[FFI] Phase out ctypes fallback in favor of cython#17714
Merged
yongwww merged 3 commits intoapache:mainfrom Mar 8, 2025
Merged
[FFI] Phase out ctypes fallback in favor of cython#17714yongwww merged 3 commits intoapache:mainfrom
yongwww merged 3 commits intoapache:mainfrom
Conversation
In the past we maintained both ctypes fallback and cython path for our FFI. One reason was cython was not that widely available then. The situation has changed, so this PR removes the ctypes fallback so we can focus on building cython based core for FFI. Notably, the cython core still leverages ctypes, especially for some of the features that do not need fast path, and primarily uses cython to speedup core FFI features with fast path. With cython, we can get most FFI costs down to level of roughtly 1e-6 sec, which aligns with most of the costs like overhead of a numpy operation.
Member
Author
|
@tvm-bot rerun |
yongwww
approved these changes
Mar 8, 2025
ShiboXing
pushed a commit
to ShiboXing/tvm
that referenced
this pull request
Aug 10, 2025
* [FFI] Phase out ctypes fallback in favor of cython In the past we maintained both ctypes fallback and cython path for our FFI. One reason was cython was not that widely available then. The situation has changed, so this PR removes the ctypes fallback so we can focus on building cython based core for FFI. Notably, the cython core still leverages ctypes, especially for some of the features that do not need fast path, and primarily uses cython to speedup core FFI features with fast path. With cython, we can get most FFI costs down to level of roughtly 1e-6 sec, which aligns with most of the costs like overhead of a numpy operation. * fix lint * Fix the dso loading issue
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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 the past we maintained both ctypes fallback and cython path for our FFI. One reason was cython was not that widely available then. The situation has changed, so this PR removes the ctypes fallback so we can focus on building cython based core for FFI.
Notably, the cython core still leverages ctypes, especially for some of the features that do not need fast path, and primarily uses cython to speedup core FFI features with fast path.
With cython, we can get most FFI costs down to level of roughtly 1e-6 sec, which aligns with most of the costs like overhead of a numpy operation.