Skip to content

Commit 7d6e8f2

Browse files
preamespreames
committed
[slp] Delete dead scalar instructions feeding vectorized instructions
If we vectorize a e.g. store, we leave around a bunch of getelementptrs for the individual scalar stores which we removed. We can go ahead and delete them as well. This is purely for test output quality and readability. It should have no effect in any sane pipeline. Differential Revision: https://reviews.llvm.org/D122493
1 parent 0217d11 commit 7d6e8f2

File tree

110 files changed

+615
-2952
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

110 files changed

+615
-2952
lines changed

‎llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Lines changed: 14 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -87,6 +87,7 @@
8787
#include "llvm/Support/MathExtras.h"
8888
#include "llvm/Support/raw_ostream.h"
8989
#include "llvm/Transforms/Utils/InjectTLIMappings.h"
90+
#include "llvm/Transforms/Utils/Local.h"
9091
#include "llvm/Transforms/Utils/LoopUtils.h"
9192
#include "llvm/Transforms/Vectorize.h"
9293
#include <algorithm>
@@ -3204,13 +3205,25 @@ template <> struct DOTGraphTraits<BoUpSLP *> : public DefaultDOTGraphTraits {
32043205
} // end namespace llvm
32053206

32063207
BoUpSLP::~BoUpSLP() {
3207-
for (auto *I : DeletedInstructions)
3208+
SmallVector<WeakTrackingVH> DeadInsts;
3209+
for (auto *I : DeletedInstructions) {
3210+
for (Use &U : I->operands()) {
3211+
auto *Op = dyn_cast<Instruction>(U.get());
3212+
if (Op && !DeletedInstructions.count(Op) && Op->hasOneUser() &&
3213+
wouldInstructionBeTriviallyDead(Op, TLI))
3214+
DeadInsts.emplace_back(Op);
3215+
}
32083216
I->dropAllReferences();
3217+
}
32093218
for (auto *I : DeletedInstructions) {
32103219
assert(I->use_empty() &&
32113220
"trying to erase instruction with users.");
32123221
I->eraseFromParent();
32133222
}
3223+
3224+
// Cleanup any dead scalar code feeding the vectorized instructions
3225+
RecursivelyDeleteTriviallyDeadInstructions(DeadInsts, TLI);
3226+
32143227
#ifdef EXPENSIVE_CHECKS
32153228
// If we could guarantee that this call is not extremely slow, we could
32163229
// remove the ifdef limitation (see PR47712).

‎llvm/test/Transforms/SLPVectorizer/AArch64/64-bit-vector.ll

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -8,9 +8,7 @@
88
define void @f(float* %r, float* %w) {
99
; CHECK-LABEL: @f(
1010
; CHECK-NEXT: [[R0:%.*]] = getelementptr inbounds float, float* [[R:%.*]], i64 0
11-
; CHECK-NEXT: [[R1:%.*]] = getelementptr inbounds float, float* [[R]], i64 1
1211
; CHECK-NEXT: [[W0:%.*]] = getelementptr inbounds float, float* [[W:%.*]], i64 0
13-
; CHECK-NEXT: [[W1:%.*]] = getelementptr inbounds float, float* [[W]], i64 1
1412
; CHECK-NEXT: [[TMP1:%.*]] = bitcast float* [[R0]] to <2 x float>*
1513
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x float>, <2 x float>* [[TMP1]], align 4
1614
; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x float> [[TMP2]], [[TMP2]]

