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: vmovaps (%eax), %ymm0
196 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm0
197 ; X32-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
198 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
201 ; X64-LABEL: load_splat_8i32_8i32_55555555:
202 ; X64: ## BB#0: ## %entry
203 ; X64-NEXT: vmovaps (%rdi), %ymm0
204 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
205 ; X64-NEXT: vpermilps {{.*#+}} xmm0 = xmm0[1,1,1,1]
206 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
209 %ld = load <8 x i32>, <8 x i32>* %ptr
210 %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>
214 define <4 x float> @load_splat_4f32_4f32_1111(<4 x float>* %ptr) nounwind uwtable readnone ssp {
215 ; X32-LABEL: load_splat_4f32_4f32_1111:
216 ; X32: ## BB#0: ## %entry
217 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
218 ; X32-NEXT: vbroadcastss 4(%eax), %xmm0
221 ; X64-LABEL: load_splat_4f32_4f32_1111:
222 ; X64: ## BB#0: ## %entry
223 ; X64-NEXT: vbroadcastss 4(%rdi), %xmm0
226 %ld = load <4 x float>, <4 x float>* %ptr
227 %ret = shufflevector <4 x float> %ld, <4 x float> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
231 define <8 x float> @load_splat_8f32_4f32_33333333(<4 x float>* %ptr) nounwind uwtable readnone ssp {
232 ; X32-LABEL: load_splat_8f32_4f32_33333333:
233 ; X32: ## BB#0: ## %entry
234 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
235 ; X32-NEXT: vbroadcastss 12(%eax), %ymm0
238 ; X64-LABEL: load_splat_8f32_4f32_33333333:
239 ; X64: ## BB#0: ## %entry
240 ; X64-NEXT: vbroadcastss 12(%rdi), %ymm0
243 %ld = load <4 x float>, <4 x float>* %ptr
244 %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>
248 define <8 x float> @load_splat_8f32_8f32_55555555(<8 x float>* %ptr) nounwind uwtable readnone ssp {
249 ; X32-LABEL: load_splat_8f32_8f32_55555555:
250 ; X32: ## BB#0: ## %entry
251 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
252 ; X32-NEXT: vbroadcastss 20(%eax), %ymm0
255 ; X64-LABEL: load_splat_8f32_8f32_55555555:
256 ; X64: ## BB#0: ## %entry
257 ; X64-NEXT: vbroadcastss 20(%rdi), %ymm0
260 %ld = load <8 x float>, <8 x float>* %ptr
261 %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>
265 define <2 x i64> @load_splat_2i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
266 ; X32-LABEL: load_splat_2i64_2i64_1111:
267 ; X32: ## BB#0: ## %entry
268 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
269 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
272 ; X64-LABEL: load_splat_2i64_2i64_1111:
273 ; X64: ## BB#0: ## %entry
274 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = mem[2,3,2,3]
277 %ld = load <2 x i64>, <2 x i64>* %ptr
278 %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
282 define <4 x i64> @load_splat_4i64_2i64_1111(<2 x i64>* %ptr) nounwind uwtable readnone ssp {
283 ; X32-LABEL: load_splat_4i64_2i64_1111:
284 ; X32: ## BB#0: ## %entry
285 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
286 ; X32-NEXT: vmovaps (%eax), %xmm0
287 ; X32-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
288 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
291 ; X64-LABEL: load_splat_4i64_2i64_1111:
292 ; X64: ## BB#0: ## %entry
293 ; X64-NEXT: vmovaps (%rdi), %xmm0
294 ; X64-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
295 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
298 %ld = load <2 x i64>, <2 x i64>* %ptr
299 %ret = shufflevector <2 x i64> %ld, <2 x i64> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
303 define <4 x i64> @load_splat_4i64_4i64_2222(<4 x i64>* %ptr) nounwind uwtable readnone ssp {
304 ; X32-LABEL: load_splat_4i64_4i64_2222:
305 ; X32: ## BB#0: ## %entry
306 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
307 ; X32-NEXT: vmovapd (%eax), %ymm0
308 ; X32-NEXT: vextractf128 $1, %ymm0, %xmm0
309 ; X32-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
310 ; X32-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
