cuda-cutlass-fmha-incremental-rebuild Skill
Use when rebuilding ONNX Runtime CUDA after editing CUTLASS fused-MHA headers (onnxruntime/contribops/cuda/bert/cutlassfmha/.h such as kernelforward.h or fmhalaunchtemplate.h), or when a header edit "passed" an incremental build but test behavior did not change. Explains the nvcc depfile gotcha that produces stale Memo Published by microsoft in onnxruntime.
What is cuda-cutlass-fmha-incremental-rebuild Skill?
Use when rebuilding ONNX Runtime CUDA after editing CUTLASS fused-MHA headers (onnxruntime/contribops/cuda/bert/cutlassfmha/.h such as kernelforward.h or fmhalaunchtemplate.h), or when a header edit "passed" an incremental build but test behavior did not change. Explains the nvcc depfile gotcha that produces stale Memo Published by microsoft in onnxruntime. This profile combines repository metadata with install, compatibility, and usage signals so developers can quickly decide whether it fits their agent workflow before opening the source repository.
Automated repository signals based on public metadata such as recency, license, installation evidence, and adoption. These are not a security audit or endorsement.
Key capabilities
- Includes SKILL.md support
- Reusable instructions support
- Testing
- Testing use cases
Technical details
- Install or run with Copy skill directory
When to use cuda-cutlass-fmha-incremental-rebuild Skill
- Use it for testing.
Built with
Editorial notes
Source
- Creator: microsoft
- Repository: microsoft/onnxruntime
- Skill file: .agents/skills/cuda-cutlass-fmha-incremental-rebuild/SKILL.md
What it does
Use when rebuilding ONNX Runtime CUDA after editing CUTLASS fused-MHA headers (onnxruntime/contribops/cuda/bert/cutlassfmha/.h such as kernelforward.h or fmhalaunchtemplate.h), or when a header edit "passed" an incremental build but test behavior did not change. Explains the nvcc depfile gotcha that produces stale Memo
Skill instructions
Incremental rebuilds silently use STALE CUTLASS fused-MHA kernels The general false-green principles (stale binary, wrong-artifact mtime) are summarised in the ort-test skill's "False-green taxonomy". This skill is the CUDA/CUTLASS-specific detail. The gotcha (verification-integrity bug) nvcc-generated depfiles do not track the CUTLASS fused-MHA headers under onnxruntime/contribops/cuda/bert/cutlassfmha/ (e.g. kernelforward.h, fmhalaunchtemplate.h). These headers are included by the fmhasm.cu translation units, but the build system does not record that dependency. Consequence: after you edit one of those headers, an incremental build.sh: - does not recompile fmhasm.cu, - reports [100%] Built target ... and exits 0, - leaves the recompiled artifacts — the fmhasm.cu.o objects and the libonnxruntimeproviderscuda.so they link into — unchanged (same mtime as the pre-edit build). (Do not use the gtest test-exe mtime as the stale symptom: in the shared-provider build the exe dlopens the .so a
Explore related resources
Frequently asked questions
What is cuda-cutlass-fmha-incremental-rebuild?
cuda-cutlass-fmha-incremental-rebuild is a open-source AI agent skill with Copy skill directory. Use when rebuilding ONNX Runtime CUDA after editing CUTLASS fused-MHA headers (onnxruntime/contribops/cuda/bert/cutlassfmha/.h such as kernelforward.h or fmhalaunchtemplate.h), or.
Who is cuda-cutlass-fmha-incremental-rebuild best for?
cuda-cutlass-fmha-incremental-rebuild is best for reusing agent instructions, scripts, and references, testing workflows.
How do I install cuda-cutlass-fmha-incremental-rebuild?
Install or run cuda-cutlass-fmha-incremental-rebuild using Copy skill directory. Check cuda-cutlass-fmha-incremental-rebuild for the latest setup command.
Is cuda-cutlass-fmha-incremental-rebuild actively maintained?
cuda-cutlass-fmha-incremental-rebuild may need a closer maintenance check before production use.
Auto-fetched from GitHub.
