x86_avx512_maskz_vpermt2var_hi_256:
1216 ; X64: # %bb.0:
1217 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1218 ; X64-NEXT: vpermi2w %ymm2, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x75,0xc2]
1219 ; X64-NEXT: retq # encoding: [0xc3]
1220 %1 = call <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16> %x1, <16 x i16> %x0, <16 x i16> %x2)
1221 %2 = bitcast i16 %x3 to <16 x i1>
1222 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
1223 ret <16 x i16> %3
1224 }
1226 declare <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16>, <8 x i16>, <8 x i16>)
1228 define <8 x i16>@test_int_x86_avx512_vpermi2var_hi_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2) {
1229 ; CHECK-LABEL: test_int_x86_avx512_vpermi2var_hi_128:
1230 ; CHECK: # %bb.0:
1231 ; CHECK-NEXT: vpermt2w %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x7d,0xc2]
1232 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1233 %1 = call <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2)
1234 ret <8 x i16> %1
1235 }
1237 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) {
1238 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
1239 ; X86: # %bb.0:
1240 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1241 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1242 ; X86-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
1243 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1244 ; X86-NEXT: retl # encoding: [0xc3]
1245 ;
1246 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_128:
1247 ; X64: # %bb.0:
1248 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1249 ; X64-NEXT: vpermi2w %xmm2, %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x75,0xca]
1250 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1251 ; X64-NEXT: retq # encoding: [0xc3]
1252 %1 = call <8 x i16> @llvm.x86.avx512.vpermi2var.hi.128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2)
1253 %2 = bitcast i8 %x3 to <8 x i1>
1254 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x1
1255 ret <8 x i16> %3
1256 }
1258 declare <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16>, <16 x i16>, <16 x i16>)
1260 define <16 x i16>@test_int_x86_avx512_vpermi2var_hi_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2) {
1261 ; CHECK-LABEL: test_int_x86_avx512_vpermi2var_hi_256:
1262 ; CHECK: # %bb.0:
1263 ; CHECK-NEXT: vpermt2w %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x7d,0xc2]
1264 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1265 %1 = call <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2)
1266 ret <16 x i16> %1
1267 }
1269 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) {
1270 ; X86-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
1271 ; X86: # %bb.0:
1272 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1273 ; X86-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
1274 ; X86-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1275 ; X86-NEXT: retl # encoding: [0xc3]
1276 ;
1277 ; X64-LABEL: test_int_x86_avx512_mask_vpermi2var_hi_256:
1278 ; X64: # %bb.0:
1279 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1280 ; X64-NEXT: vpermi2w %ymm2, %ymm0, %ymm1 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x75,0xca]
1281 ; X64-NEXT: vmovdqa %ymm1, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc1]
1282 ; X64-NEXT: retq # encoding: [0xc3]
1283 %1 = call <16 x i16> @llvm.x86.avx512.vpermi2var.hi.256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2)
1284 %2 = bitcast i16 %x3 to <16 x i1>
1285 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x1
1286 ret <16 x i16> %3
1287 }
1289 declare <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8>, <16 x i8>)
1291 define <16 x i8> @test_int_x86_avx512_mask_pavg_b_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
1292 ; X86-LABEL: test_int_x86_avx512_mask_pavg_b_128:
1293 ; X86: # %bb.0:
1294 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1295 ; X86-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
1296 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1297 ; X86-NEXT: retl # encoding: [0xc3]
1298 ;
1299 ; X64-LABEL: test_int_x86_avx512_mask_pavg_b_128:
1300 ; X64: # %bb.0:
1301 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1302 ; X64-NEXT: vpavgb %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe0,0xd1]
1303 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1304 ; X64-NEXT: retq # encoding: [0xc3]
1305 %1 = call <16 x i8> @llvm.x86.sse2.pavg.b(<16 x i8> %x0, <16 x i8> %x1)
1306 %2 = bitcast i16 %x3 to <16 x i1>
1307 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %x2
1308 ret <16 x i8> %3
1309 }
1311 declare <32 x i8> @llvm.x86.avx2.pavg.b(<32 x i8>, <32 x i8>)
1313 define <32 x i8> @test_int_x86_avx512_mask_pavg_b_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
1314 ; X86-LABEL: test_int_x86_avx512_mask_pavg_b_256:
1315 ; X86: # %bb.0:
1316 ; X86-NEXT: kmovd {{[0-9]+}}(%esp), %k1 # encoding: [0xc4,0xe1,0xf9,0x90,0x4c,0x24,0x04]
1317 ; X86-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
1318 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1319 ; X86-NEXT: retl # encoding: [0xc3]
1320 ;
1321 ; X64-LABEL: test_int_x86_avx512_mask_pavg_b_256:
1322 ; X64: # %bb.0:
1323 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1324 ; X64-NEXT: vpavgb %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe0,0xd1]
1325 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1326 ; X64-NEXT: retq # encoding: [0xc3]
1327 %1 = call <32 x i8> @llvm.x86.avx2.pavg.b(<32 x i8> %x0, <32 x i8> %x1)
1328 %2 = bitcast i32 %x3 to <32 x i1>
1329 %3 = select <32 x i1> %2, <32 x i8> %1, <32 x i8> %x2
1330 ret <32 x i8> %3
1331 }
1333 declare <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16>, <8 x i16>)
1335 define <8 x i16> @test_int_x86_avx512_mask_pavg_w_128(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 %x3) {
1336 ; X86-LABEL: test_int_x86_avx512_mask_pavg_w_128:
1337 ; X86: # %bb.0:
1338 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1339 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1340 ; X86-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
1341 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1342 ; X86-NEXT: retl # encoding: [0xc3]
1343 ;
1344 ; X64-LABEL: test_int_x86_avx512_mask_pavg_w_128:
1345 ; X64: # %bb.0:
1346 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1347 ; X64-NEXT: vpavgw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe3,0xd1]
1348 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1349 ; X64-NEXT: retq # encoding: [0xc3]
1350 %1 = call <8 x i16> @llvm.x86.sse2.pavg.w(<8 x i16> %x0, <8 x i16> %x1)
1351 %2 = bitcast i8 %x3 to <8 x i1>
1352 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
1353 ret <8 x i16> %3
1354 }
1356 declare <16 x i16> @llvm.x86.avx2.pavg.w(<16 x i16>, <16 x i16>)
1358 define <16 x i16> @test_int_x86_avx512_mask_pavg_w_256(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 %x3) {
1359 ; X86-LABEL: test_int_x86_avx512_mask_pavg_w_256:
1360 ; X86: # %bb.0:
1361 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1362 ; X86-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
1363 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1364 ; X86-NEXT: retl # encoding: [0xc3]
1365 ;
1366 ; X64-LABEL: test_int_x86_avx512_mask_pavg_w_256:
1367 ; X64: # %bb.0:
1368 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1369 ; X64-NEXT: vpavgw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe3,0xd1]
1370 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1371 ; X64-NEXT: retq # encoding: [0xc3]
1372 %1 = call <16 x i16> @llvm.x86.avx2.pavg.w(<16 x i16> %x0, <16 x i16> %x1)
1373 %2 = bitcast i16 %x3 to <16 x i1>
1374 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
1375 ret <16 x i16> %3
1376 }
1378 declare <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16>, <8 x i16>)
1380 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) {
1381 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
1382 ; X86: # %bb.0:
1383 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1384 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1385 ; X86-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
1386 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1387 ; X86-NEXT: retl # encoding: [0xc3]
1388 ;
1389 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_128:
1390 ; X64: # %bb.0:
