[clang] 457dc72 - Reland [InstCombine] Infer inbounds for more GEPs of dereferenceable pointers
Arthur Eubanks via cfe-commits
cfe-commits at lists.llvm.org
Tue Jun 27 09:31:30 PDT 2023
Author: Arthur Eubanks
Date: 2023-06-27T09:31:20-07:00
New Revision: 457dc72fdd2eefe364548e721619cc1951854372
URL: https://github.com/llvm/llvm-project/commit/457dc72fdd2eefe364548e721619cc1951854372
DIFF: https://github.com/llvm/llvm-project/commit/457dc72fdd2eefe364548e721619cc1951854372.diff
LOG: Reland [InstCombine] Infer inbounds for more GEPs of dereferenceable pointers
Use Value::getPointerDereferenceableBytes() instead of hardcoding dereferenceable only for allocas. Allows us to infer inbounds GEPs for other Values like CallInsts and Arguments.
Fixed clang test broken in initial land.
Reviewed By: nikic
Differential Revision: https://reviews.llvm.org/D153815
Added:
Modified:
clang/test/CodeGenOpenCL/builtins-amdgcn.cl
llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
llvm/test/Transforms/InstCombine/compare-unescaped.ll
llvm/test/Transforms/InstCombine/inbounds-gep.ll
llvm/test/Transforms/PhaseOrdering/X86/vec-load-combine.ll
llvm/test/Transforms/PhaseOrdering/bitcast-store-branch.ll
llvm/test/Transforms/PhaseOrdering/dce-after-argument-promotion-loads.ll
Removed:
################################################################################
diff --git a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
index 18725d4193efb..74f3343f79e4a 100644
--- a/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
+++ b/clang/test/CodeGenOpenCL/builtins-amdgcn.cl
@@ -598,11 +598,11 @@ void test_get_local_id(int d, global int *out)
// CHECK-LABEL: @test_get_workgroup_size(
// CHECK: call align 4 dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 4
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 4
// CHECK: load i16, ptr addrspace(4) %{{.*}}, align 4, !range [[$WS_RANGE:![0-9]*]], !invariant.load{{.*}}, !noundef
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 6
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 6
// CHECK: load i16, ptr addrspace(4) %{{.*}}, align 2, !range [[$WS_RANGE:![0-9]*]], !invariant.load{{.*}}, !noundef
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 8
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 8
// CHECK: load i16, ptr addrspace(4) %{{.*}}, align 4, !range [[$WS_RANGE:![0-9]*]], !invariant.load{{.*}}, !noundef
void test_get_workgroup_size(int d, global int *out)
{
@@ -616,11 +616,11 @@ void test_get_workgroup_size(int d, global int *out)
// CHECK-LABEL: @test_get_grid_size(
// CHECK: call align 4 dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 12
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 12
// CHECK: load i32, ptr addrspace(4) %{{.*}}, align 4, !invariant.load
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 16
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 16
// CHECK: load i32, ptr addrspace(4) %{{.*}}, align 4, !invariant.load
-// CHECK: getelementptr i8, ptr addrspace(4) %{{.*}}, i64 20
+// CHECK: getelementptr inbounds i8, ptr addrspace(4) %{{.*}}, i64 20
// CHECK: load i32, ptr addrspace(4) %{{.*}}, align 4, !invariant.load
void test_get_grid_size(int d, global int *out)
{
diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