313 ; X64-LABEL: load_splat_4i64_4i64_2222:
314 ; X64: ## BB#0: ## %entry
315 ; X64-NEXT: vmovapd (%rdi), %ymm0
316 ; X64-NEXT: vextractf128 $1, %ymm0, %xmm0
317 ; X64-NEXT: vmovddup {{.*#+}} xmm0 = xmm0[0,0]
318 ; X64-NEXT: vinsertf128 $1, %xmm0, %ymm0, %ymm0
321 %ld = load <4 x i64>, <4 x i64>* %ptr
322 %ret = shufflevector <4 x i64> %ld, <4 x i64> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
326 define <2 x double> @load_splat_2f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
327 ; X32-LABEL: load_splat_2f64_2f64_1111:
328 ; X32: ## BB#0: ## %entry
329 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
330 ; X32-NEXT: vmovaps (%eax), %xmm0
331 ; X32-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
334 ; X64-LABEL: load_splat_2f64_2f64_1111:
335 ; X64: ## BB#0: ## %entry
336 ; X64-NEXT: vmovaps (%rdi), %xmm0
337 ; X64-NEXT: vmovhlps {{.*#+}} xmm0 = xmm0[1,1]
340 %ld = load <2 x double>, <2 x double>* %ptr
341 %ret = shufflevector <2 x double> %ld, <2 x double> undef, <2 x i32> <i32 1, i32 1>
342 ret <2 x double> %ret
345 define <4 x double> @load_splat_4f64_2f64_1111(<2 x double>* %ptr) nounwind uwtable readnone ssp {
346 ; X32-LABEL: load_splat_4f64_2f64_1111:
347 ; X32: ## BB#0: ## %entry
348 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
349 ; X32-NEXT: vbroadcastsd 8(%eax), %ymm0
352 ; X64-LABEL: load_splat_4f64_2f64_1111:
353 ; X64: ## BB#0: ## %entry
354 ; X64-NEXT: vbroadcastsd 8(%rdi), %ymm0
357 %ld = load <2 x double>, <2 x double>* %ptr
358 %ret = shufflevector <2 x double> %ld, <2 x double> undef, <4 x i32> <i32 1, i32 1, i32 1, i32 1>
359 ret <4 x double> %ret
362 define <4 x double> @load_splat_4f64_4f64_2222(<4 x double>* %ptr) nounwind uwtable readnone ssp {
363 ; X32-LABEL: load_splat_4f64_4f64_2222:
364 ; X32: ## BB#0: ## %entry
365 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
366 ; X32-NEXT: vbroadcastsd 16(%eax), %ymm0
369 ; X64-LABEL: load_splat_4f64_4f64_2222:
370 ; X64: ## BB#0: ## %entry
371 ; X64-NEXT: vbroadcastsd 16(%rdi), %ymm0
374 %ld = load <4 x double>, <4 x double>* %ptr
375 %ret = shufflevector <4 x double> %ld, <4 x double> undef, <4 x i32> <i32 2, i32 2, i32 2, i32 2>
376 ret <4 x double> %ret
379 ; Unsupported vbroadcasts
381 define <2 x i64> @G(i64* %ptr) nounwind uwtable readnone ssp {
383 ; X32: ## BB#0: ## %entry
384 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
385 ; X32-NEXT: movl (%eax), %ecx
386 ; X32-NEXT: movl 4(%eax), %eax
387 ; X32-NEXT: vmovd %ecx, %xmm0
388 ; X32-NEXT: vpinsrd $1, %eax, %xmm0, %xmm0
389 ; X32-NEXT: vpinsrd $2, %ecx, %xmm0, %xmm0
390 ; X32-NEXT: vpinsrd $3, %eax, %xmm0, %xmm0
394 ; X64: ## BB#0: ## %entry
395 ; X64-NEXT: vmovq {{.*#+}} xmm0 = mem[0],zero
396 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
399 %q = load i64, i64* %ptr, align 8
400 %vecinit.i = insertelement <2 x i64> undef, i64 %q, i32 0
401 %vecinit2.i = insertelement <2 x i64> %vecinit.i, i64 %q, i32 1
402 ret <2 x i64> %vecinit2.i
405 define <4 x i32> @H(<4 x i32> %a) {
407 ; X32: ## BB#0: ## %entry
408 ; X32-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
412 ; X64: ## BB#0: ## %entry
413 ; X64-NEXT: vpshufd {{.*#+}} xmm0 = xmm0[1,1,2,3]
416 %x = shufflevector <4 x i32> %a, <4 x i32> undef, <4 x i32> <i32 1, i32 undef, i32 undef, i32 undef>
420 define <2 x double> @I(double* %ptr) nounwind uwtable readnone ssp {
422 ; X32: ## BB#0: ## %entry
423 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
424 ; X32-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
428 ; X64: ## BB#0: ## %entry
429 ; X64-NEXT: vmovddup {{.*#+}} xmm0 = mem[0,0]
432 %q = load double, double* %ptr, align 4
433 %vecinit.i = insertelement <2 x double> undef, double %q, i32 0
434 %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1
435 ret <2 x double> %vecinit2.i
438 define <4 x float> @_RR(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
440 ; X32: ## BB#0: ## %entry
441 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