1391 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1392 ; X64-NEXT: vpmulhuw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe4,0xd1]
1393 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1394 ; X64-NEXT: retq # encoding: [0xc3]
1395 %1 = call <8 x i16> @llvm.x86.sse2.pmulhu.w(<8 x i16> %x0, <8 x i16> %x1)
1396 %2 = bitcast i8 %x3 to <8 x i1>
1397 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
1398 ret <8 x i16> %3
1399 }
1401 declare <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16>, <16 x i16>)
1403 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) {
1404 ; X86-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
1405 ; X86: # %bb.0:
1406 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1407 ; X86-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
1408 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1409 ; X86-NEXT: retl # encoding: [0xc3]
1410 ;
1411 ; X64-LABEL: test_int_x86_avx512_mask_pmulhu_w_256:
1412 ; X64: # %bb.0:
1413 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1414 ; X64-NEXT: vpmulhuw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe4,0xd1]
1415 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1416 ; X64-NEXT: retq # encoding: [0xc3]
1417 %1 = call <16 x i16> @llvm.x86.avx2.pmulhu.w(<16 x i16> %x0, <16 x i16> %x1)
1418 %2 = bitcast i16 %x3 to <16 x i1>
1419 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
1420 ret <16 x i16> %3
1421 }
1423 declare <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16>, <8 x i16>)
1425 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) {
1426 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
1427 ; X86: # %bb.0:
1428 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1429 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1430 ; X86-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
1431 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1432 ; X86-NEXT: retl # encoding: [0xc3]
1433 ;
1434 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_128:
1435 ; X64: # %bb.0:
1436 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1437 ; X64-NEXT: vpmulhw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xe5,0xd1]
1438 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1439 ; X64-NEXT: retq # encoding: [0xc3]
1440 %1 = call <8 x i16> @llvm.x86.sse2.pmulh.w(<8 x i16> %x0, <8 x i16> %x1)
1441 %2 = bitcast i8 %x3 to <8 x i1>
1442 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
1443 ret <8 x i16> %3
1444 }
1446 declare <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16>, <16 x i16>)
1448 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) {
1449 ; X86-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
1450 ; X86: # %bb.0:
1451 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1452 ; X86-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
1453 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1454 ; X86-NEXT: retl # encoding: [0xc3]
1455 ;
1456 ; X64-LABEL: test_int_x86_avx512_mask_pmulh_w_256:
1457 ; X64: # %bb.0:
1458 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1459 ; X64-NEXT: vpmulhw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xe5,0xd1]
1460 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1461 ; X64-NEXT: retq # encoding: [0xc3]
1462 %1 = call <16 x i16> @llvm.x86.avx2.pmulh.w(<16 x i16> %x0, <16 x i16> %x1)
1463 %2 = bitcast i16 %x3 to <16 x i1>
1464 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
1465 ret <16 x i16> %3
1466 }
1468 declare <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16>, <8 x i16>)
1470 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) {
1471 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
1472 ; X86: # %bb.0:
1473 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1474 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1475 ; X86-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
1476 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1477 ; X86-NEXT: retl # encoding: [0xc3]
1478 ;
1479 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_128:
1480 ; X64: # %bb.0:
1481 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1482 ; X64-NEXT: vpmulhrsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x0b,0xd1]
1483 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1484 ; X64-NEXT: retq # encoding: [0xc3]
1485 %1 = call <8 x i16> @llvm.x86.ssse3.pmul.hr.sw.128(<8 x i16> %x0, <8 x i16> %x1)
1486 %2 = bitcast i8 %x3 to <8 x i1>
1487 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
1488 ret <8 x i16> %3
1489 }
1491 declare <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16>, <16 x i16>)
1493 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) {
1494 ; X86-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
1495 ; X86: # %bb.0:
1496 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1497 ; X86-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
1498 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1499 ; X86-NEXT: retl # encoding: [0xc3]
1500 ;
1501 ; X64-LABEL: test_int_x86_avx512_mask_pmulhr_sw_256:
1502 ; X64: # %bb.0:
1503 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1504 ; X64-NEXT: vpmulhrsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x0b,0xd1]
1505 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1506 ; X64-NEXT: retq # encoding: [0xc3]
1507 %1 = call <16 x i16> @llvm.x86.avx2.pmul.hr.sw(<16 x i16> %x0, <16 x i16> %x1)
1508 %2 = bitcast i16 %x3 to <16 x i1>
1509 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
1510 ret <16 x i16> %3
1511 }
1513 declare <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16>, <16 x i8>, i8)
1515 define { <16 x i8>, <16 x i8>, <16 x i8> } @test_int_x86_avx512_mask_pmov_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
1516 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_128:
1517 ; X86: # %bb.0:
1518 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1519 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1520 ; X86-NEXT: vpmovwb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x30,0xc3]
1521 ; X86-NEXT: vpmovwb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x30,0xc1]
1522 ; X86-NEXT: vpmovwb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x30,0xc2]
1523 ; X86-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1524 ; X86-NEXT: retl # encoding: [0xc3]
1525 ;
1526 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_128:
1527 ; X64: # %bb.0:
1528 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1529 ; X64-NEXT: vpmovwb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x30,0xc3]
1530 ; X64-NEXT: vpmovwb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x30,0xc1]
1531 ; X64-NEXT: vpmovwb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x30,0xc2]
1532 ; X64-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1533 ; X64-NEXT: retq # encoding: [0xc3]
1534 %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
1535 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
1536 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmov.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
1537 %res3 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } poison, <16 x i8> %res0, 0
1538 %res4 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res3, <16 x i8> %res1, 1
1539 %res5 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res4, <16 x i8> %res2, 2
1540 ret { <16 x i8>, <16 x i8>, <16 x i8> } %res5
1541 }
1543 declare void @llvm.x86.avx512.mask.pmov.wb.mem.128(ptr %ptr, <8 x i16>, i8)
1545 define void @test_int_x86_avx512_mask_pmov_wb_mem_128(ptr %ptr, <8 x i16> %x1, i8 %x2) {
1546 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_128:
1547 ; X86: # %bb.0:
1548 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
1549 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1550 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1551 ; X86-NEXT: vpmovwb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x30,0x00]
1552 ; X86-NEXT: vpmovwb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x30,0x00]
1553 ; X86-NEXT: retl # encoding: [0xc3]
1554 ;
1555 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_128:
1556 ; X64: # %bb.0:
1557 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1558 ; X64-NEXT: vpmovwb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x30,0x07]
1559 ; X64-NEXT: vpmovwb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x30,0x07]
1560 ; X64-NEXT: retq # encoding: [0xc3]
1561 call void @llvm.x86.avx512.mask.pmov.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 -1)
