Skip to content

Commit 8d8f2b1

Browse files
davemgreendavemgreen
committed
Add one-use check and rebase
1 parent 7b17239 commit 8d8f2b1

File tree

8 files changed

+53
-84
lines changed

8 files changed

+53
-84
lines changed

‎llvm/lib/Transforms/InstCombine/InstructionCombining.cpp

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -2795,8 +2795,8 @@ Instruction *InstCombinerImpl::visitGetElementPtrInst(GetElementPtrInst &GEP) {
27952795
if (GEPEltType->isScalableTy() ||
27962796
(!GEPEltType->isIntegerTy(8) && GEP.getNumIndices() == 1 &&
27972797
match(GEP.getOperand(1),
2798-
m_CombineOr(m_Mul(m_Value(), m_ConstantInt()),
2799-
m_Shl(m_Value(), m_ConstantInt()))))) {
2798+
m_OneUse(m_CombineOr(m_Mul(m_Value(), m_ConstantInt()),
2799+
m_Shl(m_Value(), m_ConstantInt())))))) {
28002800
Value *Offset = EmitGEPOffset(cast<GEPOperator>(&GEP));
28012801
return replaceInstUsesWith(
28022802
GEP, Builder.CreatePtrAdd(PtrOp, Offset, "", GEP.isInBounds()));

‎llvm/test/Transforms/InstCombine/canonicalize-gep-mul.ll

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -48,8 +48,7 @@ define ptr @usemul(ptr %p, i64 %x) {
4848
; CHECK-SAME: ptr [[P:%.*]], i64 [[X:%.*]]) {
4949
; CHECK-NEXT: entry:
5050
; CHECK-NEXT: [[MUL:%.*]] = mul i64 [[X]], 5
51-
; CHECK-NEXT: [[GEP_IDX:%.*]] = mul i64 [[X]], 20
52-
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[GEP_IDX]]
51+
; CHECK-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[P]], i64 [[MUL]]
5352
; CHECK-NEXT: call void @use(i64 [[MUL]])
5453
; CHECK-NEXT: ret ptr [[GEP]]
5554
;

‎llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll

Lines changed: 11 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -480,18 +480,18 @@ define void @even_load_static_tc(ptr noalias nocapture readonly %A, ptr noalias
480480
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
481481
; CHECK: vector.body:
482482
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
483-
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = and i64 [[INDEX]], 9223372036854775804
484483
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
485484
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
486485
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP4]], align 4
487486
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]])
488487
; CHECK-NEXT: [[TMP5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0
489488
; CHECK-NEXT: [[TMP6:%.*]] = shl nsw <vscale x 4 x i32> [[TMP5]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
490-
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
491-
; CHECK-NEXT: store <vscale x 4 x i32> [[TMP6]], ptr [[TMP7]], align 4
489+
; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[INDEX]], 9223372036854775804
490+
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP7]]
491+
; CHECK-NEXT: store <vscale x 4 x i32> [[TMP6]], ptr [[TMP8]], align 4
492492
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP3]]
493-
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
494-
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
493+
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
494+
; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
495495
; CHECK: middle.block:
496496
; CHECK-NEXT: br label [[SCALAR_PH]]
497497
; CHECK: scalar.ph:
@@ -566,18 +566,18 @@ define void @even_load_dynamic_tc(ptr noalias nocapture readonly %A, ptr noalias
566566
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
567567
; CHECK: vector.body:
568568
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
569-
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = and i64 [[INDEX]], 9223372036854775804
570569
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
571570
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
572571
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP12]], align 4
573572
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]])
574573
; CHECK-NEXT: [[TMP13:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0
575574
; CHECK-NEXT: [[TMP14:%.*]] = shl nsw <vscale x 4 x i32> [[TMP13]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 1, i64 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
576-
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
577-
; CHECK-NEXT: store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4
575+
; CHECK-NEXT: [[TMP15:%.*]] = and i64 [[INDEX]], 9223372036854775804
576+
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP15]]
577+
; CHECK-NEXT: store <vscale x 4 x i32> [[TMP14]], ptr [[TMP16]], align 4
578578
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
579-
; CHECK-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
580-
; CHECK-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
579+
; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
580+
; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
581581
; CHECK: middle.block:
582582
; CHECK-NEXT: br label [[SCALAR_PH]]
583583
; CHECK: scalar.ph:
@@ -715,8 +715,7 @@ define void @mixed_load2_store2(ptr noalias nocapture readonly %A, ptr noalias n
715715
; CHECK: vector.body:
716716
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
717717
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 1
718-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
719-
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
718+
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
720719
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP2]], align 4
721720
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]])
722721
; CHECK-NEXT: [[TMP3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0

‎llvm/test/Transforms/LoopVectorize/X86/interleaving.ll

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -158,8 +158,7 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
158158
; ATOM: for.body:
159159
; ATOM-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
160160
; ATOM-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[INDVARS_IV]], 1
161-
; ATOM-NEXT: [[ARRAYIDX_IDX:%.*]] = shl nsw i64 [[INDVARS_IV]], 3
162-
; ATOM-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[ARRAYIDX_IDX]]
161+
; ATOM-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP0]]
163162
; ATOM-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
164163
; ATOM-NEXT: [[TMP2:%.*]] = or disjoint i64 [[TMP0]], 1
165164
; ATOM-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP2]]

