1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefixes=CHECK,X86,AVX,X86-AVX
3 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefixes=CHECK,X86,AVX2,X86-AVX2
4 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx,+xop | FileCheck %s --check-prefixes=CHECK,X64,AVX,X64-AVX
5 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+xop | FileCheck %s --check-prefixes=CHECK,X64,AVX2,X64-AVX2
7 declare <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double>, <2 x double>, <2 x i64>, i8) nounwind readnone
8 declare <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double>, <4 x double>, <4 x i64>, i8) nounwind readnone
10 declare <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float>, <4 x float>, <4 x i32>, i8) nounwind readnone
11 declare <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float>, <8 x float>, <8 x i32>, i8) nounwind readnone
13 declare <16 x i8> @llvm.x86.xop.vpperm(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
15 define <2 x double> @combine_vpermil2pd_identity(<2 x double> %a0, <2 x double> %a1) {
16 ; CHECK-LABEL: combine_vpermil2pd_identity:
18 ; CHECK-NEXT: vmovaps %xmm1, %xmm0
19 ; CHECK-NEXT: ret{{[l|q]}}
20 %res0 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a1, <2 x double> %a0, <2 x i64> <i64 2, i64 0>, i8 0)
21 %res1 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %res0, <2 x double> undef, <2 x i64> <i64 2, i64 0>, i8 0)
22 ret <2 x double> %res1
25 define <4 x double> @combine_vpermil2pd256_identity(<4 x double> %a0, <4 x double> %a1) {
26 ; CHECK-LABEL: combine_vpermil2pd256_identity:
28 ; CHECK-NEXT: vmovaps %ymm1, %ymm0
29 ; CHECK-NEXT: ret{{[l|q]}}
30 %res0 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a1, <4 x double> %a0, <4 x i64> <i64 2, i64 0, i64 2, i64 0>, i8 0)
31 %res1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %res0, <4 x double> undef, <4 x i64> <i64 2, i64 0, i64 2, i64 0>, i8 0)
32 ret <4 x double> %res1
35 define <4 x double> @combine_vpermil2pd256_0z73(<4 x double> %a0, <4 x double> %a1) {
36 ; CHECK-LABEL: combine_vpermil2pd256_0z73:
38 ; CHECK-NEXT: vpermil2pd {{.*#+}} ymm0 = ymm0[0],zero,ymm1[3],ymm0[3]
39 ; CHECK-NEXT: ret{{[l|q]}}
40 %res0 = shufflevector <4 x double> %a0, <4 x double> %a1, <4 x i32> <i32 0, i32 undef, i32 7, i32 3>
41 %res1 = shufflevector <4 x double> %res0, <4 x double> zeroinitializer, <4 x i32> <i32 0, i32 7, i32 2, i32 3>
42 ret <4 x double> %res1
45 define <4 x float> @combine_vpermil2ps_identity(<4 x float> %a0, <4 x float> %a1) {
46 ; CHECK-LABEL: combine_vpermil2ps_identity:
48 ; CHECK-NEXT: vmovaps %xmm1, %xmm0
49 ; CHECK-NEXT: ret{{[l|q]}}
50 %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a1, <4 x float> %a0, <4 x i32> <i32 3, i32 2, i32 1, i32 0>, i8 0)
51 %res1 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %res0, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>, i8 0)
55 define <4 x float> @combine_vpermil2ps_1z74(<4 x float> %a0, <4 x float> %a1) {
