[profcheck] Add unknown branch weights to expanded cmpxchg loop. #165841
+18
−4
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
|
@llvm/pr-subscribers-backend-powerpc @llvm/pr-subscribers-backend-risc-v Author: Jin Huang (jinhuang1102) ChangesThe AtomicExpandPass is responsible for lowering high-level atomic operations (like Given that we cannot empirically prove the precision branch weights, It uses the Patch is 2.90 MiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/165841.diff 23 Files Affected:
diff --git a/llvm/lib/CodeGen/AtomicExpandPass.cpp b/llvm/lib/CodeGen/AtomicExpandPass.cpp
index 53f1cfe24a68d..dffb69425bb31 100644
--- a/llvm/lib/CodeGen/AtomicExpandPass.cpp
+++ b/llvm/lib/CodeGen/AtomicExpandPass.cpp
@@ -38,6 +38,7 @@
#include "llvm/IR/MDBuilder.h"
#include "llvm/IR/MemoryModelRelaxationAnnotations.h"
#include "llvm/IR/Module.h"
+#include "llvm/IR/ProfDataUtils.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
@@ -1259,8 +1260,7 @@ Value *AtomicExpandImpl::insertRMWLLSCLoop(
BasicBlock *BB = Builder.GetInsertBlock();
Function *F = BB->getParent();
- assert(AddrAlign >=
- F->getDataLayout().getTypeStoreSize(ResultTy) &&
+ assert(AddrAlign >= F->getDataLayout().getTypeStoreSize(ResultTy) &&
"Expected at least natural alignment at this point.");
// Given: atomicrmw some_op iN* %addr, iN %incr ordering
@@ -1680,7 +1680,12 @@ Value *AtomicExpandImpl::insertRMWCmpXchgLoop(
Loaded->addIncoming(NewLoaded, LoopBB);
- Builder.CreateCondBr(Success, ExitBB, LoopBB);
+ Instruction *CondBr = Builder.CreateCondBr(Success, ExitBB, LoopBB);
+
+ // Atomic RMW expands to a cmpxchg loop, Since precise branch weights
+ // cannot be easily determined here, we mark the branch as "unknown" (50/50)
+ // to prevent misleading optimizations.
+ setExplicitlyUnknownBranchWeightsIfProfiled(*CondBr, *F, DEBUG_TYPE);
Builder.SetInsertPoint(ExitBB, ExitBB->begin());
return NewLoaded;
diff --git a/llvm/test/Transforms/AtomicExpand/AArch64/atomicrmw-fp.ll b/llvm/test/Transforms/AtomicExpand/AArch64/atomicrmw-fp.ll
index 8ffacb9bdd5f6..fe42a5439857c 100644
--- a/llvm/test/Transforms/AtomicExpand/AArch64/atomicrmw-fp.ll
+++ b/llvm/test/Transforms/AtomicExpand/AArch64/atomicrmw-fp.ll
@@ -14,7 +14,7 @@ define float @test_atomicrmw_fadd_f32(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
; CHECK-NEXT: [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1:![0-9]+]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP5]]
;
@@ -35,7 +35,7 @@ define float @test_atomicrmw_fsub_f32(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP4]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP4]], 0
; CHECK-NEXT: [[TMP5]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP5]]
;
@@ -56,7 +56,7 @@ define float @atomicrmw_fmin_float(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP6]]
;
@@ -77,7 +77,7 @@ define float @atomicrmw_fmax_float(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP6]]
;
@@ -98,7 +98,7 @@ define double @atomicrmw_fmin_double(ptr %ptr, double %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret double [[TMP6]]
;
@@ -119,7 +119,7 @@ define double @atomicrmw_fmax_double(ptr %ptr, double %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret double [[TMP6]]
;
@@ -140,7 +140,7 @@ define float @atomicrmw_fminimum_float(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP6]]
;
@@ -161,7 +161,7 @@ define float @atomicrmw_fmaximum_float(ptr %ptr, float %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to float
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret float [[TMP6]]
;
@@ -182,7 +182,7 @@ define double @atomicrmw_fminimum_double(ptr %ptr, double %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret double [[TMP6]]
;
@@ -203,7 +203,7 @@ define double @atomicrmw_fmaximum_double(ptr %ptr, double %value) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i64, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i64, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i64 [[NEWLOADED]] to double
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret double [[TMP6]]
;
@@ -224,7 +224,7 @@ define bfloat @atomicrmw_fmaximum_bfloat(ptr %ptr, bfloat %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i16, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i16, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i16 [[NEWLOADED]] to bfloat
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret bfloat [[TMP6]]
;
@@ -245,7 +245,7 @@ define half @atomicrmw_fmaximum_half(ptr %ptr, half %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i16, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i16, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i16 [[NEWLOADED]] to half
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret half [[TMP6]]
;
