[Flang][Driver]Implemented the support for option -f[no-]optimize-sibling-calls in Flang - #216650
Open
kaviya2510 wants to merge 1 commit into
Open
[Flang][Driver]Implemented the support for option -f[no-]optimize-sibling-calls in Flang#216650kaviya2510 wants to merge 1 commit into
kaviya2510 wants to merge 1 commit into
Conversation
kaviya2510
requested review from
NimishMishra,
Thirumalai-Shaktivel,
kiranktp,
tarunprabhu and
tblah
August 17, 2026 07:41
llvmorg-github-actions
Bot
added
clang:driver
|
@llvm/pr-subscribers-flang-fir-hlfir @llvm/pr-subscribers-clang-driver Author: Kaviya Rajendiran (kaviya2510) ChangesAdded support for -foptimize-sibling-calls and -fno-optimize-sibling-calls in Flang.
Full diff: https://github.com/llvm/llvm-project/pull/216650.diff 15 Files Affected:
diff --git a/clang/include/clang/Options/Options.td b/clang/include/clang/Options/Options.td
index adc4224dd561c..71504c806d886 100644
--- a/clang/include/clang/Options/Options.td
+++ b/clang/include/clang/Options/Options.td
@@ -4342,11 +4342,15 @@ def fopenmp_new_driver : Flag<["-"], "fopenmp-new-driver">, Flags<[HelpHidden]>,
def fno_openmp_new_driver : Flag<["-"], "fno-openmp-new-driver">,
Flags<[HelpHidden]>,
HelpText<"Don't use the new driver for OpenMP offloading.">;
-def fno_optimize_sibling_calls : Flag<["-"], "fno-optimize-sibling-calls">, Group<f_Group>,
- Visibility<[ClangOption, CC1Option]>,
- HelpText<"Disable tail call optimization, keeping the call stack accurate">,
- MarshallingInfoFlag<CodeGenOpts<"DisableTailCalls">>;
-def foptimize_sibling_calls : Flag<["-"], "foptimize-sibling-calls">, Group<f_Group>;
+def fno_optimize_sibling_calls
+ : Flag<["-"], "fno-optimize-sibling-calls">,
+ Group<f_Group>,
+ Visibility<[ClangOption, CC1Option, FlangOption, FC1Option]>,
+ HelpText<
+ "Disable tail call optimization, keeping the call stack accurate">,
+ MarshallingInfoFlag<CodeGenOpts<"DisableTailCalls">>;
+def foptimize_sibling_calls : Flag<["-"], "foptimize-sibling-calls">,
+ Group<f_Group>, Visibility<[ClangOption, FlangOption, FC1Option]>;
defm escaping_block_tail_calls : BoolFOption<"escaping-block-tail-calls",
CodeGenOpts<"NoEscapingBlockTailCalls">, DefaultFalse,
NegFlag<SetTrue, [], [ClangOption, CC1Option]>,
diff --git a/clang/lib/Driver/ToolChains/Flang.cpp b/clang/lib/Driver/ToolChains/Flang.cpp
index a48e41159f367..0c74f2b8365db 100644
--- a/clang/lib/Driver/ToolChains/Flang.cpp
+++ b/clang/lib/Driver/ToolChains/Flang.cpp
@@ -367,6 +367,10 @@ void Flang::addCodegenOptions(const ArgList &Args,
options::OPT_funroll_loops, options::OPT_fno_unroll_loops,
options::OPT_relaxed_c_loc});
+ if (Arg *A = Args.getLastArg(options::OPT_foptimize_sibling_calls,
+ options::OPT_fno_optimize_sibling_calls))
+ A->render(Args, CmdArgs);
+
const llvm::Triple &Triple = getToolChain().getEffectiveTriple();
addSeparateSectionFlags(Triple, Args, CmdArgs);
diff --git a/flang/include/flang/Frontend/CodeGenOptions.def b/flang/include/flang/Frontend/CodeGenOptions.def
index d49a7f3647eec..537a54f917d16 100644
--- a/flang/include/flang/Frontend/CodeGenOptions.def
+++ b/flang/include/flang/Frontend/CodeGenOptions.def
@@ -32,6 +32,7 @@ ENUM_CODEGENOPT(ProfileUse, llvm::driver::ProfileInstrKind, 2, llvm::driver::Pro
CODEGENOPT(InstrumentFunctions, 1, 0) ///< Set when -finstrument_functions is
///< enabled on the compile step.
