# Kebab Codebase Navigation > Navigate and modify the Kebab CUDA codebase safely. Use when asked to locate GEMM, CuTe, CUDA baseline, benchmark, microbenchmark, profiling, or build logic in this repository, and when deciding where a change should be implemented. Use when this capability is needed. - Skill: `tomevault-io/kebab-codebase-navigation` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/kebab-codebase-navigation` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/kebab-codebase-navigation/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: AI & ML - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/kebab-codebase-navigation --- # Kebab Codebase Navigation ## When to Use This Skill - User asks where to change GEMM or elementwise kernels - User asks where build flags, CUDA architecture, or CMake targets are defined - User asks where benchmark/microbenchmark binaries are wired - User asks where profiling/reporting scripts live ## Project Map - Main CMake entry: `kebab/CMakeLists.txt` - Library target wiring: `kebab/lib/CMakeLists.txt` - Benchmark executables: `kebab/lib/benchmark/CMakeLists.txt` - Microbenchmark executables: `kebab/lib/microbench/CMakeLists.txt` - Top-level workflow automation: `Makefile` - Runtime/config knobs: `config.yaml` - Performance docs and design notes: `docs/` ## Step-by-Step Workflow 1. Determine requested domain first: build, runtime kernel, benchmark, microbenchmark, profiling, or report. 2. For build/toolchain issues, inspect `Makefile` and `kebab/CMakeLists.txt` first. 3. For kernel implementation issues: - CuTe path: `kebab/lib/cute/` - CUDA baseline path: `kebab/lib/cuda/` 4. For executable mapping, verify target names in: - `kebab/lib/benchmark/CMakeLists.txt` - `kebab/lib/microbench/CMakeLists.txt` 5. Cross-check that proposed commands exist in `Makefile` target templates. ## Troubleshooting | Issue | What to check | |---|---| | Command exists in docs but fails | Confirm target name in `Makefile` and generated binary path in `build/lib/...` | | GEMM variant confusion | Verify CUDA vs CuTe file under `kebab/lib/cuda/` vs `kebab/lib/cute/` | | Architecture mismatch | Inspect `CUDA_ARCH` handling in `Makefile` and `kebab/CMakeLists.txt` | ## References - `kebab/CMakeLists.txt` - `kebab/lib/CMakeLists.txt` - `kebab/lib/benchmark/CMakeLists.txt` - `kebab/lib/microbench/CMakeLists.txt` - `Makefile` --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/lancerlab) — claim your Tome and manage your conversions.