56 ; CHECK-LABEL: combine_vpermil2ps_1z74:
58 ; CHECK-NEXT: vshufps {{.*#+}} xmm0 = xmm0[1,1],xmm1[3,0]
59 ; CHECK-NEXT: vxorps %xmm1, %xmm1, %xmm1
60 ; CHECK-NEXT: vblendps {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3]
61 ; CHECK-NEXT: ret{{[l|q]}}
62 %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 1, i32 1, i32 7, i32 4>, i8 0)
63 %res1 = shufflevector <4 x float> %res0, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 7, i32 2, i32 3>
67 define <4 x float> @combine_vpermil2ps_02zu(<4 x float> %a0, <4 x float> %a1) {
68 ; CHECK-LABEL: combine_vpermil2ps_02zu:
70 ; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0,2],zero,zero
71 ; CHECK-NEXT: ret{{[l|q]}}
72 %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a0, <4 x float> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 undef>, i8 0)
76 define <8 x float> @combine_vpermil2ps256_identity(<8 x float> %a0, <8 x float> %a1) {
77 ; CHECK-LABEL: combine_vpermil2ps256_identity:
79 ; CHECK-NEXT: vmovaps %ymm1, %ymm0
80 ; CHECK-NEXT: ret{{[l|q]}}
81 %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a1, <8 x float> %a0, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 1, i32 0, i32 3, i32 2>, i8 0)
82 %res1 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %res0, <8 x float> undef, <8 x i32> <i32 3, i32 2, i32 1, i32 0, i32 1, i32 0, i32 3, i32 2>, i8 0)
86 define <8 x float> @combine_vpermil2ps256_08z945Az(<8 x float> %a0, <8 x float> %a1) {
87 ; CHECK-LABEL: combine_vpermil2ps256_08z945Az:
89 ; CHECK-NEXT: vpermil2ps {{.*#+}} ymm0 = ymm0[0],ymm1[0],zero,ymm1[1],ymm0[4,5],ymm1[6],zero
90 ; CHECK-NEXT: ret{{[l|q]}}
91 %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a0, <8 x float> %a1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 0, i32 1, i32 6, i32 7>, i8 0)
92 %res1 = shufflevector <8 x float> %res0, <8 x float> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 8, i32 3, i32 4, i32 5, i32 6, i32 8>
96 define <8 x float> @combine_vpermil2ps256_zero(<8 x float> %a0, <8 x float> %a1) {
97 ; CHECK-LABEL: combine_vpermil2ps256_zero:
99 ; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
100 ; CHECK-NEXT: ret{{[l|q]}}
101 %res0 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> %a1, <8 x float> %a0, <8 x i32> <i32 8, i32 9, i32 10, i32 11, i32 8, i32 9, i32 10, i32 11>, i8 2)
102 ret <8 x float> %res0
105 define <4 x float> @combine_vpermil2ps_blend_with_zero(<4 x float> %a0, <4 x float> %a1) {
106 ; CHECK-LABEL: combine_vpermil2ps_blend_with_zero:
108 ; CHECK-NEXT: vxorps %xmm1, %xmm1, %xmm1
109 ; CHECK-NEXT: vblendps {{.*#+}} xmm0 = xmm1[0],xmm0[1,2,3]
110 ; CHECK-NEXT: ret{{[l|q]}}
111 %res0 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> %a0, <4 x float> %a1, <4 x i32> <i32 8, i32 1, i32 2, i32 3>, i8 2)
112 ret <4 x float> %res0
115 define <2 x double> @combine_vpermil2pd_as_shufpd(<2 x double> %a0, <2 x double> %a1) {
