Skip to content

[CUDA] Use --image3 to construct fat binary #151760

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 1 commit into from
Aug 1, 2025
+30 −36
Merged

[CUDA] Use --image3 to construct fat binary #151760

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
20 changes: 7 additions & 13 deletions clang/lib/Driver/ToolChains/Cuda.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -549,22 +549,16 @@ void NVPTX::FatBinary::ConstructJob(Compilation &C, const JobAction &JA,
auto *A = II.getAction();
assert(A->getInputs().size() == 1 &&
"Device offload action is expected to have a single input");
const char *gpu_arch_str = A->getOffloadingArch();
assert(gpu_arch_str &&
StringRef GpuArch = A->getOffloadingArch();
assert(!GpuArch.empty() &&
"Device action expected to have associated a GPU architecture!");
OffloadArch gpu_arch = StringToOffloadArch(gpu_arch_str);

if (II.getType() == types::TY_PP_Asm &&
!shouldIncludePTX(Args, gpu_arch_str))
if (II.getType() == types::TY_PP_Asm && !shouldIncludePTX(Args, GpuArch))
continue;
// We need to pass an Arch of the form "sm_XX" for cubin files and
// "compute_XX" for ptx.
const char *Arch = (II.getType() == types::TY_PP_Asm)
? OffloadArchToVirtualArchString(gpu_arch)
: gpu_arch_str;
CmdArgs.push_back(
Args.MakeArgString(llvm::Twine("--image=profile=") + Arch +
",file=" + getToolChain().getInputFilename(II)));
StringRef Kind = (II.getType() == types::TY_PP_Asm) ? "ptx" : "elf";
CmdArgs.push_back(Args.MakeArgString(
"--image3=kind=" + Kind + ",sm=" + GpuArch.drop_front(3) +
",file=" + getToolChain().getInputFilename(II)));
}

for (const auto &A : Args.getAllArgValues(options::OPT_Xcuda_fatbinary))
Expand Down
26 changes: 13 additions & 13 deletions clang/test/Driver/cuda-arch-translation.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -68,19 +68,19 @@

// HIP: clang-offload-bundler

// SM20:--image=profile=sm_20{{.*}}
// SM21:--image=profile=sm_21{{.*}}
// SM30:--image=profile=sm_30{{.*}}
// SM32:--image=profile=sm_32{{.*}}
// SM35:--image=profile=sm_35{{.*}}
// SM37:--image=profile=sm_37{{.*}}
// SM50:--image=profile=sm_50{{.*}}
// SM52:--image=profile=sm_52{{.*}}
// SM53:--image=profile=sm_53{{.*}}
// SM60:--image=profile=sm_60{{.*}}
// SM61:--image=profile=sm_61{{.*}}
// SM62:--image=profile=sm_62{{.*}}
// SM70:--image=profile=sm_70{{.*}}
// SM20:--image3=kind=elf,sm=20{{.*}}
// SM21:--image3=kind=elf,sm=21{{.*}}
// SM30:--image3=kind=elf,sm=30{{.*}}
// SM32:--image3=kind=elf,sm=32{{.*}}
// SM35:--image3=kind=elf,sm=35{{.*}}
// SM37:--image3=kind=elf,sm=37{{.*}}
// SM50:--image3=kind=elf,sm=50{{.*}}
// SM52:--image3=kind=elf,sm=52{{.*}}
// SM53:--image3=kind=elf,sm=53{{.*}}
// SM60:--image3=kind=elf,sm=60{{.*}}
// SM61:--image3=kind=elf,sm=61{{.*}}
// SM62:--image3=kind=elf,sm=62{{.*}}
// SM70:--image3=kind=elf,sm=70{{.*}}
// GFX600:-targets=host-x86_64-unknown-linux-gnu,hipv4-amdgcn-amd-amdhsa--gfx600
// GFX601:-targets=host-x86_64-unknown-linux-gnu,hipv4-amdgcn-amd-amdhsa--gfx601
// GFX602:-targets=host-x86_64-unknown-linux-gnu,hipv4-amdgcn-amd-amdhsa--gfx602
Expand Down
20 changes: 10 additions & 10 deletions clang/test/Driver/cuda-options.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -243,10 +243,10 @@

// INCLUDES-DEVICE:fatbinary
// INCLUDES-DEVICE-DAG: "--create" "[[FATBINARY:[^"]*]]"
// INCLUDES-DEVICE-DAG: "--image=profile=sm_{{[0-9]+}},file=[[CUBINFILE]]"
// INCLUDES-DEVICE-DAG: "--image=profile=compute_{{[0-9]+}},file=[[PTXFILE]]"
// INCLUDES-DEVICE2-DAG: "--image=profile=sm_{{[0-9]+}},file=[[CUBINFILE2]]"
// INCLUDES-DEVICE2-DAG: "--image=profile=compute_{{[0-9]+}},file=[[PTXFILE2]]"
// INCLUDES-DEVICE-DAG: "--image3=kind=elf,sm={{[0-9]+}},file=[[CUBINFILE]]"
// INCLUDES-DEVICE-DAG: "--image3=kind=ptx,sm={{[0-9]+}},file=[[PTXFILE]]"
// INCLUDES-DEVICE2-DAG: "--image3=kind=elf,sm={{[0-9]+}},file=[[CUBINFILE2]]"
// INCLUDES-DEVICE2-DAG: "--image3=kind=ptx,sm={{[0-9]+}},file=[[PTXFILE2]]"

// Match host-side preprocessor job with -save-temps.
// HOST-SAVE: "-cc1" "-triple" "x86_64-unknown-linux-gnu"
Expand Down Expand Up @@ -288,9 +288,9 @@

// FATBIN-COMMON:fatbinary
// FATBIN-COMMON: "--create" "[[FATBINARY:[^"]*]]"
// FATBIN-COMMON: "--image=profile=sm_52,file=
// PTX-SM52: "--image=profile=compute_52,file=
// NOPTX-SM52-NOT: "--image=profile=compute_52,file=
// FATBIN-COMMON: "--image=profile=sm_60,file=
// PTX-SM60: "--image=profile=compute_60,file=
// NOPTX-SM60-NOT: "--image=profile=compute_60,file=
// FATBIN-COMMON: "--image3=kind=elf,sm=52,file=
// PTX-SM52: "--image3=kind=ptx,sm=52,file=
// NOPTX-SM52-NOT: "--image3=kind=ptx,sm=52,file=
// FATBIN-COMMON: "--image3=kind=elf,sm=60,file=
// PTX-SM60: "--image3=kind=ptx,sm=60,file=
// NOPTX-SM60-NOT: "--image3=kind=ptx,sm=60,file=