Deep DRC can miss violations when a cell has many shapes

未关闭
  • 哪里不对？
#2,546 5 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。

  • #2547 来自 @nikosavola —— 已关闭，未合并

评估

难度
4/5
预计耗时
3-5 天
新手友好度
52/100
Issue 类型
缺陷
描述清晰度
描述清楚
活跃度
活跃
技术栈
cpp
领域
computer-graphics, testing

调研方向

Read the box_scanner2 sweep and the instance to shape interaction scan in dbHierProcessor.cc, focusing on how reported pairs are keyed and how layer ids enter the scan. Reproduce the issue with the small box_scanner2 test described in the report, using duplicate insertions with different properties and enough boxes to cross the threshold. Done means the sweep reports both pairs and the deep DRC result retains all 4296 markers.

由索引模型根据 Issue 内容生成。

描述

On a clipped deep DRC run of the vexriscv layout, a via1 in metal1 enclosure check loses 2127 of its 4296 markers. Forcing the brute force path everywhere gives all 4296. Almost every via either keeps all of its markers or loses all of them.

The cause is in the box_scanner2 sweep. It remembers the pairs it has already reported in a set, and that set is keyed on the two object pointers only. In the instance to shape interaction scan in dbHierProcessor.cc the same shape can be inserted twice with two different intruder layer ids. The second pair then looks like a duplicate and is dropped. Small cells don't use the sweep, so there both pairs are reported and the result depends on how many shapes a cell has. A small box_scanner2 test shows the same thing when one object is inserted twice with different properties and there are enough other boxes to go over the threshold.

主要语言
C++
星标
1.2k
派生
305
平均合并
3 天 2 小时
30 天内合并 PR
11

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

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。

KLayout/klayout 的其他 Issue

查看 KLayout/klayout 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。