wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A collection of practical techniques for improving Triton Ascend kernel performance, including tile tuning, memory loading, reductions, broadcasting, and multi-pass fusion.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An example of a Triton Ascend program that launches two kernels in sequence, using a buffer to hold the first kernel's result.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A complete Triton Ascend example of LayerNorm, which normalizes values using statistics calculated in one stage and applied in a second stage.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A complete Triton Ascend example of matrix multiplication, which multiplies rows of one matrix by columns of another. It demonstrates two-dimensional blocks, accumulation across the shared dimension, masking, and Cube Core use.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A complete Triton Ascend example of ReLU, an element-by-element operation that replaces negative numbers with zero. It shows blocked iteration and masking at input boundaries.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A complete Triton Ascend example of softmax, an operation that turns a group of numbers into normalized scores. It shows finding the maximum, summing exponentials, and dividing by that sum.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
Integration examples for using Triton Ascend kernels inside MindSpore, a machine-learning framework. They show how to register a custom operation and pass tensors in and out.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
Triton Ascend hard API restrictions and forbidden syntax. MUST-follow rules that apply to every kernel: forbidden control flow (return/break/continue/lambda/while), tensor slice/index restrictions, scalar conversion rules, BLOCKSIZE upper bound. Violating any of these produces a compile or runtime error on Ascend.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A reference for Ascend hardware limits and compiler restrictions, including storage budgets, compiler limitations, and the performance effects of strided memory access.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An introduction to Triton Ascend kernel programming, including program IDs, blocks, grids, decorators, and common function structures. Triton is a language for writing parallel numerical kernels.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A troubleshooting checklist for Triton Ascend kernels, covering compiler errors, runtime failures, incorrect results, and slow performance.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
Guidance for choosing grid and block settings when launching Triton Ascend kernels. These settings control how work is divided across the NPU's processing units.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
Memory-access guidance for Triton kernels running on Ascend NPUs, Huawei's neural-network processors. It covers how data is stored, moved, combined, and prefetched.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A general performance guide for Triton kernels on Ascend hardware, covering block-size selection, grid setup, memory alignment, automatic tuning, and numeric precision.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An optimization guide for Transformer-style attention operations in Triton on Ascend hardware, including QKV computation, online softmax, and causal masking.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An optimization guide for Triton kernels that first transform each tensor element and then combine the results with a sum, mean, or other reduction.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An optimization guide for Triton kernels where every tensor element receives the same independent operation, such as addition, activation, casting, or clamping.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An optimization guide for matrix multiplication in Triton on Ascend hardware, covering tiled computation, work distribution, and large inner dimensions.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An optimization guide for reduction operations in Triton on Ascend hardware, including sums, maxima, cumulative operations, softmax, and normalization.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A troubleshooting guide for fixing mathematical function calls in Triton CUDA kernels, including the correct way to call tanh through CUDA's device library.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A reference for the main Triton CUDA functions, including loading and storing data, reductions, dot products, and atomic operations, with signatures and examples.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
An implementation guide for attention operations in Triton on CUDA, with a complete Flash Attention example and changes for causal, grouped-query, multi-query, and rotary-position variants.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A beginner's guide to programming Triton CUDA kernels, covering program identifiers, blocks, grids, kernel structure, decorators, and common code patterns.
wenyi-li/awesome-agent-kernel-skills
Skill Claude CodeCodex
A troubleshooting checklist for Triton CUDA kernels covering compilation errors, runtime errors, numerical-accuracy problems, and performance issues.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: