Skip to content

Commit bdd0159

Browse files
rotaterightmemfrob
rotateright
authored and
memfrob
committed
[VectorCombine] add tests for load+insert; NFC
1 parent f593edb commit bdd0159

File tree

1 file changed

+170
-0
lines changed
  • llvm/test/Transforms/VectorCombine/X86

1 file changed

+170
-0
lines changed

‎llvm/test/Transforms/VectorCombine/X86/load.ll

Lines changed: 170 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -171,3 +171,173 @@ define double @larger_fp_scalar_256bit_vec(<8 x float>* align 32 dereferenceable
171171
%r = load double, double* %bc, align 32
172172
ret double %r
173173
}
174+
175+
define <4 x float> @load_f32_insert_v4f32(float* align 16 dereferenceable(16) %p) {
176+
; CHECK-LABEL: @load_f32_insert_v4f32(
177+
; CHECK-NEXT: [[S:%.*]] = load float, float* [[P:%.*]], align 4
178+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i32 0
179+
; CHECK-NEXT: ret <4 x float> [[R]]
180+
;
181+
%s = load float, float* %p, align 4
182+
%r = insertelement <4 x float> undef, float %s, i32 0
183+
ret <4 x float> %r
184+
}
185+
186+
define <4 x float> @casted_load_f32_insert_v4f32(<4 x float>* align 4 dereferenceable(16) %p) {
187+
; CHECK-LABEL: @casted_load_f32_insert_v4f32(
188+
; CHECK-NEXT: [[B:%.*]] = bitcast <4 x float>* [[P:%.*]] to float*
189+
; CHECK-NEXT: [[S:%.*]] = load float, float* [[B]], align 4
190+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i32 0
191+
; CHECK-NEXT: ret <4 x float> [[R]]
192+
;
193+
%b = bitcast <4 x float>* %p to float*
194+
%s = load float, float* %b, align 4
195+
%r = insertelement <4 x float> undef, float %s, i32 0
196+
ret <4 x float> %r
197+
}
198+
199+
define <4 x i32> @load_i32_insert_v4i32(i32* align 16 dereferenceable(16) %p) {
200+
; CHECK-LABEL: @load_i32_insert_v4i32(
201+
; CHECK-NEXT: [[S:%.*]] = load i32, i32* [[P:%.*]], align 4
202+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x i32> undef, i32 [[S]], i32 0
203+
; CHECK-NEXT: ret <4 x i32> [[R]]
204+
;
205+
%s = load i32, i32* %p, align 4
206+
%r = insertelement <4 x i32> undef, i32 %s, i32 0
207+
ret <4 x i32> %r
208+
}
209+
210+
define <4 x i32> @casted_load_i32_insert_v4i32(<16 x i8>* align 4 dereferenceable(16) %p) {
211+
; CHECK-LABEL: @casted_load_i32_insert_v4i32(
212+
; CHECK-NEXT: [[B:%.*]] = bitcast <16 x i8>* [[P:%.*]] to i32*
213+
; CHECK-NEXT: [[S:%.*]] = load i32, i32* [[B]], align 4
214+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x i32> undef, i32 [[S]], i32 0
215+
; CHECK-NEXT: ret <4 x i32> [[R]]
216+
;
217+
%b = bitcast <16 x i8>* %p to i32*
218+
%s = load i32, i32* %b, align 4
219+
%r = insertelement <4 x i32> undef, i32 %s, i32 0
220+
ret <4 x i32> %r
221+
}
222+
223+
define <4 x float> @gep00_load_f32_insert_v4f32(<4 x float>* align 16 dereferenceable(16) %p) {
224+
; CHECK-LABEL: @gep00_load_f32_insert_v4f32(
225+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds <4 x float>, <4 x float>* [[P:%.*]], i64 0, i64 0
226+
; CHECK-NEXT: [[S:%.*]] = load float, float* [[GEP]], align 16
227+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i64 0
228+
; CHECK-NEXT: ret <4 x float> [[R]]
229+
;
230+
%gep = getelementptr inbounds <4 x float>, <4 x float>* %p, i64 0, i64 0
231+
%s = load float, float* %gep, align 16
232+
%r = insertelement <4 x float> undef, float %s, i64 0
233+
ret <4 x float> %r
234+
}
235+
236+
define <8 x i16> @gep01_load_i16_insert_v8i16(<8 x i16>* align 16 dereferenceable(18) %p) {
237+
; CHECK-LABEL: @gep01_load_i16_insert_v8i16(
238+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[P:%.*]], i64 0, i64 1
239+
; CHECK-NEXT: [[S:%.*]] = load i16, i16* [[GEP]], align 2
240+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i16> undef, i16 [[S]], i64 0
241+
; CHECK-NEXT: ret <8 x i16> [[R]]
242+
;
243+
%gep = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i64 0, i64 1
244+
%s = load i16, i16* %gep, align 2
245+
%r = insertelement <8 x i16> undef, i16 %s, i64 0
246+
ret <8 x i16> %r
247+
}
248+
249+
define <8 x i16> @gep01_load_i16_insert_v8i16_deref(<8 x i16>* align 16 dereferenceable(17) %p) {
250+
; CHECK-LABEL: @gep01_load_i16_insert_v8i16_deref(
251+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[P:%.*]], i64 0, i64 1
