1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=X32
3 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=X64
5 define <2 x double> @signbits_sext_v2i64_sitofp_v2f64(i32 %a0, i32 %a1) nounwind {
6 ; X32-LABEL: signbits_sext_v2i64_sitofp_v2f64:
8 ; X32-NEXT: vcvtdq2pd {{[0-9]+}}(%esp), %xmm0
11 ; X64-LABEL: signbits_sext_v2i64_sitofp_v2f64:
13 ; X64-NEXT: vmovd %edi, %xmm0
14 ; X64-NEXT: vpinsrd $1, %esi, %xmm0, %xmm0
15 ; X64-NEXT: vcvtdq2pd %xmm0, %xmm0
17 %1 = sext i32 %a0 to i64
18 %2 = sext i32 %a1 to i64
19 %3 = insertelement <2 x i64> undef, i64 %1, i32 0
20 %4 = insertelement <2 x i64> %3, i64 %2, i32 1
21 %5 = sitofp <2 x i64> %4 to <2 x double>
25 define <4 x float> @signbits_sext_v4i64_sitofp_v4f32(i8 signext %a0, i16 signext %a1, i32 %a2, i32 %a3) nounwind {
26 ; X32-LABEL: signbits_sext_v4i64_sitofp_v4f32:
28 ; X32-NEXT: movswl {{[0-9]+}}(%esp), %eax
29 ; X32-NEXT: movsbl {{[0-9]+}}(%esp), %ecx
30 ; X32-NEXT: vmovd %ecx, %xmm0
31 ; X32-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0
32 ; X32-NEXT: vpinsrd $2, {{[0-9]+}}(%esp), %xmm0, %xmm0
33 ; X32-NEXT: vpinsrd $3, {{[0-9]+}}(%esp), %xmm0, %xmm0
34 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
37 ; X64-LABEL: signbits_sext_v4i64_sitofp_v4f32:
39 ; X64-NEXT: vmovd %edi, %xmm0
40 ; X64-NEXT: vpinsrd $1, %esi, %xmm0, %xmm0
41 ; X64-NEXT: vpinsrd $2, %edx, %xmm0, %xmm0
42 ; X64-NEXT: vpinsrd $3, %ecx, %xmm0, %xmm0
43 ; X64-NEXT: vcvtdq2ps %xmm0, %xmm0
45 %1 = sext i8 %a0 to i64
46 %2 = sext i16 %a1 to i64
47 %3 = sext i32 %a2 to i64
48 %4 = sext i32 %a3 to i64
49 %5 = insertelement <4 x i64> undef, i64 %1, i32 0
50 %6 = insertelement <4 x i64> %5, i64 %2, i32 1
51 %7 = insertelement <4 x i64> %6, i64 %3, i32 2
52 %8 = insertelement <4 x i64> %7, i64 %4, i32 3
53 %9 = sitofp <4 x i64> %8 to <4 x float>
57 define float @signbits_ashr_extract_sitofp_0(<2 x i64> %a0) nounwind {
58 ; X32-LABEL: signbits_ashr_extract_sitofp_0:
60 ; X32-NEXT: pushl %eax
61 ; X32-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,3]
62 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
63 ; X32-NEXT: vmovss %xmm0, (%esp)
64 ; X32-NEXT: flds (%esp)
68 ; X64-LABEL: signbits_ashr_extract_sitofp_0:
70 ; X64-NEXT: vmovq %xmm0, %rax
71 ; X64-NEXT: shrq $32, %rax
72 ; X64-NEXT: vcvtsi2ss %eax, %xmm1, %xmm0
74 %1 = ashr <2 x i64> %a0, <i64 32, i64 32>
75 %2 = extractelement <2 x i64> %1, i32 0
76 %3 = sitofp i64 %2 to float
80 define float @signbits_ashr_extract_sitofp_1(<2 x i64> %a0) nounwind {
81 ; X32-LABEL: signbits_ashr_extract_sitofp_1:
83 ; X32-NEXT: pushl %eax
84 ; X32-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,2,3]
85 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
86 ; X32-NEXT: vmovss %xmm0, (%esp)
87 ; X32-NEXT: flds (%esp)
91 ; X64-LABEL: signbits_ashr_extract_sitofp_1:
93 ; X64-NEXT: vmovq %xmm0, %rax
94 ; X64-NEXT: shrq $32, %rax
95 ; X64-NEXT: vcvtsi2ss %eax, %xmm1, %xmm0
97 %1 = ashr <2 x i64> %a0, <i64 32, i64 63>
98 %2 = extractelement <2 x i64> %1, i32 0
99 %3 = sitofp i64 %2 to float