‎llvm/test/Transforms/SLPVectorizer/AArch64/PR38339.ll

Lines changed: 0 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -5,9 +5,6 @@ define void @f1(<2 x i16> %x, i16* %a) {
55
; CHECK-LABEL: @f1(
66
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i16> [[X:%.*]], <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 0>
77
; CHECK-NEXT: [[PTR0:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 0
8-
; CHECK-NEXT: [[PTR1:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 1
9-
; CHECK-NEXT: [[PTR2:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 2
10-
; CHECK-NEXT: [[PTR3:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 3
118
; CHECK-NEXT: [[TMP1:%.*]] = extractelement <2 x i16> [[X]], i32 0
129
; CHECK-NEXT: store i16 [[TMP1]], i16* [[A:%.*]], align 2
1310
; CHECK-NEXT: [[TMP2:%.*]] = bitcast i16* [[PTR0]] to <4 x i16>*
@@ -37,9 +34,6 @@ define void @f2(<2 x i16> %x, i16* %a) {
3734
; CHECK-NEXT: [[AA:%.*]] = phi i16* [ [[A:%.*]], [[ENTRY]] ], [ undef, [[CONT]] ]
3835
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i16> [[XX]], <2 x i16> poison, <4 x i32> <i32 0, i32 1, i32 1, i32 0>
3936
; CHECK-NEXT: [[PTR0:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 0
40-
; CHECK-NEXT: [[PTR1:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 1
41-
; CHECK-NEXT: [[PTR2:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 2
42-
; CHECK-NEXT: [[PTR3:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 3
4337
; CHECK-NEXT: [[TMP0:%.*]] = extractelement <2 x i16> [[XX]], i32 0
4438
; CHECK-NEXT: store i16 [[TMP0]], i16* [[A]], align 2
4539
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i16* [[PTR0]] to <4 x i16>*
@@ -84,9 +78,6 @@ define void @f3(<2 x i16> %x, i16* %a) {
8478
; CHECK-NEXT: [[AA:%.*]] = phi i16* [ [[A:%.*]], [[ENTRY]] ], [ undef, [[CONT]] ]
8579
; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <2 x i16> [[XX]], <2 x i16> poison, <4 x i32> <i32 1, i32 0, i32 0, i32 1>
8680
; CHECK-NEXT: [[PTR0:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 0
87-
; CHECK-NEXT: [[PTR1:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 1
88-
; CHECK-NEXT: [[PTR2:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 2
89-
; CHECK-NEXT: [[PTR3:%.*]] = getelementptr inbounds [4 x i16], [4 x i16]* undef, i16 0, i16 3
9081
; CHECK-NEXT: [[TMP0:%.*]] = extractelement <2 x i16> [[XX]], i32 1
9182
; CHECK-NEXT: store i16 [[TMP0]], i16* [[A]], align 2
9283
; CHECK-NEXT: [[TMP1:%.*]] = bitcast i16* [[PTR0]] to <4 x i16>*

‎llvm/test/Transforms/SLPVectorizer/AArch64/commute.ll

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -14,7 +14,6 @@ define void @test1(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
1414
; CHECK: for.body3.lr.ph:
1515
; CHECK-NEXT: [[TMP2:%.*]] = sitofp <2 x i32> [[TMP1]] to <2 x float>
1616
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [[STRUCTA:%.*]], %structA* [[J:%.*]], i64 0, i32 0, i64 0
17-
; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCTA]], %structA* [[J]], i64 0, i32 0, i64 1
1817
; CHECK-NEXT: [[TMP3:%.*]] = bitcast float* [[ARRAYIDX4]] to <2 x float>*
1918
; CHECK-NEXT: [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
2019
; CHECK-NEXT: [[TMP5:%.*]] = fsub fast <2 x float> [[TMP2]], [[TMP4]]
@@ -59,7 +58,6 @@ define void @test2(%structA* nocapture readonly %J, i32 %xmin, i32 %ymin) {
5958
; CHECK: for.body3.lr.ph:
6059
; CHECK-NEXT: [[TMP2:%.*]] = sitofp <2 x i32> [[TMP1]] to <2 x float>
6160
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [[STRUCTA:%.*]], %structA* [[J:%.*]], i64 0, i32 0, i64 0
62-
; CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds [[STRUCTA]], %structA* [[J]], i64 0, i32 0, i64 1
6361
; CHECK-NEXT: [[TMP3:%.*]] = bitcast float* [[ARRAYIDX4]] to <2 x float>*
6462
; CHECK-NEXT: [[TMP4:%.*]] = load <2 x float>, <2 x float>* [[TMP3]], align 4
6563
; CHECK-NEXT: [[TMP5:%.*]] = fsub fast <2 x float> [[TMP2]], [[TMP4]]

