[Mlir-commits] [llvm] [libc] [flang] [openmp] [lldb] [libcxx] [mlir] [clang] Fix Logical expression used for merged conditional if in FlattenCFG pass (PR #72522)
llvmlistbot at llvm.org
llvmlistbot at llvm.org
Thu Nov 16 09:53:29 PST 2023
https://github.com/mahtohappy updated https://github.com/llvm/llvm-project/pull/72522
>From ce4f4fc02e25359c8f38ff9ecf2a2d82aa90df72 Mon Sep 17 00:00:00 2001
From: mahtohappy <Happy.Kumar at windriver.com>
Date: Thu, 16 Nov 2023 06:53:24 -0800
Subject: [PATCH 1/2] Fix Logical expression used for merged conditional if
---
llvm/lib/Transforms/Utils/FlattenCFG.cpp | 8 ++++----
llvm/test/Transforms/Util/flattencfg.ll | 22 +++++++++++-----------
2 files changed, 15 insertions(+), 15 deletions(-)
diff --git a/llvm/lib/Transforms/Utils/FlattenCFG.cpp b/llvm/lib/Transforms/Utils/FlattenCFG.cpp
index 1925b91c4da7ec1..5f8cd12c1d50856 100644
--- a/llvm/lib/Transforms/Utils/FlattenCFG.cpp
+++ b/llvm/lib/Transforms/Utils/FlattenCFG.cpp
@@ -436,9 +436,9 @@ bool FlattenCFGOpt::MergeIfRegion(BasicBlock *BB, IRBuilder<> &Builder) {
bool InvertCond2 = false;
BinaryOperator::BinaryOps CombineOp;
if (IfFalse1 == FirstEntryBlock) {
- // The else-path is empty, so we must use "or" operation to combine the
+ // The then-path is empty, so we must use "and" operation to combine the
// conditions.
- CombineOp = BinaryOperator::Or;
+ CombineOp = BinaryOperator::And;
if (IfFalse2 != SecondEntryBlock) {
if (IfTrue2 != SecondEntryBlock)
return false;
@@ -450,9 +450,9 @@ bool FlattenCFGOpt::MergeIfRegion(BasicBlock *BB, IRBuilder<> &Builder) {
if (!CompareIfRegionBlock(IfTrue1, IfTrue2, SecondEntryBlock))
return false;
} else if (IfTrue1 == FirstEntryBlock) {
- // The then-path is empty, so we must use "and" operation to combine the
+ // The else-path is empty, so we must use "or" operation to combine the
// conditions.
- CombineOp = BinaryOperator::And;
+ CombineOp = BinaryOperator::Or;
if (IfTrue2 != SecondEntryBlock) {
if (IfFalse2 != SecondEntryBlock)
return false;
diff --git a/llvm/test/Transforms/Util/flattencfg.ll b/llvm/test/Transforms/Util/flattencfg.ll
index 4a4d4279f360d6a..0fd285be2715b80 100644
--- a/llvm/test/Transforms/Util/flattencfg.ll
+++ b/llvm/test/Transforms/Util/flattencfg.ll
@@ -10,12 +10,12 @@ define void @test_not_crash(i32 %in_a) #0 {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[CMP0:%.*]] = icmp eq i32 [[IN_A]], -1
; CHECK-NEXT: [[CMP1:%.*]] = icmp ne i32 [[IN_A]], 0
-; CHECK-NEXT: [[COND0:%.*]] = and i1 [[CMP0]], [[CMP1]]
+; CHECK-NEXT: [[COND0:%.*]] = or i1 [[CMP0]], [[CMP1]]