@@ -266,7 +266,7 @@ define <2 x half> @atomicrmw_fmaximum_2_x_half(ptr %ptr, <2 x half> %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to <2 x half>
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret <2 x half> [[TMP6]]
;
@@ -287,7 +287,7 @@ define bfloat @atomicrmw_fminimum_bfloat(ptr %ptr, bfloat %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i16, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i16, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i16 [[NEWLOADED]] to bfloat
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret bfloat [[TMP6]]
;
@@ -308,7 +308,7 @@ define half @atomicrmw_fminimum_half(ptr %ptr, half %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i16, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i16, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i16 [[NEWLOADED]] to half
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret half [[TMP6]]
;
@@ -329,7 +329,7 @@ define <2 x half> @atomicrmw_fminimum_2_x_half(ptr %ptr, <2 x half> %val) {
; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP5]], 1
; CHECK-NEXT: [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP5]], 0
; CHECK-NEXT: [[TMP6]] = bitcast i32 [[NEWLOADED]] to <2 x half>
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret <2 x half> [[TMP6]]
;
diff --git a/llvm/test/Transforms/AtomicExpand/AArch64/pcsections.ll b/llvm/test/Transforms/AtomicExpand/AArch64/pcsections.ll
index c5c890559152d..5c60f21c207b9 100644
--- a/llvm/test/Transforms/AtomicExpand/AArch64/pcsections.ll
+++ b/llvm/test/Transforms/AtomicExpand/AArch64/pcsections.ll
@@ -4,7 +4,7 @@
define i8 @atomic8_load_unordered(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_load_unordered(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] unordered, align 1, !pcsections [[META0:![0-9]+]]
+; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] unordered, align 1, !pcsections [[META1:![0-9]+]]
; CHECK-NEXT: ret i8 [[TMP0]]
;
entry:
@@ -15,7 +15,7 @@ entry:
define i8 @atomic8_load_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_load_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] monotonic, align 1, !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] monotonic, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret i8 [[TMP0]]
;
entry:
@@ -26,7 +26,7 @@ entry:
define i8 @atomic8_load_acquire(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_load_acquire(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] acquire, align 1, !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] acquire, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret i8 [[TMP0]]
;
entry:
@@ -37,7 +37,7 @@ entry:
define i8 @atomic8_load_seq_cst(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_load_seq_cst(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] seq_cst, align 1, !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load atomic i8, ptr [[A:%.*]] seq_cst, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret i8 [[TMP0]]
;
entry:
@@ -48,7 +48,7 @@ entry:
define void @atomic8_store_unordered(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_store_unordered(
; CHECK-NEXT: entry:
-; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] unordered, align 1, !pcsections [[META0]]
+; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] unordered, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret void
;
entry:
@@ -59,7 +59,7 @@ entry:
define void @atomic8_store_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_store_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] monotonic, align 1, !pcsections [[META0]]
+; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] monotonic, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret void
;
entry:
@@ -70,7 +70,7 @@ entry:
define void @atomic8_store_release(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_store_release(
; CHECK-NEXT: entry:
-; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] release, align 1, !pcsections [[META0]]
+; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] release, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret void
;
entry:
@@ -81,7 +81,7 @@ entry:
define void @atomic8_store_seq_cst(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_store_seq_cst(
; CHECK-NEXT: entry:
-; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] seq_cst, align 1, !pcsections [[META0]]
+; CHECK-NEXT: store atomic i8 0, ptr [[A:%.*]] seq_cst, align 1, !pcsections [[META1]]
; CHECK-NEXT: ret void
;
entry:
@@ -92,14 +92,14 @@ entry:
define void @atomic8_xchg_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_xchg_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META0]]
-; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META1]]
+; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META1]]
