[AutoDiff] SILGen derivative function thunks must not be transparent. #27752
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
SILGen generates thunks for derivative functions registered via `@differentiable` and `@differentiating` attributes. The thunks have a canonical, consistent naming (currently based on the original function name and parameter indices) that TBDGen can also generate via `SILDeclRef::mangle`. These SILGen derivative function thunks must not be transparent; otherwise, the functions will be inlined during MandatoryInlining and will not be exposed publicly. This bug was discovered as a DeadFunctionElimination crash: `function_ref` instructions referencing transparent derivative thunks had their referenced functions set to null by MandatoryInling and became invalid.
rxwei
approved these changes
Oct 17, 2019
rxwei
reviewed
Oct 17, 2019
TBDGen behavior is not expected to change, testing in case.
dan-zheng
force-pushed
the
autodiff-thunk-must-not-transparent
branch
from
f337af9 to
1e6bfbd
Compare
|
@swift-ci Please test tensorflow |
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Outdated suggestions cannot be applied.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.
SILGen generates thunks for derivative functions registered via
@differentiableand@differentiatingattributes.The thunks have a canonical, consistent naming (currently based on
the original function name and parameter indices) that TBDGen can also
generate via
SILDeclRef::mangle.These SILGen derivative function thunks must not be transparent; otherwise,
the functions will be inlined during MandatoryInlining and will not be
exposed publicly.
This bug was discovered as a DeadFunctionElimination crash:
function_refinstructions referencing transparent derivative thunkshad their referenced functions set to null by MandatoryInling and became
invalid.
More SILGen derivative function thunking documentation coming soon.