xref: /llvm-project/llvm/test/CodeGen/X86/avx512vl_vnni-intrinsics-upgrade.ll (revision 2f448bf509432c1a19ec46ab8cbc7353c03c6280) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vnni,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X86
3; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vnni,+avx512vl --show-mc-encoding | FileCheck %s --check-prefixes=CHECK,X64
4
5declare <8 x i32> @llvm.x86.avx512.mask.vpdpbusd.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
6declare <8 x i32> @llvm.x86.avx512.maskz.vpdpbusd.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
7
8define <8 x i32>@test_int_x86_avx512_vpdpbusd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2) {
9; CHECK-LABEL: test_int_x86_avx512_vpdpbusd_256:
10; CHECK:       # %bb.0:
11; CHECK-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x50,0xc2]
12; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
13  %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
14  ret <8 x i32> %res
15}
16
17define { <8 x i32>, <8 x i32> } @test_int_x86_avx512_mask_vpdpbusd_256(<8 x i32> %x0, <8 x i32> %x1, ptr %x2p, <8 x i32> %x4, i8 %x3) {
18; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
19; X86:       # %bb.0:
20; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
21; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
22; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
23; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
24; X86-NEXT:    vpdpbusd (%eax), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x00]
25; X86-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xda]
26; X86-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
27; X86-NEXT:    retl # encoding: [0xc3]
28;
29; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_256:
30; X64:       # %bb.0:
31; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
32; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
33; X64-NEXT:    vpdpbusd (%rdi), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x50,0x07]
34; X64-NEXT:    vpdpbusd %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x50,0xda]
35; X64-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
36; X64-NEXT:    retq # encoding: [0xc3]
37  %x2 = load <8 x i32>, ptr %x2p
38  %res0 = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
39  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpbusd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
40  %res2 = insertvalue { <8 x i32>, <8 x i32> } poison, <8 x i32> %res0, 0
41  %res3 = insertvalue { <8 x i32>, <8 x i32> }  %res2, <8 x i32> %res1, 1
42  ret { <8 x i32>, <8 x i32> } %res3
43}
44
45declare <4 x i32> @llvm.x86.avx512.mask.vpdpbusd.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
46declare <4 x i32> @llvm.x86.avx512.maskz.vpdpbusd.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
47
48define <4 x i32>@test_int_x86_avx512_vpdpbusd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2) {
49; CHECK-LABEL: test_int_x86_avx512_vpdpbusd_128:
50; CHECK:       # %bb.0:
51; CHECK-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x50,0xc2]
52; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
53  %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
54  ret <4 x i32> %res
55}
56
57define { <4 x i32>, <4 x i32> } @test_int_x86_avx512_mask_vpdpbusd_128(<4 x i32> %x0, <4 x i32> %x1, ptr %x2p, <4 x i32> %x4, i8 %x3) {
58; X86-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
59; X86:       # %bb.0:
60; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
61; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
62; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
63; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
64; X86-NEXT:    vpdpbusd (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x00]
65; X86-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xda]
66; X86-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
67; X86-NEXT:    retl # encoding: [0xc3]
68;
69; X64-LABEL: test_int_x86_avx512_mask_vpdpbusd_128:
70; X64:       # %bb.0:
71; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
72; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
73; X64-NEXT:    vpdpbusd (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x50,0x07]
74; X64-NEXT:    vpdpbusd %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x50,0xda]
75; X64-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
76; X64-NEXT:    retq # encoding: [0xc3]
77  %x2 = load <4 x i32>, ptr %x2p
78  %res0 = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
79  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpbusd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
80  %res2 = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> %res0, 0
81  %res3 = insertvalue { <4 x i32>, <4 x i32> }  %res2, <4 x i32> %res1, 1