+CODEGENOPT(DisableTailCalls , 1, 0) ///< Do not emit tail calls
CODEGENOPT(DisableIntegratedAS, 1, 0) ///< -no-integrated-as
diff --git a/flang/include/flang/Optimizer/Transforms/Passes.td b/flang/include/flang/Optimizer/Transforms/Passes.td
index e7bb8ae9bb9bf..e770144dfc69e 100644
--- a/flang/include/flang/Optimizer/Transforms/Passes.td
+++ b/flang/include/flang/Optimizer/Transforms/Passes.td
@@ -489,6 +489,10 @@ def FunctionAttr : Pass<"function-attr", "mlir::func::FuncOp"> {
/*default=*/"false",
"Set the use-sample-profile attribute on functions in the "
"module.">,
+ Option<"disableTailCalls", "disable-tail-calls", "bool",
+ /*default=*/"false",
+ "Set the disable-tail-calls attribute on functions to prevent "
+ "tail call optimization.">,
Option<"tuneCPU", "tune-cpu", "std::string", /*default=*/"",
"Set the tune-cpu attribute on functions in the module.">,
Option<"setNoCapture", "set-nocapture", "bool", /*default=*/"false",
diff --git a/flang/include/flang/Tools/CrossToolHelpers.h b/flang/include/flang/Tools/CrossToolHelpers.h
index fb8007637b114..3b49354a8ff96 100644
--- a/flang/include/flang/Tools/CrossToolHelpers.h
+++ b/flang/include/flang/Tools/CrossToolHelpers.h
@@ -143,6 +143,7 @@ struct MLIRToLLVMPassPipelineConfig : public FlangEPCallBacks {
InstrumentFunctionEntry = "__cyg_profile_func_enter";
InstrumentFunctionExit = "__cyg_profile_func_exit";
}
+ DisableTailCalls = opts.DisableTailCalls;
DwarfVersion = opts.DwarfVersion;
SplitDwarfFile = opts.SplitDwarfFile;
DwarfDebugFlags = opts.DwarfDebugFlags;
@@ -177,6 +178,7 @@ struct MLIRToLLVMPassPipelineConfig : public FlangEPCallBacks {
false; ///< Compiling for an OpenMP target device.
bool UseSampleProfile = false; ///< Enable sample based profiling
bool DebugInfoForProfiling = false; ///< Enable extra debugging info
+ bool DisableTailCalls = false; ///< Disable tail call optimization
bool EnableOpenMPSimd = false; ///< Enable OpenMP simd-only mode.
bool SkipConvertComplexPow = false; ///< Do not run complex pow conversion.
std::string InstrumentFunctionEntry =
diff --git a/flang/lib/Frontend/CompilerInvocation.cpp b/flang/lib/Frontend/CompilerInvocation.cpp
index b57bc4583be38..b20ecc3000cfd 100644
--- a/flang/lib/Frontend/CompilerInvocation.cpp
+++ b/flang/lib/Frontend/CompilerInvocation.cpp
@@ -376,6 +376,12 @@ static void parseCodeGenArgs(Fortran::frontend::CodeGenOptions &opts,
if (args.hasArg(clang::options::OPT_finstrument_functions))
opts.InstrumentFunctions = 1;
+ if (const llvm::opt::Arg *a = args.getLastArg(
+ clang::options::OPT_foptimize_sibling_calls,
+ clang::options::OPT_fno_optimize_sibling_calls))
+ opts.DisableTailCalls =
+ a->getOption().matches(clang::options::OPT_fno_optimize_sibling_calls);
+
// -fno-integrated-as: emit GNU Assembler compatible assembly.