1562 call void @llvm.x86.avx512.mask.pmov.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 %x2)
1563 ret void
1564 }
1566 declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16>, <16 x i8>, i8)
1568 define { <16 x i8>, <16 x i8>, <16 x i8> } @test_int_x86_avx512_mask_pmovs_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
1569 ; X86-LABEL: test_int_x86_avx512_mask_pmovs_wb_128:
1570 ; X86: # %bb.0:
1571 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1572 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1573 ; X86-NEXT: vpmovswb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x20,0xc3]
1574 ; X86-NEXT: vpmovswb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x20,0xc1]
1575 ; X86-NEXT: vpmovswb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x20,0xc2]
1576 ; X86-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1577 ; X86-NEXT: retl # encoding: [0xc3]
1578 ;
1579 ; X64-LABEL: test_int_x86_avx512_mask_pmovs_wb_128:
1580 ; X64: # %bb.0:
1581 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1582 ; X64-NEXT: vpmovswb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x20,0xc3]
1583 ; X64-NEXT: vpmovswb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x20,0xc1]
1584 ; X64-NEXT: vpmovswb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x20,0xc2]
1585 ; X64-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1586 ; X64-NEXT: retq # encoding: [0xc3]
1587 %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
1588 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
1589 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
1590 %res3 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } poison, <16 x i8> %res0, 0
1591 %res4 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res3, <16 x i8> %res1, 1
1592 %res5 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res4, <16 x i8> %res2, 2
1593 ret { <16 x i8>, <16 x i8>, <16 x i8> } %res5
1594 }
1596 declare void @llvm.x86.avx512.mask.pmovs.wb.mem.128(ptr %ptr, <8 x i16>, i8)
1598 define void @test_int_x86_avx512_mask_pmovs_wb_mem_128(ptr %ptr, <8 x i16> %x1, i8 %x2) {
1599 ; X86-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_128:
1600 ; X86: # %bb.0:
1601 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
1602 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1603 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1604 ; X86-NEXT: vpmovswb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x20,0x00]
1605 ; X86-NEXT: vpmovswb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x20,0x00]
1606 ; X86-NEXT: retl # encoding: [0xc3]
1607 ;
1608 ; X64-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_128:
1609 ; X64: # %bb.0:
1610 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1611 ; X64-NEXT: vpmovswb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x20,0x07]
1612 ; X64-NEXT: vpmovswb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x20,0x07]
1613 ; X64-NEXT: retq # encoding: [0xc3]
1614 call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 -1)
1615 call void @llvm.x86.avx512.mask.pmovs.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 %x2)
1616 ret void
1617 }
1619 declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16>, <16 x i8>, i8)
1621 define { <16 x i8>, <16 x i8>, <16 x i8> } @test_int_x86_avx512_mask_pmovus_wb_128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2) {
1622 ; X86-LABEL: test_int_x86_avx512_mask_pmovus_wb_128:
1623 ; X86: # %bb.0:
1624 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1625 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1626 ; X86-NEXT: vpmovuswb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x10,0xc3]
1627 ; X86-NEXT: vpmovuswb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x10,0xc1]
1628 ; X86-NEXT: vpmovuswb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x10,0xc2]
1629 ; X86-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1630 ; X86-NEXT: retl # encoding: [0xc3]
1631 ;
1632 ; X64-LABEL: test_int_x86_avx512_mask_pmovus_wb_128:
1633 ; X64: # %bb.0:
1634 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1635 ; X64-NEXT: vpmovuswb %xmm0, %xmm3 # encoding: [0x62,0xf2,0x7e,0x08,0x10,0xc3]
1636 ; X64-NEXT: vpmovuswb %xmm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x10,0xc1]
1637 ; X64-NEXT: vpmovuswb %xmm0, %xmm2 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0x89,0x10,0xc2]
1638 ; X64-NEXT: vmovdqa %xmm3, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc3]
1639 ; X64-NEXT: retq # encoding: [0xc3]
1640 %res0 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 -1)
1641 %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> %x1, i8 %x2)
1642 %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.128(<8 x i16> %x0, <16 x i8> zeroinitializer, i8 %x2)
1643 %res3 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } poison, <16 x i8> %res0, 0
1644 %res4 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res3, <16 x i8> %res1, 1
1645 %res5 = insertvalue { <16 x i8>, <16 x i8>, <16 x i8> } %res4, <16 x i8> %res2, 2
1646 ret { <16 x i8>, <16 x i8>, <16 x i8> } %res5
1647 }
1649 declare void @llvm.x86.avx512.mask.pmovus.wb.mem.128(ptr %ptr, <8 x i16>, i8)
1651 define void @test_int_x86_avx512_mask_pmovus_wb_mem_128(ptr %ptr, <8 x i16> %x1, i8 %x2) {
1652 ; X86-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_128:
1653 ; X86: # %bb.0:
1654 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x08]
1655 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1656 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1657 ; X86-NEXT: vpmovuswb %xmm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x08,0x10,0x00]
1658 ; X86-NEXT: vpmovuswb %xmm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x10,0x00]
1659 ; X86-NEXT: retl # encoding: [0xc3]
1660 ;
1661 ; X64-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_128:
1662 ; X64: # %bb.0:
1663 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1664 ; X64-NEXT: vpmovuswb %xmm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x08,0x10,0x07]
1665 ; X64-NEXT: vpmovuswb %xmm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x09,0x10,0x07]
1666 ; X64-NEXT: retq # encoding: [0xc3]
1667 call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 -1)
1668 call void @llvm.x86.avx512.mask.pmovus.wb.mem.128(ptr %ptr, <8 x i16> %x1, i8 %x2)
1669 ret void
1670 }
1672 define <16 x i8>@test_int_x86_avx512_pmov_wb_256(<16 x i16> %x0) {
1673 ; CHECK-LABEL: test_int_x86_avx512_pmov_wb_256:
1674 ; CHECK: # %bb.0:
1675 ; CHECK-NEXT: vpmovwb %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x30,0xc0]
1676 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1677 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1678 %1 = trunc <16 x i16> %x0 to <16 x i8>
1679 ret <16 x i8> %1
1680 }
1682 define <16 x i8>@test_int_x86_avx512_mask_pmov_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
1683 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
1684 ; X86: # %bb.0:
1685 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1686 ; X86-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
1687 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1688 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1689 ; X86-NEXT: retl # encoding: [0xc3]
1690 ;
1691 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_256:
1692 ; X64: # %bb.0:
1693 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1694 ; X64-NEXT: vpmovwb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0xc1]
1695 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1696 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1697 ; X64-NEXT: retq # encoding: [0xc3]
1698 %1 = trunc <16 x i16> %x0 to <16 x i8>
1699 %2 = bitcast i16 %x2 to <16 x i1>
1700 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> %x1
1701 ret <16 x i8> %3
1702 }
1704 define <16 x i8>@test_int_x86_avx512_maskz_pmov_wb_256(<16 x i16> %x0, i16 %x2) {
1705 ; X86-LABEL: test_int_x86_avx512_maskz_pmov_wb_256:
1706 ; X86: # %bb.0:
1707 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1708 ; X86-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
1709 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1710 ; X86-NEXT: retl # encoding: [0xc3]
1711 ;
1712 ; X64-LABEL: test_int_x86_avx512_maskz_pmov_wb_256:
1713 ; X64: # %bb.0:
1714 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1715 ; X64-NEXT: vpmovwb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x30,0xc0]