; CHECK-NEXT: br i1 [[COND0]], label [[B0:%.*]], label [[B1:%.*]]
; CHECK: b0:
; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[IN_A]], 0
; CHECK-NEXT: [[CMP3:%.*]] = icmp ne i32 [[IN_A]], 1
-; CHECK-NEXT: [[COND1:%.*]] = or i1 [[CMP2]], [[CMP3]]
+; CHECK-NEXT: [[COND1:%.*]] = and i1 [[CMP2]], [[CMP3]]
; CHECK-NEXT: br i1 [[COND1]], label [[EXIT:%.*]], label [[B1]]
; CHECK: b1:
; CHECK-NEXT: br label [[EXIT]]
@@ -25,13 +25,13 @@ define void @test_not_crash(i32 %in_a) #0 {
entry:
%cmp0 = icmp eq i32 %in_a, -1
%cmp1 = icmp ne i32 %in_a, 0
- %cond0 = and i1 %cmp0, %cmp1
+ %cond0 = or i1 %cmp0, %cmp1
br i1 %cond0, label %b0, label %b1
b0: ; preds = %entry
%cmp2 = icmp eq i32 %in_a, 0
%cmp3 = icmp ne i32 %in_a, 1
- %cond1 = or i1 %cmp2, %cmp3
+ %cond1 = and i1 %cmp2, %cmp3
br i1 %cond1, label %exit, label %b1
b1: ; preds = %entry, %b0
@@ -47,7 +47,7 @@ define void @test_not_crash2(float %a, float %b) #0 {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = fcmp ult float [[A]], 1.000000e+00
; CHECK-NEXT: [[TMP1:%.*]] = fcmp ult float [[B]], 1.000000e+00
-; CHECK-NEXT: [[TMP2:%.*]] = and i1 [[TMP0]], [[TMP1]]
+; CHECK-NEXT: [[TMP2:%.*]] = or i1 [[TMP0]], [[TMP1]]
; CHECK-NEXT: br i1 [[TMP2]], label [[BB4:%.*]], label [[BB3:%.*]]
; CHECK: bb3:
; CHECK-NEXT: br label [[BB4]]
@@ -78,7 +78,7 @@ define void @test_not_crash3(i32 %a) #0 {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[A_EQ_0:%.*]] = icmp eq i32 [[A]], 0
; CHECK-NEXT: [[A_EQ_1:%.*]] = icmp eq i32 [[A]], 1
-; CHECK-NEXT: [[TMP0:%.*]] = or i1 [[A_EQ_0]], [[A_EQ_1]]
+; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[A_EQ_0]], [[A_EQ_1]]
; CHECK-NEXT: br i1 [[TMP0]], label [[BB2:%.*]], label [[BB3:%.*]]
; CHECK: bb2:
; CHECK-NEXT: br label [[BB3]]
@@ -114,7 +114,7 @@ define void @test_then(i32 %x, i32 %y, i32 %z) {
; CHECK-NEXT: entry.x:
; CHECK-NEXT: [[CMP_X:%.*]] = icmp ne i32 [[X]], 0
; CHECK-NEXT: [[CMP_Y:%.*]] = icmp ne i32 [[Y]], 0
-; CHECK-NEXT: [[TMP0:%.*]] = or i1 [[CMP_X]], [[CMP_Y]]
+; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[CMP_X]], [[CMP_Y]]
; CHECK-NEXT: br i1 [[TMP0]], label [[IF_THEN_Y:%.*]], label [[EXIT:%.*]]
; CHECK: if.then.y:
; CHECK-NEXT: store i32 [[Z]], ptr @g, align 4
@@ -148,7 +148,7 @@ define void @test_else(i32 %x, i32 %y, i32 %z) {
; CHECK-NEXT: entry.x:
; CHECK-NEXT: [[CMP_X:%.*]] = icmp eq i32 [[X]], 0
; CHECK-NEXT: [[CMP_Y:%.*]] = icmp eq i32 [[Y]], 0
-; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[CMP_X]], [[CMP_Y]]
+; CHECK-NEXT: [[TMP0:%.*]] = or i1 [[CMP_X]], [[CMP_Y]]
; CHECK-NEXT: br i1 [[TMP0]], label [[EXIT:%.*]], label [[IF_ELSE_Y:%.*]]
