Improve specialization optimization #2944
Merged
Conversation
This file contains 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
lukaszcz
force-pushed
the
improve-spec-opt
branch
from
fe97d3b to
7b938d9
Compare
lukaszcz
force-pushed
the
improve-spec-opt
branch
from
1fe86ad to
fcd64b8
Compare
lukaszcz
force-pushed
the
improve-spec-opt
branch
from
db472c2 to
833ef24
Compare
paulcadman
force-pushed
the
improve-spec-opt
branch
from
833ef24 to
7f3fbfb
Compare
paulcadman
reviewed
Aug 13, 2024
| @@ -269,7 +271,7 @@ convertNode = dmapLRM go | |||
|
|
|||
| shiftSpecargs :: [Int] -> [Int] -> [Int] | |||
| shiftSpecargs specargs = | |||
| map (\argNum -> argNum - length (filter (argNum <) specargs)) | |||
| map (\argNum -> argNum - length (filter (argNum >) specargs)) | |||
There was a problem hiding this comment.
Does the change from < -> > fix a bug or is it related to another change?
There was a problem hiding this comment.
That was incorrect, but we didn't see it because:
- This adjusts the indices of arguments marked for specialization which are not specialized because the provided argument has the wrong form, e.g., it is a variable (makes no sense to specialize by an unknown). They can be specialized later in another iteration if, e.g., inlining substitutes a concrete value for the variable, but this is not very common.
- If specialization fails nothing really bad happens -- the function application is just not specialized and thus less efficient.
paulcadman
reviewed
Aug 13, 2024
| @@ -19,7 +19,7 @@ optimize' :: CoreOptions -> Module -> Module | |||
| optimize' opts@CoreOptions {..} md = | |||
| filterUnreachable | |||
| . compose | |||
| (4 * _optOptimizationLevel) | |||
| (6 * _optOptimizationLevel) | |||
There was a problem hiding this comment.
Why has this changed from 4 to 6?
There was a problem hiding this comment.
Because with maps and folds inside record definitions 4 iterations are not enough to inline & specialize everything in a typical case (without further nestings).
paulcadman
previously approved these changes
Aug 13, 2024
lukaszcz
force-pushed
the
improve-spec-opt
branch
from
d5f4a76 to
e396a01
Compare
paulcadman
approved these changes
Aug 13, 2024
paulcadman
added a commit
to anoma/juvix-stdlib
that referenced
this pull request
Aug 14, 2024
* Closes #121 * Depends on anoma/juvix#2944 * Includes #123 * Adds `Stdlib.Data.Range` to the Prelude. Since now `for` and `rfor` are trait fields, there is no name clash anymore. --------- Co-authored-by: Paul Cadman <git@paulcadman.dev>
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.
Checklist