perf: curate guarded string and closure gains - #1387
Merged
Merged
Conversation
Port the retained BMP substring offset scan and integer literal unary-minus lowering from performance research, with permanent Perl regressions and the delivery-selection manifest for the remaining closure and method groups. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Add the retained closure and method semantic regressions, including a PadWalker capture-rebind case that any direct closure entry must preserve. Record the validation evidence and minimal-port constraints in the delivery selection manifest. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Mark compiler-proven captured-scalar addition closures and bypass the ordinary call boundary only for scalar, zero-argument native-integer invocations. Preserve ordinary behavior for debugger, lvalue, overflow, tainted, overloaded, and PadWalker-rebound cases. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Record the integrated closure fast path and defer the two-slot method recognizer in favor of investigating costs shared by ordinary methods. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Record the source-matched pilot, matching tooling identities, and unstable host conditions without treating its directional throughput as delivery evidence. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Add the required work-in-progress changelog entry for the selected guarded compiler performance improvements without making an unmeasured claim. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Record the current-source control run, checksum validation, and the non-authoritative interpretation of the curated portfolio. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Distinguish the passing current-target correctness gate from the earlier non-authoritative portfolio artifacts. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Record the current master base and passing immutable validation gate without transferring a throughput claim from the earlier inconclusive portfolio. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <noreply@openai.com>
Recycle tagged one-scalar return lists only after direct JVM call sites extract their scalar result. Preserve ordinary list-context and multi-value result lifetime, add return-context regression coverage, and record bounded closure JFR allocation evidence. Generated with [Codex](https://openai.com/codex/) Co-Authored-By: Codex <codex@openai.com>
Recycle the private one-scalar RuntimeList wrapper when compound assignment scalarizes a subroutine or method result, without changing ordinary list identity or call-frame ownership. Add a cross-backend regression and record the allocation boundary and benchmark evidence in the performance handoff. Generated with Codex (https://openai.com/codex) Co-Authored-By: Codex <codex@openai.com>
Route JVM coderef and method scalar-result conversions through the existing private RuntimeList recycler. Add opt-in lifecycle counters that prove the method hot path now recycles its scalar return wrappers, and record the allocation evidence and measurement limits in the performance handoff. Generated with Codex (https://openai.com/codex) Co-Authored-By: Codex <codex@openai.com>
Keep Perl's required scalar return copy, then mark only the private one-element scalar-context wrapper for runtime-local recycling. Generated with Codex (https://openai.com/codex) Co-Authored-By: Codex <noreply@openai.com>
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
1 participant
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.
Summary
Curates two independently guarded compiler/runtime improvements from the
performance research branch without merging its experimental history:
lowering paths.
The branch intentionally does not include the benchmark-shaped plain-hash
method specialization. The retained changes preserve the ordinary fallback for
unsupported, observable, debugger, lvalue, overflow, taint, and overload
cases.
Validation
passed (recorded in
dev/design/performance-delivery-selection.md).makepassed at21081cf4b, rebased onto current master300954833(/tmp/make-perf-curated-rebase-300954833-20260915.log).Performance evidence and scope
Earlier source-matched seven-pair portfolio artifacts have matching semantic
checksums, but their analyzer marked them inconclusive because of host load and
warmup instability. This PR therefore makes no portfolio-wide throughput or
Perl-parity claim. The selection manifest preserves the artifacts and their
qualification. A fresh stable-host portfolio is required before any such claim.
Follow-up
String, Regex, and Life remain below the parity objective. Future method work
will target broadly shared costs—argument-frame allocation, cached dispatch,
scalar-result handling, and hash access—rather than recognizing a particular
method implementation.
Related benchmark tooling is in #1295 and remains a separate PR.