1716 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1717 ; X64-NEXT: retq # encoding: [0xc3]
1718 %1 = trunc <16 x i16> %x0 to <16 x i8>
1719 %2 = bitcast i16 %x2 to <16 x i1>
1720 %3 = select <16 x i1> %2, <16 x i8> %1, <16 x i8> zeroinitializer
1721 ret <16 x i8> %3
1722 }
1724 declare void @llvm.x86.avx512.mask.pmov.wb.mem.256(ptr %ptr, <16 x i16>, i16)
1726 define void @test_int_x86_avx512_mask_pmov_wb_mem_256(ptr %ptr, <16 x i16> %x1, i16 %x2) {
1727 ; X86-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_256:
1728 ; X86: # %bb.0:
1729 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1730 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1731 ; X86-NEXT: vpmovwb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x30,0x00]
1732 ; X86-NEXT: vpmovwb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0x00]
1733 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1734 ; X86-NEXT: retl # encoding: [0xc3]
1735 ;
1736 ; X64-LABEL: test_int_x86_avx512_mask_pmov_wb_mem_256:
1737 ; X64: # %bb.0:
1738 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1739 ; X64-NEXT: vpmovwb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x30,0x07]
1740 ; X64-NEXT: vpmovwb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x30,0x07]
1741 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1742 ; X64-NEXT: retq # encoding: [0xc3]
1743 call void @llvm.x86.avx512.mask.pmov.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 -1)
1744 call void @llvm.x86.avx512.mask.pmov.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 %x2)
1745 ret void
1746 }
1748 declare <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16>, <16 x i8>, i16)
1750 define <16 x i8>@test_int_x86_avx512_pmovs_wb_256(<16 x i16> %x0, <16 x i8> %x1) {
1751 ; CHECK-LABEL: test_int_x86_avx512_pmovs_wb_256:
1752 ; CHECK: # %bb.0:
1753 ; CHECK-NEXT: vpmovswb %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x20,0xc0]
1754 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1755 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1756 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
1757 ret <16 x i8> %res
1758 }
1760 define <16 x i8>@test_int_x86_avx512_mask_pmovs_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
1761 ; X86-LABEL: test_int_x86_avx512_mask_pmovs_wb_256:
1762 ; X86: # %bb.0:
1763 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1764 ; X86-NEXT: vpmovswb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x20,0xc1]
1765 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1766 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1767 ; X86-NEXT: retl # encoding: [0xc3]
1768 ;
1769 ; X64-LABEL: test_int_x86_avx512_mask_pmovs_wb_256:
1770 ; X64: # %bb.0:
1771 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1772 ; X64-NEXT: vpmovswb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x20,0xc1]
1773 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1774 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1775 ; X64-NEXT: retq # encoding: [0xc3]
1776 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
1777 ret <16 x i8> %res
1778 }
1780 define <16 x i8>@test_int_x86_avx512_maskz_pmovs_wb_256(<16 x i16> %x0, i16 %x2) {
1781 ; X86-LABEL: test_int_x86_avx512_maskz_pmovs_wb_256:
1782 ; X86: # %bb.0:
1783 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1784 ; X86-NEXT: vpmovswb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x20,0xc0]
1785 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1786 ; X86-NEXT: retl # encoding: [0xc3]
1787 ;
1788 ; X64-LABEL: test_int_x86_avx512_maskz_pmovs_wb_256:
1789 ; X64: # %bb.0:
1790 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1791 ; X64-NEXT: vpmovswb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x20,0xc0]
1792 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1793 ; X64-NEXT: retq # encoding: [0xc3]
1794 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovs.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
1795 ret <16 x i8> %res
1796 }
1798 declare void @llvm.x86.avx512.mask.pmovs.wb.mem.256(ptr %ptr, <16 x i16>, i16)
1800 define void @test_int_x86_avx512_mask_pmovs_wb_mem_256(ptr %ptr, <16 x i16> %x1, i16 %x2) {
1801 ; X86-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_256:
1802 ; X86: # %bb.0:
1803 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1804 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1805 ; X86-NEXT: vpmovswb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x20,0x00]
1806 ; X86-NEXT: vpmovswb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x20,0x00]
1807 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1808 ; X86-NEXT: retl # encoding: [0xc3]
1809 ;
1810 ; X64-LABEL: test_int_x86_avx512_mask_pmovs_wb_mem_256:
1811 ; X64: # %bb.0:
1812 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1813 ; X64-NEXT: vpmovswb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x20,0x07]
1814 ; X64-NEXT: vpmovswb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x20,0x07]
1815 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1816 ; X64-NEXT: retq # encoding: [0xc3]
1817 call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 -1)
1818 call void @llvm.x86.avx512.mask.pmovs.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 %x2)
1819 ret void
1820 }
1822 declare <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16>, <16 x i8>, i16)
1824 define <16 x i8>@test_int_x86_avx512_pmovus_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
1825 ; CHECK-LABEL: test_int_x86_avx512_pmovus_wb_256:
1826 ; CHECK: # %bb.0:
1827 ; CHECK-NEXT: vpmovuswb %ymm0, %xmm0 # encoding: [0x62,0xf2,0x7e,0x28,0x10,0xc0]
1828 ; CHECK-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1829 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
1830 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 -1)
1831 ret <16 x i8> %res
1832 }
1834 define <16 x i8>@test_int_x86_avx512_mask_pmovus_wb_256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2) {
1835 ; X86-LABEL: test_int_x86_avx512_mask_pmovus_wb_256:
1836 ; X86: # %bb.0:
1837 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1838 ; X86-NEXT: vpmovuswb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x10,0xc1]
1839 ; X86-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1840 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1841 ; X86-NEXT: retl # encoding: [0xc3]
1842 ;
1843 ; X64-LABEL: test_int_x86_avx512_mask_pmovus_wb_256:
1844 ; X64: # %bb.0:
1845 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1846 ; X64-NEXT: vpmovuswb %ymm0, %xmm1 {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x10,0xc1]
1847 ; X64-NEXT: vmovdqa %xmm1, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc1]
1848 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1849 ; X64-NEXT: retq # encoding: [0xc3]
1850 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> %x1, i16 %x2)
1851 ret <16 x i8> %res
1852 }
1854 define <16 x i8>@test_int_x86_avx512_maskz_pmovus_wb_256(<16 x i16> %x0, i16 %x2) {
1855 ; X86-LABEL: test_int_x86_avx512_maskz_pmovus_wb_256:
1856 ; X86: # %bb.0:
1857 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1858 ; X86-NEXT: vpmovuswb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x10,0xc0]
1859 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1860 ; X86-NEXT: retl # encoding: [0xc3]
1861 ;
1862 ; X64-LABEL: test_int_x86_avx512_maskz_pmovus_wb_256:
1863 ; X64: # %bb.0:
1864 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1865 ; X64-NEXT: vpmovuswb %ymm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0x7e,0xa9,0x10,0xc0]
1866 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1867 ; X64-NEXT: retq # encoding: [0xc3]
1868 %res = call <16 x i8> @llvm.x86.avx512.mask.pmovus.wb.256(<16 x i16> %x0, <16 x i8> zeroinitializer, i16 %x2)
1869 ret <16 x i8> %res
1870 }
1872 declare void @llvm.x86.avx512.mask.pmovus.wb.mem.256(ptr %ptr, <16 x i16>, i16)
1874 define void @test_int_x86_avx512_mask_pmovus_wb_mem_256(ptr %ptr, <16 x i16> %x1, i16 %x2) {
1875 ; X86-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_256:
1876 ; X86: # %bb.0:
1877 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x08]
1878 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
1879 ; X86-NEXT: vpmovuswb %ymm0, (%eax) # encoding: [0x62,0xf2,0x7e,0x28,0x10,0x00]
1880 ; X86-NEXT: vpmovuswb %ymm0, (%eax) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x10,0x00]