if (!args.hasFlag(clang::options::OPT_fintegrated_as,
clang::options::OPT_fno_integrated_as, true))
diff --git a/flang/lib/Optimizer/Passes/Pipelines.cpp b/flang/lib/Optimizer/Passes/Pipelines.cpp
index 15a342e10fc7f..98b7574d0849a 100644
--- a/flang/lib/Optimizer/Passes/Pipelines.cpp
+++ b/flang/lib/Optimizer/Passes/Pipelines.cpp
@@ -441,7 +441,8 @@ void createDefaultFIRCodeGenPassPipeline(mlir::PassManager &pm,
config.InstrumentFunctionExit, config.NoInfsFPMath, config.NoNaNsFPMath,
config.ApproxFuncFPMath, config.NoSignedZerosFPMath, config.UnsafeFPMath,
config.Reciprocals, config.PreferVectorWidth, config.UseSampleProfile,
- /*tuneCPU=*/"", setNoCapture, setNoAlias, setReadOnly}));
+ config.DisableTailCalls, /*tuneCPU=*/"", setNoCapture, setNoAlias,
+ setReadOnly}));
if (config.EnableOpenMP) {
pm.addNestedPass<mlir::func::FuncOp>(
diff --git a/flang/lib/Optimizer/Transforms/FunctionAttr.cpp b/flang/lib/Optimizer/Transforms/FunctionAttr.cpp
index 45b32d13ad62e..1aadd16fe1cf4 100644
--- a/flang/lib/Optimizer/Transforms/FunctionAttr.cpp
+++ b/flang/lib/Optimizer/Transforms/FunctionAttr.cpp
@@ -145,5 +145,12 @@ void FunctionAttrPass::runOnOperation() {
llvmFuncOpName)),
mlir::BoolAttr::get(context, true));
+ if (disableTailCalls)
+ func->setAttr(
+ getLlvmFuncPropertyAttrName(
+ context, mlir::LLVM::LLVMFuncOp::getDisableTailCallsAttrName(
+ llvmFuncOpName)),
+ mlir::BoolAttr::get(context, true));
+
LLVM_DEBUG(llvm::dbgs() << "=== End " DEBUG_TYPE " ===\n");
}
diff --git a/flang/test/Driver/fno-optimize-sibling-calls.f90 b/flang/test/Driver/fno-optimize-sibling-calls.f90
new file mode 100644
index 0000000000000..163803fdff759
--- /dev/null
+++ b/flang/test/Driver/fno-optimize-sibling-calls.f90
@@ -0,0 +1,20 @@
+! Test -f[no-]optimize-sibling-calls driver forwarding to flang -fc1.