252+
; CHECK-NEXT: [[S:%.*]] = load i16, i16* [[GEP]], align 2
253+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i16> undef, i16 [[S]], i64 0
254+
; CHECK-NEXT: ret <8 x i16> [[R]]
255+
;
256+
%gep = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i64 0, i64 1
257+
%s = load i16, i16* %gep, align 2
258+
%r = insertelement <8 x i16> undef, i16 %s, i64 0
259+
ret <8 x i16> %r
260+
}
261+
262+
define <8 x i16> @gep10_load_i16_insert_v8i16(<8 x i16>* align 16 dereferenceable(32) %p) {
263+
; CHECK-LABEL: @gep10_load_i16_insert_v8i16(
264+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[P:%.*]], i64 1, i64 0
265+
; CHECK-NEXT: [[S:%.*]] = load i16, i16* [[GEP]], align 16
266+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i16> undef, i16 [[S]], i64 0
267+
; CHECK-NEXT: ret <8 x i16> [[R]]
268+
;
269+
%gep = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i64 1, i64 0
270+
%s = load i16, i16* %gep, align 16
271+
%r = insertelement <8 x i16> undef, i16 %s, i64 0
272+
ret <8 x i16> %r
273+
}
274+
275+
define <8 x i16> @gep10_load_i16_insert_v8i16_deref(<8 x i16>* align 16 dereferenceable(31) %p) {
276+
; CHECK-LABEL: @gep10_load_i16_insert_v8i16_deref(
277+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds <8 x i16>, <8 x i16>* [[P:%.*]], i64 1, i64 0
278+
; CHECK-NEXT: [[S:%.*]] = load i16, i16* [[GEP]], align 16
279+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i16> undef, i16 [[S]], i64 0
280+
; CHECK-NEXT: ret <8 x i16> [[R]]
281+
;
282+
%gep = getelementptr inbounds <8 x i16>, <8 x i16>* %p, i64 1, i64 0
283+
%s = load i16, i16* %gep, align 16
284+
%r = insertelement <8 x i16> undef, i16 %s, i64 0
285+
ret <8 x i16> %r
286+
}
287+
288+
define <4 x float> @load_f32_insert_v4f32_volatile(float* align 16 dereferenceable(16) %p) {
289+
; CHECK-LABEL: @load_f32_insert_v4f32_volatile(
290+
; CHECK-NEXT: [[S:%.*]] = load volatile float, float* [[P:%.*]], align 4
291+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i32 0
292+
; CHECK-NEXT: ret <4 x float> [[R]]
293+
;
294+
%s = load volatile float, float* %p, align 4
295+
%r = insertelement <4 x float> undef, float %s, i32 0
296+
ret <4 x float> %r
297+
}
298+
299+
define <4 x float> @load_f32_insert_v4f32_align(float* align 1 dereferenceable(16) %p) {
300+
; CHECK-LABEL: @load_f32_insert_v4f32_align(
301+
; CHECK-NEXT: [[S:%.*]] = load float, float* [[P:%.*]], align 4
302+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i32 0
303+
; CHECK-NEXT: ret <4 x float> [[R]]
304+
;
305+
%s = load float, float* %p, align 4
306+
%r = insertelement <4 x float> undef, float %s, i32 0
307+
ret <4 x float> %r
308+
}
309+
310+
define <4 x float> @load_f32_insert_v4f32_deref(float* align 4 dereferenceable(15) %p) {
311+
; CHECK-LABEL: @load_f32_insert_v4f32_deref(
312+
; CHECK-NEXT: [[S:%.*]] = load float, float* [[P:%.*]], align 4
313+
; CHECK-NEXT: [[R:%.*]] = insertelement <4 x float> undef, float [[S]], i32 0
314+
; CHECK-NEXT: ret <4 x float> [[R]]
315+
;
316+
%s = load float, float* %p, align 4
317+
%r = insertelement <4 x float> undef, float %s, i32 0
318+
ret <4 x float> %r
319+
}
320+
321+
define <8 x i32> @load_i32_insert_v8i32(i32* align 16 dereferenceable(16) %p) {
322+
; CHECK-LABEL: @load_i32_insert_v8i32(
323+
; CHECK-NEXT: [[S:%.*]] = load i32, i32* [[P:%.*]], align 4
324+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i32> undef, i32 [[S]], i32 0
325+
; CHECK-NEXT: ret <8 x i32> [[R]]
326+
;
327+
%s = load i32, i32* %p, align 4
328+
%r = insertelement <8 x i32> undef, i32 %s, i32 0
329+
ret <8 x i32> %r
330+
}
331+
332+
define <8 x i32> @casted_load_i32_insert_v8i32(<4 x i32>* align 4 dereferenceable(16) %p) {
333+
; CHECK-LABEL: @casted_load_i32_insert_v8i32(
334+
; CHECK-NEXT: [[B:%.*]] = bitcast <4 x i32>* [[P:%.*]] to i32*
335+
; CHECK-NEXT: [[S:%.*]] = load i32, i32* [[B]], align 4
336+
; CHECK-NEXT: [[R:%.*]] = insertelement <8 x i32> undef, i32 [[S]], i32 0
337+
; CHECK-NEXT: ret <8 x i32> [[R]]
338+
;
339+
%b = bitcast <4 x i32>* %p to i32*
340+
%s = load i32, i32* %b, align 4
341+
%r = insertelement <8 x i32> undef, i32 %s, i32 0
342+
ret <8 x i32> %r
343+
}

0 commit comments

Comments
 (0)