1; NOTE: Assertions have been autogenerated by update_llc_test_checks.py
2; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx -mattr=+avx512vl -mattr=+avx512vbmi --show-mc-encoding| FileCheck %s
3declare <16 x i8> @llvm.x86.avx512.mask.permvar.qi.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
4
5define <16 x i8>@test_int_x86_avx512_mask_permvar_qi_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
6; CHECK-LABEL: test_int_x86_avx512_mask_permvar_qi_128:
7; CHECK:       ## BB#0:
8; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
9; CHECK-NEXT:    vpermb %xmm0, %xmm1, %xmm2 {%k1} ## encoding: [0x62,0xf2,0x75,0x09,0x8d,0xd0]
10; CHECK-NEXT:    vpermb %xmm0, %xmm1, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0x89,0x8d,0xd8]
11; CHECK-NEXT:    vpermb %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf2,0x75,0x08,0x8d,0xc0]
12; CHECK-NEXT:    vpaddb %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xcb]
13; CHECK-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfc,0xc0]
14; CHECK-NEXT:    retq ## encoding: [0xc3]
15  %res = call <16 x i8> @llvm.x86.avx512.mask.permvar.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
16  %res1 = call <16 x i8> @llvm.x86.avx512.mask.permvar.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %x3)
17  %res2 = call <16 x i8> @llvm.x86.avx512.mask.permvar.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
18  %res3 = add <16 x i8> %res, %res1
19  %res4 = add <16 x i8> %res3, %res2
20  ret <16 x i8> %res4
21}
22
23declare <32 x i8> @llvm.x86.avx512.mask.permvar.qi.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
24
25define <32 x i8>@test_int_x86_avx512_mask_permvar_qi_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
26; CHECK-LABEL: test_int_x86_avx512_mask_permvar_qi_256:
27; CHECK:       ## BB#0:
28; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
29; CHECK-NEXT:    vpermb %ymm0, %ymm1, %ymm2 {%k1} ## encoding: [0x62,0xf2,0x75,0x29,0x8d,0xd0]
30; CHECK-NEXT:    vpermb %ymm0, %ymm1, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0x75,0xa9,0x8d,0xd8]
31; CHECK-NEXT:    vpermb %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf2,0x75,0x28,0x8d,0xc0]
32; CHECK-NEXT:    vpaddb %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xcb]
33; CHECK-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfc,0xc0]
34; CHECK-NEXT:    retq ## encoding: [0xc3]
35  %res = call <32 x i8> @llvm.x86.avx512.mask.permvar.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
36  %res1 = call <32 x i8> @llvm.x86.avx512.mask.permvar.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> zeroinitializer, i32 %x3)
37  %res2 = call <32 x i8> @llvm.x86.avx512.mask.permvar.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
38  %res3 = add <32 x i8> %res, %res1
39  %res4 = add <32 x i8> %res3, %res2
40  ret <32 x i8> %res4
41}
42
43declare <16 x i8> @llvm.x86.avx512.mask.pmultishift.qb.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
44
45define <16 x i8>@test_int_x86_avx512_mask_pmultishift_qb_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
46; CHECK-LABEL: test_int_x86_avx512_mask_pmultishift_qb_128:
47; CHECK:       ## BB#0:
48; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
49; CHECK-NEXT:    vpmultishiftqb %xmm1, %xmm0, %xmm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x09,0x83,0xd1]
50; CHECK-NEXT:    vpmultishiftqb %xmm1, %xmm0, %xmm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0x89,0x83,0xd9]
51; CHECK-NEXT:    vpmultishiftqb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf2,0xfd,0x08,0x83,0xc1]
52; CHECK-NEXT:    vpaddb %xmm3, %xmm2, %xmm1 ## encoding: [0x62,0xf1,0x6d,0x08,0xfc,0xcb]
53; CHECK-NEXT:    vpaddb %xmm0, %xmm1, %xmm0 ## encoding: [0x62,0xf1,0x75,0x08,0xfc,0xc0]
54; CHECK-NEXT:    retq ## encoding: [0xc3]
55  %res = call <16 x i8> @llvm.x86.avx512.mask.pmultishift.qb.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
56  %res1 = call <16 x i8> @llvm.x86.avx512.mask.pmultishift.qb.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> zeroinitializer, i16 %x3)
57  %res2 = call <16 x i8> @llvm.x86.avx512.mask.pmultishift.qb.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
58  %res3 = add <16 x i8> %res, %res1
59  %res4 = add <16 x i8> %res3, %res2
60  ret <16 x i8> %res4
61}
62
63declare <32 x i8> @llvm.x86.avx512.mask.pmultishift.qb.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