442 ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx
443 ; X32-NEXT: vbroadcastss (%ecx), %xmm0
444 ; X32-NEXT: movl (%eax), %eax
445 ; X32-NEXT: movl %eax, (%eax)
449 ; X64: ## BB#0: ## %entry
450 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
451 ; X64-NEXT: movl (%rsi), %eax
452 ; X64-NEXT: movl %eax, (%rax)
455 %q = load float, float* %ptr, align 4
456 %vecinit.i = insertelement <4 x float> undef, float %q, i32 0
457 %vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
458 %vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
459 %vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
461 %j = load i32, i32* %k, align 4
462 store i32 %j, i32* undef
463 ret <4 x float> %vecinit6.i
466 define <4 x float> @_RR2(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
468 ; X32: ## BB#0: ## %entry
469 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
470 ; X32-NEXT: vbroadcastss (%eax), %xmm0
474 ; X64: ## BB#0: ## %entry
475 ; X64-NEXT: vbroadcastss (%rdi), %xmm0
478 %q = load float, float* %ptr, align 4
479 %v = insertelement <4 x float> undef, float %q, i32 0
480 %t = shufflevector <4 x float> %v, <4 x float> undef, <4 x i32> zeroinitializer
484 ; These tests check that a vbroadcast instruction is used when we have a splat
485 ; formed from a concat_vectors (via the shufflevector) of two BUILD_VECTORs
486 ; (via the insertelements).
488 define <8 x float> @splat_concat1(float* %p) {
489 ; X32-LABEL: splat_concat1:
491 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
492 ; X32-NEXT: vbroadcastss (%eax), %ymm0
495 ; X64-LABEL: splat_concat1:
497 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
499 %1 = load float, float* %p, align 4
500 %2 = insertelement <4 x float> undef, float %1, i32 0
501 %3 = insertelement <4 x float> %2, float %1, i32 1
502 %4 = insertelement <4 x float> %3, float %1, i32 2
503 %5 = insertelement <4 x float> %4, float %1, i32 3
504 %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>
508 define <8 x float> @splat_concat2(float* %p) {
509 ; X32-LABEL: splat_concat2:
511 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
512 ; X32-NEXT: vbroadcastss (%eax), %ymm0
515 ; X64-LABEL: splat_concat2:
517 ; X64-NEXT: vbroadcastss (%rdi), %ymm0
519 %1 = load float, float* %p, align 4
520 %2 = insertelement <4 x float> undef, float %1, i32 0
521 %3 = insertelement <4 x float> %2, float %1, i32 1
522 %4 = insertelement <4 x float> %3, float %1, i32 2
523 %5 = insertelement <4 x float> %4, float %1, i32 3
524 %6 = insertelement <4 x float> undef, float %1, i32 0
525 %7 = insertelement <4 x float> %6, float %1, i32 1
526 %8 = insertelement <4 x float> %7, float %1, i32 2
527 %9 = insertelement <4 x float> %8, float %1, i32 3
528 %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>
532 define <4 x double> @splat_concat3(double* %p) {
533 ; X32-LABEL: splat_concat3:
535 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
536 ; X32-NEXT: vbroadcastsd (%eax), %ymm0
539 ; X64-LABEL: splat_concat3:
541 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
543 %1 = load double, double* %p, align 8
544 %2 = insertelement <2 x double> undef, double %1, i32 0
545 %3 = insertelement <2 x double> %2, double %1, i32 1
546 %4 = shufflevector <2 x double> %3, <2 x double> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
550 define <4 x double> @splat_concat4(double* %p) {
551 ; X32-LABEL: splat_concat4:
553 ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax
554 ; X32-NEXT: vbroadcastsd (%eax), %ymm0
557 ; X64-LABEL: splat_concat4:
559 ; X64-NEXT: vbroadcastsd (%rdi), %ymm0
561 %1 = load double, double* %p, align 8
562 %2 = insertelement <2 x double> undef, double %1, i32 0
563 %3 = insertelement <2 x double> %2, double %1, i32 1
564 %4 = insertelement <2 x double> undef, double %1, i32 0
565 %5 = insertelement <2 x double> %2, double %1, i32 1
566 %6 = shufflevector <2 x double> %3, <2 x double> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>