116 ; CHECK-LABEL: combine_vpermil2pd_as_shufpd:
118 ; CHECK-NEXT: vshufpd {{.*#+}} xmm0 = xmm0[1],xmm1[0]
119 ; CHECK-NEXT: ret{{[l|q]}}
120 %res0 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> %a0, <2 x double> %a1, <2 x i64> <i64 2, i64 4>, i8 0)
121 ret <2 x double> %res0
124 define <4 x double> @combine_vpermil2pd256_as_shufpd(<4 x double> %a0, <4 x double> %a1) {
125 ; CHECK-LABEL: combine_vpermil2pd256_as_shufpd:
127 ; CHECK-NEXT: vshufpd {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[3],ymm1[3]
128 ; CHECK-NEXT: ret{{[l|q]}}
129 %res0 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x i64> <i64 0, i64 4, i64 2, i64 7>, i8 0)
130 ret <4 x double> %res0
133 define <4 x double> @demandedelts_vpermil2pd256_as_shufpd(<4 x double> %a0, <4 x double> %a1, i64 %a2) {
134 ; X86-AVX-LABEL: demandedelts_vpermil2pd256_as_shufpd:
136 ; X86-AVX-NEXT: movl $4, %eax
137 ; X86-AVX-NEXT: vmovd %eax, %xmm2
138 ; X86-AVX-NEXT: vmovq {{.*#+}} xmm3 = mem[0],zero
139 ; X86-AVX-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
140 ; X86-AVX-NEXT: vinsertf128 $1, {{\.LCPI.*}}, %ymm2, %ymm2
141 ; X86-AVX-NEXT: vpermil2pd $0, %ymm2, %ymm1, %ymm0, %ymm0
142 ; X86-AVX-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
145 ; X86-AVX2-LABEL: demandedelts_vpermil2pd256_as_shufpd:
147 ; X86-AVX2-NEXT: movl $4, %eax
148 ; X86-AVX2-NEXT: vmovd %eax, %xmm2
149 ; X86-AVX2-NEXT: vmovq {{.*#+}} xmm3 = mem[0],zero
150 ; X86-AVX2-NEXT: vpunpcklqdq {{.*#+}} xmm2 = xmm3[0],xmm2[0]
151 ; X86-AVX2-NEXT: vinserti128 $1, {{\.LCPI.*}}, %ymm2, %ymm2
152 ; X86-AVX2-NEXT: vpermil2pd $0, %ymm2, %ymm1, %ymm0, %ymm0
153 ; X86-AVX2-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,1,2,3]
154 ; X86-AVX2-NEXT: retl
156 ; X64-LABEL: demandedelts_vpermil2pd256_as_shufpd:
158 ; X64-NEXT: vpermil2pd {{.*#+}} ymm0 = ymm1[0,0],ymm0[3],ymm1[3]
160 %res0 = insertelement <4 x i64> <i64 0, i64 4, i64 2, i64 7>, i64 %a2, i32 0
161 %res1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> %a0, <4 x double> %a1, <4 x i64> %res0, i8 0)
162 %res2 = shufflevector <4 x double> %res1, <4 x double> undef, <4 x i32> <i32 1, i32 1, i32 2, i32 3>
163 ret <4 x double> %res2
166 define <16 x i8> @combine_vpperm_identity(<16 x i8> %a0, <16 x i8> %a1) {
167 ; CHECK-LABEL: combine_vpperm_identity:
169 ; CHECK-NEXT: vmovaps %xmm1, %xmm0
170 ; CHECK-NEXT: ret{{[l|q]}}
171 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 31, i8 30, i8 29, i8 28, i8 27, i8 26, i8 25, i8 24, i8 23, i8 22, i8 21, i8 20, i8 19, i8 18, i8 17, i8 16>)
172 %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>)
176 define <16 x i8> @combine_vpperm_zero(<16 x i8> %a0, <16 x i8> %a1) {
177 ; CHECK-LABEL: combine_vpperm_zero:
179 ; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
180 ; CHECK-NEXT: ret{{[l|q]}}
181 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 128, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>)
