-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[WIP] on-disk caching of CUDA kernels #8079
Closed
Closed
Commits on May 19, 2022
-
CUDA caching: enough hacks to save empty function
The following: ```python from numba import cuda @cuda.jit(cache=True) def f(): pass f[1, 1]() ``` runs and saves cache files to the disk. The subsequent run loading from the cache still fails.
Configuration menu - View commit details
-
Copy full SHA for ea4fc08 - Browse repository at this point
Copy the full SHA ea4fc08View commit details -
[WIP] Attempt to make the load work
Presently fails because the codegen can't be pickled due to ctypes pointer.
Configuration menu - View commit details
-
Copy full SHA for b0fcf79 - Browse repository at this point
Copy the full SHA b0fcf79View commit details -
Successful pickle and unpickle of empty kernel
This incorporates fixes to the PTX pickling from @c200chromebook: https://gist.github.com/c200chromebook/1ee304161a39b247e8a029ff9d2b3cd7 For the simple test: ```python from numba import cuda @cuda.jit(cache=True) def f(): pass f[1, 1]() ``` The following CUDA Driver API calls are seen on the first invocation: ``` == CUDA [427] DEBUG -- call driver api: cuLinkCreate_v2 == CUDA [428] DEBUG -- call driver api: cuLinkAddData_v2 == CUDA [428] DEBUG -- call driver api: cuLinkComplete == CUDA [428] DEBUG -- call driver api: cuLinkDestroy ``` but are absent from a subsequent run.
Configuration menu - View commit details
-
Copy full SHA for cdede56 - Browse repository at this point
Copy the full SHA cdede56View commit details -
Configuration menu - View commit details
-
Copy full SHA for a4c85fe - Browse repository at this point
Copy the full SHA a4c85feView commit details -
Configuration menu - View commit details
-
Copy full SHA for 89362c9 - Browse repository at this point
Copy the full SHA 89362c9View commit details -
Revert "Pickle dispatcher overloads too - issue with pickling device …
…functions though" This reverts commit 65b9615.
Configuration menu - View commit details
-
Copy full SHA for 62a15d0 - Browse repository at this point
Copy the full SHA 62a15d0View commit details -
Configuration menu - View commit details
-
Copy full SHA for 49c8fcf - Browse repository at this point
Copy the full SHA 49c8fcfView commit details -
Configuration menu - View commit details
-
Copy full SHA for 88423a2 - Browse repository at this point
Copy the full SHA 88423a2View commit details -
Configuration menu - View commit details
-
Copy full SHA for 926627e - Browse repository at this point
Copy the full SHA 926627eView commit details -
Configuration menu - View commit details
-
Copy full SHA for f8dc8ee - Browse repository at this point
Copy the full SHA f8dc8eeView commit details -
Configuration menu - View commit details
-
Copy full SHA for 0d4e447 - Browse repository at this point
Copy the full SHA 0d4e447View commit details -
Configuration menu - View commit details
-
Copy full SHA for 4864bfa - Browse repository at this point
Copy the full SHA 4864bfaView commit details
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.