1881 ; X86-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1882 ; X86-NEXT: retl # encoding: [0xc3]
1883 ;
1884 ; X64-LABEL: test_int_x86_avx512_mask_pmovus_wb_mem_256:
1885 ; X64: # %bb.0:
1886 ; X64-NEXT: kmovd %esi, %k1 # encoding: [0xc5,0xfb,0x92,0xce]
1887 ; X64-NEXT: vpmovuswb %ymm0, (%rdi) # encoding: [0x62,0xf2,0x7e,0x28,0x10,0x07]
1888 ; X64-NEXT: vpmovuswb %ymm0, (%rdi) {%k1} # encoding: [0x62,0xf2,0x7e,0x29,0x10,0x07]
1889 ; X64-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77]
1890 ; X64-NEXT: retq # encoding: [0xc3]
1891 call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 -1)
1892 call void @llvm.x86.avx512.mask.pmovus.wb.mem.256(ptr %ptr, <16 x i16> %x1, i16 %x2)
1893 ret void
1894 }
1896 declare <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16>, <8 x i16>)
1898 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) {
1899 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
1900 ; X86: # %bb.0:
1901 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1902 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1903 ; X86-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
1904 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1905 ; X86-NEXT: retl # encoding: [0xc3]
1906 ;
1907 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_128:
1908 ; X64: # %bb.0:
1909 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1910 ; X64-NEXT: vpmaddwd %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x09,0xf5,0xd1]
1911 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1912 ; X64-NEXT: retq # encoding: [0xc3]
1913 %1 = call <4 x i32> @llvm.x86.sse2.pmadd.wd(<8 x i16> %x0, <8 x i16> %x1)
1914 %2 = bitcast i8 %x3 to <8 x i1>
1915 %extract = shufflevector <8 x i1> %2, <8 x i1> %2, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
1916 %3 = select <4 x i1> %extract, <4 x i32> %1, <4 x i32> %x2
1917 ret <4 x i32> %3
1918 }
1920 declare <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16>, <16 x i16>)
1922 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) {
1923 ; X86-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
1924 ; X86: # %bb.0:
1925 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1926 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1927 ; X86-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
1928 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1929 ; X86-NEXT: retl # encoding: [0xc3]
1930 ;
1931 ; X64-LABEL: test_int_x86_avx512_mask_pmaddw_d_256:
1932 ; X64: # %bb.0:
1933 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1934 ; X64-NEXT: vpmaddwd %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf1,0x7d,0x29,0xf5,0xd1]
1935 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1936 ; X64-NEXT: retq # encoding: [0xc3]
1937 %1 = call <8 x i32> @llvm.x86.avx2.pmadd.wd(<16 x i16> %x0, <16 x i16> %x1)
1938 %2 = bitcast i8 %x3 to <8 x i1>
1939 %3 = select <8 x i1> %2, <8 x i32> %1, <8 x i32> %x2
1940 ret <8 x i32> %3
1941 }
1943 declare <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<16 x i8>, <16 x i8>)
1945 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) {
1946 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
1947 ; X86: # %bb.0:
1948 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1949 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1950 ; X86-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
1951 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1952 ; X86-NEXT: retl # encoding: [0xc3]
1953 ;
1954 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_128:
1955 ; X64: # %bb.0:
1956 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1957 ; X64-NEXT: vpmaddubsw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x09,0x04,0xd1]
1958 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
1959 ; X64-NEXT: retq # encoding: [0xc3]
1960 %1 = call <8 x i16> @llvm.x86.ssse3.pmadd.ub.sw.128(<16 x i8> %x0, <16 x i8> %x1)
1961 %2 = bitcast i8 %x3 to <8 x i1>
1962 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
1963 ret <8 x i16> %3
1964 }
1966 declare <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8>, <32 x i8>)
1968 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) {
1969 ; X86-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
1970 ; X86: # %bb.0:
1971 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
1972 ; X86-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
1973 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1974 ; X86-NEXT: retl # encoding: [0xc3]
1975 ;
1976 ; X64-LABEL: test_int_x86_avx512_mask_pmaddubs_w_256:
1977 ; X64: # %bb.0:
1978 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
1979 ; X64-NEXT: vpmaddubsw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0x7d,0x29,0x04,0xd1]
1980 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
1981 ; X64-NEXT: retq # encoding: [0xc3]
1982 %1 = call <16 x i16> @llvm.x86.avx2.pmadd.ub.sw(<32 x i8> %x0, <32 x i8> %x1)
1983 %2 = bitcast i16 %x3 to <16 x i1>
1984 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
1985 ret <16 x i16> %3
1986 }
1988 declare <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8>, <16 x i8>, i32)
1990 define { <8 x i16>, <8 x i16>, <8 x i16> } @test_int_x86_avx512_mask_dbpsadbw_128(<16 x i8> %x0, <16 x i8> %x1, <8 x i16> %x3, i8 %x4) {
1991 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
1992 ; X86: # %bb.0:
1993 ; X86-NEXT: vmovdqa %xmm2, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xe2]
1994 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
1995 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
1996 ; X86-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xe1,0x02]
1997 ; X86-NEXT: vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
1998 ; X86-NEXT: vdbpsadbw $4, %xmm1, %xmm0, %xmm2 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd1,0x04]
1999 ; X86-NEXT: vmovdqa %xmm4, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc4]
2000 ; X86-NEXT: vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
2001 ; X86-NEXT: retl # encoding: [0xc3]
2002 ;
2003 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_128:
2004 ; X64: # %bb.0:
2005 ; X64-NEXT: vmovdqa %xmm2, %xmm4 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xe2]
2006 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2007 ; X64-NEXT: vdbpsadbw $2, %xmm1, %xmm0, %xmm4 {%k1} # encoding: [0x62,0xf3,0x7d,0x09,0x42,0xe1,0x02]
2008 ; X64-NEXT: vdbpsadbw $3, %xmm1, %xmm0, %xmm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0x89,0x42,0xd9,0x03]
2009 ; X64-NEXT: vdbpsadbw $4, %xmm1, %xmm0, %xmm2 # encoding: [0x62,0xf3,0x7d,0x08,0x42,0xd1,0x04]
2010 ; X64-NEXT: vmovdqa %xmm4, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc4]
2011 ; X64-NEXT: vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
2012 ; X64-NEXT: retq # encoding: [0xc3]
2013 %1 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 2)
2014 %2 = bitcast i8 %x4 to <8 x i1>
2015 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x3
2016 %4 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 3)
2017 %5 = bitcast i8 %x4 to <8 x i1>
2018 %6 = select <8 x i1> %5, <8 x i16> %4, <8 x i16> zeroinitializer
2019 %7 = call <8 x i16> @llvm.x86.avx512.dbpsadbw.128(<16 x i8> %x0, <16 x i8> %x1, i32 4)
2020 %res3 = insertvalue { <8 x i16>, <8 x i16>, <8 x i16> } poison, <8 x i16> %3, 0
2021 %res4 = insertvalue { <8 x i16>, <8 x i16>, <8 x i16> } %res3, <8 x i16> %6, 1
2022 %res5 = insertvalue { <8 x i16>, <8 x i16>, <8 x i16> } %res4, <8 x i16> %7, 2
2023 ret { <8 x i16>, <8 x i16>, <8 x i16> } %res5
2024 }
2026 declare <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8>, <32 x i8>, i32)
2028 define { <16 x i16>, <16 x i16>, <16 x i16> } @test_int_x86_avx512_mask_dbpsadbw_256(<32 x i8> %x0, <32 x i8> %x1, <16 x i16> %x3, i16 %x4) {
2029 ; X86-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
2030 ; X86: # %bb.0:
2031 ; X86-NEXT: vmovdqa %ymm2, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xe2]
2032 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2033 ; X86-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xe1,0x02]
2034 ; X86-NEXT: vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
2035 ; X86-NEXT: vdbpsadbw $4, %ymm1, %ymm0, %ymm2 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd1,0x04]
2036 ; X86-NEXT: vmovdqa %ymm4, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc4]
2037 ; X86-NEXT: vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
2038 ; X86-NEXT: retl # encoding: [0xc3]
2039 ;