; CHECK: atomicrmw.start:
-; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META0]]
-; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 0 monotonic monotonic, align 1, !pcsections [[META0]]
-; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META0]]
-; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META0]]
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !pcsections [[META0]]
+; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META1]]
+; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 0 monotonic monotonic, align 1, !pcsections [[META1]]
+; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META1]]
+; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META1]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF2:![0-9]+]], !pcsections [[META1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret void
;
@@ -111,14 +111,14 @@ entry:
define void @atomic8_add_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_add_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META0]]
-; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META1]]
+; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META1]]
; CHECK: atomicrmw.start:
-; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META0]]
-; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 [[LOADED]] monotonic monotonic, align 1, !pcsections [[META0]]
-; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META0]]
-; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META0]]
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !pcsections [[META0]]
+; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META1]]
+; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 [[LOADED]] monotonic monotonic, align 1, !pcsections [[META1]]
+; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META1]]
+; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META1]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF2]], !pcsections [[META1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret void
;
@@ -130,14 +130,14 @@ entry:
define void @atomic8_sub_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_sub_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META0]]
-; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META1]]
+; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META1]]
; CHECK: atomicrmw.start:
-; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META0]]
-; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 [[LOADED]] monotonic monotonic, align 1, !pcsections [[META0]]
-; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META0]]
-; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META0]]
-; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !pcsections [[META0]]
+; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META1]]
+; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 [[LOADED]] monotonic monotonic, align 1, !pcsections [[META1]]
+; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META1]]
+; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcsections [[META1]]
+; CHECK-NEXT: br i1 [[SUCCESS]], label [[ATOMICRMW_END:%.*]], label [[ATOMICRMW_START]], !prof [[PROF2]], !pcsections [[META1]]
; CHECK: atomicrmw.end:
; CHECK-NEXT: ret void
;
@@ -149,14 +149,14 @@ entry:
define void @atomic8_and_monotonic(ptr %a) nounwind uwtable {
; CHECK-LABEL: @atomic8_and_monotonic(
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META0]]
-; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META0]]
+; CHECK-NEXT: [[TMP0:%.*]] = load i8, ptr [[A:%.*]], align 1, !pcsections [[META1]]
+; CHECK-NEXT: br label [[ATOMICRMW_START:%.*]], !pcsections [[META1]]
; CHECK: atomicrmw.start:
-; CHECK-NEXT: [[LOADED:%.*]] = phi i8 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[NEWLOADED:%.*]], [[ATOMICRMW_START]] ], !pcsections [[META0]]
-; CHECK-NEXT: [[TMP1:%.*]] = cmpxchg ptr [[A]], i8 [[LOADED]], i8 0 monotonic monotonic, align 1, !pcsections [[META0]]
-; CHECK-NEXT: [[SUCCESS:%.*]] = extractvalue { i8, i1 } [[TMP1]], 1, !pcsections [[META0]]
-; CHECK-NEXT: [[NEWLOADED]] = extractvalue { i8, i1 } [[TMP1]], 0, !pcse...
[truncated]
|
alanzhao1
approved these changes
Oct 31, 2025
arsenm
requested changes
Oct 31, 2025
|
|
||
| assert(AddrAlign >= | ||
| F->getDataLayout().getTypeStoreSize(ResultTy) && | ||
| assert(AddrAlign >= F->getDataLayout().getTypeStoreSize(ResultTy) && |
Contributor
Author
There was a problem hiding this comment.
restore the changes. Thx
Member
see https://discourse.llvm.org/t/rfc-profile-information-propagation-unittesting/73595 |
jinhuang1102
force-pushed
the
fixup/prof-missing-atomic-expand
branch
2 times, most recently
from
edf9bfe to
8140097
Compare
Contributor
boomanaiden154
left a comment
boomanaiden154
left a comment
There was a problem hiding this comment.
