[SYCL][RTC] Add minimum amount of post-link functionality to extract symbol table and properties #16109
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
Signed-off-by: Julian Oppermann <[email protected]>
Signed-off-by: Julian Oppermann <[email protected]>
jopperm
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Signed-off-by: Julian Oppermann <[email protected]>
Signed-off-by: Julian Oppermann <[email protected]>
jopperm
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
Signed-off-by: Julian Oppermann <[email protected]>
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
changed the title
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
commented
Nov 20, 2024
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Contributor
|
I think @cperkinsintel is much more familiar with this code than I am, so I'll leave this review to him. |
sommerlukas
reviewed
Nov 21, 2024
Contributor
sommerlukas
left a comment
sommerlukas
left a comment
There was a problem hiding this comment.
Some stylistic nits.
Comment on lines
41
to
48
| static JITResult errorToFusionResult(llvm::Error &&Err, | ||
| const std::string &Msg) { | ||
| return wrapError<JITResult>(std::move(Err), Msg); | ||
| } | ||
|
|
||
| static RTCResult errorToRTCResult(llvm::Error &&Err, const std::string &Msg) { | ||
| return wrapError<RTCResult>(std::move(Err), Msg); | ||
| } |
Contributor
There was a problem hiding this comment.
Why don't we use wrapError directly (maybe with a different name)?
So, instead of errorToFusionResult or errorToRTCResult, we would call errorToResult<JITResult> and errorToResult<RTCResult>?
Contributor
Author
There was a problem hiding this comment.
errorToResult<T> makes sense, I initially decided against that to keep the diff a bit cleaner.
Co-authored-by: Lukas Sommer <[email protected]>
jopperm
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
Signed-off-by: Julian Oppermann <[email protected]>
Signed-off-by: Julian Oppermann <[email protected]>
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
…ect. Signed-off-by: Julian Oppermann <[email protected]>
jopperm
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
jopperm
had a problem deploying
to
WindowsCILock
GitHub Actions
Error
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
sommerlukas
reviewed
Nov 22, 2024
| !F->use_empty()) { | ||
| // The element in "llvm.used" array has other users. That is Ok for | ||
| // specialization constants, but is wrong for kernels. | ||
| llvm::report_fatal_error("Unexpected usage of SYCL kernel"); |
Contributor
There was a problem hiding this comment.
Do we ever expect to hit this during user code compilation? If yes, we should probably fail more gracefully and report back through a failed RTCResult.
Contributor
Author
There was a problem hiding this comment.
Good point, however, I wasn't able to trigger the generation of llvm.used and llvm.compiler.used with the feature set we currently support. Hence I'm proposing to just assert that they're not present, and to drop the copied code from sycl-post-link for now.
Signed-off-by: Julian Oppermann <[email protected]>
Signed-off-by: Julian Oppermann <[email protected]>
Signed-off-by: Julian Oppermann <[email protected]>
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
jopperm
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
sommerlukas
approved these changes
Nov 25, 2024
cperkinsintel
approved these changes
Nov 25, 2024
Contributor
Author
|
@intel/llvm-gatekeepers Please merge this, thanks. |
Contributor
Author
|
Working on a fix for the post-commit failure. |
Contributor
|
Thanks, tag me when it's ready |
sarnex
pushed a commit
that referenced
this pull request
Nov 25, 2024
Fixes post-commit failure due to an unused variable introduced by #16109. --------- Signed-off-by: Julian Oppermann <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
This PR implements symbol table and property set extraction from the runtime-compiled module, based on the skeleton used by
sycl-post-link, assuming that no splitting of the module is required and/or requested. We add the necessary plumbing to pass this information to the runtime, which then reuses existingsycl-jitinfrastructure to create the necessary data structures for device binary images.We don't use the properties yet during the creating of the kernel bundles, but the necessary information to feed the images into the program manager is there.