+
+! RUN: %flang -### %s 2>&1 | FileCheck %s --check-prefix=CHECK-OPTIMIZE-SIBLING
+! RUN: %flang -### -foptimize-sibling-calls %s 2>&1 | FileCheck %s --check-prefix=CHECK-OPTIMIZE-SIBLING-FORWARD
+! RUN: %flang -### -fno-optimize-sibling-calls %s 2>&1 | FileCheck %s --check-prefix=CHECK-OPTIMIZE-NOSIBLING
+! RUN: %flang -### -fno-optimize-sibling-calls -foptimize-sibling-calls %s 2>&1 | FileCheck %s --check-prefix=CHECK-OPTIMIZE-SIBLING-FORWARD
+! RUN: %flang -### -foptimize-sibling-calls -fno-optimize-sibling-calls %s 2>&1 | FileCheck %s --check-prefix=CHECK-OPTIMIZE-NOSIBLING
+
+! CHECK-OPTIMIZE-SIBLING: "-fc1"
+! CHECK-OPTIMIZE-SIBLING-NOT: "-fno-optimize-sibling-calls"
+! CHECK-OPTIMIZE-SIBLING-NOT: "-foptimize-sibling-calls"
+
+! CHECK-OPTIMIZE-SIBLING-FORWARD: "-fc1"{{.*}}"-foptimize-sibling-calls"
+! CHECK-OPTIMIZE-SIBLING-FORWARD-NOT: "-fno-optimize-sibling-calls"
+
+! CHECK-OPTIMIZE-NOSIBLING: "-fc1"{{.*}}"-fno-optimize-sibling-calls"
+! CHECK-OPTIMIZE-NOSIBLING-NOT: "-foptimize-sibling-calls"
+
+subroutine test
+end subroutine test
diff --git a/flang/test/Integration/disable-tail-calls.f90 b/flang/test/Integration/disable-tail-calls.f90
new file mode 100644
index 0000000000000..117512bc7739c
--- /dev/null
+++ b/flang/test/Integration/disable-tail-calls.f90
@@ -0,0 +1,12 @@
+! test -fno-optimize-sibling-calls flag disables tail call optimization
+
+! RUN: %flang_fc1 -emit-llvm -O2 -fno-optimize-sibling-calls -o - %s | FileCheck %s
+
+recursive subroutine f(n)
+ integer, intent(in) :: n
+ if (n > 0) call f(n - 1)
+end subroutine f
+
+! CHECK: define void @f_{{.*}}#[[ATTRS:[0-9]+]]
+! CHECK: call void @f_
+! CHECK: attributes #[[ATTRS]]{{.*}}"disable-tail-calls"="true"
diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
index e670e6699e57d..e578d5c22d36c 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
@@ -2106,7 +2106,8 @@ def LLVM_LLVMFuncOp : LLVM_Op<"func", [
OptionalAttr<DenseI32ArrayAttr>:$reqd_work_group_size,
OptionalAttr<I32Attr>:$intel_reqd_sub_group_size,
OptionalAttr<UWTableKindAttr>:$uwtable_kind,
- OptionalAttr<BoolAttr>:$use_sample_profile
+ OptionalAttr<BoolAttr>:$use_sample_profile,
+ OptionalAttr<BoolAttr>:$disable_tail_calls
);
let regions = (region AnyRegion:$body);
diff --git a/mlir/lib/Target/LLVMIR/ModuleImport.cpp b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
index 2ab4529ddef53..d061397d38edc 100644
--- a/mlir/lib/Target/LLVMIR/ModuleImport.cpp
+++ b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
@@ -2833,6 +2833,7 @@ static constexpr std::array kExplicitLLVMFuncOpAttributes{
StringLiteral("alwaysinline"),
StringLiteral("cold"),
StringLiteral("convergent"),
+ StringLiteral("disable-tail-calls"),
StringLiteral("fp-contract"),
StringLiteral("frame-pointer"),
StringLiteral("hot"),
@@ -3020,6 +3021,10 @@ void ModuleImport::processFunctionAttributes(llvm::Function *func,
if (func->hasFnAttribute("use-sample-profile"))
funcOp.setUseSampleProfile(true);
+ if (llvm::Attribute attr = func->getFnAttribute("disable-tail-calls");
+ attr.isStringAttribute())
+ funcOp.setDisableTailCalls(attr.getValueAsString() == "true");
+
if (llvm::Attribute attr = func->getFnAttribute("target-cpu");
attr.isStringAttribute())
funcOp.setTargetCpuAttr(StringAttr::get(context, attr.getValueAsString()));
diff --git a/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp b/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
index b87a581a5185e..c5fd545bbbaae 100644
--- a/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
+++ b/mlir/lib/Target/LLVMIR/ModuleTranslation.cpp