103 define float @signbits_ashr_shl_extract_sitofp(<2 x i64> %a0) nounwind {
104 ; X32-LABEL: signbits_ashr_shl_extract_sitofp:
106 ; X32-NEXT: pushl %eax
107 ; X32-NEXT: vpsrad $29, %xmm0, %xmm0
108 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
109 ; X32-NEXT: vpsllq $20, %xmm0, %xmm0
110 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
111 ; X32-NEXT: vmovss %xmm0, (%esp)
112 ; X32-NEXT: flds (%esp)
113 ; X32-NEXT: popl %eax
116 ; X64-LABEL: signbits_ashr_shl_extract_sitofp:
118 ; X64-NEXT: vmovq %xmm0, %rax
119 ; X64-NEXT: sarq $61, %rax
120 ; X64-NEXT: shll $20, %eax
121 ; X64-NEXT: vcvtsi2ss %eax, %xmm1, %xmm0
123 %1 = ashr <2 x i64> %a0, <i64 61, i64 60>
124 %2 = shl <2 x i64> %1, <i64 20, i64 16>
125 %3 = extractelement <2 x i64> %2, i32 0
126 %4 = sitofp i64 %3 to float
130 define float @signbits_ashr_insert_ashr_extract_sitofp(i64 %a0, i64 %a1) nounwind {
131 ; X32-LABEL: signbits_ashr_insert_ashr_extract_sitofp:
133 ; X32-NEXT: pushl %eax
134 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
135 ; X32-NEXT: movl %eax, %ecx
136 ; X32-NEXT: sarl $30, %ecx
137 ; X32-NEXT: shll $2, %eax
138 ; X32-NEXT: vmovd %eax, %xmm0
139 ; X32-NEXT: vpinsrd $1, %ecx, %xmm0, %xmm0
140 ; X32-NEXT: vpsrlq $3, %xmm0, %xmm0
141 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
142 ; X32-NEXT: vmovss %xmm0, (%esp)
143 ; X32-NEXT: flds (%esp)
144 ; X32-NEXT: popl %eax
147 ; X64-LABEL: signbits_ashr_insert_ashr_extract_sitofp:
149 ; X64-NEXT: sarq $30, %rdi
150 ; X64-NEXT: shrq $3, %rdi
151 ; X64-NEXT: vcvtsi2ss %edi, %xmm0, %xmm0
153 %1 = ashr i64 %a0, 30
154 %2 = insertelement <2 x i64> undef, i64 %1, i32 0
155 %3 = insertelement <2 x i64> %2, i64 %a1, i32 1
156 %4 = ashr <2 x i64> %3, <i64 3, i64 3>
157 %5 = extractelement <2 x i64> %4, i32 0
158 %6 = sitofp i64 %5 to float
162 define <4 x double> @signbits_sext_shuffle_sitofp(<4 x i32> %a0, <4 x i64> %a1) nounwind {
163 ; X32-LABEL: signbits_sext_shuffle_sitofp:
165 ; X32-NEXT: vpmovsxdq %xmm0, %xmm1
166 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
167 ; X32-NEXT: vpmovsxdq %xmm0, %xmm0
168 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
169 ; X32-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
170 ; X32-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
171 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm1
172 ; X32-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
173 ; X32-NEXT: vcvtdq2pd %xmm0, %ymm0
176 ; X64-LABEL: signbits_sext_shuffle_sitofp:
178 ; X64-NEXT: vpmovsxdq %xmm0, %xmm1
179 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
180 ; X64-NEXT: vpmovsxdq %xmm0, %xmm0
181 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
182 ; X64-NEXT: vpermilpd {{.*#+}} ymm0 = ymm0[1,0,3,2]
183 ; X64-NEXT: vperm2f128 {{.*#+}} ymm0 = ymm0[2,3,0,1]
184 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
185 ; X64-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
186 ; X64-NEXT: vcvtdq2pd %xmm0, %ymm0
188 %1 = sext <4 x i32> %a0 to <4 x i64>
