Skills
.Work. You
Rest

Optimize Triton Block Parameters

Official

Tune Triton block sizes for peak kernel speed.

Software Engineering#throughput#triton#occupancy#gemm#autotuning#block-size#gpu-kernels
Authortensormux
Version1.0.0
Installs0

System Documentation

What problem does it solve?

This skill guides the agent to systematically select and tune Triton launch parameters (BLOCK_M, BLOCK_N, BLOCK_K, num_warps, and num_stages) to maximize GEMM-style kernel throughput while respecting hardware constraints.

Core Features & Use Cases

  • Systematic autotune for block sizes and parallelism to improve performance on GPUs such as A100 and H100.
  • Shape- and dtype-aware configuration that preserves correctness and efficiency across representative problem instances.
  • Reproducible benchmark-driven workflow that documents the winning configuration and expected throughput.

Quick Start

Provide a representative Triton GEMM kernel and hardware target, then run the autotune workflow to discover the optimal BLOCK_M, BLOCK_N, BLOCK_K, and related parameters.

Dependency Matrix

Required Modules

None required

Components

Standard package

💻 Claude Code Installation

Recommended: Let Claude install automatically. Simply copy and paste the text below to Claude Code.

Please help me install this Skill:
Name: Optimize Triton Block Parameters
Download link: https://github.com/tensormux/kernel-skills/archive/main.zip#optimize-triton-block-parameters

Please download this .zip file, extract it, and install it in the .claude/skills/ directory.
View Source Repository

Agent Skills Search Helper

Install a tiny helper to your Agent, search and equip skill from 510,000+ vetted skills library on demand.