@@ -1706,6 +1706,10 @@ LogicalResult ModuleTranslation::convertOneFunction(LLVMFuncOp func) {
if (func.getUseSampleProfile())
llvmFunc->addFnAttr("use-sample-profile");
+ if (auto disableTailCalls = func.getDisableTailCalls())
+ llvmFunc->addFnAttr("disable-tail-calls",
+ llvm::toStringRef(*disableTailCalls));
+
if (auto attr = func.getVscaleRange())
llvmFunc->addFnAttr(llvm::Attribute::getWithVScaleRangeArgs(
getLLVMContext(), attr->getMinRange().getInt(),
diff --git a/mlir/test/Target/LLVMIR/Import/disable-tail-calls.ll b/mlir/test/Target/LLVMIR/Import/disable-tail-calls.ll
new file mode 100644
index 0000000000000..3b1b894532d99
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/Import/disable-tail-calls.ll
@@ -0,0 +1,9 @@
+; RUN: mlir-translate -import-llvm %s | FileCheck %s
+
+; CHECK-LABEL: llvm.func @disable_tail_calls()
+; CHECK-SAME: disable_tail_calls = true
+define void @disable_tail_calls() #0 {
+ ret void
+}
+
+attributes #0 = { "disable-tail-calls"="true" }
diff --git a/mlir/test/Target/LLVMIR/disable-tail-calls.mlir b/mlir/test/Target/LLVMIR/disable-tail-calls.mlir
new file mode 100644
index 0000000000000..c28f248443c42
--- /dev/null
+++ b/mlir/test/Target/LLVMIR/disable-tail-calls.mlir
@@ -0,0 +1,7 @@
+// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
+
+// CHECK: define void @disable_tail_calls() #[[ATTRS:.*]] {
+// CHECK: attributes #[[ATTRS]] = { "disable-tail-calls"="true" }
+llvm.func @disable_tail_calls() attributes {disable_tail_calls = true} {
+ llvm.return
+}
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
…ing-calls in Flang
kaviya2510
force-pushed
the
optimize-sibling-calls
branch
from
August 17, 2026 08:47
45f4e1c to
aac8799
Compare
tblah
reviewed
Aug 17, 2026
|
|
||
| CODEGENOPT(InstrumentFunctions, 1, 0) ///< Set when -finstrument_functions is | ||
| ///< enabled on the compile step. | ||
| CODEGENOPT(DisableTailCalls , 1, 0) ///< Do not emit tail calls |
|
|
||
| if (llvm::Attribute attr = func->getFnAttribute("disable-tail-calls"); | ||
| attr.isStringAttribute()) | ||
| funcOp.setDisableTailCalls(attr.getValueAsString() == "true"); |
Contributor
There was a problem hiding this comment.
nit: maybe this should generate an error for values which are neither "true" nor "false"
vzakhari
reviewed
Aug 17, 2026
Comment on lines
+379
to
+384
| if (const llvm::opt::Arg *a = | ||
| args.getLastArg(clang::options::OPT_foptimize_sibling_calls, | ||
| clang::options::OPT_fno_optimize_sibling_calls)) | ||
| opts.DisableTailCalls = | ||
| a->getOption().matches(clang::options::OPT_fno_optimize_sibling_calls); | ||
|
|
Contributor
There was a problem hiding this comment.
Can you please use the established code pattern that uses arg.hasFlag(true-flag, false-flag, /*Default=*/true)?
| @@ -0,0 +1,12 @@ | |||
| ! test -fno-optimize-sibling-calls flag disables tail call optimization | |||
|
|
|||
| ! RUN: %flang_fc1 -emit-llvm -O2 -fno-optimize-sibling-calls -o - %s | FileCheck %s | |||
Contributor
There was a problem hiding this comment.
Can you please add more cases, e.g. no option, combinations of true/false flags?
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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.
Added support for -foptimize-sibling-calls and -fno-optimize-sibling-calls in Flang.
-f[no-]optimize-sibling-callsto flang -fc1.-fno-optimize-sibling-callsflag is passed, it sets boolenDisableTailCallsand this results in adding the LLVM IR attribute"disable-tail-calls"="true"which disables sibling call optimization.