82  ret { <4 x i32>, <4 x i32> } %res3
83}
84
85declare <8 x i32> @llvm.x86.avx512.mask.vpdpbusds.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
86declare <8 x i32> @llvm.x86.avx512.maskz.vpdpbusds.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
87
88define <8 x i32>@test_int_x86_avx512_vpdpbusds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2) {
89; CHECK-LABEL: test_int_x86_avx512_vpdpbusds_256:
90; CHECK:       # %bb.0:
91; CHECK-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x51,0xc2]
92; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
93  %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
94  ret <8 x i32> %res
95}
96
97define { <8 x i32>, <8 x i32> } @test_int_x86_avx512_mask_vpdpbusds_256(<8 x i32> %x0, <8 x i32> %x1, ptr %x2p, <8 x i32> %x4, i8 %x3) {
98; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
99; X86:       # %bb.0:
100; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
101; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
102; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
103; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
104; X86-NEXT:    vpdpbusds (%eax), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x00]
105; X86-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xda]
106; X86-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
107; X86-NEXT:    retl # encoding: [0xc3]
108;
109; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_256:
110; X64:       # %bb.0:
111; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
112; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
113; X64-NEXT:    vpdpbusds (%rdi), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x51,0x07]
114; X64-NEXT:    vpdpbusds %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x51,0xda]
115; X64-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
116; X64-NEXT:    retq # encoding: [0xc3]
117  %x2 = load <8 x i32>, ptr %x2p
118  %res0 = call <8 x i32> @llvm.x86.avx512.mask.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
119  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpbusds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
120  %res2 = insertvalue { <8 x i32>, <8 x i32> } poison, <8 x i32> %res0, 0
121  %res3 = insertvalue { <8 x i32>, <8 x i32> }  %res2, <8 x i32> %res1, 1
122  ret { <8 x i32>, <8 x i32> } %res3
123}
124
125declare <4 x i32> @llvm.x86.avx512.mask.vpdpbusds.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
126declare <4 x i32> @llvm.x86.avx512.maskz.vpdpbusds.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
127
128define <4 x i32>@test_int_x86_avx512_vpdpbusds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2) {
129; CHECK-LABEL: test_int_x86_avx512_vpdpbusds_128:
130; CHECK:       # %bb.0:
131; CHECK-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x51,0xc2]
132; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
133  %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
134  ret <4 x i32> %res
135}
136
137define { <4 x i32>, <4 x i32> } @test_int_x86_avx512_mask_vpdpbusds_128(<4 x i32> %x0, <4 x i32> %x1, ptr %x2p, <4 x i32> %x4, i8 %x3) {
138; X86-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
139; X86:       # %bb.0:
140; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
141; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
142; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
143; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
144; X86-NEXT:    vpdpbusds (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x00]
145; X86-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xda]
146; X86-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
147; X86-NEXT:    retl # encoding: [0xc3]
148;
149; X64-LABEL: test_int_x86_avx512_mask_vpdpbusds_128:
150; X64:       # %bb.0:
151; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
152; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
153; X64-NEXT:    vpdpbusds (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x51,0x07]
154; X64-NEXT:    vpdpbusds %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x51,0xda]
155; X64-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
156; X64-NEXT:    retq # encoding: [0xc3]
157  %x2 = load <4 x i32>, ptr %x2p
158  %res0 = call <4 x i32> @llvm.x86.avx512.mask.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
159  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpbusds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
160  %res2 = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> %res0, 0
161  %res3 = insertvalue { <4 x i32>, <4 x i32> }  %res2, <4 x i32> %res1, 1
162  ret { <4 x i32>, <4 x i32> } %res3
163}
164