This should probably remove the test case(s) it fixed from the profcheck-xfail.txt list?
jinhuang1102
force-pushed
the
fixup/prof-missing-atomic-expand
branch
from
8140097 to
015201c
Compare
Contributor
Author
Thanks for letting me know! Yes, I will remove all of these expected failed tests after the PR is merged! |
jinhuang1102
force-pushed
the
fixup/prof-missing-atomic-expand
branch
from
015201c to
0366285
Compare
arsenm
approved these changes
Nov 5, 2025
Contributor
Is there a reason you can't update it as part of this PR? |
jinhuang1102
added a commit
that referenced
this pull request
Nov 5, 2025
As a follow-up to PR#165841, this change addresses `prof_md` metadata loss in AtomicExpandPass when lowering `atomicrmw xchg` to a Load-Linked/Store-Exclusive (LL/SC) loop. This path is distinct from the LSE path addressed previously: PR #165841 (and its tests) used `-mtriple=aarch64-linux-gnu`, which targets a modern **ARMv8.1+** architecture. This architecture supports **Large System Extensions (LSE)**, allowing `atomicrmw` to be lowered directly to a more efficient hardware instruction. This PR (and its tests) uses `-mtriple=aarch64--` or `-mtriple=armv8-linux-gnueabihf`. This indicates an `ARMv8.0 or lower architecture that does not support LSE`. On these targets, the pass must fall back to synthesizing a manual LL/SC loop using the `ldaxr/stxr` instruction pair. Similar to previous issue, the new conditional branch was failin to inherit the `prof_md` metadata. Theis PR correctly fix the branch weights to the newly created branch within the LL/SC loop, ensuring profile information is preserved. Co-authored-by: Jin Huang <[email protected]>
jinhuang1102
force-pushed
the
fixup/prof-missing-atomic-expand
branch
from
0cb1083 to
ed77198
Compare
Contributor
Author
Because not all 45 failed tests are fixed in this single PR. I will update them after all PR merged. |
jinhuang1102
commented
Nov 5, 2025
|
|
||
| !0 = !{!"function_entry_count", i64 1000} | ||
| ;. | ||
| ; CHECK: attributes #[[ATTR0:[0-9]+]] = { nocallback nocreateundeforpoison nofree nosync nounwind speculatable willreturn memory(none) } |
Contributor
Author
There was a problem hiding this comment.
Manually add nocreateundeforpoison to pass the upstream test checks.
Contributor
There was a problem hiding this comment.
I'm guessing something else was changed to infer the new attribute and your local opt binary was too old.
Contributor
Can you not add the ones that do get fixed? |
Contributor
Author
Sure! Added the fixed tests in the description. |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/159/builds/34516 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/177/builds/23762 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/190/builds/30375 Here is the relevant piece of the build log for the reference |
tgymnich
reviewed
Nov 5, 2025
| // Atomic RMW expands to a cmpxchg loop, Since precise branch weights | ||
| // cannot be easily determined here, we mark the branch as "unknown" (50/50) | ||
| // to prevent misleading optimizations. | ||
| setExplicitlyUnknownBranchWeightsIfProfiled(*CondBr, *F, DEBUG_TYPE); |
Member
There was a problem hiding this comment.
Suggested change
| setExplicitlyUnknownBranchWeightsIfProfiled(*CondBr, *F, DEBUG_TYPE); | |
| setExplicitlyUnknownBranchWeightsIfProfiled(*CondBr, DEBUG_TYPE, *F); |
The argument order seems wrong. (was changed 2h ago #166032)
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/204/builds/27054 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/203/builds/28242 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/205/builds/27033 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/2/builds/37654 Here is the relevant piece of the build log for the reference |
Contributor
|
#165841 (review) (earlier on this PR) links to what fixed the build failures. |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/144/builds/39438 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/137/builds/28490 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/33/builds/25998 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/185/builds/28312 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/129/builds/32530 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/160/builds/27787 Here is the relevant piece of the build log for the reference |
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/180/builds/27927 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
9 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The AtomicExpandPass is responsible for lowering high-level atomic operations (like
atomicrmw fadd) that are unsupported by the target hardware into a cmpxchg retry loop.Given that we cannot empirically prove the precision branch weights, It uses the
setExplicitlyUnknownBranchWeightsIfProfiledfunction to explicitly add "unknown" (50/50) branch weights to this branch.This PR includes fies for the following tests: