make NVRTC work with Wave headers #11500
Open
+7,906
−260
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
facebook-github-bot
added
CLA Signed
✅ Deploy Preview for meta-velox canceled.
|
|
@oerling has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
1 similar comment
|
@oerling has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@oerling has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
This pull request was exported from Phabricator. Differential Revision: D65760296 |
oerling
pushed a commit
to oerling/velox-1
that referenced
this pull request
Nov 13, 2024
Summary: NVRTC requires headers to be extracted into text strings and passed to the compiler. This is done by the jitify utility added from Nvidia, added to external. Extracting the headers takes a long time. This is therefore done on first use and the headers are subsequently kept as strings. NVRTC further requires some header substitutions that are provided by jitify. jitify also provides defaults for --gpu-architecture for NVRTC. Even with jitify, NVRTC fails when including <cuda/atomic> and cub at the same time. We depend on <cuda/atomic> and <cuda/semaphore> and extract warp scan from cub as jit/Scan.cuh. Other cub items may be similarly extracted if needed. All Wave headers that may be needed by NVRTC are extracted into jit/Headers.h. A make_headers.sh script is added for regenerating Headers.h from actual headers. We add the stringify utility from NVIDIA to make escapes when changing headers to string literals. We add tests that verify use of <cuda/semaphore> for updating aggregates where the updating kernel is compiled at run time. We also add a test for the warp scan replacement in jit/. Pull Request resolved: facebookincubator#11500 Differential Revision: D65760296 Pulled By: oerling
Summary: NVRTC requires headers to be extracted into text strings and passed to the compiler. This is done by the jitify utility added from Nvidia, added to external. Extracting the headers takes a long time. This is therefore done on first use and the headers are subsequently kept as strings. NVRTC further requires some header substitutions that are provided by jitify. jitify also provides defaults for --gpu-architecture for NVRTC. Even with jitify, NVRTC fails when including <cuda/atomic> and cub at the same time. We depend on <cuda/atomic> and <cuda/semaphore> and extract warp scan from cub as jit/Scan.cuh. Other cub items may be similarly extracted if needed. All Wave headers that may be needed by NVRTC are extracted into jit/Headers.h. A make_headers.sh script is added for regenerating Headers.h from actual headers. We add the stringify utility from NVIDIA to make escapes when changing headers to string literals. We add tests that verify use of <cuda/semaphore> for updating aggregates where the updating kernel is compiled at run time. We also add a test for the warp scan replacement in jit/. Pull Request resolved: facebookincubator#11500 Differential Revision: D65760296 Pulled By: oerling
|
This pull request was exported from Phabricator. Differential Revision: D65760296 |
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.
NVRTC requires headers to be extracted into text strings and passed to the compiler. This is done by the jitify utility added from Nvidia, added to external. Extracting the headers takes a long time. This is therefore done on first use and the headers are subsequently kept as strings. NVRTC further requires some header substitutions that are provided by jitify. jitify also provides defaults for --gpu-architecture for NVRTC.
Even with jitify, NVRTC fails when including <cuda/atomic> and cub at the same time. We depend on <cuda/atomic> and <cuda/semaphore> and extract warp scan from cub as jit/Scan.cuh. Other cub items may be similarly extracted if needed.
All Wave headers that may be needed by NVRTC are extracted into jit/Headers.h. A make_headers.sh script is added for regenerating Headers.h from actual headers. We add the stringify utility from NVIDIA to make escapes when changing headers to string literals.
We add tests that verify use of <cuda/semaphore> for updating aggregates where the updating kernel is compiled at run time. We also add a test for the warp scan replacement in jit/.