Hacktoberfest 2026:维护者为十月标记出来的 issue,仍然开放、适合新手。 浏览 Hacktoberfest issue

An exact whole power with a huge exponent hangs evaluation: 2^(10^9), 2^(3^20)

未关闭
#1,448 7 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
32/100
Issue 类型
缺陷
描述清晰度
需要澄清
活跃度
活跃
技术栈
csharp
领域
performance

调研方向

Start by reproducing the two expressions through ToEntity().Evaled and trace the Number.Pow/BinaryIntPow path described in the issue. Review the quantifier witness-search path and related issue #1338, then establish the chosen behavior for oversized exact powers and confirm that witness searches remain bounded without silently losing exactness.

由索引模型根据 Issue 内容生成。

描述

"2^(10^9)".ToEntity().Evaled and "2^(3^20)".ToEntity().Evaled do not return (killed at 60 s each; 2^(3^12) returns in 1 s). The result of the first is a 10⁹-bit integer (125 MB) and of the second a 3.5·10⁹-bit one; Number.Pow's BinaryIntPow builds them by squaring in EInteger, and the last squarings are Toom-4 multiplications of numbers hundreds of megabytes long. Measured on 8fcfaf42 (a branch of 345914f6); the code path is unchanged since long before.

Found through a quantifier: forall n in ZZ* : 3^(n + 1) divides 2^(3^n) + 1 (Sullivan and Mackey's Prob 5.7.13) hangs because a witness search evaluates the body at n = 20 or so, where 2^(3^n) is the second number above. Any consumer that evaluates a tower at a large argument hits the same.

What the answer should be is the decision to make, and it belongs with #1338's numerics: an exact power whose result would exceed some size — say 2²⁴ bits, which is 2 MB and a fraction of a second — is either (a) left as written (2^(10^9) evaluates to itself, the way an unsolved statement stays a statement), or (b) refused with an exception naming the limit, or (c) evaluated approximately as Real at the working precision, which is what a decimal calculator does and loses exactness silently. (a) is the honest one under not answering is legitimate; answering wrongly is not, and it composes: a witness search that gets an unevaluated power back moves on. Until then, a quantifier's witness search should skip a sample it cannot evaluate in bounded time, which is the narrower fix I would take first.

Part of #1409.

🤖 Generated with Claude Code

https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura

主要语言
C#
星标
831
派生
79
平均合并
2 小时 17 分钟
30 天内合并 PR
443

环境准备

  • 没有 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 阅读贡献指南

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。

ASC-Community/AngouriMath 的其他 Issue

查看 ASC-Community/AngouriMath 的全部 Issue

相似的 Issue

更多 C# Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。