Skip to content

[LLVM][VecLib] Refactor LIBMVEC integration to be target neutral. #138262

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
May 7, 2025
Merged

[LLVM][VecLib] Refactor LIBMVEC integration to be target neutral. #138262

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5da6f3e
Add test showing x86 functions when targeting AArch64.
paulwalker-arm May 2, 2025
c1e632b
[LLVM][VecLib] Refactor LIBMVEC integration to be target neutral.
paulwalker-arm May 2, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion clang/lib/Driver/ToolChains/CommonArgs.cpp
View file
Original file line number Diff line number Diff line change
@@ -934,7 +934,7 @@ void tools::addLTOOptions(const ToolChain &ToolChain, const ArgList &Args,
std::optional<StringRef> OptVal =
llvm::StringSwitch<std::optional<StringRef>>(ArgVecLib->getValue())
.Case("Accelerate", "Accelerate")
.Case("libmvec", "LIBMVEC-X86")
.Case("libmvec", "LIBMVEC")
.Case("MASSV", "MASSV")
.Case("SVML", "SVML")
.Case("SLEEF", "sleefgnuabi")
2 changes: 1 addition & 1 deletion clang/test/Driver/fveclib.c
View file
Original file line number Diff line number Diff line change
@@ -38,7 +38,7 @@
/* Verify that the correct vector library is passed to LTO flags. */

// RUN: %clang -### --target=x86_64-unknown-linux-gnu -fveclib=libmvec -flto %s 2>&1 | FileCheck --check-prefix=CHECK-LTO-LIBMVEC %s
// CHECK-LTO-LIBMVEC: "-plugin-opt=-vector-library=LIBMVEC-X86"
// CHECK-LTO-LIBMVEC: "-plugin-opt=-vector-library=LIBMVEC"

// RUN: %clang -### --target=powerpc64-unknown-linux-gnu -fveclib=MASSV -flto %s 2>&1 | FileCheck --check-prefix=CHECK-LTO-MASSV %s
// CHECK-LTO-MASSV: "-plugin-opt=-vector-library=MASSV"
2 changes: 1 addition & 1 deletion llvm/include/llvm/Analysis/TargetLibraryInfo.h
View file
Original file line number Diff line number Diff line change
@@ -127,7 +127,7 @@ class TargetLibraryInfoImpl {
NoLibrary, // Don't use any vector library.
Accelerate, // Use Accelerate framework.
DarwinLibSystemM, // Use Darwin's libsystem_m.
LIBMVEC_X86, // GLIBC Vector Math library.
LIBMVEC, // GLIBC Vector Math library.
MASSV, // IBM MASS vector library.
SVML, // Intel short vector math library.
SLEEFGNUABI, // SLEEF - SIMD Library for Evaluating Elementary Functions.
13 changes: 10 additions & 3 deletions llvm/lib/Analysis/TargetLibraryInfo.cpp
View file
Original file line number Diff line number Diff line change
@@ -29,7 +29,7 @@ static cl::opt<TargetLibraryInfoImpl::VectorLibrary> ClVectorLibrary(
"Accelerate framework"),
clEnumValN(TargetLibraryInfoImpl::DarwinLibSystemM,
"Darwin_libsystem_m", "Darwin libsystem_m"),
clEnumValN(TargetLibraryInfoImpl::LIBMVEC_X86, "LIBMVEC-X86",
clEnumValN(TargetLibraryInfoImpl::LIBMVEC, "LIBMVEC",
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this needs changes in clang too as "LIBMVEC" is currently mapped to "LIBMVEC-X86":

llvm-project/clang/lib/Driver/ToolChains/CommonArgs.cpp

Lines 932 to 937 in 7d01b85

// Map the vector library names from clang front-end to opt front-end. The
// values are taken from the TargetLibraryInfo class command line options.
std::optional<StringRef> OptVal =
llvm::StringSwitch<std::optional<StringRef>>(ArgVecLib->getValue())
.Case("Accelerate", "Accelerate")
.Case("LIBMVEC", "LIBMVEC-X86")

Copy link
Member

@MacDue MacDue May 2, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm. There seems to be a bug in clang's checks too, as it's checking for "LIBMVEC-X86" before this remapping, it should actually output err_drv_unsupported_opt_for_target for LIBMVEC on AArch64. I think once you rename all uses of LIBMVEC-X86 to LIBMVEC in clang that error will start working.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, I've just noticed the same.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I've raised #138288 to fix the driver bug. I'll update this PR when that lands and request a re-review.

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done.