182 %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> <i8 0, i8 128, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>)
183 %res2 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res1, <16 x i8> undef, <16 x i8> <i8 0, i8 1, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128, i8 128>)
187 define <16 x i8> @combine_vpperm_identity_bitcast(<16 x i8> %a0, <16 x i8> %a1) {
188 ; X86-LABEL: combine_vpperm_identity_bitcast:
190 ; X86-NEXT: vpaddq {{\.LCPI.*}}, %xmm0, %xmm0
193 ; X64-LABEL: combine_vpperm_identity_bitcast:
195 ; X64-NEXT: vpaddq {{.*}}(%rip), %xmm0, %xmm0
197 %mask = bitcast <2 x i64> <i64 1084818905618843912, i64 506097522914230528> to <16 x i8>
198 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> %mask)
199 %res1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res0, <16 x i8> undef, <16 x i8> %mask)
200 %res2 = bitcast <16 x i8> %res1 to <2 x i64>
201 %res3 = add <2 x i64> %res2, <i64 1084818905618843912, i64 506097522914230528>
202 %res4 = bitcast <2 x i64> %res3 to <16 x i8>
206 define <16 x i8> @combine_vpperm_as_blend_with_zero(<16 x i8> %a0, <16 x i8> %a1) {
207 ; CHECK-LABEL: combine_vpperm_as_blend_with_zero:
209 ; CHECK-NEXT: vpxor %xmm1, %xmm1, %xmm1
210 ; CHECK-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0],xmm1[1],xmm0[2,3],xmm1[4,5,6,7]
211 ; CHECK-NEXT: ret{{[l|q]}}
212 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 0, i8 1, i8 128, i8 129, i8 4, i8 5, i8 6, i8 7, i8 130, i8 131, i8 132, i8 133, i8 134, i8 135, i8 136, i8 137>)
216 define <16 x i8> @combine_vpperm_as_unary_unpckhbw(<16 x i8> %a0, <16 x i8> %a1) {
217 ; CHECK-LABEL: combine_vpperm_as_unary_unpckhbw:
219 ; CHECK-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
220 ; CHECK-NEXT: ret{{[l|q]}}
221 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a0, <16 x i8> <i8 8, i8 undef, i8 9, i8 25, i8 10, i8 26, i8 11, i8 27, i8 12, i8 28, i8 13, i8 29, i8 14, i8 30, i8 15, i8 31>)
225 define <16 x i8> @combine_vpperm_as_unpckhbw(<16 x i8> %a0, <16 x i8> %a1) {
226 ; CHECK-LABEL: combine_vpperm_as_unpckhbw:
228 ; CHECK-NEXT: vpunpckhbw {{.*#+}} xmm0 = xmm0[8],xmm1[8],xmm0[9],xmm1[9],xmm0[10],xmm1[10],xmm0[11],xmm1[11],xmm0[12],xmm1[12],xmm0[13],xmm1[13],xmm0[14],xmm1[14],xmm0[15],xmm1[15]
229 ; CHECK-NEXT: ret{{[l|q]}}
230 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 8, i8 24, i8 9, i8 25, i8 10, i8 26, i8 11, i8 27, i8 12, i8 28, i8 13, i8 29, i8 14, i8 30, i8 15, i8 31>)
234 define <16 x i8> @combine_vpperm_as_unpcklbw(<16 x i8> %a0, <16 x i8> %a1) {
235 ; CHECK-LABEL: combine_vpperm_as_unpcklbw:
237 ; CHECK-NEXT: vpunpcklbw {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3],xmm1[4],xmm0[4],xmm1[5],xmm0[5],xmm1[6],xmm0[6],xmm1[7],xmm0[7]
238 ; CHECK-NEXT: ret{{[l|q]}}
239 %res0 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %a0, <16 x i8> %a1, <16 x i8> <i8 16, i8 0, i8 17, i8 1, i8 18, i8 2, i8 19, i8 3, i8 20, i8 4, i8 21, i8 5, i8 22, i8 6, i8 23, i8 7>)
