A set of small optimizer improvements in preparation for the early SIL module serialization #11561
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
…n calls of big functions It is more profitable in most cases to inline the big callee rather than unroll the loop, because unrolling it would create a lot of calls which cannot be further optimized due to increased size of the function containing a loop.
This is very beneficial because many class_method/witness_method instructions may get concrete types after specialization. Doing it this way improves compilation times.
…ins are sunk in one round.
|
@swift-ci please smoke test |
|
@swift-ci please smoke benchmark |
|
@aschwaighofer Arnold, you may be interested in this change: 2e5b820 |
A simple call is an invocation of a function without any side-effects, where all passed parameters are constants.
…onstant for the purposes of the pure invocations checks
… in a specialized function, just don't perform the copy propagation in such a case Do not report any constant propagation issues in (generic) function specializations, because these issues are eventually not visible in the original function and thus would be very misleading and difficult to understand.
Build comment file:Optimized (O)Regression (1)
Improvement (2)
No Changes (324)
Unoptimized (Onone)Regression (1)
No Changes (326)
Hardware Overview |
swiftix
force-pushed
the
optimizer-improvements
branch
from
9d6b906 to
7bda199
Compare
|
@swift-ci please smoke test |
1 similar comment
|
@swift-ci please smoke test |
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Outdated suggestions cannot be applied.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.
Do not unroll loops if their bodies contain function calls of big functions. It is more profitable in most cases to inline the big callee rather than unroll the loop, because unrolling it would create a lot of calls which cannot be further optimized due to increased size of the function containing a loop.
Run the devirtualizer after the generic specializer. This is very beneficial because many class_method/witness_method instructions may get concrete types after specialization. Doing it this way improves compilation times.
Add another round of retain sinking, because not all of possible retains are sunk in one round.
[sil-inliner] It is always OK to inline a simple call. A simple call is an invocation of a function without any side-effects, where all passed parameters are constants.
[sil-constant-propagation] If an overflow/underflow would happen only in a specialized function, just don't perform the copy propagation in such a case. Do not report any constant propagation issues in (generic) function specializations, because these issues are eventually not visible in the original function and thus would be very misleading and difficult to understand.