"GLIBC Vector Math library"),
clEnumValN(TargetLibraryInfoImpl::MASSV, "MASSV",
"IBM MASS vector library"),
@@ -1360,8 +1360,15 @@ void TargetLibraryInfoImpl::addVectorizableFunctionsFromVecLib(
addVectorizableFunctions(VecFuncs_DarwinLibSystemM);
break;
}
case LIBMVEC_X86: {
addVectorizableFunctions(VecFuncs_LIBMVEC_X86);
case LIBMVEC: {
switch (TargetTriple.getArch()) {
default:
break;
case llvm::Triple::x86:
case llvm::Triple::x86_64:
addVectorizableFunctions(VecFuncs_LIBMVEC_X86);
break;
}
Comment on lines +1364 to +1371
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
switch (TargetTriple.getArch()) {
default:
break;
case llvm::Triple::x86:
case llvm::Triple::x86_64:
addVectorizableFunctions(VecFuncs_LIBMVEC_X86);
break;
}
if (TargetTriple.isX86())
addVectorizableFunctions(VecFuncs_LIBMVEC_X86);

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This PR is prep work for adding support for the AArch64 build of libmvec. We'll want this layout then anyway so I'm going to reject this suggestion.

break;
}
case MASSV: {
2 changes: 1 addition & 1 deletion llvm/lib/Frontend/Driver/CodeGenOptions.cpp
View file
Original file line number Diff line number Diff line change
@@ -23,7 +23,7 @@ TargetLibraryInfoImpl *createTLII(const llvm::Triple &TargetTriple,
TargetTriple);
break;
case VectorLibrary::LIBMVEC:
TLII->addVectorizableFunctionsFromVecLib(TargetLibraryInfoImpl::LIBMVEC_X86,
TLII->addVectorizableFunctionsFromVecLib(TargetLibraryInfoImpl::LIBMVEC,
TargetTriple);
break;
case VectorLibrary::MASSV:
2 changes: 1 addition & 1 deletion llvm/test/CodeGen/Generic/replace-intrinsics-with-veclib.ll
View file
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-attributes
; RUN: opt -vector-library=SVML -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,SVML
; RUN: opt -vector-library=AMDLIBM -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,AMDLIBM
; RUN: opt -vector-library=LIBMVEC-X86 -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,LIBMVEC-X86
; RUN: opt -vector-library=LIBMVEC -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,LIBMVEC-X86
; RUN: opt -vector-library=MASSV -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,MASSV
; RUN: opt -vector-library=Accelerate -replace-with-veclib -S < %s | FileCheck %s --check-prefixes=COMMON,ACCELERATE

2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-VF2-VF8.ll
View file
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s
; RUN: opt -vector-library=LIBMVEC -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s

target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls-finite.ll
View file
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s
; RUN: opt -vector-library=LIBMVEC -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"

2 changes: 1 addition & 1 deletion llvm/test/Transforms/LoopVectorize/X86/libm-vector-calls.ll
View file
Original file line number Diff line number Diff line change
@@ -1,4 +1,4 @@
; RUN: opt -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s
; RUN: opt -vector-library=LIBMVEC -passes=inject-tli-mappings,loop-vectorize -S < %s | FileCheck %s

target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-unknown-linux-gnu"
8 changes: 7 additions & 1 deletion llvm/test/Transforms/Util/add-TLI-mappings.ll
View file
Original file line number Diff line number Diff line change
@@ -1,12 +1,15 @@
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -vector-library=SVML -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,SVML
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -vector-library=AMDLIBM -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,AMDLIBM
; RUN: opt -mtriple=powerpc64-unknown-linux-gnu -vector-library=MASSV -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,MASSV
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -vector-library=LIBMVEC-X86 -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,LIBMVEC-X86
; RUN: opt -mtriple=aarch64-unknown-linux-gnu -vector-library=LIBMVEC -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=LIBMVEC-AARCH64
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -vector-library=LIBMVEC -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,LIBMVEC-X86
; RUN: opt -mtriple=x86_64-unknown-linux-gnu -vector-library=Accelerate -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,ACCELERATE
; RUN: opt -mtriple=aarch64-unknown-linux-gnu -vector-library=sleefgnuabi -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,SLEEFGNUABI
; RUN: opt -mtriple=riscv64-unknown-linux-gnu -vector-library=sleefgnuabi -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,SLEEFGNUABI_RISCV
; RUN: opt -mtriple=aarch64-unknown-linux-gnu -vector-library=ArmPL -passes=inject-tli-mappings -S < %s | FileCheck %s --check-prefixes=COMMON,ARMPL

; LIBMVEC-AARCH64-NOT: llvm.compiler.used

; COMMON-LABEL: @llvm.compiler.used = appending global
; SVML-SAME: [6 x ptr] [
; SVML-SAME: ptr @__svml_sin2,
@@ -193,6 +196,9 @@ declare float @llvm.log10.f32(float) #0
; MASSV: declare <2 x double> @__sind2(<2 x double>)
; MASSV: declare <4 x float> @__log10f4(<4 x float>)

; LIBMVEC-AARCH64-NOT: declare <2 x double> @_ZGVbN2v_sin(<2 x double>)
; LIBMVEC-AARCH64-NOT: declare <4 x double> @_ZGVdN4v_sin(<4 x double>)

; LIBMVEC-X86: declare <2 x double> @_ZGVbN2v_sin(<2 x double>)
; LIBMVEC-X86: declare <4 x double> @_ZGVdN4v_sin(<4 x double>)