Skip to content

Commit 11a73e7

Browse files
omarahmed1111omarahmed1111
authored
Pass -foffload-lto instead of -flto for cuda/hip kernels in clangLinkerWrapper (#16605)
ClangLinkerWrapper tool in one of its clang commands to generate ptx kernel binary from llvm bitcode kernel was using `-flto` option which should be only used for cpu code not gpu kernel code. This PR fixes that by changing that to `-foffload-lto` for cuda/hip kernels. This fixes [16413](#16413) issue.
1 parent d142923 commit 11a73e7

File tree

5 files changed

+15
-17
lines changed

5 files changed

+15
-17
lines changed

‎clang/test/Driver/linker-wrapper.c

Lines changed: 9 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -21,7 +21,7 @@ __attribute__((visibility("protected"), used)) int x;
2121
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
2222
// RUN: --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefix=NVPTX-LINK
2323

24-
// NVPTX-LINK: clang{{.*}} -o {{.*}}.img --target=nvptx64-nvidia-cuda -march=sm_70 -O2 -flto {{.*}}.o {{.*}}.o
24+
// NVPTX-LINK: clang{{.*}} -o {{.*}}.img --target=nvptx64-nvidia-cuda -march=sm_70 -O2 -foffload-lto {{.*}}.o {{.*}}.o
2525

2626
// RUN: clang-offload-packager -o %t.out \
2727
// RUN: --image=file=%t.elf.o,kind=openmp,triple=nvptx64-nvidia-cuda,arch=sm_70 \
@@ -30,7 +30,7 @@ __attribute__((visibility("protected"), used)) int x;
3030
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run --device-debug -O0 \
3131
// RUN: --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefix=NVPTX-LINK-DEBUG
3232

33-
// NVPTX-LINK-DEBUG: clang{{.*}} -o {{.*}}.img --target=nvptx64-nvidia-cuda -march=sm_70 -O2 -flto {{.*}}.o {{.*}}.o -g
33+
// NVPTX-LINK-DEBUG: clang{{.*}} -o {{.*}}.img --target=nvptx64-nvidia-cuda -march=sm_70 -O2 -foffload-lto {{.*}}.o {{.*}}.o -g
3434

3535
// RUN: clang-offload-packager -o %t.out \
3636
// RUN: --image=file=%t.elf.o,kind=openmp,triple=amdgcn-amd-amdhsa,arch=gfx908 \
@@ -39,7 +39,7 @@ __attribute__((visibility("protected"), used)) int x;
3939
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
4040
// RUN: --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefix=AMDGPU-LINK
4141

42-
// AMDGPU-LINK: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -flto -Wl,--no-undefined {{.*}}.o {{.*}}.o
42+
// AMDGPU-LINK: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o {{.*}}.o
4343

4444
// RUN: clang-offload-packager -o %t.out \
4545
// RUN: --image=file=%t.amdgpu.bc,kind=openmp,triple=amdgcn-amd-amdhsa,arch=gfx1030 \
@@ -48,7 +48,7 @@ __attribute__((visibility("protected"), used)) int x;
4848
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run --save-temps -O2 \
4949
// RUN: --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefix=AMDGPU-LTO-TEMPS
5050

51-
// AMDGPU-LTO-TEMPS: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -O2 -flto -Wl,--no-undefined {{.*}} -save-temps
51+
// AMDGPU-LTO-TEMPS: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx1030 -O2 -foffload-lto -Wl,--no-undefined {{.*}} -save-temps
5252

5353
// RUN: clang-offload-packager -o %t.out \
5454
// RUN: --image=file=%t.elf.o,kind=openmp,triple=x86_64-unknown-linux-gnu \
@@ -148,7 +148,7 @@ __attribute__((visibility("protected"), used)) int x;
148148
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run --clang-backend \
149149
// RUN: --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefix=CLANG-BACKEND
150150

151-
// CLANG-BACKEND: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -flto -Wl,--no-undefined {{.*}}.o
151+
// CLANG-BACKEND: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o
152152

153153
// RUN: clang-offload-packager -o %t.out \
154154
// RUN: --image=file=%t.elf.o,kind=openmp,triple=nvptx64-nvidia-cuda,arch=sm_70
@@ -171,8 +171,8 @@ __attribute__((visibility("protected"), used)) int x;
171171
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
172172
// RUN: --linker-path=/usr/bin/ld %t-on.o %t-off.o %t.a -o a.out 2>&1 | FileCheck %s --check-prefix=AMD-TARGET-ID
173173