189 %2 = shufflevector <4 x i64> %1, <4 x i64>%a1, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
190 %3 = sitofp <4 x i64> %2 to <4 x double>
194 ; TODO: Fix vpshufd+vpsrlq -> vpshufd/vpermilps
195 define <2 x double> @signbits_ashr_concat_ashr_extract_sitofp(<2 x i64> %a0, <4 x i64> %a1) nounwind {
196 ; X32-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
198 ; X32-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,3,2,3]
199 ; X32-NEXT: vcvtdq2pd %xmm0, %xmm0
202 ; X64-LABEL: signbits_ashr_concat_ashr_extract_sitofp:
204 ; X64-NEXT: vpsrlq $32, %xmm0, %xmm0
205 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
206 ; X64-NEXT: vcvtdq2pd %xmm0, %xmm0
208 %1 = ashr <2 x i64> %a0, <i64 16, i64 16>
209 %2 = shufflevector <2 x i64> %1, <2 x i64> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef>
210 %3 = shufflevector <4 x i64> %a1, <4 x i64> %2, <4 x i32> <i32 0, i32 1, i32 4, i32 5>
211 %4 = ashr <4 x i64> %3, <i64 16, i64 16, i64 16, i64 16>
212 %5 = shufflevector <4 x i64> %4, <4 x i64> undef, <2 x i32> <i32 2, i32 3>
213 %6 = sitofp <2 x i64> %5 to <2 x double>
217 define float @signbits_ashr_sext_sextinreg_and_extract_sitofp(<2 x i64> %a0, <2 x i64> %a1, i32 %a2) nounwind {
218 ; X32-LABEL: signbits_ashr_sext_sextinreg_and_extract_sitofp:
220 ; X32-NEXT: pushl %eax
221 ; X32-NEXT: vpsrad $29, %xmm0, %xmm0
222 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
223 ; X32-NEXT: vmovd {{.*#+}} xmm1 = mem[0],zero,zero,zero
224 ; X32-NEXT: vpand %xmm1, %xmm0, %xmm0
225 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
226 ; X32-NEXT: vmovss %xmm0, (%esp)
227 ; X32-NEXT: flds (%esp)
228 ; X32-NEXT: popl %eax
231 ; X64-LABEL: signbits_ashr_sext_sextinreg_and_extract_sitofp:
233 ; X64-NEXT: vpsrad $29, %xmm0, %xmm0
234 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
235 ; X64-NEXT: vmovd %edi, %xmm1
236 ; X64-NEXT: vpand %xmm1, %xmm0, %xmm0
237 ; X64-NEXT: vmovq %xmm0, %rax
238 ; X64-NEXT: vcvtsi2ss %eax, %xmm2, %xmm0
240 %1 = ashr <2 x i64> %a0, <i64 61, i64 60>
241 %2 = sext i32 %a2 to i64
242 %3 = insertelement <2 x i64> %a1, i64 %2, i32 0
243 %4 = shl <2 x i64> %3, <i64 20, i64 20>
244 %5 = ashr <2 x i64> %4, <i64 20, i64 20>
245 %6 = and <2 x i64> %1, %5
246 %7 = extractelement <2 x i64> %6, i32 0
247 %8 = sitofp i64 %7 to float
251 define float @signbits_ashr_sextvecinreg_bitops_extract_sitofp(<2 x i64> %a0, <4 x i32> %a1) nounwind {
252 ; X32-LABEL: signbits_ashr_sextvecinreg_bitops_extract_sitofp:
254 ; X32-NEXT: pushl %eax
255 ; X32-NEXT: vpsrlq $60, %xmm0, %xmm2
256 ; X32-NEXT: vpsrlq $61, %xmm0, %xmm0
257 ; X32-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
258 ; X32-NEXT: vmovdqa {{.*#+}} xmm2 = [4,0,8,0]
259 ; X32-NEXT: vpxor %xmm2, %xmm0, %xmm0
260 ; X32-NEXT: vpsubq %xmm2, %xmm0, %xmm0
261 ; X32-NEXT: vpmovsxdq %xmm1, %xmm1
262 ; X32-NEXT: vpand %xmm1, %xmm0, %xmm2
263 ; X32-NEXT: vpor %xmm1, %xmm2, %xmm1
264 ; X32-NEXT: vpxor %xmm0, %xmm1, %xmm0
265 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
266 ; X32-NEXT: vmovss %xmm0, (%esp)
267 ; X32-NEXT: flds (%esp)
