1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s -mtriple=i686-apple-darwin -mattr=+avx | FileCheck %s --check-prefix=X32
3 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -mattr=+avx | FileCheck %s --check-prefix=X64
5 define <4 x i64> @A(i64* %ptr) nounwind uwtable readnone ssp {
7 ; X32: ## BB#0: ## %entry
8 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
9 ; X32-NEXT: movl (%eax), %ecx
10 ; X32-NEXT: movl 4(%eax), %eax
11 ; X32-NEXT: vmovd %ecx, %xmm0
12 ; X32-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0
13 ; X32-NEXT: vpinsrd $2, %ecx, %xmm0, %xmm0
14 ; X32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0
15 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
19 ; X64: ## BB#0: ## %entry
20 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
23 %q = load i64, i64* %ptr, align 8
24 %vecinit.i = insertelement <4 x i64> undef, i64 %q, i32 0
25 %vecinit2.i = insertelement <4 x i64> %vecinit.i, i64 %q, i32 1
26 %vecinit4.i = insertelement <4 x i64> %vecinit2.i, i64 %q, i32 2
27 %vecinit6.i = insertelement <4 x i64> %vecinit4.i, i64 %q, i32 3
28 ret <4 x i64> %vecinit6.i
31 define <8 x i32> @B(i32* %ptr) nounwind uwtable readnone ssp {
33 ; X32: ## BB#0: ## %entry
34 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
35 ; X32-NEXT: vbroadcastss (%eax), %ymm0
39 ; X64: ## BB#0: ## %entry
40 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
43 %q = load i32, i32* %ptr, align 4
44 %vecinit.i = insertelement <8 x i32> undef, i32 %q, i32 0
45 %vecinit2.i = insertelement <8 x i32> %vecinit.i, i32 %q, i32 1
46 %vecinit4.i = insertelement <8 x i32> %vecinit2.i, i32 %q, i32 2
47 %vecinit6.i = insertelement <8 x i32> %vecinit4.i, i32 %q, i32 3
48 ret <8 x i32> %vecinit6.i
51 define <4 x double> @C(double* %ptr) nounwind uwtable readnone ssp {
53 ; X32: ## BB#0: ## %entry
54 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
55 ; X32-NEXT: vbroadcastsd (%eax), %ymm0
59 ; X64: ## BB#0: ## %entry
60 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
63 %q = load double, double* %ptr, align 8
64 %vecinit.i = insertelement <4 x double> undef, double %q, i32 0
65 %vecinit2.i = insertelement <4 x double> %vecinit.i, double %q, i32 1
66 %vecinit4.i = insertelement <4 x double> %vecinit2.i, double %q, i32 2
67 %vecinit6.i = insertelement <4 x double> %vecinit4.i, double %q, i32 3
68 ret <4 x double> %vecinit6.i
71 define <8 x float> @D(float* %ptr) nounwind uwtable readnone ssp {
73 ; X32: ## BB#0: ## %entry
74 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
75 ; X32-NEXT: vbroadcastss (%eax), %ymm0
79 ; X64: ## BB#0: ## %entry
80 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
83 %q = load float, float* %ptr, align 4
84 %vecinit.i = insertelement <8 x float> undef, float %q, i32 0
85 %vecinit2.i = insertelement <8 x float> %vecinit.i, float %q, i32 1
86 %vecinit4.i = insertelement <8 x float> %vecinit2.i, float %q, i32 2
87 %vecinit6.i = insertelement <8 x float> %vecinit4.i, float %q, i32 3
88 ret <8 x float> %vecinit6.i
93 define <4 x float> @e(float* %ptr) nounwind uwtable readnone ssp {
95 ; X32: ## BB#0: ## %entry
96 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
97 ; X32-NEXT: vbroadcastss (%eax), %xmm0
101 ; X64: ## BB#0: ## %entry
102 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
105 %q = load float, float* %ptr, align 4
106 %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
107 %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
108 %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
109 %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
110 ret <4 x float> %vecinit6.i
113 ; Don't broadcast constants on pre-AVX2 hardware.