243 define <4 x i32> @combine_vpperm_10zz32BA(<4 x i32> %a0, <4 x i32> %a1) {
244 ; CHECK-LABEL: combine_vpperm_10zz32BA:
246 ; CHECK-NEXT: vpperm {{.*#+}} xmm0 = xmm0[2,3,0,1],zero,zero,zero,zero,xmm0[6,7,4,5],xmm1[6,7,4,5]
247 ; CHECK-NEXT: ret{{[l|q]}}
248 %res0 = shufflevector <4 x i32> %a0, <4 x i32> %a1, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
249 %res1 = bitcast <4 x i32> %res0 to <16 x i8>
250 %res2 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> %res1, <16 x i8> undef, <16 x i8> <i8 2, i8 3, i8 0, i8 1, i8 128, i8 128, i8 128, i8 128, i8 10, i8 11, i8 8, i8 9, i8 14, i8 15, i8 12, i8 13>)
251 %res3 = bitcast <16 x i8> %res2 to <4 x i32>
255 define void @buildvector_v4f32_0404(float %a, float %b, <4 x float>* %ptr) {
256 ; X86-LABEL: buildvector_v4f32_0404:
258 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
259 ; X86-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
260 ; X86-NEXT: vmovaps %xmm0, (%eax)
263 ; X64-AVX-LABEL: buildvector_v4f32_0404:
265 ; X64-AVX-NEXT: vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[0],xmm1[0]
266 ; X64-AVX-NEXT: vmovaps %xmm0, (%rdi)
269 ; X64-AVX2-LABEL: buildvector_v4f32_0404:
271 ; X64-AVX2-NEXT: vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
272 ; X64-AVX2-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
273 ; X64-AVX2-NEXT: vmovaps %xmm0, (%rdi)
274 ; X64-AVX2-NEXT: retq
275 %v0 = insertelement <4 x float> undef, float %a, i32 0
276 %v1 = insertelement <4 x float> %v0, float %b, i32 1
277 %v2 = insertelement <4 x float> %v1, float %a, i32 2
278 %v3 = insertelement <4 x float> %v2, float %b, i32 3
279 store <4 x float> %v3, <4 x float>* %ptr
283 define void @buildvector_v4f32_07z6(float %a, <4 x float> %b, <4 x float>* %ptr) {
284 ; X86-LABEL: buildvector_v4f32_07z6:
286 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
287 ; X86-NEXT: vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
288 ; X86-NEXT: vpermil2ps {{.*#+}} xmm0 = xmm1[0],xmm0[3],zero,xmm0[2]
289 ; X86-NEXT: vmovaps %xmm0, (%eax)
292 ; X64-LABEL: buildvector_v4f32_07z6:
294 ; X64-NEXT: vpermil2ps {{.*#+}} xmm0 = xmm0[0],xmm1[3],zero,xmm1[2]
295 ; X64-NEXT: vmovaps %xmm0, (%rdi)
297 %b2 = extractelement <4 x float> %b, i32 2
298 %b3 = extractelement <4 x float> %b, i32 3
299 %v0 = insertelement <4 x float> undef, float %a, i32 0
300 %v1 = insertelement <4 x float> %v0, float %b3, i32 1
301 %v2 = insertelement <4 x float> %v1, float 0.0, i32 2
302 %v3 = insertelement <4 x float> %v2, float %b2, i32 3
303 store <4 x float> %v3, <4 x float>* %ptr
307 define <2 x double> @constant_fold_vpermil2pd() {
308 ; CHECK-LABEL: constant_fold_vpermil2pd:
310 ; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [-2.0E+0,2.0E+0]
311 ; CHECK-NEXT: ret{{[l|q]}}
312 %1 = call <2 x double> @llvm.x86.xop.vpermil2pd(<2 x double> <double 1.0, double 2.0>, <2 x double> <double -2.0, double -1.0>, <2 x i64> <i64 4, i64 2>, i8 2)
