[CUDA] Use --image3 to construct fat binary #151760
Merged
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
llvmbot
added
clang
|
@llvm/pr-subscribers-clang @llvm/pr-subscribers-clang-driver Author: Artem Belevich (Artem-B) ChangesCUDA-12.9 has removed fatbinary tool's --image3 has been supported since cuda-9, so we do not need CUDA SDK version checks. Full diff: https://github.com/llvm/llvm-project/pull/151760.diff 3 Files Affected:
diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
index 1f0b478c02b25..fdfcea852b4f2 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -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))
diff --git a/clang/test/Driver/cuda-arch-translation.cu b/clang/test/Driver/cuda-arch-translation.cu
index e4f83740a92eb..b4a521df2e6f5 100644
--- a/clang/test/Driver/cuda-arch-translation.cu
+++ b/clang/test/Driver/cuda-arch-translation.cu
@@ -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
diff --git a/clang/test/Driver/cuda-options.cu b/clang/test/Driver/cuda-options.cu
index db6536ca9e03b..fc8e83a2bb279 100644
--- a/clang/test/Driver/cuda-options.cu
+++ b/clang/test/Driver/cuda-options.cu
@@ -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"
@@ -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=
|
jhuber6
approved these changes
Aug 1, 2025
CUDA-12.9 has removed fatbinary tool's `--image` argument we've been using till now. --image3 has been supported since cuda-9, so we do not need CUDA SDK version checks.
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/59/builds/21980 Here is the relevant piece of the build log for the reference |
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/46/builds/21078 Here is the relevant piece of the build log for the reference |
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.
CUDA-12.9 has removed fatbinary tool's
--imageargument we've been using till now.--image3 has been supported since cuda-9, so we do not need CUDA SDK version checks.