64
65define <32 x i8>@test_int_x86_avx512_mask_pmultishift_qb_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
66; CHECK-LABEL: test_int_x86_avx512_mask_pmultishift_qb_256:
67; CHECK:       ## BB#0:
68; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
69; CHECK-NEXT:    vpmultishiftqb %ymm1, %ymm0, %ymm2 {%k1} ## encoding: [0x62,0xf2,0xfd,0x29,0x83,0xd1]
70; CHECK-NEXT:    vpmultishiftqb %ymm1, %ymm0, %ymm3 {%k1} {z} ## encoding: [0x62,0xf2,0xfd,0xa9,0x83,0xd9]
71; CHECK-NEXT:    vpmultishiftqb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf2,0xfd,0x28,0x83,0xc1]
72; CHECK-NEXT:    vpaddb %ymm3, %ymm2, %ymm1 ## encoding: [0x62,0xf1,0x6d,0x28,0xfc,0xcb]
73; CHECK-NEXT:    vpaddb %ymm0, %ymm1, %ymm0 ## encoding: [0x62,0xf1,0x75,0x28,0xfc,0xc0]
74; CHECK-NEXT:    retq ## encoding: [0xc3]
75  %res = call <32 x i8> @llvm.x86.avx512.mask.pmultishift.qb.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
76  %res1 = call <32 x i8> @llvm.x86.avx512.mask.pmultishift.qb.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> zeroinitializer, i32 %x3)
77  %res2 = call <32 x i8> @llvm.x86.avx512.mask.pmultishift.qb.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
78  %res3 = add <32 x i8> %res, %res1
79  %res4 = add <32 x i8> %res3, %res2
80  ret <32 x i8> %res4
81}
82
83declare <16 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
84
85define <16 x i8>@test_int_x86_avx512_mask_vpermi2var_qi_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
86; CHECK-LABEL: test_int_x86_avx512_mask_vpermi2var_qi_128:
87; CHECK:       ## BB#0:
88; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
89; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
90; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x7d,0xda]
91; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm1 ## encoding: [0x62,0xf2,0x7d,0x08,0x7d,0xca]
92; CHECK-NEXT:    vpxord %xmm4, %xmm4, %xmm4 ## encoding: [0x62,0xf1,0x5d,0x08,0xef,0xe4]
93; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm4 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x7d,0xe2]
94; CHECK-NEXT:    vpaddb %xmm4, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfc,0xc4]
95; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
96; CHECK-NEXT:    retq ## encoding: [0xc3]
97  %res = call <16 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
98  %res1 = call <16 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.128(<16 x i8> %x0, <16 x i8> zeroinitializer, <16 x i8> %x2, i16 %x3)
99  %res2 = call <16 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
100  %res3 = add <16 x i8> %res, %res1
101  %res4 = add <16 x i8> %res3, %res2
102  ret <16 x i8> %res4
103}
104
105declare <32 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
106
107define <32 x i8>@test_int_x86_avx512_mask_vpermi2var_qi_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
108; CHECK-LABEL: test_int_x86_avx512_mask_vpermi2var_qi_256:
109; CHECK:       ## BB#0:
110; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
111; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
112; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x7d,0xda]
113; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm1 ## encoding: [0x62,0xf2,0x7d,0x28,0x7d,0xca]
114; CHECK-NEXT:    vpxord %ymm4, %ymm4, %ymm4 ## encoding: [0x62,0xf1,0x5d,0x28,0xef,0xe4]
115; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm4 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x7d,0xe2]
116; CHECK-NEXT:    vpaddb %ymm4, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x65,0x28,0xfc,0xc4]
117; CHECK-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfc,0xc1]
118; CHECK-NEXT:    retq ## encoding: [0xc3]
119  %res = call <32 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
120  %res1 = call <32 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.256(<32 x i8> %x0, <32 x i8> zeroinitializer, <32 x i8> %x2, i32 %x3)
121  %res2 = call <32 x i8> @llvm.x86.avx512.mask.vpermi2var.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
122  %res3 = add <32 x i8> %res, %res1
123  %res4 = add <32 x i8> %res3, %res2
124  ret <32 x i8> %res4
125}
126
127declare <16 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
128
129define <16 x i8>@test_int_x86_avx512_mask_vpermt2var_qi_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