‎llvm/test/Transforms/SLPVectorizer/AArch64/horizontal.ll

Lines changed: 0 additions & 26 deletions
Original file line numberDiff line numberDiff line change
@@ -32,12 +32,6 @@ define i32 @test_select(i32* noalias nocapture readonly %blk1, i32* noalias noca
3232
; CHECK-NEXT: [[J_025:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
3333
; CHECK-NEXT: [[P2_024:%.*]] = phi i32* [ [[BLK2:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR29:%.*]], [[FOR_BODY]] ]
3434
; CHECK-NEXT: [[P1_023:%.*]] = phi i32* [ [[BLK1:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR:%.*]], [[FOR_BODY]] ]
35-
; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, i32* [[P1_023]], i64 1
36-
; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, i32* [[P2_024]], i64 1
37-
; CHECK-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds i32, i32* [[P1_023]], i64 2
38-
; CHECK-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds i32, i32* [[P2_024]], i64 2
39-
; CHECK-NEXT: [[ARRAYIDX20:%.*]] = getelementptr inbounds i32, i32* [[P1_023]], i64 3
40-
; CHECK-NEXT: [[ARRAYIDX21:%.*]] = getelementptr inbounds i32, i32* [[P2_024]], i64 3
4135
; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[P1_023]] to <4 x i32>*
4236
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
4337
; CHECK-NEXT: [[TMP2:%.*]] = bitcast i32* [[P2_024]] to <4 x i32>*
@@ -158,12 +152,6 @@ define i32 @reduction_with_br(i32* noalias nocapture readonly %blk1, i32* noalia
158152
; CHECK-NEXT: [[J_019:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[IF_END]] ]
159153
; CHECK-NEXT: [[P2_018:%.*]] = phi i32* [ [[BLK2:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR16:%.*]], [[IF_END]] ]
160154
; CHECK-NEXT: [[P1_017:%.*]] = phi i32* [ [[BLK1:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR:%.*]], [[IF_END]] ]
161-
; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, i32* [[P1_017]], i64 1
162-
; CHECK-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, i32* [[P2_018]], i64 1
163-
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i32, i32* [[P1_017]], i64 2
164-
; CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds i32, i32* [[P2_018]], i64 2
165-
; CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds i32, i32* [[P1_017]], i64 3
166-
; CHECK-NEXT: [[ARRAYIDX11:%.*]] = getelementptr inbounds i32, i32* [[P2_018]], i64 3
167155
; CHECK-NEXT: [[TMP0:%.*]] = bitcast i32* [[P1_017]] to <4 x i32>*
168156
; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* [[TMP0]], align 4
169157
; CHECK-NEXT: [[TMP2:%.*]] = bitcast i32* [[P2_018]] to <4 x i32>*
@@ -261,20 +249,6 @@ define i32 @test_unrolled_select(i8* noalias nocapture readonly %blk1, i8* noali
261249
; CHECK-NEXT: [[J_046:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH]] ], [ [[INC:%.*]], [[IF_END_86]] ]
262250
; CHECK-NEXT: [[P2_045:%.*]] = phi i8* [ [[BLK2:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR88:%.*]], [[IF_END_86]] ]
263251
; CHECK-NEXT: [[P1_044:%.*]] = phi i8* [ [[BLK1:%.*]], [[FOR_BODY_LR_PH]] ], [ [[ADD_PTR:%.*]], [[IF_END_86]] ]
264-
; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 1
265-
; CHECK-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 1
266-
; CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 2
267-
; CHECK-NEXT: [[ARRAYIDX19:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 2
268-
; CHECK-NEXT: [[ARRAYIDX28:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 3
269-
; CHECK-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 3
270-
; CHECK-NEXT: [[ARRAYIDX39:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 4
271-
; CHECK-NEXT: [[ARRAYIDX41:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 4
272-
; CHECK-NEXT: [[ARRAYIDX50:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 5
273-
; CHECK-NEXT: [[ARRAYIDX52:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 5
274-
; CHECK-NEXT: [[ARRAYIDX61:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 6
275-
; CHECK-NEXT: [[ARRAYIDX63:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 6
276-
; CHECK-NEXT: [[ARRAYIDX72:%.*]] = getelementptr inbounds i8, i8* [[P1_044]], i64 7
277-
; CHECK-NEXT: [[ARRAYIDX74:%.*]] = getelementptr inbounds i8, i8* [[P2_045]], i64 7
278252
; CHECK-NEXT: [[TMP0:%.*]] = bitcast i8* [[P1_044]] to <8 x i8>*
279253
; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, <8 x i8>* [[TMP0]], align 1
280254
; CHECK-NEXT: [[TMP2:%.*]] = zext <8 x i8> [[TMP1]] to <8 x i32>