2040 ; X64-LABEL: test_int_x86_avx512_mask_dbpsadbw_256:
2041 ; X64: # %bb.0:
2042 ; X64-NEXT: vmovdqa %ymm2, %ymm4 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xe2]
2043 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2044 ; X64-NEXT: vdbpsadbw $2, %ymm1, %ymm0, %ymm4 {%k1} # encoding: [0x62,0xf3,0x7d,0x29,0x42,0xe1,0x02]
2045 ; X64-NEXT: vdbpsadbw $3, %ymm1, %ymm0, %ymm3 {%k1} {z} # encoding: [0x62,0xf3,0x7d,0xa9,0x42,0xd9,0x03]
2046 ; X64-NEXT: vdbpsadbw $4, %ymm1, %ymm0, %ymm2 # encoding: [0x62,0xf3,0x7d,0x28,0x42,0xd1,0x04]
2047 ; X64-NEXT: vmovdqa %ymm4, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc4]
2048 ; X64-NEXT: vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
2049 ; X64-NEXT: retq # encoding: [0xc3]
2050 %1 = call <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 2)
2051 %2 = bitcast i16 %x4 to <16 x i1>
2052 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x3
2053 %4 = call <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 3)
2054 %5 = bitcast i16 %x4 to <16 x i1>
2055 %6 = select <16 x i1> %5, <16 x i16> %4, <16 x i16> zeroinitializer
2056 %7 = call <16 x i16> @llvm.x86.avx512.dbpsadbw.256(<32 x i8> %x0, <32 x i8> %x1, i32 4)
2057 %res3 = insertvalue { <16 x i16>, <16 x i16>, <16 x i16> } poison, <16 x i16> %3, 0
2058 %res4 = insertvalue { <16 x i16>, <16 x i16>, <16 x i16> } %res3, <16 x i16> %6, 1
2059 %res5 = insertvalue { <16 x i16>, <16 x i16>, <16 x i16> } %res4, <16 x i16> %7, 2
2060 ret { <16 x i16>, <16 x i16>, <16 x i16> } %res5
2061 }
2063 declare <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16>, <16 x i16>, <16 x i16>, i16)
2065 define <16 x i16>@test_int_x86_avx512_psrlv16_hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2) {
2066 ; CHECK-LABEL: test_int_x86_avx512_psrlv16_hi:
2067 ; CHECK: # %bb.0:
2068 ; CHECK-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0xc1]
2069 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2070 %res = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> %x2, i16 -1)
2071 ret <16 x i16> %res
2072 }
2074 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) {
2075 ; X86-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
2076 ; X86: # %bb.0:
2077 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2078 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
2079 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2080 ; X86-NEXT: retl # encoding: [0xc3]
2081 ;
2082 ; X64-LABEL: test_int_x86_avx512_mask_psrlv16_hi:
2083 ; X64: # %bb.0:
2084 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2085 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x10,0xd1]
2086 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2087 ; X64-NEXT: retq # encoding: [0xc3]
2088 %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)
2089 ret <16 x i16> %res
2090 }
2092 define <16 x i16>@test_int_x86_avx512_maskz_psrlv16_hi(<16 x i16> %x0, <16 x i16> %x1, i16 %x3) {
2093 ; X86-LABEL: test_int_x86_avx512_maskz_psrlv16_hi:
2094 ; X86: # %bb.0:
2095 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2096 ; X86-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
2097 ; X86-NEXT: retl # encoding: [0xc3]
2098 ;
2099 ; X64-LABEL: test_int_x86_avx512_maskz_psrlv16_hi:
2100 ; X64: # %bb.0:
2101 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2102 ; X64-NEXT: vpsrlvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x10,0xc1]
2103 ; X64-NEXT: retq # encoding: [0xc3]
2104 %res = call <16 x i16> @llvm.x86.avx512.mask.psrlv16.hi(<16 x i16> %x0, <16 x i16> %x1, <16 x i16> zeroinitializer, i16 %x3)
2105 ret <16 x i16> %res
2106 }
2108 declare <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16>, <8 x i16>, <8 x i16>, i8)
2110 define <8 x i16>@test_int_x86_avx512_psrlv8_hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2) {
2111 ; CHECK-LABEL: test_int_x86_avx512_psrlv8_hi:
2112 ; CHECK: # %bb.0:
2113 ; CHECK-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0xc1]
2114 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2115 %res = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> %x2, i8 -1)
2116 ret <8 x i16> %res
2117 }
2119 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) {
2120 ; X86-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
2121 ; X86: # %bb.0:
2122 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2123 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2124 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
2125 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2126 ; X86-NEXT: retl # encoding: [0xc3]
2127 ;
2128 ; X64-LABEL: test_int_x86_avx512_mask_psrlv8_hi:
2129 ; X64: # %bb.0:
2130 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2131 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x10,0xd1]
2132 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2133 ; X64-NEXT: retq # encoding: [0xc3]
2134 %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)
2135 ret <8 x i16> %res
2136 }
2138 define <8 x i16>@test_int_x86_avx512_maskz_psrlv8_hi(<8 x i16> %x0, <8 x i16> %x1, i8 %x3) {
2139 ; X86-LABEL: test_int_x86_avx512_maskz_psrlv8_hi:
2140 ; X86: # %bb.0:
2141 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2142 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2143 ; X86-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
2144 ; X86-NEXT: retl # encoding: [0xc3]
2145 ;
2146 ; X64-LABEL: test_int_x86_avx512_maskz_psrlv8_hi:
2147 ; X64: # %bb.0:
2148 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2149 ; X64-NEXT: vpsrlvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x10,0xc1]
2150 ; X64-NEXT: retq # encoding: [0xc3]
2151 %res = call <8 x i16> @llvm.x86.avx512.mask.psrlv8.hi(<8 x i16> %x0, <8 x i16> %x1, <8 x i16> zeroinitializer, i8 %x3)
2152 ret <8 x i16> %res
2153 }
2156 define <8 x i16> @test_int_x86_avx512_psrlv_w_128_const() optsize {
2157 ; X86-LABEL: test_int_x86_avx512_psrlv_w_128_const:
2158 ; X86: # %bb.0:
2159 ; X86-NEXT: vmovdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,65535]
2160 ; X86-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
2161 ; X86-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2162 ; X86-NEXT: vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0x05,A,A,A,A]
2163 ; X86-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2164 ; X86-NEXT: retl # encoding: [0xc3]
2165 ;
2166 ; X64-LABEL: test_int_x86_avx512_psrlv_w_128_const:
2167 ; X64: # %bb.0:
2168 ; X64-NEXT: vmovdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,65535]
2169 ; X64-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
2170 ; X64-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2171 ; X64-NEXT: vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x10,0x05,A,A,A,A]
2172 ; X64-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2173 ; X64-NEXT: retq # encoding: [0xc3]
2174 %res = call <8 x i16> @llvm.x86.avx512.psrlv.w.128(<8 x i16> <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 -1>, <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 -1>)
2175 ret <8 x i16> %res
2176 }
2178 declare <8 x i16> @llvm.x86.avx512.psrlv.w.128(<8 x i16>, <8 x i16>)
2180 define <16 x i16> @test_int_x86_avx512_psrlv_w_256_const() optsize {
2181 ; X86-LABEL: test_int_x86_avx512_psrlv_w_256_const:
2182 ; X86: # %bb.0:
2183 ; X86-NEXT: vmovdqa {{.*#+}} ymm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,65535]
2184 ; X86-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
2185 ; X86-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2186 ; X86-NEXT: vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0x05,A,A,A,A]
2187 ; X86-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2188 ; X86-NEXT: retl # encoding: [0xc3]
2189 ;
2190 ; X64-LABEL: test_int_x86_avx512_psrlv_w_256_const:
2191 ; X64: # %bb.0:
2192 ; X64-NEXT: vmovdqa {{.*#+}} ymm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,65535]
2193 ; X64-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
2194 ; X64-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2195 ; X64-NEXT: vpsrlvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x10,0x05,A,A,A,A]
2196 ; X64-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2197 ; X64-NEXT: retq # encoding: [0xc3]
2198 %res = call <16 x i16> @llvm.x86.avx512.psrlv.w.256(<16 x i16> <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 -1>, <16 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 -1>)
2199 ret <16 x i16> %res
2200 }