174-
// AMD-TARGET-ID: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a:xnack+ -O2 -flto -Wl,--no-undefined {{.*}}.o {{.*}}.o
175-
// AMD-TARGET-ID: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a:xnack- -O2 -flto -Wl,--no-undefined {{.*}}.o {{.*}}.o
174+
// AMD-TARGET-ID: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a:xnack+ -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o {{.*}}.o
175+
// AMD-TARGET-ID: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a:xnack- -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o {{.*}}.o
176176

177177
// RUN: clang-offload-packager -o %t-lib.out \
178178
// RUN: --image=file=%t.elf.o,kind=openmp,triple=amdgcn-amd-amdhsa,arch=generic
@@ -187,8 +187,8 @@ __attribute__((visibility("protected"), used)) int x;
187187
// RUN: clang-linker-wrapper --host-triple=x86_64-unknown-linux-gnu --dry-run \
188188
// RUN: --linker-path=/usr/bin/ld %t1.o %t2.o %t.a -o a.out 2>&1 | FileCheck %s --check-prefix=ARCH-ALL
189189

190-
// ARCH-ALL: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a -O2 -flto -Wl,--no-undefined {{.*}}.o {{.*}}.o
191-
// ARCH-ALL: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -flto -Wl,--no-undefined {{.*}}.o {{.*}}.o
190+
// ARCH-ALL: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx90a -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o {{.*}}.o
191+
// ARCH-ALL: clang{{.*}} -o {{.*}}.img --target=amdgcn-amd-amdhsa -mcpu=gfx908 -O2 -foffload-lto -Wl,--no-undefined {{.*}}.o {{.*}}.o
192192

193193
// RUN: clang-offload-packager -o %t.out \
194194
// RUN: --image=file=%t.elf.o,kind=openmp,triple=x86_64-unknown-linux-gnu \

‎clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp

Lines changed: 6 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1533,12 +1533,17 @@ Expected<StringRef> clang(ArrayRef<StringRef> InputFiles, const ArgList &Args,
15331533
};
15341534

15351535
// Forward all of the `--offload-opt` and similar options to the device.
1536-
CmdArgs.push_back("-flto");
15371536
for (auto &Arg : Args.filtered(OPT_offload_opt_eq_minus, OPT_mllvm))
15381537
CmdArgs.append(
15391538
{"-Xlinker",
15401539
Args.MakeArgString("--plugin-opt=" + StringRef(Arg->getValue()))});
15411540

1541+
if (Triple.isNVPTX() || Triple.isAMDGPU()) {
1542+
CmdArgs.push_back("-foffload-lto");
1543+
} else {
1544+
CmdArgs.push_back("-flto");
1545+
}
1546+
15421547
if (!Triple.isNVPTX() && !Triple.isSPIRV())
15431548
CmdArgs.push_back("-Wl,--no-undefined");
15441549

‎sycl/test-e2e/NewOffloadDriver/multisource.cpp

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -5,8 +5,6 @@
55
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
66
//
77
//===----------------------------------------------------------------------===//
8-
// XFAIL: cuda
9-
// XFAIL-TRACKER: https://github.com/intel/llvm/issues/16413
108
// Separate kernel sources and host code sources
119
// Test with `--offload-new-driver`
1210
// RUN: %{build} --offload-new-driver -c -o %t.kernel.o -DINIT_KERNEL -DCALC_KERNEL

‎sycl/test-e2e/NewOffloadDriver/split-per-source-main.cpp

Lines changed: 0 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,6 @@
11
// RUN: %{build} -Wno-error=unused-command-line-argument -fsycl-device-code-split=per_source -I %S/Inputs -o %t.out %S/Inputs/split-per-source-second-file.cpp \
22
// RUN: --offload-new-driver -fsycl-dead-args-optimization
33
// RUN: %{run} %t.out
4-
//
5-
// XFAIL: cuda
6-
// XFAIL-TRACKER: https://github.com/intel/llvm/issues/16413
74

85
#include "Inputs/split-per-source.h"
96

‎sycl/test-e2e/NewOffloadDriver/sycl-external-with-optional-features.cpp

Lines changed: 0 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3,8 +3,6 @@
33
// RUN: %{build} -DSOURCE2 --offload-new-driver -c -o %t2.o
44
// RUN: %clangxx -Wno-error=unused-command-line-argument -fsycl %{sycl_target_opts} --offload-new-driver %t1.o %t2.o -o %t.exe
55
// RUN: %{run} %t.exe
6-
// XFAIL: cuda
7-
// XFAIL-TRACKER: https://github.com/intel/llvm/issues/16413
86

97
// XFAIL: spirv-backend
108
// XFAIL-TRACKER: CMPLRLLVM-64059

0 commit comments

Comments
 (0)