Hacktoberfest 2026: as issues que os mantenedores marcaram para outubro, abertas e boas para iniciantes. Ver issues do Hacktoberfest

perf: close method-call throughput gap to blead Perl

Aberta
  • Algo errado?
#1,846 1 comentário 0 reações 0 responsáveis Ver no GitHub

Mantenedores costumam responder em até 1 dia

@fglock já está trabalhando nisso.

Desde 19/10/2026.

  • #1850 de @fglock — aberto

Avaliação

Dificuldade
5/5
Tempo estimado
Mais de uma semana
Facilidade para iniciantes
25/100
Tipo de issue
Funcionalidade
Clareza
Claramente especificada
Status de atividade
Ativa
Stack de tecnologia
java, perl
Domínio
performance

Direção de pesquisa

Start with the measurement contract in issue #1196 and the existing dev/bench/benchmark_method.pl anchor; establish a pinned blead, JDK, source and JAR hashes, and JVM flags before comparing performance. Build diagnostic workloads, then use CPU, allocation, GC, and HotSpot evidence to identify bottlenecks. Done means the anchor reaches at least 1.00x pinned blead with the 95% interval lower bound at or above 1.00x, while meeting issue #1196’s portfolio contract.

Escrita pelo modelo de indexação a partir do texto da issue.

Descrição

area:performance enhancement

Current state

PR #1843 improves the dev/bench/benchmark_method.pl workload through guarded argument unpacking, cached hash-field cells, reduced runtime and frame lookups, and scalar-result reuse. The remaining parity goal is open.

A fresh sequential run from PR head 0d4f883b8437b843c62b30c49d26ffac2c31a231 produced:

  • perl dev/bench/benchmark_method.pl: 340.14 iterations/s, checksum 399600080000.
  • ./jperl dev/bench/benchmark_method.pl: 267.42 iterations/s, checksum 399600080000.
  • PerlOnJava was 0.786x of system Perl in this single run, so it needs about 27% more throughput to reach parity. Treat these as a current snapshot, not a stable estimate; the host was also running test work.

The comparison used the locally installed Perl, not a pinned optimized blead build. The next phase must establish that reference before making parity claims. Existing profiling has sampled RuntimeCode.apply/applyCachedMethod, generated method bodies, hash-field access, warning-bit lookup, and integer scalar construction. These samples identify candidates; they do not prove which cost limits end-to-end throughput.

Next steps

  1. Build a reproducible method benchmark runner following the measurement contract in issue #1196: pin the Perl blead revision, JDK, JAR/source hashes and JVM flags; use deterministic checksums and operation counts; warm to stability; then collect randomized paired measurement windows and confidence intervals.
  2. Split the anchor into diagnostic workloads for no-op dispatch, one-argument dispatch, blessed-hash reads, integer updates, scalar return, and an equivalent inline body. Keep the original method benchmark as the acceptance anchor.
  3. Profile the stable anchor with CPU, allocation, GC and HotSpot inlining/deoptimization evidence. Report exclusive and inclusive costs and allocated bytes per operation. Continue only with bottlenecks that explain a material share of the anchor or portfolio.
  4. Prioritize general call-boundary or primitive-result improvements from that evidence. Keep guarded fallbacks for @_, caller, context, warnings, ties/overload, eval/debugger, exceptions, lvalue returns, non-local control flow, and escaping values.
  5. Add focused semantic coverage before each runtime change; validate Perl-level tests with system Perl and run JVM/interpreter parity checks. Retain exact-parent failure evidence for each regression.
  6. Measure each candidate against pinned blead and the exact parent. Require the method anchor to reach at least 1.00x blead with a 95% interval whose lower bound is at least 1.00x; also meet the portfolio acceptance contract already recorded in issue #1196 before claiming project-wide performance parity.

Related work

Linguagem predominante
Perl
Estrelas
64
Forks
8
Merge médio
5h 28min
PRs com merge (30d)
199

Preparar o ambiente

Primeiros passos

  1. Leia a issue inteira e depois o guia de contribuição do projeto.
  2. Comente na issue dizendo que vai assumir — evita que duas pessoas façam o mesmo trabalho.
  3. Faça um fork do repositório e trabalhe em uma branch.
  4. Abra um pull request que referencie o número da issue.

Mais de fglock/PerlOnJava

Todas as issues de fglock/PerlOnJava

Issues semelhantes

Mais issues de Perl

Receba novas issues na sua caixa de entrada

Um resumo curto de issues do GitHub para quem está começando.