How do I install the Kebab Microbench And Dump skill?

Run npx skillmds@latest add tomevault-io/kebab-microbench-and-dump in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Kebab Microbench And Dump safe to use?

This skill has not completed SkillMD's automated safety review yet. Independent scanners report: SkillSpector: PASS, Skill Scanner: PASS. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Kebab Microbench And Dump?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Kebab Microbench And Dump free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Kebab Microbench And Dump?

tomevault-io (@tomevault-io) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Kebab Microbench And Dump by tomevault-io

Name: Kebab Microbench And Dump
Author: tomevault-io

Kebab Microbench And Dump

Execute Kebab microbenchmarks and generate PTX/SASS dumps for kernel-level analysis. Use when asked to run mbench targets, inspect assembly output, compare copy or MMA kernels, or analyze low-level code generation. Use when this capability is needed.

tomevault-io Updated Sep 17, 2026

File contents

Kebab Microbenchmark and Dump

When to Use This Skill

User requests microbenchmark execution (mbench-*)
User needs SASS/PTX dumps for kernels
User wants to compare implementations like native/vectorized/PTX/CuTe copy paths

Prerequisites

Successful build (make build)
cuobjdump available in CUDA toolkit

Step-by-Step Workflows

Workflow A: Run Microbenchmark

Build:
- make build
Run one microbenchmark:
- make mbench-copy-gmem-to-smem
- make mbench-mma-wgmma
- make mbench-hgemm

Workflow B: Generate Kernel Dumps

Build:
- make build
Dump one microbenchmark binary:
- make mdump-copy-gmem-to-smem
Inspect outputs under:
- dump/microbench/<microbench_name>/

Workflow C: Operator Dumps

make dump-gemm-cute
make dump-gemm-cuda
make dump-gemm-ref

Outputs are under dump/operator/.

Troubleshooting

Issue	Mitigation
`cuobjdump not found`	Ensure CUDA toolkit bin directory is available and retry
No PTX generated	Some binaries may not embed PTX; SASS output is still usable
Dump files too large	Focus on split kernel files instead of `all_kernels.*`

References

Makefile (mbench-*, mdump-*, dump-*-cute/cuda/ref)
kebab/lib/microbench/CMakeLists.txt
dump/

Converted and distributed by TomeVault — claim your Tome and manage your conversions.

Frequently asked questions