316 define <4 x double> @constant_fold_vpermil2pd_256() {
317 ; CHECK-LABEL: constant_fold_vpermil2pd_256:
319 ; CHECK-NEXT: vmovaps {{.*#+}} ymm0 = [-4.0E+0,0.0E+0,4.0E+0,3.0E+0]
320 ; CHECK-NEXT: ret{{[l|q]}}
321 %1 = call <4 x double> @llvm.x86.xop.vpermil2pd.256(<4 x double> <double 1.0, double 2.0, double 3.0, double 4.0>, <4 x double> <double -4.0, double -3.0, double -2.0, double -1.0>, <4 x i64> <i64 4, i64 8, i64 2, i64 0>, i8 2)
325 define <4 x float> @constant_fold_vpermil2ps() {
326 ; CHECK-LABEL: constant_fold_vpermil2ps:
328 ; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [-4.0E+0,1.0E+0,3.0E+0,0.0E+0]
329 ; CHECK-NEXT: ret{{[l|q]}}
330 %1 = call <4 x float> @llvm.x86.xop.vpermil2ps(<4 x float> <float 1.0, float 2.0, float 3.0, float 4.0>, <4 x float> <float -4.0, float -3.0, float -2.0, float -1.0>, <4 x i32> <i32 4, i32 0, i32 2, i32 8>, i8 2)
334 define <8 x float> @constant_fold_vpermil2ps_256() {
335 ; CHECK-LABEL: constant_fold_vpermil2ps_256:
337 ; CHECK-NEXT: vmovaps {{.*#+}} ymm0 = [-8.0E+0,1.0E+0,3.0E+0,0.0E+0,5.0E+0,0.0E+0,5.0E+0,7.0E+0]
338 ; CHECK-NEXT: ret{{[l|q]}}
339 %1 = call <8 x float> @llvm.x86.xop.vpermil2ps.256(<8 x float> <float 1.0, float 2.0, float 3.0, float 4.0, float 5.0, float 6.0, float 7.0, float 8.0>, <8 x float> <float -8.0, float -7.0, float -6.0, float -5.0, float -4.0, float -3.0, float -2.0, float -1.0>, <8 x i32> <i32 4, i32 0, i32 2, i32 8, i32 0, i32 8, i32 0, i32 2>, i8 2)
343 define <16 x i8> @constant_fold_vpperm() {
344 ; CHECK-LABEL: constant_fold_vpperm:
346 ; CHECK-NEXT: vmovaps {{.*#+}} xmm0 = [0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]
347 ; CHECK-NEXT: ret{{[l|q]}}
348 %1 = call <16 x i8> @llvm.x86.xop.vpperm(<16 x i8> <i8 0, i8 -1, i8 -2, i8 -3, i8 -4, i8 -5, i8 -6, i8 -7, i8 -8, i8 -9, i8 -10, i8 -11, i8 -12, i8 -13, i8 -14, i8 -15>, <16 x i8> <i8 15, i8 14, i8 13, i8 12, i8 11, i8 10, i8 9, i8 8, i8 7, i8 6, i8 5, i8 4, i8 3, i8 2, i8 1, i8 0>, <16 x i8> <i8 31, i8 30, i8 29, i8 28, i8 27, i8 26, i8 25, i8 24, i8 23, i8 22, i8 21, i8 20, i8 19, i8 18, i8 17, i8 16>)
352 define <4 x float> @PR31296(i8* %in) {
353 ; X86-LABEL: PR31296:
354 ; X86: # %bb.0: # %entry
355 ; X86-NEXT: movl {{[0-9]+}}(%esp), %eax
356 ; X86-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
357 ; X86-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],zero,zero,mem[0]
360 ; X64-LABEL: PR31296:
361 ; X64: # %bb.0: # %entry
362 ; X64-NEXT: movl (%rdi), %eax
363 ; X64-NEXT: vmovq %rax, %xmm0
364 ; X64-NEXT: vinsertps {{.*#+}} xmm0 = xmm0[0],zero,zero,mem[0]
367 %0 = getelementptr i8, i8* %in, i32 0
368 %1 = bitcast i8* %0 to i32*
369 %2 = load i32, i32* %1
370 %3 = zext i32 %2 to i128
371 %4 = bitcast i128 %3 to <4 x float>
372 %5 = shufflevector <4 x float> %4, <4 x float> <float 0.000000e+00, float 1.000000e+00, float undef, float undef>, <4 x i32> <i32 0, i32 4, i32 4, i32 5>