2202 declare <16 x i16> @llvm.x86.avx512.psrlv.w.256(<16 x i16>, <16 x i16>)
2204 declare <16 x i16> @llvm.x86.avx512.psrav.w.256(<16 x i16>, <16 x i16>)
2206 define <16 x i16>@test_int_x86_avx512_psrav16_hi(<16 x i16> %x0, <16 x i16> %x1) {
2207 ; CHECK-LABEL: test_int_x86_avx512_psrav16_hi:
2208 ; CHECK: # %bb.0:
2209 ; CHECK-NEXT: vpsravw %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x11,0xc1]
2210 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2211 %1 = call <16 x i16> @llvm.x86.avx512.psrav.w.256(<16 x i16> %x0, <16 x i16> %x1)
2212 ret <16 x i16> %1
2213 }
2215 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) {
2216 ; X86-LABEL: test_int_x86_avx512_mask_psrav16_hi:
2217 ; X86: # %bb.0:
2218 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2219 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
2220 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2221 ; X86-NEXT: retl # encoding: [0xc3]
2222 ;
2223 ; X64-LABEL: test_int_x86_avx512_mask_psrav16_hi:
2224 ; X64: # %bb.0:
2225 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2226 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x11,0xd1]
2227 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2228 ; X64-NEXT: retq # encoding: [0xc3]
2229 %1 = call <16 x i16> @llvm.x86.avx512.psrav.w.256(<16 x i16> %x0, <16 x i16> %x1)
2230 %2 = bitcast i16 %x3 to <16 x i1>
2231 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
2232 ret <16 x i16> %3
2233 }
2235 define <16 x i16>@test_int_x86_avx512_maskz_psrav16_hi(<16 x i16> %x0, <16 x i16> %x1, i16 %x3) {
2236 ; X86-LABEL: test_int_x86_avx512_maskz_psrav16_hi:
2237 ; X86: # %bb.0:
2238 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2239 ; X86-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
2240 ; X86-NEXT: retl # encoding: [0xc3]
2241 ;
2242 ; X64-LABEL: test_int_x86_avx512_maskz_psrav16_hi:
2243 ; X64: # %bb.0:
2244 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2245 ; X64-NEXT: vpsravw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x11,0xc1]
2246 ; X64-NEXT: retq # encoding: [0xc3]
2247 %1 = call <16 x i16> @llvm.x86.avx512.psrav.w.256(<16 x i16> %x0, <16 x i16> %x1)
2248 %2 = bitcast i16 %x3 to <16 x i1>
2249 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
2250 ret <16 x i16> %3
2251 }
2253 declare <8 x i16> @llvm.x86.avx512.psrav.w.128(<8 x i16>, <8 x i16>)
2255 define <8 x i16>@test_int_x86_avx512_psrav8_hi(<8 x i16> %x0, <8 x i16> %x1) {
2256 ; CHECK-LABEL: test_int_x86_avx512_psrav8_hi:
2257 ; CHECK: # %bb.0:
2258 ; CHECK-NEXT: vpsravw %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x11,0xc1]
2259 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2260 %1 = call <8 x i16> @llvm.x86.avx512.psrav.w.128(<8 x i16> %x0, <8 x i16> %x1)
2261 ret <8 x i16> %1
2262 }
2264 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) {
2265 ; X86-LABEL: test_int_x86_avx512_mask_psrav8_hi:
2266 ; X86: # %bb.0:
2267 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2268 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2269 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
2270 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2271 ; X86-NEXT: retl # encoding: [0xc3]
2272 ;
2273 ; X64-LABEL: test_int_x86_avx512_mask_psrav8_hi:
2274 ; X64: # %bb.0:
2275 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2276 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x11,0xd1]
2277 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2278 ; X64-NEXT: retq # encoding: [0xc3]
2279 %1 = call <8 x i16> @llvm.x86.avx512.psrav.w.128(<8 x i16> %x0, <8 x i16> %x1)
2280 %2 = bitcast i8 %x3 to <8 x i1>
2281 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
2282 ret <8 x i16> %3
2283 }
2285 define <8 x i16>@test_int_x86_avx512_maskz_psrav8_hi(<8 x i16> %x0, <8 x i16> %x1, i8 %x3) {
2286 ; X86-LABEL: test_int_x86_avx512_maskz_psrav8_hi:
2287 ; X86: # %bb.0:
2288 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2289 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2290 ; X86-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
2291 ; X86-NEXT: retl # encoding: [0xc3]
2292 ;
2293 ; X64-LABEL: test_int_x86_avx512_maskz_psrav8_hi:
2294 ; X64: # %bb.0:
2295 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2296 ; X64-NEXT: vpsravw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x11,0xc1]
2297 ; X64-NEXT: retq # encoding: [0xc3]
2298 %1 = call <8 x i16> @llvm.x86.avx512.psrav.w.128(<8 x i16> %x0, <8 x i16> %x1)
2299 %2 = bitcast i8 %x3 to <8 x i1>
2300 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
2301 ret <8 x i16> %3
2302 }
2304 define <16 x i16>@test_int_x86_avx512_psllv16_hi(<16 x i16> %x0, <16 x i16> %x1) {
2305 ; CHECK-LABEL: test_int_x86_avx512_psllv16_hi:
2306 ; CHECK: # %bb.0:
2307 ; CHECK-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0xc1]
2308 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2309 %1 = call <16 x i16> @llvm.x86.avx512.psllv.w.256(<16 x i16> %x0, <16 x i16> %x1)
2310 ret <16 x i16> %1
2311 }
2313 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) {
2314 ; X86-LABEL: test_int_x86_avx512_mask_psllv16_hi:
2315 ; X86: # %bb.0:
2316 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2317 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
2318 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2319 ; X86-NEXT: retl # encoding: [0xc3]
2320 ;
2321 ; X64-LABEL: test_int_x86_avx512_mask_psllv16_hi:
2322 ; X64: # %bb.0:
2323 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2324 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x29,0x12,0xd1]
2325 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2326 ; X64-NEXT: retq # encoding: [0xc3]
2327 %1 = call <16 x i16> @llvm.x86.avx512.psllv.w.256(<16 x i16> %x0, <16 x i16> %x1)
2328 %2 = bitcast i16 %x3 to <16 x i1>
2329 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
2330 ret <16 x i16> %3
2331 }
2333 define <16 x i16>@test_int_x86_avx512_maskz_psllv16_hi(<16 x i16> %x0, <16 x i16> %x1, i16 %x3) {
2334 ; X86-LABEL: test_int_x86_avx512_maskz_psllv16_hi:
2335 ; X86: # %bb.0:
2336 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2337 ; X86-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
2338 ; X86-NEXT: retl # encoding: [0xc3]
2339 ;
2340 ; X64-LABEL: test_int_x86_avx512_maskz_psllv16_hi:
2341 ; X64: # %bb.0:
2342 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2343 ; X64-NEXT: vpsllvw %ymm1, %ymm0, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0xa9,0x12,0xc1]
2344 ; X64-NEXT: retq # encoding: [0xc3]
2345 %1 = call <16 x i16> @llvm.x86.avx512.psllv.w.256(<16 x i16> %x0, <16 x i16> %x1)
2346 %2 = bitcast i16 %x3 to <16 x i1>
2347 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
2348 ret <16 x i16> %3
2349 }
2351 define <8 x i16>@test_int_x86_avx512_psllv8_hi(<8 x i16> %x0, <8 x i16> %x1) {
2352 ; CHECK-LABEL: test_int_x86_avx512_psllv8_hi:
2353 ; CHECK: # %bb.0:
2354 ; CHECK-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0xc1]
2355 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2356 %1 = call <8 x i16> @llvm.x86.avx512.psllv.w.128(<8 x i16> %x0, <8 x i16> %x1)
2357 ret <8 x i16> %1
2358 }
2360 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) {
2361 ; X86-LABEL: test_int_x86_avx512_mask_psllv8_hi:
2362 ; X86: # %bb.0:
2363 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2364 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2365 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
2366 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2367 ; X86-NEXT: retl # encoding: [0xc3]
2368 ;
2369 ; X64-LABEL: test_int_x86_avx512_mask_psllv8_hi:
2370 ; X64: # %bb.0:
2371 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2372 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm2 {%k1} # encoding: [0x62,0xf2,0xfd,0x09,0x12,0xd1]
2373 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2374 ; X64-NEXT: retq # encoding: [0xc3]
2375 %1 = call <8 x i16> @llvm.x86.avx512.psllv.w.128(<8 x i16> %x0, <8 x i16> %x1)
2376 %2 = bitcast i8 %x3 to <8 x i1>
2377 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
2378 ret <8 x i16> %3
2379 }
2381 define <8 x i16>@test_int_x86_avx512_maskz_psllv8_hi(<8 x i16> %x0, <8 x i16> %x1, i8 %x3) {
2382 ; X86-LABEL: test_int_x86_avx512_maskz_psllv8_hi:
2383 ; X86: # %bb.0:
2384 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2385 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2386 ; X86-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