165declare <8 x i32> @llvm.x86.avx512.mask.vpdpwssd.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
166declare <8 x i32> @llvm.x86.avx512.maskz.vpdpwssd.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
167
168define <8 x i32>@test_int_x86_avx512_vpdpwssd_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2) {
169; CHECK-LABEL: test_int_x86_avx512_vpdpwssd_256:
170; CHECK:       # %bb.0:
171; CHECK-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x52,0xc2]
172; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
173  %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
174  ret <8 x i32> %res
175}
176
177define { <8 x i32>, <8 x i32> } @test_int_x86_avx512_mask_vpdpwssd_256(<8 x i32> %x0, <8 x i32> %x1, ptr %x2p, <8 x i32> %x4, i8 %x3) {
178; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
179; X86:       # %bb.0:
180; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
181; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
182; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
183; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
184; X86-NEXT:    vpdpwssd (%eax), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x00]
185; X86-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xda]
186; X86-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
187; X86-NEXT:    retl # encoding: [0xc3]
188;
189; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_256:
190; X64:       # %bb.0:
191; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
192; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
193; X64-NEXT:    vpdpwssd (%rdi), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x52,0x07]
194; X64-NEXT:    vpdpwssd %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x52,0xda]
195; X64-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
196; X64-NEXT:    retq # encoding: [0xc3]
197  %x2 = load <8 x i32>, ptr %x2p
198  %res0 = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
199  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpwssd.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
200  %res2 = insertvalue { <8 x i32>, <8 x i32> } poison, <8 x i32> %res0, 0
201  %res3 = insertvalue { <8 x i32>, <8 x i32> }  %res2, <8 x i32> %res1, 1
202  ret { <8 x i32>, <8 x i32> } %res3
203}
204
205declare <4 x i32> @llvm.x86.avx512.mask.vpdpwssd.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
206declare <4 x i32> @llvm.x86.avx512.maskz.vpdpwssd.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
207
208define <4 x i32>@test_int_x86_avx512_vpdpwssd_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2) {
209; CHECK-LABEL: test_int_x86_avx512_vpdpwssd_128:
210; CHECK:       # %bb.0:
211; CHECK-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x52,0xc2]
212; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
213  %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
214  ret <4 x i32> %res
215}
216
217define { <4 x i32>, <4 x i32> } @test_int_x86_avx512_mask_vpdpwssd_128(<4 x i32> %x0, <4 x i32> %x1, ptr %x2p, <4 x i32> %x4, i8 %x3) {
218; X86-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
219; X86:       # %bb.0:
220; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
221; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
222; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
223; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
224; X86-NEXT:    vpdpwssd (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x00]
225; X86-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xda]
226; X86-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
227; X86-NEXT:    retl # encoding: [0xc3]
228;
229; X64-LABEL: test_int_x86_avx512_mask_vpdpwssd_128:
230; X64:       # %bb.0:
231; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
232; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
233; X64-NEXT:    vpdpwssd (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x52,0x07]
234; X64-NEXT:    vpdpwssd %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x52,0xda]
235; X64-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
236; X64-NEXT:    retq # encoding: [0xc3]
237  %x2 = load <4 x i32>, ptr %x2p
238  %res0 = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
239  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpwssd.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
240  %res2 = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> %res0, 0
241  %res3 = insertvalue { <4 x i32>, <4 x i32> }  %res2, <4 x i32> %res1, 1
242  ret { <4 x i32>, <4 x i32> } %res3
243}
244
245
246declare <8 x i32> @llvm.x86.avx512.mask.vpdpwssds.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
247declare <8 x i32> @llvm.x86.avx512.maskz.vpdpwssds.256(<8 x i32>, <8 x i32>, <8 x i32>, i8)
248
