[RUNTIME] Implement dynamic loading with defineGetFunctionHandle for CUDA version compatibility #2771
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
|
LGTM, thanks! |
davidberard98
pushed a commit
to davidberard98/triton
that referenced
this pull request
Dec 12, 2023
…onHandle for CUDA version compatibility (triton-lang#2771)" This is needed for CUDA 11 support, which we'd like to have in the PyTorch 2.2 release. Original commit message: In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues.
malfet
pushed a commit
that referenced
this pull request
Dec 13, 2023
…onHandle for CUDA version compatibility (#2771)" (#2789) This is needed for CUDA 11 support, which we'd like to have in the PyTorch 2.2 release. Original commit message: In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues. Co-authored-by: Keren Zhou <kerenzhou@openai.com>
malfet
added a commit
to pytorch/pytorch
that referenced
this pull request
Dec 13, 2023
To include a cherry-pick of triton-lang/triton#2771 that should fix cuda-11.8 runtime issues
pytorchmergebot
pushed a commit
to pytorch/pytorch
that referenced
this pull request
Dec 14, 2023
To include a cherry-pick of triton-lang/triton#2771 that should fix cuda-11.8 runtime issues Also, tweak build wheel script to update both ROCm and vanilla Trition builds version to 2.2 (even though on trunk it should probably be 3.3 already) TODO: Remove `ROCM_TRITION_VERSION` once both trunk and ROCM version are in sync again Pull Request resolved: #115743 Approved by: https://github.com/davidberard98
guilhermeleobas
pushed a commit
to guilhermeleobas/pytorch
that referenced
this pull request
Dec 18, 2023
To include a cherry-pick of triton-lang/triton#2771 that should fix cuda-11.8 runtime issues Also, tweak build wheel script to update both ROCm and vanilla Trition builds version to 2.2 (even though on trunk it should probably be 3.3 already) TODO: Remove `ROCM_TRITION_VERSION` once both trunk and ROCM version are in sync again Pull Request resolved: pytorch#115743 Approved by: https://github.com/davidberard98
dmenig
pushed a commit
to dmenig/pytorch
that referenced
this pull request
Dec 21, 2023
To include a cherry-pick of triton-lang/triton#2771 that should fix cuda-11.8 runtime issues Also, tweak build wheel script to update both ROCm and vanilla Trition builds version to 2.2 (even though on trunk it should probably be 3.3 already) TODO: Remove `ROCM_TRITION_VERSION` once both trunk and ROCM version are in sync again Pull Request resolved: pytorch#115743 Approved by: https://github.com/davidberard98
feihugis
pushed a commit
to feihugis/triton
that referenced
this pull request
Feb 13, 2024
…CUDA version compatibility (triton-lang#2771) In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues.
pingzhuu
pushed a commit
to siliconflow/triton
that referenced
this pull request
Apr 2, 2024
…onHandle for CUDA version compatibility (triton-lang#2771)" (triton-lang#2789) This is needed for CUDA 11 support, which we'd like to have in the PyTorch 2.2 release. Original commit message: In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues. Co-authored-by: Keren Zhou <kerenzhou@openai.com>
binarman
pushed a commit
to binarman/triton
that referenced
this pull request
Apr 2, 2024
…CUDA version compatibility (triton-lang#2771) In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues.
Merged
malfet
added a commit
that referenced
this pull request
Jul 16, 2024
That is only present in CUDA-12 compatible drivers, and is missing in CUDA-11 ones Spiritual follow up after #2771
malfet
added a commit
that referenced
this pull request
Jul 16, 2024
That is only present in CUDA-12 compatible drivers, and is missing in CUDA-11 ones Spiritual follow up after #2771
7 tasks
Jokeren
pushed a commit
that referenced
this pull request
Jul 16, 2024
That is only present in CUDA-12 compatible drivers, and is missing in CUDA-11 ones Spiritual follow up after #2771 allows for dynamic query of the symbol and if run on an older driver, it will return an error. Also, fix `occupancyMaxActiveClusters` behavior when symbol is not found (before this change it would crash with null pointer deref, now it should return a structured exception)
atalman
pushed a commit
to atalman/triton
that referenced
this pull request
Jul 17, 2024
That is only present in CUDA-12 compatible drivers, and is missing in CUDA-11 ones Spiritual follow up after triton-lang#2771 allows for dynamic query of the symbol and if run on an older driver, it will return an error. Also, fix `occupancyMaxActiveClusters` behavior when symbol is not found (before this change it would crash with null pointer deref, now it should return a structured exception)
bertmaher
pushed a commit
to bertmaher/triton
that referenced
this pull request
Dec 10, 2024
That is only present in CUDA-12 compatible drivers, and is missing in CUDA-11 ones Spiritual follow up after triton-lang#2771 allows for dynamic query of the symbol and if run on an older driver, it will return an error. Also, fix `occupancyMaxActiveClusters` behavior when symbol is not found (before this change it would crash with null pointer deref, now it should return a structured exception)
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.
In case cuda 11 drivers are still used on some systems, we shouldn't call TMA and block cluster related functions directly. Instead, we can dynamically lookup the handles to avoid compatibility issues.