130; CHECK-LABEL: test_int_x86_avx512_mask_vpermt2var_qi_128:
131; CHECK:       ## BB#0:
132; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
133; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
134; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x09,0x7d,0xda]
135; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm1 ## encoding: [0x62,0xf2,0x7d,0x08,0x7d,0xca]
136; CHECK-NEXT:    vpxord %xmm4, %xmm4, %xmm4 ## encoding: [0x62,0xf1,0x5d,0x08,0xef,0xe4]
137; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm4 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x7d,0xe2]
138; CHECK-NEXT:    vpaddb %xmm4, %xmm3, %xmm0 ## encoding: [0x62,0xf1,0x65,0x08,0xfc,0xc4]
139; CHECK-NEXT:    vpaddb %xmm1, %xmm0, %xmm0 ## encoding: [0x62,0xf1,0x7d,0x08,0xfc,0xc1]
140; CHECK-NEXT:    retq ## encoding: [0xc3]
141  %res = call <16 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
142  %res1 = call <16 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.128(<16 x i8> %x0, <16 x i8> zeroinitializer, <16 x i8> %x2, i16 %x3)
143  %res2 = call <16 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 -1)
144  %res3 = add <16 x i8> %res, %res1
145  %res4 = add <16 x i8> %res3, %res2
146  ret <16 x i8> %res4
147}
148
149declare <32 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
150
151define <32 x i8>@test_int_x86_avx512_mask_vpermt2var_qi_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
152; CHECK-LABEL: test_int_x86_avx512_mask_vpermt2var_qi_256:
153; CHECK:       ## BB#0:
154; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
155; CHECK-NEXT:    vmovaps %zmm1, %zmm3 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xd9]
156; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm3 {%k1} ## encoding: [0x62,0xf2,0x7d,0x29,0x7d,0xda]
157; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm1 ## encoding: [0x62,0xf2,0x7d,0x28,0x7d,0xca]
158; CHECK-NEXT:    vpxord %ymm4, %ymm4, %ymm4 ## encoding: [0x62,0xf1,0x5d,0x28,0xef,0xe4]
159; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm4 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x7d,0xe2]
160; CHECK-NEXT:    vpaddb %ymm4, %ymm3, %ymm0 ## encoding: [0x62,0xf1,0x65,0x28,0xfc,0xc4]
161; CHECK-NEXT:    vpaddb %ymm1, %ymm0, %ymm0 ## encoding: [0x62,0xf1,0x7d,0x28,0xfc,0xc1]
162; CHECK-NEXT:    retq ## encoding: [0xc3]
163  %res = call <32 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
164  %res1 = call <32 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.256(<32 x i8> %x0, <32 x i8> zeroinitializer, <32 x i8> %x2, i32 %x3)
165  %res2 = call <32 x i8> @llvm.x86.avx512.mask.vpermt2var.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 -1)
166  %res3 = add <32 x i8> %res, %res1
167  %res4 = add <32 x i8> %res3, %res2
168  ret <32 x i8> %res4
169}
170
171declare <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8>, <16 x i8>, <16 x i8>, i16)
172
173define <16 x i8>@test_int_x86_avx512_maskz_vpermt2var_qi_128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3) {
174; CHECK-LABEL: test_int_x86_avx512_maskz_vpermt2var_qi_128:
175; CHECK:       ## BB#0:
176; CHECK-NEXT:    kmovw %edi, %k1 ## encoding: [0xc5,0xf8,0x92,0xcf]
177; CHECK-NEXT:    vpermt2b %xmm2, %xmm0, %xmm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0x89,0x7d,0xca]
178; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
179; CHECK-NEXT:    retq ## encoding: [0xc3]
180  %res = call <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8> %x0, <16 x i8> %x1, <16 x i8> %x2, i16 %x3)
181  ret <16 x i8> %res
182}
183
184declare <32 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.256(<32 x i8>, <32 x i8>, <32 x i8>, i32)
185
186define <32 x i8>@test_int_x86_avx512_maskz_vpermt2var_qi_256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3) {
187; CHECK-LABEL: test_int_x86_avx512_maskz_vpermt2var_qi_256:
188; CHECK:       ## BB#0:
189; CHECK-NEXT:    kmovd %edi, %k1 ## encoding: [0xc5,0xfb,0x92,0xcf]
190; CHECK-NEXT:    vpermt2b %ymm2, %ymm0, %ymm1 {%k1} {z} ## encoding: [0x62,0xf2,0x7d,0xa9,0x7d,0xca]
191; CHECK-NEXT:    vmovaps %zmm1, %zmm0 ## encoding: [0x62,0xf1,0x7c,0x48,0x28,0xc1]
192; CHECK-NEXT:    retq ## encoding: [0xc3]
193  %res = call <32 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.256(<32 x i8> %x0, <32 x i8> %x1, <32 x i8> %x2, i32 %x3)
194  ret <32 x i8> %res
195}
196