114 define <4 x float> @_e2(float* %ptr) nounwind uwtable readnone ssp {
116 ; X32: ## BB#0: ## %entry
117 ; X32-NEXT: vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
121 ; X64: ## BB#0: ## %entry
122 ; X64-NEXT: vmovaps {{.*#+}} xmm0 = [-7.812500e-03,-7.812500e-03,-7.812500e-03,-7.812500e-03]
125 %vecinit.i = insertelement <4 x float> undef, float 0xbf80000000000000, i32 0
126 %vecinit2.i = insertelement <4 x float> %vecinit.i, float 0xbf80000000000000, i32 1
127 %vecinit4.i = insertelement <4 x float> %vecinit2.i, float 0xbf80000000000000, i32 2
128 %vecinit6.i = insertelement <4 x float> %vecinit4.i, float 0xbf80000000000000, i32 3
129 ret <4 x float> %vecinit6.i
133 define <4 x i32> @F(i32* %ptr) nounwind uwtable readnone ssp {
135 ; X32: ## BB#0: ## %entry
136 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
137 ; X32-NEXT: vbroadcastss (%eax), %xmm0
141 ; X64: ## BB#0: ## %entry
142 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
145 %q = load i32, i32* %ptr, align 4
146 %vecinit.i = insertelement <4 x i32> undef, i32 %q, i32 0
147 %vecinit2.i = insertelement <4 x i32> %vecinit.i, i32 %q, i32 1
148 %vecinit4.i = insertelement <4 x i32> %vecinit2.i, i32 %q, i32 2
149 %vecinit6.i = insertelement <4 x i32> %vecinit4.i, i32 %q, i32 3
150 ret <4 x i32> %vecinit6.i
153 ; FIXME: Pointer adjusted broadcasts
155 define <4 x i32> @load_splat_4i32_4i32_1111(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
156 ; X32-LABEL: load_splat_4i32_4i32_1111:
157 ; X32: ## BB#0: ## %entry
158 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
159 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,1,1]
162 ; X64-LABEL: load_splat_4i32_4i32_1111:
163 ; X64: ## BB#0: ## %entry
164 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = mem[1,1,1,1]
167 %ld = load <4 x i32>, <4 x i32>* %ptr
168 %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
172 define <8 x i32> @load_splat_8i32_4i32_33333333(<4 x i32>* %ptr) nounwind uwtable readnone ssp {
173 ; X32-LABEL: load_splat_8i32_4i32_33333333:
174 ; X32: ## BB#0: ## %entry
175 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
176 ; X32-NEXT: vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
177 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
180 ; X64-LABEL: load_splat_8i32_4i32_33333333:
181 ; X64: ## BB#0: ## %entry
182 ; X64-NEXT: vpermilps {{.*#+}} xmm0 = mem[3,3,3,3]
183 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
186 %ld = load <4 x i32>, <4 x i32>* %ptr
187 %ret = shufflevector <4 x i32> %ld, <4 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
191 define <8 x i32> @load_splat_8i32_8i32_55555555(<8 x i32>* %ptr) nounwind uwtable readnone ssp {
192 ; X32-LABEL: load_splat_8i32_8i32_55555555:
193 ; X32: ## BB#0: ## %entry
194 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
195 ; X32-NEXT: vbroadcastss 20(%eax), %ymm0
198 ; X64-LABEL: load_splat_8i32_8i32_55555555:
199 ; X64: ## BB#0: ## %entry
200 ; X64-NEXT: vbroadcastss 20(%rdi), %ymm0
203 %ld = load <8 x i32>, <8 x i32>* %ptr
204 %ret = shufflevector <8 x i32> %ld, <8 x i32> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
208 define <4 x float> @load_splat_4f32_4f32_1111(<4 x float>* %ptr) nounwind uwtable readnone ssp {
209 ; X32-LABEL: load_splat_4f32_4f32_1111:
210 ; X32: ## BB#0: ## %entry
211 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
212 ; X32-NEXT: vbroadcastss 4(%eax), %xmm0
215 ; X64-LABEL: load_splat_4f32_4f32_1111:
216 ; X64: ## BB#0: ## %entry
217 ; X64-NEXT: vbroadcastss 4(%rdi), %xmm0
220 %ld = load <4 x float>, <4 x float>* %ptr
221 %ret = shufflevector <4 x float> %ld, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
225 define <8 x float> @load_splat_8f32_4f32_33333333(<4 x float>* %ptr) nounwind uwtable readnone ssp {
226 ; X32-LABEL: load_splat_8f32_4f32_33333333:
227 ; X32: ## BB#0: ## %entry
228 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
229 ; X32-NEXT: vbroadcastss 12(%eax), %ymm0
232 ; X64-LABEL: load_splat_8f32_4f32_33333333:
233 ; X64: ## BB#0: ## %entry
234 ; X64-NEXT: vbroadcastss 12(%rdi), %ymm0
237 %ld = load <4 x float>, <4 x float>* %ptr
238 %ret = shufflevector <4 x float> %ld, <4 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
242 define <8 x float> @load_splat_8f32_8f32_55555555(<8 x float>* %ptr) nounwind uwtable readnone ssp {
243 ; X32-LABEL: load_splat_8f32_8f32_55555555:
244 ; X32: ## BB#0: ## %entry
245 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
246 ; X32-NEXT: vbroadcastss 20(%eax), %ymm0
249 ; X64-LABEL: load_splat_8f32_8f32_55555555:
250 ; X64: ## BB#0: ## %entry
251 ; X64-NEXT: vbroadcastss 20(%rdi), %ymm0
254 %ld = load <8 x float>, <8 x float>* %ptr