268 ; X32-NEXT: popl %eax
271 ; X64-LABEL: signbits_ashr_sextvecinreg_bitops_extract_sitofp:
273 ; X64-NEXT: vpsrlq $60, %xmm0, %xmm2
274 ; X64-NEXT: vpsrlq $61, %xmm0, %xmm0
275 ; X64-NEXT: vpblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm2[4,5,6,7]
276 ; X64-NEXT: vmovdqa {{.*#+}} xmm2 = [4,8]
277 ; X64-NEXT: vpxor %xmm2, %xmm0, %xmm0
278 ; X64-NEXT: vpsubq %xmm2, %xmm0, %xmm0
279 ; X64-NEXT: vpmovsxdq %xmm1, %xmm1
280 ; X64-NEXT: vpand %xmm1, %xmm0, %xmm2
281 ; X64-NEXT: vpor %xmm1, %xmm2, %xmm1
282 ; X64-NEXT: vpxor %xmm0, %xmm1, %xmm0
283 ; X64-NEXT: vmovq %xmm0, %rax
284 ; X64-NEXT: vcvtsi2ss %eax, %xmm3, %xmm0
286 %1 = ashr <2 x i64> %a0, <i64 61, i64 60>
287 %2 = shufflevector <4 x i32> %a1, <4 x i32> undef, <2 x i32> <i32 0, i32 1>
288 %3 = sext <2 x i32> %2 to <2 x i64>
289 %4 = and <2 x i64> %1, %3
290 %5 = or <2 x i64> %4, %3
291 %6 = xor <2 x i64> %5, %1
292 %7 = extractelement <2 x i64> %6, i32 0
293 %8 = sitofp i64 %7 to float
297 define <4 x float> @signbits_ashr_sext_select_shuffle_sitofp(<4 x i64> %a0, <4 x i64> %a1, <4 x i64> %a2, <4 x i32> %a3) nounwind {
298 ; X32-LABEL: signbits_ashr_sext_select_shuffle_sitofp:
300 ; X32-NEXT: pushl %ebp
301 ; X32-NEXT: movl %esp, %ebp
302 ; X32-NEXT: andl $-16, %esp
303 ; X32-NEXT: subl $16, %esp
304 ; X32-NEXT: vpmovsxdq 8(%ebp), %xmm3
305 ; X32-NEXT: vpmovsxdq 16(%ebp), %xmm4
306 ; X32-NEXT: vpsrad $31, %xmm2, %xmm5
307 ; X32-NEXT: vpsrad $1, %xmm2, %xmm6
308 ; X32-NEXT: vpshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
309 ; X32-NEXT: vpblendw {{.*#+}} xmm5 = xmm6[0,1],xmm5[2,3],xmm6[4,5],xmm5[6,7]
310 ; X32-NEXT: vextractf128 $1, %ymm2, %xmm2
311 ; X32-NEXT: vpsrad $31, %xmm2, %xmm6
312 ; X32-NEXT: vpsrad $1, %xmm2, %xmm2
313 ; X32-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
314 ; X32-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm6[2,3],xmm2[4,5],xmm6[6,7]
315 ; X32-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm6
316 ; X32-NEXT: vblendvpd %xmm6, %xmm5, %xmm3, %xmm3
317 ; X32-NEXT: vextractf128 $1, %ymm1, %xmm1
318 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm0
319 ; X32-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
320 ; X32-NEXT: vblendvpd %xmm0, %xmm2, %xmm4, %xmm0
321 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm3, %ymm0
322 ; X32-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
323 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm1
324 ; X32-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
325 ; X32-NEXT: vcvtdq2ps %xmm0, %xmm0
326 ; X32-NEXT: movl %ebp, %esp
327 ; X32-NEXT: popl %ebp
328 ; X32-NEXT: vzeroupper
331 ; X64-LABEL: signbits_ashr_sext_select_shuffle_sitofp:
333 ; X64-NEXT: vpsrad $31, %xmm2, %xmm4
334 ; X64-NEXT: vpsrad $1, %xmm2, %xmm5
335 ; X64-NEXT: vpshufd {{.*#+}} xmm5 = xmm5[1,1,3,3]
336 ; X64-NEXT: vpblendw {{.*#+}} xmm4 = xmm5[0,1],xmm4[2,3],xmm5[4,5],xmm4[6,7]
337 ; X64-NEXT: vextractf128 $1, %ymm2, %xmm2
338 ; X64-NEXT: vpsrad $31, %xmm2, %xmm5
339 ; X64-NEXT: vpsrad $1, %xmm2, %xmm2
