[Flang][Pass]Disable memory intrinsics expansions #171650
Merged
+46
−0
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
Created using spr 1.3.7
llvmbot
added
flang:driver
flang
Member
|
@llvm/pr-subscribers-flang-fir-hlfir Author: Alexey Bataev (alexey-bataev) ChangesPatch disables memory intrinsics expansion, enabled by default in The expansion causes massive perf regressions, up to 2x times in Full diff: https://github.com/llvm/llvm-project/pull/171650.diff 2 Files Affected:
diff --git a/flang/lib/Frontend/FrontendActions.cpp b/flang/lib/Frontend/FrontendActions.cpp
index ddf125f9bb216..d99c44d5aa1ea 100644
--- a/flang/lib/Frontend/FrontendActions.cpp
+++ b/flang/lib/Frontend/FrontendActions.cpp
@@ -42,9 +42,11 @@
#include "clang/Driver/DriverDiagnostic.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/RuntimeLibcallInfo.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Bitcode/BitcodeWriterPass.h"
+#include "llvm/CodeGen/LibcallLoweringInfo.h"
#include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
#include "llvm/IR/LLVMRemarkStreamer.h"
#include "llvm/IR/LegacyPassManager.h"
@@ -902,6 +904,9 @@ static void generateMachineCodeOrAssemblyImpl(clang::DiagnosticsEngine &diags,
llvm::TargetLibraryInfoImpl *tlii =
llvm::driver::createTLII(triple, codeGenOpts.getVecLib());
codeGenPasses.add(new llvm::TargetLibraryInfoWrapperPass(*tlii));
+ codeGenPasses.add(new llvm::RuntimeLibraryInfoWrapper(
+ triple, tm.Options.ExceptionModel, tm.Options.FloatABIType,
+ tm.Options.EABIVersion, tm.Options.MCOptions.ABIName, tm.Options.VecLib));
llvm::CodeGenFileType cgft = (act == BackendActionTy::Backend_EmitAssembly)
? llvm::CodeGenFileType::AssemblyFile
@@ -1009,6 +1014,14 @@ void CodeGenAction::runOptimizationPipeline(llvm::raw_pwrite_stream &os) {
llvm::TargetLibraryInfoImpl *tlii =
llvm::driver::createTLII(triple, opts.getVecLib());
fam.registerPass([&] { return llvm::TargetLibraryAnalysis(*tlii); });
+ mam.registerPass([&] {
+ return llvm::RuntimeLibraryAnalysis(
+ triple, targetMachine->Options.ExceptionModel,
+ targetMachine->Options.FloatABIType, targetMachine->Options.EABIVersion,
+ targetMachine->Options.MCOptions.ABIName,
+ targetMachine->Options.VecLib);
+ });
+ mam.registerPass([&] { return llvm::LibcallLoweringModuleAnalysis(); });
// Register all the basic analyses with the managers.
pb.registerModuleAnalyses(mam);
diff --git a/flang/test/Lower/memory-intrinsics-expansion.F90 b/flang/test/Lower/memory-intrinsics-expansion.F90
new file mode 100644
index 0000000000000..6d9eee429f487
--- /dev/null
+++ b/flang/test/Lower/memory-intrinsics-expansion.F90
@@ -0,0 +1,33 @@
+! REQUIRES: aarch64-registered-target
+
+! RUN: %flang_fc1 -S -O1 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O0 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -S -O2 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O2 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -S -O3 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O3 2>&1 | FileCheck %s
+! RUN: FileCheck --input-file=%t_O0 --check-prefix=CALL %s
+! RUN: FileCheck --input-file=%t_O2 --check-prefix=CALL %s
+! RUN: FileCheck --input-file=%t_O3 --check-prefix=CALL %s
+
+! CHECK: Target Library Information
+! CHECK: Runtime Library Function Analysis
+! CHECK: Library Function Lowering Analysis
+
+! CALL: {{callq|bl}} memcpy
+program memcpy_test
+ implicit none
+ integer, parameter :: n = 100
+ real :: a(n), b(n)
+ integer :: i
+
+ ! Initialize array a
+ do i = 1, n
+ a(i) = real(i)
+ end do
+
+ ! Array assignment - this should generate memcpy
+ b = a
+
+ ! Use array b to prevent optimization
+ print *, b(1), b(n)
+
+end program memcpy_test
+
|
Member
|
@llvm/pr-subscribers-flang-driver Author: Alexey Bataev (alexey-bataev) ChangesPatch disables memory intrinsics expansion, enabled by default in The expansion causes massive perf regressions, up to 2x times in Full diff: https://github.com/llvm/llvm-project/pull/171650.diff 2 Files Affected:
diff --git a/flang/lib/Frontend/FrontendActions.cpp b/flang/lib/Frontend/FrontendActions.cpp
index ddf125f9bb216..d99c44d5aa1ea 100644