255 %ret = shufflevector <8 x float> %ld, <8 x float> undef, <8 x i32> <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
259 define <2 x i64> @load_splat_2i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
260 ; X32-LABEL: load_splat_2i64_2i64_1111:
261 ; X32: ## BB#0: ## %entry
262 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
263 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
266 ; X64-LABEL: load_splat_2i64_2i64_1111:
267 ; X64: ## BB#0: ## %entry
268 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
271 %ld = load <2 x i64>, <2 x i64>* %ptr
272 %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
276 define <4 x i64> @load_splat_4i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
277 ; X32-LABEL: load_splat_4i64_2i64_1111:
278 ; X32: ## BB#0: ## %entry
279 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
280 ; X32-NEXT: vmovaps (%eax), %xmm0
281 ; X32-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
282 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
285 ; X64-LABEL: load_splat_4i64_2i64_1111:
286 ; X64: ## BB#0: ## %entry
287 ; X64-NEXT: vmovaps (%rdi), %xmm0
288 ; X64-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
289 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
292 %ld = load <2 x i64>, <2 x i64>* %ptr
293 %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
297 define <4 x i64> @load_splat_4i64_4i64_2222(<4 x i64>* %ptr) nounwind uwtable readnone ssp {
298 ; X32-LABEL: load_splat_4i64_4i64_2222:
299 ; X32: ## BB#0: ## %entry
300 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
301 ; X32-NEXT: vbroadcastsd 16(%eax), %ymm0
304 ; X64-LABEL: load_splat_4i64_4i64_2222:
305 ; X64: ## BB#0: ## %entry
306 ; X64-NEXT: vbroadcastsd 16(%rdi), %ymm0
309 %ld = load <4 x i64>, <4 x i64>* %ptr
310 %ret = shufflevector <4 x i64> %ld, <4 x i64> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
314 define <2 x double> @load_splat_2f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
315 ; X32-LABEL: load_splat_2f64_2f64_1111:
316 ; X32: ## BB#0: ## %entry
317 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
318 ; X32-NEXT: vmovaps (%eax), %xmm0
319 ; X32-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
322 ; X64-LABEL: load_splat_2f64_2f64_1111:
323 ; X64: ## BB#0: ## %entry
324 ; X64-NEXT: vmovaps (%rdi), %xmm0
325 ; X64-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
328 %ld = load <2 x double>, <2 x double>* %ptr
329 %ret = shufflevector <2 x double> %ld, <2 x double> undef, <2 x i32> <i32 1, i32 1>
330 ret <2 x double> %ret
333 define <4 x double> @load_splat_4f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
334 ; X32-LABEL: load_splat_4f64_2f64_1111:
335 ; X32: ## BB#0: ## %entry
336 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
337 ; X32-NEXT: vbroadcastsd 8(%eax), %ymm0
340 ; X64-LABEL: load_splat_4f64_2f64_1111:
341 ; X64: ## BB#0: ## %entry
342 ; X64-NEXT: vbroadcastsd 8(%rdi), %ymm0
345 %ld = load <2 x double>, <2 x double>* %ptr
346 %ret = shufflevector <2 x double> %ld, <2 x double> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
347 ret <4 x double> %ret
350 define <4 x double> @load_splat_4f64_4f64_2222(<4 x double>* %ptr) nounwind uwtable readnone ssp {
351 ; X32-LABEL: load_splat_4f64_4f64_2222:
352 ; X32: ## BB#0: ## %entry
353 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
354 ; X32-NEXT: vbroadcastsd 16(%eax), %ymm0
357 ; X64-LABEL: load_splat_4f64_4f64_2222:
358 ; X64: ## BB#0: ## %entry
359 ; X64-NEXT: vbroadcastsd 16(%rdi), %ymm0
362 %ld = load <4 x double>, <4 x double>* %ptr
363 %ret = shufflevector <4 x double> %ld, <4 x double> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
364 ret <4 x double> %ret
367 ; Unsupported vbroadcasts
369 define <2 x i64> @G(i64* %ptr) nounwind uwtable readnone ssp {
371 ; X32: ## BB#0: ## %entry
372 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
373 ; X32-NEXT: movl (%eax), %ecx
374 ; X32-NEXT: movl 4(%eax), %eax
375 ; X32-NEXT: vmovd %ecx, %xmm0
376 ; X32-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0
377 ; X32-NEXT: vpinsrd $2, %ecx, %xmm0, %xmm0
378 ; X32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0
382 ; X64: ## BB#0: ## %entry
383 ; X64-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
384 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
387 %q = load i64, i64* %ptr, align 8
388 %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
389 %vecinit2.i = insertelement <2 x i64> %vecinit.i, i64 %q, i32 1
390 ret <2 x i64> %vecinit2.i
393 define <4 x i32> @H(<4 x i32> %a) {
395 ; X32: ## BB#0: ## %entry
396 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
400 ; X64: ## BB#0: ## %entry
401 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
404 %x = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
408 define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
410 ; X32: ## BB#0: ## %entry
411 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
412 ; X32-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
416 ; X64: ## BB#0: ## %entry
417 ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
420 %q = load double, double* %ptr, align 4
421 %vecinit.i = insertelement <2 x double> undef, double %q, i32 0
422 %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1
423 ret <2 x double> %vecinit2.i
426 define <4 x float> @_RR(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
428 ; X32: ## BB#0: ## %entry
429 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
430 ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
431 ; X32-NEXT: vbroadcastss (%ecx), %xmm0
432 ; X32-NEXT: movl (%eax), %eax
433 ; X32-NEXT: movl %eax, (%eax)
437 ; X64: ## BB#0: ## %entry
438 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
439 ; X64-NEXT: movl (%rsi), %eax
440 ; X64-NEXT: movl %eax, (%rax)
443 %q = load float, float* %ptr, align 4
444 %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
445 %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
446 %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
447 %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
449 %j = load i32, i32* %k, align 4
450 store i32 %j, i32* undef
451 ret <4 x float> %vecinit6.i
454 define <4 x float> @_RR2(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
456 ; X32: ## BB#0: ## %entry
457 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
458 ; X32-NEXT: vbroadcastss (%eax), %xmm0
462 ; X64: ## BB#0: ## %entry
463 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
466 %q = load float, float* %ptr, align 4
467 %v = insertelement <4 x float> undef, float %q, i32 0
468 %t = shufflevector <4 x float> %v, <4 x float> undef, <4 x i32> zeroinitializer
472 ; These tests check that a vbroadcast instruction is used when we have a splat
473 ; formed from a concat_vectors (via the shufflevector) of two BUILD_VECTORs
474 ; (via the insertelements).
476 define <8 x float> @splat_concat1(float* %p) {
477 ; X32-LABEL: splat_concat1:
479 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
480 ; X32-NEXT: vbroadcastss (%eax), %ymm0
483 ; X64-LABEL: splat_concat1:
485 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
487 %1 = load float, float* %p, align 4
488 %2 = insertelement <4 x float> undef, float %1, i32 0
489 %3 = insertelement <4 x float> %2, float %1, i32 1
490 %4 = insertelement <4 x float> %3, float %1, i32 2
491 %5 = insertelement <4 x float> %4, float %1, i32 3
492 %6 = shufflevector <4 x float> %5, <4 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
496 define <8 x float> @splat_concat2(float* %p) {
497 ; X32-LABEL: splat_concat2:
499 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
500 ; X32-NEXT: vbroadcastss (%eax), %ymm0
503 ; X64-LABEL: splat_concat2:
505 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
507 %1 = load float, float* %p, align 4
508 %2 = insertelement <4 x float> undef, float %1, i32 0
509 %3 = insertelement <4 x float> %2, float %1, i32 1
510 %4 = insertelement <4 x float> %3, float %1, i32 2
511 %5 = insertelement <4 x float> %4, float %1, i32 3
512 %6 = insertelement <4 x float> undef, float %1, i32 0
513 %7 = insertelement <4 x float> %6, float %1, i32 1
514 %8 = insertelement <4 x float> %7, float %1, i32 2
515 %9 = insertelement <4 x float> %8, float %1, i32 3
516 %10 = shufflevector <4 x float> %5, <4 x float> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
520 define <4 x double> @splat_concat3(double* %p) {
521 ; X32-LABEL: splat_concat3:
523 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
524 ; X32-NEXT: vbroadcastsd (%eax), %ymm0
527 ; X64-LABEL: splat_concat3:
529 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
531 %1 = load double, double* %p, align 8
532 %2 = insertelement <2 x double> undef, double %1, i32 0
533 %3 = insertelement <2 x double> %2, double %1, i32 1
534 %4 = shufflevector <2 x double> %3, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
538 define <4 x double> @splat_concat4(double* %p) {
539 ; X32-LABEL: splat_concat4:
541 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
542 ; X32-NEXT: vbroadcastsd (%eax), %ymm0
545 ; X64-LABEL: splat_concat4:
547 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
549 %1 = load double, double* %p, align 8
550 %2 = insertelement <2 x double> undef, double %1, i32 0
551 %3 = insertelement <2 x double> %2, double %1, i32 1
552 %4 = insertelement <2 x double> undef, double %1, i32 0
553 %5 = insertelement <2 x double> %2, double %1, i32 1
554 %6 = shufflevector <2 x double> %3, <2 x double> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>