[SYCL][Driver] Fix regression that enabled Cuda-mode in cc1 and defined __CUDA_ARCH__ #15441
Conversation
This file contains 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
GeorgeWeb
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
GeorgeWeb
force-pushed
the
georgi/cuda-arch-define
branch
from
186d6bd to
a188a0b
Compare
GeorgeWeb
changed the title
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
frasercrmck
reviewed
Sep 19, 2024
frasercrmck
reviewed
Sep 19, 2024
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
mdtoguchi
reviewed
Sep 19, 2024
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
mdtoguchi
approved these changes
Sep 19, 2024
frasercrmck
reviewed
Sep 20, 2024
frasercrmck
approved these changes
Sep 20, 2024
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
GeorgeWeb
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
elizabethandrews
approved these changes
Sep 20, 2024
smanna12
approved these changes
Sep 20, 2024
|
@intel/llvm-gatekeepers This looks good to merge now. Thanks. |
bader
reviewed
Sep 25, 2024
| // RUN: %clang_cc1 %s -triple nvptx64-nvidia-cuda -target-cpu sm_80 -fsycl-is-device -E -dM | FileCheck --check-prefix=CHECK-CUDA %s | ||
| // RUN: %clang_cc1 %s -triple nvptx64-nvidia-cuda -target-cpu sm_80 -fsycl-is-device -E -dM | FileCheck \ | ||
| // RUN: --check-prefix=CHECK-CUDA %s -DARCH_CODE=800 | ||
| // RUN: %clangxx %s -fsycl -nocudalib -fsycl-targets=nvptx64-nvidia-cuda -Xsycl-target-backend --offload-arch=sm_80 -E -dM | FileCheck \ |
There was a problem hiding this comment.
This is the preprocessor test, so we should only call %clang_cc1. Please, move the driver test to clang/test/Driver/ directory.
While you are doing this, please, make sure you either have libspirv input files or pass -fno-sycl-libspirv option. I don't build NVPTX target and this test fails with:
clang: error: cannot find 'remangled-l64-signed_char.libspirv-nvptx64-nvidia-cuda.bc'; provide path to libspirv library via '-fsycl-libspirv-path', or pass '-fno-sycl-libspirv' to build without linking with libspirv
clang: error: cannot find 'remangled-l64-signed_char.libspirv-nvptx64-nvidia-cuda.bc'; provide path to libspirv library via '-fsycl-libspirv-path', or pass '-fno-sycl-libspirv' to build without linking with libspirv
There was a problem hiding this comment.
Noted. I'll sort that out asap once I am back on a work PC.
There was a problem hiding this comment.
Addressed in #15521
bader
pushed a commit
that referenced
this pull request
Oct 2, 2024
… Preprocessor (#15521) This PR moves the driver invocation test that checks `__CUDA_ARCH__` does not get defined and ensures that it doesn't require the `libspirv-nvptx64-nvidia-cuda` bitcode files by passing `-fno-sycl-libspirv` to the `%clangxx` command. Link to the comment in related PR that reported this issue: #15441 (comment) Additionally, an extra test is added to check that the `-fcuda-is-device` option is not supplied in the CC1 invocation targeting `nvptx64-nvidia-cuda`, which enables `LangOptions.CudaIsDevice` and was the cause of defining the `__CUDA_ARCH__` macro.
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.
The
CudaToolChainset-fcuda-is-deviceunconditionally which madeInitializePredefinedMacros(called fromclang::InitializePreprocessor) to define__CUDA_ARCH__(default-init to 1). As such, the driver assumed Cuda mode while in also SYCL mode, but we don't properly support Cuda device-code compatibility and we want to avoid having the__CUDA_ARCH__macro defined altogether for SYCL offload.