Link devicelib during device code compilation #15114
Closed
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
added 2 commits
August 13, 2024 14:39
Create one single bitcode library for NVPTX by compiling each libdev file into bitcode first, linking these together and running opt on them. Strip away metadata by reusing prepare_builtins from libclc. Remove NVPTX bundles from the libdev object files and remove any unbundling action spawned by the Clang driver for the SYCL toolchain when compiling for the NVPTX backend. Make the driver link against the single bitcode libraries for NVPTX for the SYCL toolchain when device library linkage is not excluded. Ensure that the clang tests check for the correctness of the new clang driver actions and check if the driver still links the device code against the itt device libraries when device library linkage has been excluded. Refactor SYCLLibdevice.cmake by creating functions and grouping e.g. the same compilation flags for a filetype together in one variable. Reuse these variables and call functions to remove redundancies.
Use the -mlink-builtin-bitcode flag to link the CUDA devicelib during device code compilation instead of generating a separate llvm-link action in the driver. Add a new flag to exclude linking of libdevice completely. Add a new error in case that the cuda devicelib file can not be found.
MartinWehking
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
MartinWehking
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
Closed, see discussions in #15083 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Note: This patch is based on 15048 (Not merged yet)
and links against the final resulting bitcode library that is created through that patch
Use the
-mlink-builtin-bitcodeflag to link the CUDA devicelibduring device code compilation instead of generating a separate
llvm-link action in the driver.
Add a new flag to exclude linking of libdevice completely.
Add a new error in case that the cuda devicelib file can not be found.