index 0fde841a0e973..b38cde63ec2f8 100644
--- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp
@@ -2206,12 +2206,13 @@ Instruction *InstCombinerImpl::visitGetElementPtrInst(GetElementPtrInst &GEP) {
Value *UnderlyingPtrOp =
PtrOp->stripAndAccumulateInBoundsConstantOffsets(DL,
BasePtrOffset);
- if (auto *AI = dyn_cast<AllocaInst>(UnderlyingPtrOp)) {
+ bool CanBeNull, CanBeFreed;
+ uint64_t DerefBytes = UnderlyingPtrOp->getPointerDereferenceableBytes(
+ DL, CanBeNull, CanBeFreed);
+ if (!CanBeNull && !CanBeFreed && DerefBytes != 0) {
if (GEP.accumulateConstantOffset(DL, BasePtrOffset) &&
BasePtrOffset.isNonNegative()) {
- APInt AllocSize(
- IdxWidth,
- DL.getTypeAllocSize(AI->getAllocatedType()).getKnownMinValue());
+ APInt AllocSize(IdxWidth, DerefBytes);
if (BasePtrOffset.ule(AllocSize)) {
return GetElementPtrInst::CreateInBounds(
GEP.getSourceElementType(), PtrOp, Indices, GEP.getName());
diff --git a/llvm/test/Transforms/InstCombine/compare-unescaped.ll b/llvm/test/Transforms/InstCombine/compare-unescaped.ll
index e661cf9d31c5a..ab380c00f8264 100644
--- a/llvm/test/Transforms/InstCombine/compare-unescaped.ll
+++ b/llvm/test/Transforms/InstCombine/compare-unescaped.ll
@@ -382,8 +382,8 @@ define i1 @two_nonnull_mallocs_hidden() {
; CHECK-LABEL: @two_nonnull_mallocs_hidden(
; CHECK-NEXT: [[M:%.*]] = call nonnull dereferenceable(4) ptr @malloc(i64 4)
; CHECK-NEXT: [[N:%.*]] = call nonnull dereferenceable(4) ptr @malloc(i64 4)
-; CHECK-NEXT: [[GEP1:%.*]] = getelementptr i8, ptr [[M]], i64 1
-; CHECK-NEXT: [[GEP2:%.*]] = getelementptr i8, ptr [[N]], i64 2
+; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i8, ptr [[M]], i64 1
+; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i8, ptr [[N]], i64 2
; CHECK-NEXT: [[CMP:%.*]] = icmp eq ptr [[GEP1]], [[GEP2]]
; CHECK-NEXT: ret i1 [[CMP]]
;
diff --git a/llvm/test/Transforms/InstCombine/inbounds-gep.ll b/llvm/test/Transforms/InstCombine/inbounds-gep.ll
index 4c2cb177340ad..f2a647c135103 100644
--- a/llvm/test/Transforms/InstCombine/inbounds-gep.ll
+++ b/llvm/test/Transforms/InstCombine/inbounds-gep.ll
@@ -19,8 +19,8 @@ define void @call1() {
define void @call2() {
; CHECK-LABEL: define void @call2() {
; CHECK-NEXT: [[A:%.*]] = call dereferenceable(8) ptr @g()
-; CHECK-NEXT: [[B:%.*]] = getelementptr i8, ptr [[A]], i64 4
-; CHECK-NEXT: call void @use(ptr [[B]])
+; CHECK-NEXT: [[B:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 4
+; CHECK-NEXT: call void @use(ptr nonnull [[B]])
; CHECK-NEXT: ret void
;
%a = call dereferenceable(8) ptr @g()
@@ -70,8 +70,8 @@ define void @arg1(ptr %a) {
define void @arg2(ptr dereferenceable(8) %a) {
; CHECK-LABEL: define void @arg2
; CHECK-SAME: (ptr dereferenceable(8) [[A:%.*]]) {
-; CHECK-NEXT: [[B:%.*]] = getelementptr i8, ptr [[A]], i64 4
-; CHECK-NEXT: call void @use(ptr [[B]])
+; CHECK-NEXT: [[B:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 4
+; CHECK-NEXT: call void @use(ptr nonnull [[B]])
; CHECK-NEXT: ret void
;
%b = getelementptr i8, ptr %a, i64 4
diff --git a/llvm/test/Transforms/PhaseOrdering/X86/vec-load-combine.ll b/llvm/test/Transforms/PhaseOrdering/X86/vec-load-combine.ll
index 393d8601cd2ae..2296531bd1de6 100644
--- a/llvm/test/Transforms/PhaseOrdering/X86/vec-load-combine.ll