‎llvm/test/Transforms/LoopVectorize/induction.ll

Lines changed: 6 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -948,15 +948,13 @@ define float @scalarize_induction_variable_02(ptr %a, ptr %b, i64 %n) {
948948
; IND-NEXT: [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
949949
; IND-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
950950
; IND-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 8
951-
; IND-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 5
952-
; IND-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
951+
; IND-NEXT: [[TMP4:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
953952
; IND-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]
954953
; IND-NEXT: [[TMP6:%.*]] = load float, ptr [[TMP4]], align 4
955954
; IND-NEXT: [[TMP7:%.*]] = load float, ptr [[TMP5]], align 4
956955
; IND-NEXT: [[TMP8:%.*]] = insertelement <2 x float> poison, float [[TMP6]], i64 0
957956
; IND-NEXT: [[TMP9:%.*]] = insertelement <2 x float> [[TMP8]], float [[TMP7]], i64 1
958-
; IND-NEXT: [[DOTIDX1:%.*]] = shl i64 [[INDEX]], 5
959-
; IND-NEXT: [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX1]]
957+
; IND-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
960958
; IND-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP3]]
961959
; IND-NEXT: [[TMP12:%.*]] = load float, ptr [[TMP10]], align 4
962960
; IND-NEXT: [[TMP13:%.*]] = load float, ptr [[TMP11]], align 4
@@ -1013,8 +1011,7 @@ define float @scalarize_induction_variable_02(ptr %a, ptr %b, i64 %n) {
10131011
; UNROLL-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 8
10141012
; UNROLL-NEXT: [[TMP4:%.*]] = or disjoint i64 [[OFFSET_IDX]], 16
10151013
; UNROLL-NEXT: [[TMP5:%.*]] = or disjoint i64 [[OFFSET_IDX]], 24
1016-
; UNROLL-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 5
1017-
; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
1014+
; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
10181015
; UNROLL-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP3]]
10191016
; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP4]]
10201017
; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]
@@ -1026,8 +1023,7 @@ define float @scalarize_induction_variable_02(ptr %a, ptr %b, i64 %n) {
10261023
; UNROLL-NEXT: [[TMP15:%.*]] = load float, ptr [[TMP9]], align 4
10271024
; UNROLL-NEXT: [[TMP16:%.*]] = insertelement <2 x float> poison, float [[TMP14]], i64 0
10281025
; UNROLL-NEXT: [[TMP17:%.*]] = insertelement <2 x float> [[TMP16]], float [[TMP15]], i64 1
1029-
; UNROLL-NEXT: [[DOTIDX2:%.*]] = shl i64 [[INDEX]], 5
1030-
; UNROLL-NEXT: [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX2]]
1026+
; UNROLL-NEXT: [[TMP18:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
10311027
; UNROLL-NEXT: [[TMP19:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP3]]
10321028
; UNROLL-NEXT: [[TMP20:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP4]]
10331029
; UNROLL-NEXT: [[TMP21:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]
@@ -1175,15 +1171,13 @@ define float @scalarize_induction_variable_02(ptr %a, ptr %b, i64 %n) {
11751171
; INTERLEAVE-NEXT: [[VEC_PHI1:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
11761172
; INTERLEAVE-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 3
11771173
; INTERLEAVE-NEXT: [[TMP5:%.*]] = or disjoint i64 [[OFFSET_IDX]], 32
1178-
; INTERLEAVE-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 5
1179-
; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
1174+
; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
11801175
; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]]
11811176
; INTERLEAVE-NEXT: [[WIDE_VEC:%.*]] = load <32 x float>, ptr [[TMP6]], align 4
11821177
; INTERLEAVE-NEXT: [[WIDE_VEC2:%.*]] = load <32 x float>, ptr [[TMP7]], align 4
11831178
; INTERLEAVE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <32 x float> [[WIDE_VEC]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
11841179
; INTERLEAVE-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <32 x float> [[WIDE_VEC2]], <32 x float> poison, <4 x i32> <i32 0, i32 8, i32 16, i32 24>
1185-
; INTERLEAVE-NEXT: [[DOTIDX8:%.*]] = shl i64 [[INDEX]], 5
1186-
; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX8]]
1180+
; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
11871181
; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[TMP5]]
11881182
; INTERLEAVE-NEXT: [[WIDE_VEC4:%.*]] = load <32 x float>, ptr [[TMP8]], align 4
11891183
; INTERLEAVE-NEXT: [[WIDE_VEC5:%.*]] = load <32 x float>, ptr [[TMP9]], align 4

‎llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll

Lines changed: 13 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -419,17 +419,17 @@ define void @even_load_static_tc(ptr noalias nocapture readonly %A, ptr noalias
419419
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
420420
; CHECK: vector.body:
421421
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
422-
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = and i64 [[INDEX]], 9223372036854775804
423422
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
424423
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
425424
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP0]], align 4
426425
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
427426
; CHECK-NEXT: [[TMP1:%.*]] = shl nsw <4 x i32> [[STRIDED_VEC]], <i32 1, i32 1, i32 1, i32 1>
428-
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
429-
; CHECK-NEXT: store <4 x i32> [[TMP1]], ptr [[TMP2]], align 4
427+
; CHECK-NEXT: [[TMP2:%.*]] = and i64 [[INDEX]], 9223372036854775804
428+
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP2]]
429+
; CHECK-NEXT: store <4 x i32> [[TMP1]], ptr [[TMP3]], align 4
430430
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
431-
; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 508
432-
; CHECK-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
431+
; CHECK-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 508
432+
; CHECK-NEXT: br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
433433
; CHECK: middle.block:
434434
; CHECK-NEXT: br label [[SCALAR_PH]]
435435
; CHECK: scalar.ph:
@@ -495,17 +495,17 @@ define void @even_load_dynamic_tc(ptr noalias nocapture readonly %A, ptr noalias
495495
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
496496
; CHECK: vector.body:
497497
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
498-
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = and i64 [[INDEX]], 9223372036854775804
499498
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
500499
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
501500
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
502501
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
503502
; CHECK-NEXT: [[TMP6:%.*]] = shl nsw <4 x i32> [[STRIDED_VEC]], <i32 1, i32 1, i32 1, i32 1>
504-
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
505-
; CHECK-NEXT: store <4 x i32> [[TMP6]], ptr [[TMP7]], align 4
503+
; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[INDEX]], 9223372036854775804
504+
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP7]]
505+
; CHECK-NEXT: store <4 x i32> [[TMP6]], ptr [[TMP8]], align 4
506506
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
507-
; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
508-
; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
507+
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
508+
; CHECK-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
509509
; CHECK: middle.block:
510510
; CHECK-NEXT: br label [[SCALAR_PH]]
511511
; CHECK: scalar.ph:
@@ -665,8 +665,7 @@ define void @mixed_load2_store2(ptr noalias nocapture readonly %A, ptr noalias n
665665
; CHECK: vector.body:
666666
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
667667
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 1
668-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
669-
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
668+
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
670669
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <8 x i32>, ptr [[TMP0]], align 4
671670
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
672671
; CHECK-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
@@ -1288,8 +1287,7 @@ define void @PR27626_4(ptr %a, i32 %x, i32 %y, i32 %z, i64 %n) {
12881287
; CHECK-NEXT: [[TMP4:%.*]] = or disjoint i64 [[OFFSET_IDX]], 4
12891288
; CHECK-NEXT: [[TMP5:%.*]] = or disjoint i64 [[OFFSET_IDX]], 6
12901289
; CHECK-NEXT: [[TMP6:%.*]] = or disjoint i64 [[OFFSET_IDX]], 1
1291-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
1292-
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX]]
1290+
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
12931291
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP3]]
12941292
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
12951293
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]]
@@ -1380,8 +1378,7 @@ define void @PR27626_5(ptr %a, i32 %x, i32 %y, i32 %z, i64 %n) {
13801378
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
13811379
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]]
13821380
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]]
1383-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
1384-
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[A]], i64 [[DOTIDX]]
1381+
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP4]]
13851382
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP12]], i64 36
13861383
; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i64> [[TMP7]], i64 0
13871384
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]

‎llvm/test/Transforms/LoopVectorize/loop-scalars.ll

Lines changed: 4 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -75,11 +75,9 @@ define void @scalar_store(ptr %a, ptr %b, i64 %n) {
7575
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
7676
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 1
7777
; CHECK-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 2
78-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 3
79-
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX]]
78+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
8079
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP3]]
81-
; CHECK-NEXT: [[DOTIDX1:%.*]] = shl i64 [[INDEX]], 4
82-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX1]]
80+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
8381
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[TMP3]]
8482
; CHECK-NEXT: store ptr [[TMP4]], ptr [[TMP6]], align 8
8583
; CHECK-NEXT: store ptr [[TMP5]], ptr [[TMP7]], align 8
@@ -137,11 +135,9 @@ define void @expansion(ptr %a, ptr %b, i64 %n) {
137135
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
138136
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = shl i64 [[INDEX]], 1
139137
; CHECK-NEXT: [[TMP3:%.*]] = or disjoint i64 [[OFFSET_IDX]], 2
140-
; CHECK-NEXT: [[DOTIDX:%.*]] = shl i64 [[INDEX]], 4
141-
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[DOTIDX]]
138+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
142139
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
143-
; CHECK-NEXT: [[DOTIDX1:%.*]] = shl i64 [[INDEX]], 4
144-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[DOTIDX1]]
140+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
145141
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[TMP3]]
146142
; CHECK-NEXT: store ptr [[TMP4]], ptr [[TMP6]], align 8
147143
; CHECK-NEXT: store ptr [[TMP5]], ptr [[TMP7]], align 8

0 commit comments

Comments
 (0)