2387 ; X86-NEXT: retl # encoding: [0xc3]
2388 ;
2389 ; X64-LABEL: test_int_x86_avx512_maskz_psllv8_hi:
2390 ; X64: # %bb.0:
2391 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2392 ; X64-NEXT: vpsllvw %xmm1, %xmm0, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xfd,0x89,0x12,0xc1]
2393 ; X64-NEXT: retq # encoding: [0xc3]
2394 %1 = call <8 x i16> @llvm.x86.avx512.psllv.w.128(<8 x i16> %x0, <8 x i16> %x1)
2395 %2 = bitcast i8 %x3 to <8 x i1>
2396 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
2397 ret <8 x i16> %3
2398 }
2400 define <8 x i16> @test_int_x86_avx512_psllv_w_128_const() optsize {
2401 ; X86-LABEL: test_int_x86_avx512_psllv_w_128_const:
2402 ; X86: # %bb.0:
2403 ; X86-NEXT: vmovdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,65535]
2404 ; X86-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
2405 ; X86-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2406 ; X86-NEXT: vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0x05,A,A,A,A]
2407 ; X86-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2408 ; X86-NEXT: retl # encoding: [0xc3]
2409 ;
2410 ; X64-LABEL: test_int_x86_avx512_psllv_w_128_const:
2411 ; X64: # %bb.0:
2412 ; X64-NEXT: vmovdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,65535]
2413 ; X64-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x05,A,A,A,A]
2414 ; X64-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2415 ; X64-NEXT: vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0 # encoding: [0x62,0xf2,0xfd,0x08,0x12,0x05,A,A,A,A]
2416 ; X64-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2417 ; X64-NEXT: retq # encoding: [0xc3]
2418 %res = call <8 x i16> @llvm.x86.avx512.psllv.w.128(<8 x i16> <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 -1>, <8 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 -1>)
2419 ret <8 x i16> %res
2420 }
2422 declare <8 x i16> @llvm.x86.avx512.psllv.w.128(<8 x i16>, <8 x i16>)
2425 define <16 x i16> @test_int_x86_avx512_psllv_w_256_const() optsize {
2426 ; X86-LABEL: test_int_x86_avx512_psllv_w_256_const:
2427 ; X86: # %bb.0:
2428 ; X86-NEXT: vmovdqa {{.*#+}} ymm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,65535]
2429 ; X86-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
2430 ; X86-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2431 ; X86-NEXT: vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}, %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0x05,A,A,A,A]
2432 ; X86-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}, kind: FK_Data_4
2433 ; X86-NEXT: retl # encoding: [0xc3]
2434 ;
2435 ; X64-LABEL: test_int_x86_avx512_psllv_w_256_const:
2436 ; X64: # %bb.0:
2437 ; X64-NEXT: vmovdqa {{.*#+}} ymm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,65535]
2438 ; X64-NEXT: # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0x05,A,A,A,A]
2439 ; X64-NEXT: # fixup A - offset: 4, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2440 ; X64-NEXT: vpsllvw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0 # encoding: [0x62,0xf2,0xfd,0x28,0x12,0x05,A,A,A,A]
2441 ; X64-NEXT: # fixup A - offset: 6, value: {{\.?LCPI[0-9]+_[0-9]+}}-4, kind: reloc_riprel_4byte
2442 ; X64-NEXT: retq # encoding: [0xc3]
2443 %res = call <16 x i16> @llvm.x86.avx512.psllv.w.256(<16 x i16> <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 -1>, <16 x i16> <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 -1>)
2444 ret <16 x i16> %res
2445 }
2447 declare <16 x i16> @llvm.x86.avx512.psllv.w.256(<16 x i16>, <16 x i16>)
2451 declare <8 x i16> @llvm.x86.avx512.permvar.hi.128(<8 x i16>, <8 x i16>)
2453 define <8 x i16>@test_int_x86_avx512_permvar_hi_128(<8 x i16> %x0, <8 x i16> %x1) {
2454 ; CHECK-LABEL: test_int_x86_avx512_permvar_hi_128:
2455 ; CHECK: # %bb.0:
2456 ; CHECK-NEXT: vpermw %xmm0, %xmm1, %xmm0 # encoding: [0x62,0xf2,0xf5,0x08,0x8d,0xc0]
2457 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2458 %1 = call <8 x i16> @llvm.x86.avx512.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1)
2459 ret <8 x i16> %1
2460 }
2462 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) {
2463 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
2464 ; X86: # %bb.0:
2465 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2466 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2467 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
2468 ; X86-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2469 ; X86-NEXT: retl # encoding: [0xc3]
2470 ;
2471 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_128:
2472 ; X64: # %bb.0:
2473 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2474 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x09,0x8d,0xd0]
2475 ; X64-NEXT: vmovdqa %xmm2, %xmm0 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xc2]
2476 ; X64-NEXT: retq # encoding: [0xc3]
2477 %1 = call <8 x i16> @llvm.x86.avx512.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1)
2478 %2 = bitcast i8 %x3 to <8 x i1>
2479 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> %x2
2480 ret <8 x i16> %3
2481 }
2483 define <8 x i16>@test_int_x86_avx512_maskz_permvar_hi_128(<8 x i16> %x0, <8 x i16> %x1, i8 %x3) {
2484 ; X86-LABEL: test_int_x86_avx512_maskz_permvar_hi_128:
2485 ; X86: # %bb.0:
2486 ; X86-NEXT: movzbl {{[0-9]+}}(%esp), %eax # encoding: [0x0f,0xb6,0x44,0x24,0x04]
2487 ; X86-NEXT: kmovd %eax, %k1 # encoding: [0xc5,0xfb,0x92,0xc8]
2488 ; X86-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
2489 ; X86-NEXT: retl # encoding: [0xc3]
2490 ;
2491 ; X64-LABEL: test_int_x86_avx512_maskz_permvar_hi_128:
2492 ; X64: # %bb.0:
2493 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2494 ; X64-NEXT: vpermw %xmm0, %xmm1, %xmm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0x89,0x8d,0xc0]
2495 ; X64-NEXT: retq # encoding: [0xc3]
2496 %1 = call <8 x i16> @llvm.x86.avx512.permvar.hi.128(<8 x i16> %x0, <8 x i16> %x1)
2497 %2 = bitcast i8 %x3 to <8 x i1>
2498 %3 = select <8 x i1> %2, <8 x i16> %1, <8 x i16> zeroinitializer
2499 ret <8 x i16> %3
2500 }
2502 declare <16 x i16> @llvm.x86.avx512.permvar.hi.256(<16 x i16>, <16 x i16>)
2504 define <16 x i16>@test_int_x86_avx512_permvar_hi_256(<16 x i16> %x0, <16 x i16> %x1) {
2505 ; CHECK-LABEL: test_int_x86_avx512_permvar_hi_256:
2506 ; CHECK: # %bb.0:
2507 ; CHECK-NEXT: vpermw %ymm0, %ymm1, %ymm0 # encoding: [0x62,0xf2,0xf5,0x28,0x8d,0xc0]
2508 ; CHECK-NEXT: ret{{[l|q]}} # encoding: [0xc3]
2509 %1 = call <16 x i16> @llvm.x86.avx512.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1)
2510 ret <16 x i16> %1
2511 }
2513 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) {
2514 ; X86-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
2515 ; X86: # %bb.0:
2516 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2517 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
2518 ; X86-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2519 ; X86-NEXT: retl # encoding: [0xc3]
2520 ;
2521 ; X64-LABEL: test_int_x86_avx512_mask_permvar_hi_256:
2522 ; X64: # %bb.0:
2523 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2524 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm2 {%k1} # encoding: [0x62,0xf2,0xf5,0x29,0x8d,0xd0]
2525 ; X64-NEXT: vmovdqa %ymm2, %ymm0 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xc2]
2526 ; X64-NEXT: retq # encoding: [0xc3]
2527 %1 = call <16 x i16> @llvm.x86.avx512.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1)
2528 %2 = bitcast i16 %x3 to <16 x i1>
2529 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> %x2
2530 ret <16 x i16> %3
2531 }
2533 define <16 x i16>@test_int_x86_avx512_maskz_permvar_hi_256(<16 x i16> %x0, <16 x i16> %x1, i16 %x3) {
2534 ; X86-LABEL: test_int_x86_avx512_maskz_permvar_hi_256:
2535 ; X86: # %bb.0:
2536 ; X86-NEXT: kmovw {{[0-9]+}}(%esp), %k1 # encoding: [0xc5,0xf8,0x90,0x4c,0x24,0x04]
2537 ; X86-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
2538 ; X86-NEXT: retl # encoding: [0xc3]
2539 ;
2540 ; X64-LABEL: test_int_x86_avx512_maskz_permvar_hi_256:
2541 ; X64: # %bb.0:
2542 ; X64-NEXT: kmovd %edi, %k1 # encoding: [0xc5,0xfb,0x92,0xcf]
2543 ; X64-NEXT: vpermw %ymm0, %ymm1, %ymm0 {%k1} {z} # encoding: [0x62,0xf2,0xf5,0xa9,0x8d,0xc0]
2544 ; X64-NEXT: retq # encoding: [0xc3]
2545 %1 = call <16 x i16> @llvm.x86.avx512.permvar.hi.256(<16 x i16> %x0, <16 x i16> %x1)
2546 %2 = bitcast i16 %x3 to <16 x i1>
2547 %3 = select <16 x i1> %2, <16 x i16> %1, <16 x i16> zeroinitializer
2548 ret <16 x i16> %3
2549 }