tilelang-vector-ascend-ops-migration
Skill43 starsChanged 8 months agoMulanPSL-2.0View on GitHub
本skill用于指导TileLang算子从GPU（CUDA）平台迁移到华为昇腾NPU平台。通过分析GPU实现，自动生成对应的NPU实现代码。 在迁移过程中，GPU 和 NPU 的 API 接口存在显著差异，强烈建议参考以下文档： 最常见错误1：使用逐元素处理而非分块处理 这是 NPU 迁移中*最致命且最容易犯的错误*。NPU 的向量化架构要求使用分块处理，而不是逐元素处理。 为什么必须使用分块处理： 推荐的 block 大小： 最常见错误2：Grid 大小计算位置错误 在迁移过程中，另一个*最容易犯且最高频的错误*是在 @T.prim_func 内部定义局部变量来计算 grid 大小。 原因： T.Kernel 的 grid 参数在编译阶段求值，此时函数内部的局部变量还不存在。 解决方案： 始终在…
In this repository66 other files it gives its agents.
- ascendc-operator-code-genSkill
- ascendc-operator-code-reviewSkill
- ascendc-operator-compile-debugSkill
- ascendc-operator-designSkill
- ascendc-operator-devSkill
- ascendc-operator-doc-genSkill
- ascendc-mssanitizerSkill
- ascendc-operator-performance-evalSkill
- ascendc-operator-performance-optimSkill
- ascendc-operator-precision-debugSkill
- ascendc-operator-precision-evalSkill
- ascendc-operator-project-initSkill
- ascendc-operator-testcase-genSkill
- ascend-dockerSkill
- ascend-inference-repos-copilotSkill
- ascend-npu-driver-installSkill
- ascend-profiling-anomalySkill
- ascend-transformer-boostSkill
- atc-model-converterSkill
- auto-bug-fixerSkill
- auto-develop-test-genSkill
- cann-nnal-installerSkill
- cann-operator-env-configSkill
- catlass-operator-code-genSkill
- catlass-operator-designSkill
- catlass-operator-devSkill
- catlass-operator-performance-optimSkill
- ascend-mmlab-install-suiteSkill
- model-migrationSkill
- model-trainingSkill
- ascend-model-migrationSkill
- ssh-dev-suiteSkill
- hccl-testSkill
- k8s-check-fixSkill
- megatron-change-analyzerSkill
- megatron-commit-trackerSkill
- megatron-impact-mapperSkill
- megatron-migration-generatorSkill
- mindspeed-fsdp2-config-migrationSkill
- mindspeed-fsdp2-data-migrationSkill
- mindspeed-fsdp2-migration-mainSkill
- mindspeed-fsdp2-model-migrationSkill
- mindspeed-fsdp2-verificationSkill
- modelscope-cliSkill
- msverl-daily-regression-triageSkill
- npu-adapter-reviewerSkill
- npu-smiSkill
- python-refactoringSkill
- security-code-reviewSkill
- simple-vector-triton-gpu-to-npuSkill
- skill-auditorSkill
- swanlab-setupSkill
- triton-operator-code-genSkill
- triton-operator-code-reviewSkill
- triton-operator-designSkill
- triton-operator-devSkill
- triton-operator-doc-genSkill
- triton-operator-env-configSkill
- triton-operator-performance-evalSkill
- triton-operator-performance-optimSkill
- The first 60 shown.
The file
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.