[
https://issues.apache.org/jira/browse/GROOVY-12181?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18097624#comment-18097624
]
ASF GitHub Bot commented on GROOVY-12181:
-----------------------------------------
github-actions[bot] commented on PR #2725:
URL: https://github.com/apache/groovy/pull/2725#issuecomment-5025454272
### JMH summary — indy (commit `c9ef5eb`)
Speedup vs trailing 90-day baseline on gh-pages. Higher = faster.
`1.00` = in line with history. Per-benchmark ratio, geomean within group.
Time-per-op units inverted so direction is consistent. The *calibrated*
column divides out this runner's speed vs the baseline hardware, as
measured by Groovy-independent pure-Java ruler benchmarks.
| Group | Speedup | Calibrated | n |
|--------|---------|------------|---|
| bench | 0.977 × | 0.997 × | 84 |
| core | 4.287 × | 3.765 × | 77 |
| grails | 2.259 × | 2.214 × | 80 |
> ⚠️ Runner speed differs ≥15% from the historical baseline hardware for:
core-hz. Raw speedups are not meaningful for those parts — use the calibrated
column.
<sub>Runner calibration (this run vs baseline hardware): bench 0.98× (26
rulers) · core-ag 0.94× (3 rulers) · core-hz 1.43× (3 rulers) · grails-ad 0.96×
(3 rulers) · grails-ez 1.07× (3 rulers)</sub>
<sub>Baseline: <code>dev/bench/jmh/<part>/indy/data.js</code> on
gh-pages, trailing 90 days. <a
href="https://apache.github.io/groovy/dev/bench/jmh/summary.html";>Daily
dashboard</a> · <a
href="https://apache.github.io/groovy/dev/bench/jmh/";>Per-suite raw
data</a></sub>
<!
> Refactor async runtime and AST helpers introduced by GROOVY-9381
> ----------------------------------------------------------------
>
> Key: GROOVY-12181
> URL: https://issues.apache.org/jira/browse/GROOVY-12181
> Project: Groovy
> Issue Type: Improvement
> Reporter: Daniel Sun
> Priority: Major
>
> h2. Context
> GROOVY-9381 landed native {{{}async{}}}/{{{}await{}}}/{{{}defer{}}}: public
> API in {{{}groovy.concurrent{}}}, runtime in
> {{{}org.apache.groovy.runtime.async{}}}, and parser desugaring via
> {{AstBuilder}} + {{{}AsyncTransformHelper{}}}.
> The first implementation correctly prioritised a working feature, but two
> types accumulated too many responsibilities:
> * {{AsyncSupport}} — executor/scheduler setup, all {{await}} overloads, task
> launch, combinators
> ({{{}all{}}}/{{{}any{}}}/{{{}first{}}}/{{{}allSettled{}}}), defer scopes, and
> generator bridges
> * {{AstBuilder}} — multi-step rewrites for {{for await}} and {{async}}
> closures inlined next to general parse-tree walking
> This ticket covers a focused refactor of that code: clearer module
> boundaries, a thinner parser visitor, small API/codegen polish, and unit
> tests for the extracted pieces. *No new language syntax.*
> h2. Goals
> # Single responsibility for runtime pieces (entry point vs pool config vs
> combinators).
> # Keep feature-specific AST rewrites out of {{{}AstBuilder{}}}; helpers own
> the desugaring.
> # Preserve the public {{groovy.concurrent}} surface and combinator
> {*}names{*}.
> # Make compiler entry points and executor reset behaviour explicit and
> tested.
> h2. Changes
> h3. Runtime split (package-private)
> ||Concern||Before||After||
> |Entry point|{{AsyncSupport}} (monolith)|{{AsyncSupport}} — facades used by
> compiler-generated code and {{Awaitable}}|
> |Executors / scheduler|embedded in {{AsyncSupport}}|{{AsyncExecutors}}|
> |Combinators|embedded in {{AsyncSupport}}|{{AwaitCombinators}} ({{{}all{}}},
> {{{}any{}}}, {{{}first{}}}, {{{}allSettled{}}})|
> Public methods on {{AsyncSupport}} / {{Awaitable}} remain the stable surface;
> algorithms and pool wiring move behind package-private types. Package-info
> under {{groovy.concurrent}} and {{org.apache.groovy.runtime.async}} updated
> to match.
> h3. AST / parser
> Higher-level rewrites move from {{AstBuilder}} into
> {{{}AsyncTransformHelper{}}}:
> * {{transformAsyncClosure}} — defer-scope wrap, generator param injection,
> {{async}} / {{asyncGenerator}} call
> * {{wrapForAwaitLoop}} — {{toIterable}} + {{{}try{}}}/{{{}finally{}}}
> {{closeIterable}}
> {{AstBuilder}} only dispatches (one call site per construct).
> Related codegen polish in the same pass:
> * Single-arg {{await}} emits {{AsyncSupport.awaitAny(Object)}} instead of
> {{(Object) expr}} cast to force the {{await(Object)}} overload when the value
> implements several async interfaces (e.g. {{CompletableFuture}} as both
> {{CompletionStage}} and {{{}Future{}}}).
> * {{for await}} temps use {{$_{_}forAwaitSource_N{_}}} _(monotonic counter)
> instead of {{}}_{{_forAwaitSource}} + parse-context {{{}hashCode(){}}},
> avoiding synthetic-name clashes across nested/repeated loops.
> h3. Small API polish
> * {{AsyncSupport.setExecutor(null)}} / {{Awaitable.setExecutor(null)}} —
> restore the platform default (same as {{{}resetExecutor(){}}}). Previously
> null was rejected.
> * {{AwaitResult.success(T)}} — typed parameter (was {{Object}} + unchecked
> cast).
> * Javadoc for {{Awaitable.first}} all-fail path aligned with the existing
> aggregate {{CompletionException}} behaviour (cause = first failure; remaining
> as suppressed; {{await}} transparency rethrows the cause). *Exception shape
> unchanged.*
> h2. Compatibility
> * Follow-up to GROOVY-9381; intended for the 6.0 async feature line.
> * {{setExecutor(null)}} is no longer an error; it resets the default
> executor.
> * Combinator *names* and public package layout ({{{}groovy.concurrent{}}})
> are unchanged.
> * {{Awaitable.first}} all-fail path remains an aggregate
> {{CompletionException}} (no change to exception type).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
