cobolrt performance: benchmark vs COBOL batch volumes, then add proven fast paths

未关闭
#4,717 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
18/100
Issue 类型
重构
描述清晰度
基本清楚
活跃度
活跃
技术栈
python
领域
performance

调研方向

Start by reading the cobolrt decimal arithmetic path that the issue calls the hot rule, and find where every operation builds a BigDecimal. Build the benchmark harness first, using the CardDemo batch, POSTTRAN, intcalc and mortgage ports the issue lists, and record baseline throughput and allocations. A fast path counts as done only when the unit vectors, the det sweep and the cobc differential fuzz show 0 diffs, and the A/B fuzz between fast and slow paths agrees.

由索引模型根据 Issue 内容生成。

描述

enhancement legacy-modernization priority: medium

What. Mainframe batch jobs process huge volumes. Exact decimal arithmetic with BigDecimal for every operation may be slow.

  1. Benchmark: representative batch ports (CardDemo batch, POSTTRAN, intcalc, mortgage) at realistic record counts. Measure throughput, allocations, and the hot rules (JMH where useful).
  2. Fast paths: e.g. long arithmetic when the digits fit, precomputed codecs, fewer allocations. Each fast path must be proven identical: unit vectors, full det sweep, and the differential fuzz against cobc gives 0 diffs, with an A/B fuzz between fast and slow paths.

Acceptance.

  • Benchmark harness committed, with baseline numbers.
  • At least the top hot rule has a proven fast path.
  • Before/after numbers.
  • No verdict change.

Part of the cobolrt-as-a-product epic (linked from the epic). Owner-approved direction, 2026-10-08.

🤖 Generated with Claude Code

主要语言
Python
星标
57
派生
11
平均合并
1 小时 55 分钟
30 天内合并 PR
973

环境准备

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

squid-protocol/gitgalaxy 的其他 Issue

查看 squid-protocol/gitgalaxy 的全部 Issue

相似的 Issue

更多 Python Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。