‎llvm/test/Transforms/SLPVectorizer/AArch64/loadi8.ll

Lines changed: 2 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -13,13 +13,7 @@ define void @f_noalias(i8* noalias nocapture %dst, i8* noalias nocapture readonl
1313
; CHECK-NEXT: [[TMP0:%.*]] = load i32, i32* [[SCALE]], align 16
1414
; CHECK-NEXT: [[OFFSET:%.*]] = getelementptr inbounds [[STRUCT_WEIGHT_T]], %struct.weight_t* [[W]], i64 0, i32 1
1515
; CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* [[OFFSET]], align 4
16-
; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds i8, i8* [[SRC:%.*]], i64 1
17-
; CHECK-NEXT: [[ARRAYIDX2_1:%.*]] = getelementptr inbounds i8, i8* [[DST:%.*]], i64 1
18-
; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds i8, i8* [[SRC]], i64 2
19-
; CHECK-NEXT: [[ARRAYIDX2_2:%.*]] = getelementptr inbounds i8, i8* [[DST]], i64 2
20-
; CHECK-NEXT: [[ARRAYIDX_3:%.*]] = getelementptr inbounds i8, i8* [[SRC]], i64 3
21-
; CHECK-NEXT: [[ARRAYIDX2_3:%.*]] = getelementptr inbounds i8, i8* [[DST]], i64 3
22-
; CHECK-NEXT: [[TMP2:%.*]] = bitcast i8* [[SRC]] to <4 x i8>*
16+
; CHECK-NEXT: [[TMP2:%.*]] = bitcast i8* [[SRC:%.*]] to <4 x i8>*
2317
; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i8>, <4 x i8>* [[TMP2]], align 1
2418
; CHECK-NEXT: [[TMP4:%.*]] = zext <4 x i8> [[TMP3]] to <4 x i32>
2519
; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i32 0
@@ -33,7 +27,7 @@ define void @f_noalias(i8* noalias nocapture %dst, i8* noalias nocapture readonl
3327
; CHECK-NEXT: [[TMP11:%.*]] = sext <4 x i1> [[TMP10]] to <4 x i32>
3428
; CHECK-NEXT: [[TMP12:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> [[TMP8]], <4 x i32> [[TMP11]]
3529
; CHECK-NEXT: [[TMP13:%.*]] = trunc <4 x i32> [[TMP12]] to <4 x i8>
36-
; CHECK-NEXT: [[TMP14:%.*]] = bitcast i8* [[DST]] to <4 x i8>*
30+
; CHECK-NEXT: [[TMP14:%.*]] = bitcast i8* [[DST:%.*]] to <4 x i8>*
3731
; CHECK-NEXT: store <4 x i8> [[TMP13]], <4 x i8>* [[TMP14]], align 1
3832
; CHECK-NEXT: ret void
3933
;