; CHECK: if.else.y:
; CHECK-NEXT: store i32 [[Z]], ptr @g, align 4
@@ -182,7 +182,7 @@ define void @test_combine_and(i32 %x, i32 %y, i32 %z) {
; CHECK-NEXT: entry.x:
; CHECK-NEXT: [[CMP_X:%.*]] = icmp eq i32 [[X]], 0
; CHECK-NEXT: [[CMP_Y:%.*]] = icmp eq i32 [[Y]], 0
-; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[CMP_X]], [[CMP_Y]]
+; CHECK-NEXT: [[TMP0:%.*]] = or i1 [[CMP_X]], [[CMP_Y]]
; CHECK-NEXT: br i1 [[TMP0]], label [[EXIT:%.*]], label [[IF_THEN_Y:%.*]]
; CHECK: if.then.y:
; CHECK-NEXT: store i32 [[Z]], ptr @g, align 4
@@ -216,7 +216,7 @@ define void @test_combine_or(i32 %x, i32 %y, i32 %z) {
; CHECK-NEXT: entry.x:
; CHECK-NEXT: [[CMP_X:%.*]] = icmp ne i32 [[X]], 0
; CHECK-NEXT: [[CMP_Y:%.*]] = icmp ne i32 [[Y]], 0
-; CHECK-NEXT: [[TMP0:%.*]] = or i1 [[CMP_X]], [[CMP_Y]]
+; CHECK-NEXT: [[TMP0:%.*]] = and i1 [[CMP_X]], [[CMP_Y]]
; CHECK-NEXT: br i1 [[TMP0]], label [[IF_ELSE_Y:%.*]], label [[EXIT:%.*]]
; CHECK: if.else.y:
; CHECK-NEXT: store i32 [[Z]], ptr @g, align 4
@@ -279,7 +279,7 @@ define i1 @test_cond_multi_use(i32 %x, i32 %y, i32 %z) {
; CHECK-NEXT: [[CMP_X:%.*]] = icmp ne i32 [[X]], 0
; CHECK-NEXT: [[CMP_Y:%.*]] = icmp eq i32 [[Y]], 0
; CHECK-NEXT: [[TMP0:%.*]] = xor i1 [[CMP_Y]], true
-; CHECK-NEXT: [[TMP1:%.*]] = or i1 [[CMP_X]], [[TMP0]]
+; CHECK-NEXT: [[TMP1:%.*]] = and i1 [[CMP_X]], [[TMP0]]
; CHECK-NEXT: br i1 [[TMP1]], label [[IF_THEN_Y:%.*]], label [[EXIT:%.*]]
; CHECK: if.then.y:
; CHECK-NEXT: store i32 [[Z]], ptr @g, align 4
>From b7bd2f4e031dda341b9eb055caa2e1ba82ca9d46 Mon Sep 17 00:00:00 2001
From: mahtohappy <Happy.Kumar at windriver.com>
Date: Thu, 16 Nov 2023 06:53:24 -0800
Subject: [PATCH 2/2] Fix Logical expression used for merged conditional if
---
llvm/test/CodeGen/AMDGPU/parallelandifcollapse.ll | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/llvm/test/CodeGen/AMDGPU/parallelandifcollapse.ll b/llvm/test/CodeGen/AMDGPU/parallelandifcollapse.ll
index 115a9a888525dc2..45331b04e544863 100644
--- a/llvm/test/CodeGen/AMDGPU/parallelandifcollapse.ll
+++ b/llvm/test/CodeGen/AMDGPU/parallelandifcollapse.ll
@@ -3,9 +3,9 @@
; CFG flattening should use parallel-and mode to generate branch conditions and
; then merge if-regions with the same bodies.
;
-; CHECK: AND_INT
-; CHECK-NEXT: AND_INT
+; CHECK: OR_INT
; CHECK-NEXT: OR_INT
+; CHECK-NEXT: AND_INT
; FIXME: For some reason having the allocas here allowed the flatten cfg pass
; to do its transformation, however now that we are using local memory for
More information about the Mlir-commits
mailing list