--- a/flang/lib/Frontend/FrontendActions.cpp
+++ b/flang/lib/Frontend/FrontendActions.cpp
@@ -42,9 +42,11 @@
#include "clang/Driver/DriverDiagnostic.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/ADT/StringRef.h"
+#include "llvm/Analysis/RuntimeLibcallInfo.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Bitcode/BitcodeWriterPass.h"
+#include "llvm/CodeGen/LibcallLoweringInfo.h"
#include "llvm/CodeGen/MachineOptimizationRemarkEmitter.h"
#include "llvm/IR/LLVMRemarkStreamer.h"
#include "llvm/IR/LegacyPassManager.h"
@@ -902,6 +904,9 @@ static void generateMachineCodeOrAssemblyImpl(clang::DiagnosticsEngine &diags,
llvm::TargetLibraryInfoImpl *tlii =
llvm::driver::createTLII(triple, codeGenOpts.getVecLib());
codeGenPasses.add(new llvm::TargetLibraryInfoWrapperPass(*tlii));
+ codeGenPasses.add(new llvm::RuntimeLibraryInfoWrapper(
+ triple, tm.Options.ExceptionModel, tm.Options.FloatABIType,
+ tm.Options.EABIVersion, tm.Options.MCOptions.ABIName, tm.Options.VecLib));
llvm::CodeGenFileType cgft = (act == BackendActionTy::Backend_EmitAssembly)
? llvm::CodeGenFileType::AssemblyFile
@@ -1009,6 +1014,14 @@ void CodeGenAction::runOptimizationPipeline(llvm::raw_pwrite_stream &os) {
llvm::TargetLibraryInfoImpl *tlii =
llvm::driver::createTLII(triple, opts.getVecLib());
fam.registerPass([&] { return llvm::TargetLibraryAnalysis(*tlii); });
+ mam.registerPass([&] {
+ return llvm::RuntimeLibraryAnalysis(
+ triple, targetMachine->Options.ExceptionModel,
+ targetMachine->Options.FloatABIType, targetMachine->Options.EABIVersion,
+ targetMachine->Options.MCOptions.ABIName,
+ targetMachine->Options.VecLib);
+ });
+ mam.registerPass([&] { return llvm::LibcallLoweringModuleAnalysis(); });
// Register all the basic analyses with the managers.
pb.registerModuleAnalyses(mam);
diff --git a/flang/test/Lower/memory-intrinsics-expansion.F90 b/flang/test/Lower/memory-intrinsics-expansion.F90
new file mode 100644
index 0000000000000..6d9eee429f487
--- /dev/null
+++ b/flang/test/Lower/memory-intrinsics-expansion.F90
@@ -0,0 +1,33 @@
+! REQUIRES: aarch64-registered-target
+
+! RUN: %flang_fc1 -S -O1 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O0 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -S -O2 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O2 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -S -O3 %s -triple aarch64-linux-gnu -mllvm -debug-pass=Structure -o %t_O3 2>&1 | FileCheck %s
+! RUN: FileCheck --input-file=%t_O0 --check-prefix=CALL %s
+! RUN: FileCheck --input-file=%t_O2 --check-prefix=CALL %s
+! RUN: FileCheck --input-file=%t_O3 --check-prefix=CALL %s
+
+! CHECK: Target Library Information
+! CHECK: Runtime Library Function Analysis
+! CHECK: Library Function Lowering Analysis
+
+! CALL: {{callq|bl}} memcpy
+program memcpy_test
+ implicit none
+ integer, parameter :: n = 100
+ real :: a(n), b(n)
+ integer :: i
+
+ ! Initialize array a
+ do i = 1, n
+ a(i) = real(i)
+ end do
+
+ ! Array assignment - this should generate memcpy
+ b = a
+
+ ! Use array b to prevent optimization
+ print *, b(1), b(n)
+
+end program memcpy_test
+
|
alexey-bataev
deleted the
users/alexey-bataev/spr/flangpassdisable-memory-intrinsics-expansions
branch
llvm-sync bot
pushed a commit
to arm/arm-toolchain
that referenced
this pull request
Dec 10, 2025
Patch disables memory intrinsics expansion, enabled by default in llvm/llvm-project#168622. This patch does the same in clang, but not in flang. The expansion causes massive perf regressions, up to 2x times in fortran code. Reviewers: jeanPerier, vzakhari Reviewed By: vzakhari Pull Request: llvm/llvm-project#171650
Contributor
|
It doesn't build :( I suspect majority doesn't build shared libs LLVM, but sadly, we need both shared and static for a complete toolchain product... |
Member
Author
Should be fixed already |
Contributor
It builds now, thanks |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Patch disables memory intrinsics expansion, enabled by default in
#168622. This patch does the
same in clang, but not in flang.
The expansion causes massive perf regressions, up to 2x times in
fortran code.