+++ b/llvm/test/Transforms/PhaseOrdering/X86/vec-load-combine.ll
@@ -48,7 +48,7 @@ define noundef <4 x float> @ConvertVectors_ByVal(ptr noundef nonnull align 16 de
; SSE-LABEL: @ConvertVectors_ByVal(
; SSE-NEXT: entry:
; SSE-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[V:%.*]], align 16
-; SSE-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[V]], i64 8
+; SSE-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[V]], i64 8
; SSE-NEXT: [[V_VAL421:%.*]] = load i64, ptr [[TMP1]], align 8
; SSE-NEXT: [[TMP2:%.*]] = trunc i64 [[V_VAL421]] to i32
; SSE-NEXT: [[TMP3:%.*]] = bitcast i32 [[TMP2]] to float
@@ -59,7 +59,7 @@ define noundef <4 x float> @ConvertVectors_ByVal(ptr noundef nonnull align 16 de
; AVX-LABEL: @ConvertVectors_ByVal(
; AVX-NEXT: entry:
; AVX-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[V:%.*]], align 16
-; AVX-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[V]], i64 8
+; AVX-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[V]], i64 8
; AVX-NEXT: [[V_VAL421:%.*]] = load i64, ptr [[TMP1]], align 8
; AVX-NEXT: [[TMP2:%.*]] = trunc i64 [[V_VAL421]] to i32
; AVX-NEXT: [[TMP3:%.*]] = bitcast i32 [[TMP2]] to float
diff --git a/llvm/test/Transforms/PhaseOrdering/bitcast-store-branch.ll b/llvm/test/Transforms/PhaseOrdering/bitcast-store-branch.ll
index f8c551a256f98..678ac59a694b3 100644
--- a/llvm/test/Transforms/PhaseOrdering/bitcast-store-branch.ll
+++ b/llvm/test/Transforms/PhaseOrdering/bitcast-store-branch.ll
@@ -14,7 +14,7 @@ define ptr @parent(ptr align 8 dereferenceable(72) %f, half %val1, i16 %val2, i3
; CHECK-LABEL: define nonnull ptr @parent
; CHECK-SAME: (ptr readonly returned align 8 dereferenceable(72) [[F:%.*]], half [[VAL1:%.*]], i16 [[VAL2:%.*]], i32 [[VAL3:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] align 2 {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i8, ptr [[F]], i64 64
+; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[F]], i64 64
; CHECK-NEXT: [[F_VAL:%.*]] = load ptr, ptr [[TMP0]], align 8
; CHECK-NEXT: [[CMP_NOT_NOT_I:%.*]] = icmp eq i32 [[VAL3]], 0
; CHECK-NEXT: [[TMP1:%.*]] = bitcast half [[VAL1]] to i16
diff --git a/llvm/test/Transforms/PhaseOrdering/dce-after-argument-promotion-loads.ll b/llvm/test/Transforms/PhaseOrdering/dce-after-argument-promotion-loads.ll
index 2fe8f39e423a5..ee913d299aaab 100644
--- a/llvm/test/Transforms/PhaseOrdering/dce-after-argument-promotion-loads.ll
+++ b/llvm/test/Transforms/PhaseOrdering/dce-after-argument-promotion-loads.ll
@@ -16,7 +16,7 @@ define ptr @parent(ptr align 8 dereferenceable(72) %f, i16 %val1, i16 %val2, i32
; CHECK-LABEL: define nonnull ptr @parent
; CHECK-SAME: (ptr readonly returned align 8 dereferenceable(72) [[F:%.*]], i16 [[VAL1:%.*]], i16 [[VAL2:%.*]], i32 [[VAL3:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] align 2 {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i8, ptr [[F]], i64 64
+; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[F]], i64 64
; CHECK-NEXT: [[F_VAL:%.*]] = load ptr, ptr [[TMP0]], align 8
; CHECK-NEXT: [[CMP_NOT_NOT_I:%.*]] = icmp eq i32 [[VAL3]], 0
; CHECK-NEXT: [[SPEC_SELECT_I:%.*]] = select i1 [[CMP_NOT_NOT_I]], i16 [[VAL1]], i16 [[VAL2]]
More information about the cfe-commits
mailing list