249define <8 x i32>@test_int_x86_avx512_vpdpwssds_256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2) {
250; CHECK-LABEL: test_int_x86_avx512_vpdpwssds_256:
251; CHECK:       # %bb.0:
252; CHECK-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm0 # encoding: [0x62,0xf2,0x75,0x28,0x53,0xc2]
253; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
254  %res = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 -1)
255  ret <8 x i32> %res
256}
257
258define { <8 x i32>, <8 x i32> } @test_int_x86_avx512_mask_vpdpwssds_256(<8 x i32> %x0, <8 x i32> %x1, ptr %x2p, <8 x i32> %x4, i8 %x3) {
259; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
260; X86:       # %bb.0:
261; X86-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
262; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
263; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
264; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
265; X86-NEXT:    vpdpwssds (%eax), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x00]
266; X86-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xda]
267; X86-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
268; X86-NEXT:    retl # encoding: [0xc3]
269;
270; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_256:
271; X64:       # %bb.0:
272; X64-NEXT:    vmovdqa %ymm0, %ymm3 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xd8]
273; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
274; X64-NEXT:    vpdpwssds (%rdi), %ymm1, %ymm0 {%k1} # encoding: [0x62,0xf2,0x75,0x29,0x53,0x07]
275; X64-NEXT:    vpdpwssds %ymm2, %ymm1, %ymm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0xa9,0x53,0xda]
276; X64-NEXT:    vmovdqa %ymm3, %ymm1 # EVEX TO VEX Compression encoding: [0xc5,0xfd,0x6f,0xcb]
277; X64-NEXT:    retq # encoding: [0xc3]
278  %x2 = load <8 x i32>, ptr %x2p
279  %res0 = call <8 x i32> @llvm.x86.avx512.mask.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x2, i8 %x3)
280  %res1 = call <8 x i32> @llvm.x86.avx512.maskz.vpdpwssds.256(<8 x i32> %x0, <8 x i32> %x1, <8 x i32> %x4, i8  %x3)
281  %res2 = insertvalue { <8 x i32>, <8 x i32> } poison, <8 x i32> %res0, 0
282  %res3 = insertvalue { <8 x i32>, <8 x i32> }  %res2, <8 x i32> %res1, 1
283  ret { <8 x i32>, <8 x i32> } %res3
284}
285
286declare <4 x i32> @llvm.x86.avx512.mask.vpdpwssds.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
287declare <4 x i32> @llvm.x86.avx512.maskz.vpdpwssds.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
288
289define <4 x i32>@test_int_x86_avx512_vpdpwssds_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2) {
290; CHECK-LABEL: test_int_x86_avx512_vpdpwssds_128:
291; CHECK:       # %bb.0:
292; CHECK-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm0 # encoding: [0x62,0xf2,0x75,0x08,0x53,0xc2]
293; CHECK-NEXT:    ret{{[l|q]}} # encoding: [0xc3]
294  %res = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 -1)
295  ret <4 x i32> %res
296}
297
298define { <4 x i32>, <4 x i32> } @test_int_x86_avx512_mask_vpdpwssds_128(<4 x i32> %x0, <4 x i32> %x1, ptr %x2p, <4 x i32> %x4, i8 %x3) {
299; X86-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
300; X86:       # %bb.0:
301; X86-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
302; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x04]
303; X86-NEXT:    movzbl {{[0-9]+}}(%esp), %ecx # encoding: [0x0f,0xb6,0x4c,0x24,0x08]
304; X86-NEXT:    kmovw %ecx, %k1 # encoding: [0xc5,0xf8,0x92,0xc9]
305; X86-NEXT:    vpdpwssds (%eax), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x00]
306; X86-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xda]
307; X86-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
308; X86-NEXT:    retl # encoding: [0xc3]
309;
310; X64-LABEL: test_int_x86_avx512_mask_vpdpwssds_128:
311; X64:       # %bb.0:
312; X64-NEXT:    vmovdqa %xmm0, %xmm3 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xd8]
313; X64-NEXT:    kmovw %esi, %k1 # encoding: [0xc5,0xf8,0x92,0xce]
314; X64-NEXT:    vpdpwssds (%rdi), %xmm1, %xmm0 {%k1} # encoding: [0x62,0xf2,0x75,0x09,0x53,0x07]
315; X64-NEXT:    vpdpwssds %xmm2, %xmm1, %xmm3 {%k1} {z} # encoding: [0x62,0xf2,0x75,0x89,0x53,0xda]
316; X64-NEXT:    vmovdqa %xmm3, %xmm1 # EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0xcb]
317; X64-NEXT:    retq # encoding: [0xc3]
318  %x2 = load <4 x i32>, ptr %x2p
319  %res0 = call <4 x i32> @llvm.x86.avx512.mask.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3)
320  %res1 = call <4 x i32> @llvm.x86.avx512.maskz.vpdpwssds.128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x4, i8  %x3)
321  %res2 = insertvalue { <4 x i32>, <4 x i32> } poison, <4 x i32> %res0, 0
322  %res3 = insertvalue { <4 x i32>, <4 x i32> }  %res2, <4 x i32> %res1, 1
323  ret { <4 x i32>, <4 x i32> } %res3
324}
325