‎llvm/test/Transforms/SLPVectorizer/AArch64/matmul.ll

Lines changed: 1 addition & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -16,17 +16,12 @@ define void @wrap_mul4(double* nocapture %Out, [2 x double]* nocapture readonly
1616
; CHECK-NEXT: [[ARRAYIDX5_I:%.*]] = getelementptr inbounds [2 x double], [2 x double]* [[A]], i64 0, i64 1
1717
; CHECK-NEXT: [[TEMP2:%.*]] = load double, double* [[ARRAYIDX5_I]], align 8
1818
; CHECK-NEXT: [[ARRAYIDX7_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 1, i64 0
19-
; CHECK-NEXT: [[ARRAYIDX13_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 0, i64 1
20-
; CHECK-NEXT: [[ARRAYIDX18_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 1, i64 1
2119
; CHECK-NEXT: [[ARRAYIDX25_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 0, i64 2
2220
; CHECK-NEXT: [[ARRAYIDX30_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 1, i64 2
23-
; CHECK-NEXT: [[ARRAYIDX37_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 0, i64 3
24-
; CHECK-NEXT: [[ARRAYIDX42_I:%.*]] = getelementptr inbounds [4 x double], [4 x double]* [[B]], i64 1, i64 3
2521
; CHECK-NEXT: [[ARRAYIDX47_I:%.*]] = getelementptr inbounds [2 x double], [2 x double]* [[A]], i64 1, i64 0
2622
; CHECK-NEXT: [[TEMP10:%.*]] = load double, double* [[ARRAYIDX47_I]], align 8
2723
; CHECK-NEXT: [[ARRAYIDX52_I:%.*]] = getelementptr inbounds [2 x double], [2 x double]* [[A]], i64 1, i64 1
2824
; CHECK-NEXT: [[TEMP11:%.*]] = load double, double* [[ARRAYIDX52_I]], align 8
29-
; CHECK-NEXT: [[RES_I_SROA_4_0_OUT2_I_SROA_IDX2:%.*]] = getelementptr inbounds double, double* [[OUT:%.*]], i64 1
3025
; CHECK-NEXT: [[TMP1:%.*]] = bitcast double* [[ARRAYIDX3_I]] to <2 x double>*
3126
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* [[TMP1]], align 8
3227
; CHECK-NEXT: [[TMP3:%.*]] = insertelement <2 x double> poison, double [[TEMP]], i32 0
@@ -38,9 +33,8 @@ define void @wrap_mul4(double* nocapture %Out, [2 x double]* nocapture readonly
3833
; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x double> [[TMP8]], double [[TEMP2]], i32 1
3934
; CHECK-NEXT: [[TMP10:%.*]] = fmul <2 x double> [[TMP9]], [[TMP7]]
4035
; CHECK-NEXT: [[TMP11:%.*]] = fadd <2 x double> [[TMP5]], [[TMP10]]
41-
; CHECK-NEXT: [[TMP12:%.*]] = bitcast double* [[OUT]] to <2 x double>*
36+
; CHECK-NEXT: [[TMP12:%.*]] = bitcast double* [[OUT:%.*]] to <2 x double>*
4237
; CHECK-NEXT: [[RES_I_SROA_5_0_OUT2_I_SROA_IDX4:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 2
43-
; CHECK-NEXT: [[RES_I_SROA_6_0_OUT2_I_SROA_IDX6:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 3
4438
; CHECK-NEXT: [[TMP13:%.*]] = bitcast double* [[ARRAYIDX25_I]] to <2 x double>*
4539
; CHECK-NEXT: [[TMP14:%.*]] = load <2 x double>, <2 x double>* [[TMP13]], align 8
4640
; CHECK-NEXT: [[TMP15:%.*]] = fmul <2 x double> [[TMP4]], [[TMP14]]
@@ -52,7 +46,6 @@ define void @wrap_mul4(double* nocapture %Out, [2 x double]* nocapture readonly
5246
; CHECK-NEXT: [[TMP20:%.*]] = bitcast double* [[RES_I_SROA_5_0_OUT2_I_SROA_IDX4]] to <2 x double>*
5347
; CHECK-NEXT: store <2 x double> [[TMP19]], <2 x double>* [[TMP20]], align 8
5448
; CHECK-NEXT: [[RES_I_SROA_7_0_OUT2_I_SROA_IDX8:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 4
55-
; CHECK-NEXT: [[RES_I_SROA_8_0_OUT2_I_SROA_IDX10:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 5
5649
; CHECK-NEXT: [[TMP21:%.*]] = insertelement <2 x double> poison, double [[TEMP10]], i32 0
5750
; CHECK-NEXT: [[TMP22:%.*]] = insertelement <2 x double> [[TMP21]], double [[TEMP10]], i32 1
5851
; CHECK-NEXT: [[TMP23:%.*]] = fmul <2 x double> [[TMP2]], [[TMP22]]
@@ -63,7 +56,6 @@ define void @wrap_mul4(double* nocapture %Out, [2 x double]* nocapture readonly
6356
; CHECK-NEXT: [[TMP28:%.*]] = bitcast double* [[RES_I_SROA_7_0_OUT2_I_SROA_IDX8]] to <2 x double>*
6457
; CHECK-NEXT: store <2 x double> [[TMP27]], <2 x double>* [[TMP28]], align 8
6558
; CHECK-NEXT: [[RES_I_SROA_9_0_OUT2_I_SROA_IDX12:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 6
66-
; CHECK-NEXT: [[RES_I_SROA_10_0_OUT2_I_SROA_IDX14:%.*]] = getelementptr inbounds double, double* [[OUT]], i64 7
6759
; CHECK-NEXT: [[TMP29:%.*]] = fmul <2 x double> [[TMP14]], [[TMP22]]
6860
; CHECK-NEXT: [[TMP30:%.*]] = fmul <2 x double> [[TMP17]], [[TMP25]]
6961
; CHECK-NEXT: [[TMP31:%.*]] = fadd <2 x double> [[TMP29]], [[TMP30]]

0 commit comments

Comments
 (0)