Skip to content
Agentmart
Submit a tool

pyqcu-cpp-containers

当任务涉及 cpp/ 下 CUDA、KUPL、CANN、DTK、MACA 后端容器结构或构建入口时使用。

Skill by ZhangXin8069

IndexedMIT★ 4
Write a review

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-cpp-containers

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 cpp/ 下 CUDA、KUPL、CANN、DTK、MACA 后端容器结构或构建入口时使用。

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-cpp-containers

What's inside

Skill name
pyqcu-cpp-containers
Description
当任务涉及 cpp/ 下 CUDA、KUPL、CANN、DTK、MACA 后端容器结构或构建入口时使用。
License
MIT
Instructions
54 lines in SKILL.md · view on GitHub

Files (2)

  • AGENTS.md
  • SKILL.md

SKILL.md (first lines)

# cpp/ — 多后端容器目录

`cpp/` 下的目录镜像同一结构:容器层只做后端分组,真实现(或 stub)在各自的
`qcu/` 子目录。新增后端时按同一形状建 `cpp/<backend>/{AGENTS.md,qcu/}`。

CUDA/DCU/GPGPU 生产运行遵循一 Python 进程 = 一 MPI rank = 一设备;容器层
不得把单进程多设备 worker 作为默认后端拓扑。`MultiGpuMultigrid` 是显式调用的
测试专用例外,规则见
[`pyqcu-cuda/SKILL.md`](../pyqcu-cuda/SKILL.md#cuda-production-topology)。

| 容器 | 状态 | 内容 |
|---|---|---|
| `cpp/cuda/` | **真实现** | `qcu/` 是 PyQCU 主 C++ CUDA 后端,构建产物 `libqcu.so` |
| `cpp/kupl/` | **真实现** | `qcu/` 是独立 MPI+OpenMP CPU 后端;Python 经 `pyqcu.kupl` 使用 |
| `cpp/cann/` | 占位 | `qcu/PASS` 空文件,无活跃昇腾 CANN C++ 后端 |
| `cpp/dtk/` | 占位 | `qcu/PASS` 空文件,无活跃 DCU/ROCm(HIP) C++ 后端 |
| `cpp/maca/` | 占位 | `qcu/PASS` 空文件,无活跃 Maca C++ 后端 |

`cpp/cuda/` 目前不放共享工具或顶层 `CMakeLists.txt`——构建入口仍由
`cpp/cuda/qcu/{env.sh,make.sh,CMakeLists-nv.txt}` 自持;容器目录存在的意义是保持
后端结构对称,便于后续 CANN/DTK/Maca 后端接入时路径不变。

## 子目录技能

各子目录维护自己的现行技能文档;细节以那些文档为准,本文**不复制**其内容(此前
每个容器 skill 内嵌全部子目录正文,2026-10-15 整理时改为链接,避免多处副本漂移):

| 子目录 | 技能 |
|---|---|
| `cpp/cuda/qcu/` | `skills/pyqcu-cpp-qcu/SKILL.md` |
| `cpp/cuda/qcu/include/` | `skills/pyqcu-cpp-qcu-include/SKILL.md` |
| `cpp/cuda/qcu/src/` | `skills/pyqcu-cpp-qcu-src/SKILL.md` |
| `cpp/cuda/qcu/python/` | `skills/pyqcu-cpp-qcu-python/SKILL.md` |
| `cpp/kupl/qcu/` | `skills/pyqcu-kupl/SKILL.md` |
| ~~`cpp/cuda/qcu/logs/`~~(2026-10-15 已删除) | `skills/pyqcu-cpp-qcu-logs/SKILL.md`(归档说明) |

## cpp/cuda/qcu 速查(详规见 `pyqcu-cpp-qcu`)

- **构建**:`source ./env.sh && bash ./make.sh`(`make.sh` 把 `CMakeLists-nv.txt`
  软链为 `CMakeLists.txt` 后 cmake+make);仓库根 `bash ./build.sh` 是同一流程的包装。
  `env.sh` 自动探测当前 NVIDIA GPU 的 compute capability,CMake 生成对应
  cubin + PTX;需要复现或跨机型构建时用 `QCU_CUDA_ARCHITECTURES` 显式覆盖,
  并把架构写进 provenance。
- **三数组 ABI**:`params`(int32[58]) / `argv`(real[8],c64 为 float32、c128 为
  float64) / `set_ptrs`(int64[100]);`_SET_PLAN_`(params[16]) 选内核路径。
- **`_SET_INDEX_` 生命周期**:普通事务为
  `applyInitQcu → operation → applyEndQcu → _SET_INDEX_ += 1`;
  Strict 生命周期内复用同一槽位、不递增。
- **Clover MultiGrid 5 stream 架构**与 **Critical Invariants**(标量只住
  `device_vals`、每轮全流同步、点积走 `_send_tmp_` 槽 7、`mpi_real_type<T>()` 分派、
  `run_mpi` 用阻塞 `MPI_Sendrecv`)见 `pyqcu-cpp-qcu/SKILL.md`。
- **`_BLOCK_SIZE_`**(`include/define.h`):小格点测试 8/16,NVIDIA 生产 128,
  AMD DCU 生产 256。

From ZhangXin8069/PyQCU, MIT.

More from ZhangXin8069/PyQCU

Signals

Stars
4
Forks
0
Open issues
0
Last commit (this folder)
Oct 22, 2026 (6 days ago)

From GitHub (fetched Oct 24, 2026).

Details

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

More from ZhangXin8069

Cite this page

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