340 ; X64-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
341 ; X64-NEXT: vpblendw {{.*#+}} xmm2 = xmm2[0,1],xmm5[2,3],xmm2[4,5],xmm5[6,7]
342 ; X64-NEXT: vpmovsxdq %xmm3, %xmm5
343 ; X64-NEXT: vpshufd {{.*#+}} xmm3 = xmm3[2,3,0,1]
344 ; X64-NEXT: vpmovsxdq %xmm3, %xmm3
345 ; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm6
346 ; X64-NEXT: vblendvpd %xmm6, %xmm4, %xmm5, %xmm4
347 ; X64-NEXT: vextractf128 $1, %ymm1, %xmm1
348 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
349 ; X64-NEXT: vpcmpeqq %xmm1, %xmm0, %xmm0
350 ; X64-NEXT: vblendvpd %xmm0, %xmm2, %xmm3, %xmm0
351 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm4, %ymm0
352 ; X64-NEXT: vmovddup {{.*#+}} ymm0 = ymm0[0,0,2,2]
353 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm1
354 ; X64-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm1[0,2]
355 ; X64-NEXT: vcvtdq2ps %xmm0, %xmm0
356 ; X64-NEXT: vzeroupper
358 %1 = ashr <4 x i64> %a2, <i64 33, i64 63, i64 33, i64 63>
359 %2 = sext <4 x i32> %a3 to <4 x i64>
360 %3 = icmp eq <4 x i64> %a0, %a1
361 %4 = select <4 x i1> %3, <4 x i64> %1, <4 x i64> %2
362 %5 = shufflevector <4 x i64> %4, <4 x i64> undef, <4 x i32> <i32 0, i32 0, i32 2, i32 2>
363 %6 = sitofp <4 x i64> %5 to <4 x float>
367 ; Make sure we can preserve sign bit information into the second basic block
368 ; so we can avoid having to shift bit 0 into bit 7 for each element due to
369 ; v32i1->v32i8 promotion and the splitting of v32i8 into 2xv16i8. This requires
370 ; ComputeNumSignBits handling for insert_subvector.
371 define void @cross_bb_signbits_insert_subvec(<32 x i8>* %ptr, <32 x i8> %x, <32 x i8> %z) {
372 ; X32-LABEL: cross_bb_signbits_insert_subvec:
374 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
375 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm2
376 ; X32-NEXT: vpxor %xmm3, %xmm3, %xmm3
377 ; X32-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm2
378 ; X32-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
379 ; X32-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
380 ; X32-NEXT: vandnps %ymm1, %ymm0, %ymm1
381 ; X32-NEXT: vandps {{\.LCPI.*}}, %ymm0, %ymm0
382 ; X32-NEXT: vorps %ymm1, %ymm0, %ymm0
383 ; X32-NEXT: vmovaps %ymm0, (%eax)
384 ; X32-NEXT: vzeroupper
387 ; X64-LABEL: cross_bb_signbits_insert_subvec:
389 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm2
390 ; X64-NEXT: vpxor %xmm3, %xmm3, %xmm3
391 ; X64-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm2
392 ; X64-NEXT: vpcmpeqb %xmm3, %xmm0, %xmm0
393 ; X64-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
394 ; X64-NEXT: vandnps %ymm1, %ymm0, %ymm1
395 ; X64-NEXT: vandps {{.*}}(%rip), %ymm0, %ymm0
396 ; X64-NEXT: vorps %ymm1, %ymm0, %ymm0
397 ; X64-NEXT: vmovaps %ymm0, (%rdi)
398 ; X64-NEXT: vzeroupper
400 %a = icmp eq <32 x i8> %x, zeroinitializer
401 %b = icmp eq <32 x i8> %x, zeroinitializer
402 %c = and <32 x i1> %a, %b
406 %d = select <32 x i1> %c, <32 x i8> <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>, <32 x i8> %z
407 store <32 x i8> %d, <32 x i8>* %ptr, align 32