Skip to content
Agentmart
Submit a tool

pyqcu-qcu-single-tests

当任务涉及 pyqcu/testing/qcu 或 quda 的单功能测试、PyTorch 参考或布局转换校验时使用。

Skill by ZhangXin8069

IndexedMIT★ 4
Write a review

Chalk Labs

Need help deploying pyqcu-qcu-single-tests?

Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.

Book a call

Install

Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install skill/zhangxin8069-pyqcu-qcu-single-tests

Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.

Get the weekly diff

The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.

We use your email only for this newsletter. Privacy

About

当任务涉及 pyqcu/testing/qcu 或 quda 的单功能测试、PyTorch 参考或布局转换校验时使用。

Use cases: testing, ml

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this skill can do and what it asks for, from its SKILL.md and code. Checked Oct 26, 2026.

Tools
No tools found
Secrets
None asked for
Install scripts
None found

Network hosts it talks to 1

  • codeload.github.com · found in install command

Files it touches

  • ~/.claude/skills/pyqcu-qcu-single-tests

What's inside

Skill name
pyqcu-qcu-single-tests
Description
当任务涉及 pyqcu/testing/qcu 或 quda 的单功能测试、PyTorch 参考或布局转换校验时使用。
License
MIT
Instructions
49 lines in SKILL.md · view on GitHub

Files (2)

  • AGENTS.md
  • SKILL.md

SKILL.md (first lines)

# QCU/QUDA 单功能测试

## 适用范围

用于维护 `pyqcu/testing/qcu/single_qcu_*.py` 与 `pyqcu/testing/quda/test_*.py`。每个文件只覆盖一个接口或
一个紧密的接口族,输入使用小格点和固定随机种子,先做纯 PyTorch 参考,再按环境选择 CUDA/QUDA
扩展运行。缺少 CUDA、`libqcu.so`、pyquda 或 QIO 资产时必须明确 skip,不能伪造通过。

当前端到端正确性格点统一为 `8x8x8x16`。QCU/QUDA 单功能入口和 simple MPI wrapper
默认必须使用该格点;`16^4`、`8x8x16x16`、`4^4` 只可保留为显式性能、容量、
合成契约或历史数据测试,并必须在报告中标注。

## QCU 生命周期

控制张量固定为 `params=int32[58]`、`argv=real[8]`(槽 7 为 `csw`)、
`set_ptrs=int64[100]`。每次普通单操作事务严格执行:

```text
applyInitQcu(set_ptrs, params, argv)
operation(..., set_ptrs, params)
applyEndQcu(set_ptrs, params)
params[_SET_INDEX_] += 1
```

`params[_SET_PLAN_]` 使用 -2 Laplacian、-1 Gauss、0 Wilson dslash、1 求解器、2 Clover;QCU
数据为 `[2,3,3,4,X,Y,Z,T/2]`、`[2,4,3,X,Y,Z,T/2]`,纯 PyTorch 参考先经
`tools.poooxyzt2oooxyzt` 还原到 `[... ,X,Y,Z,T]`。

单接口数值语义:Wilson dslash 的输入/输出是相反 parity 的 compact 场,C++ 返回未乘 `kappa`
的 hopping;Clover dslash 随后作用 `(I + T_clover)^{-1}`;Laplacian 是内部 `T=1` 的三维颜色场,
gauge 轴为 `[colour_out, colour_in, direction, X, Y, Z]`,三者分别使用
`single_function_common.py` 中的对应参考函数,禁止用 full-site matvec 直接比较。

## QUDA 排布

PyQUDA/QDP 规范场为 `[4,T,Z,Y,X,3,3]`,费米场为 `[2,T,Z,Y,X/2,4,3]`,棋盘奇偶按
`(x+y+z+t)%2`,与 PyQCU 的 t 压缩布局不同。所有对照脚本必须通过
`pyqcu/testing/quda/common.py` 的显式往返转换,禁止直接 reshape 冒充转换。

## 验收

```bash
python pyqcu/testing/qcu/single_qcu_api.py
python pyqcu/testing/qcu/single_qcu_wilson_dslash.py
pytest pyqcu/testing/quda
git diff --check
```

设置 `QCU_STRICT_NUMERIC=1` 才把跨架构数值偏差升级为失败;默认运行仍记录相对误差,便于诊断
不同 GPU/精度下的归一化差异。新增接口或独立项目完成后,必须依照 `pyqcu/testing/AGENTS.md` 整理
测试入口并更新本技能的接口表。

From ZhangXin8069/PyQCU, MIT.

More from ZhangXin8069/PyQCU

Signals

Stars
4
Forks
0
Open issues
0
Last commit (this folder)
Oct 22, 2026 (17 days ago)

From GitHub (fetched Oct 24, 2026).

Details

License
MIT
First seen
Oct 24, 2026
Last checked
Oct 24, 2026

More from ZhangXin8069

Done for you

Rather not set it up yourself?

Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.

See how we build AI agents →Read the case studies

Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.

launches shipped
50+
launches shipped
media placements
500+
media placements
founders on every account
2
founders on every account

Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block

Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.

Get pyqcu-qcu-single-tests running in your business

Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.

Goes to the Chalk Labs founders, not the newsletter. Privacy

Need help deploying pyqcu-qcu-single-tests?

CallGet help
Cite this page

pyqcu-qcu-single-tests skill. Agent Mart. https://agentmart.dev/skills/zhangxin8069-pyqcu-qcu-single-tests (last updated 2026-10-24)