665,000+ skills indexed

MindSpore

Official

@mindspore-ai

0Followers

|

44Public Repos

|

123Published Skills

Mirror repo for gitee.com/mindspore, PR and ISSUE Open

🐙 GitHub ↗🌐 Website ↗

Skills Distribution

Agent Skills by MindSpore

Showing 123 vetted skills indexed across 2 GitHub repositories.

task-constructor

Extract PyTorch/Triton operator implementations into single-file KernelBench tasks.

performance-summary

Summarize and compare operator performance results from result.json files.

kernel-workflow

Automate end-to-end AI kernel generation and optimization workflows across Triton, CUDA C, C++, and TileLang.

cuda-c-basics

Explain and implement basic CUDA-C kernel programming for GPU acceleration.

cuda-c-examples-torch

Compile and call CUDA C kernels from PyTorch using load_inline.

cuda-c-api

Reference CUDA C APIs for kernel launches and memory management.

cuda-c-optimization

Optimize CUDA C kernels for performance, stability, and debugging.

cuda-c-patterns

Provide CUDA-C kernel pattern templates for vector operations, reductions, and matrix multiplication.

pypto-case-elemwise-gelu

Generate a 1D GELU elementwise kernel with a tanh-free formula using PyPto.

pypto-case-matmul-2d

Generate tiled 2D matmul kernels with tail handling using the pypto DSL.

pypto-case-loss-crossentropy

Compute cross-entropy loss for batched Pypto kernels with two-stage tiling.

pypto-case-norm-batchnorm

Normalizes 3D tensors per-channel using Pypto BatchNorm-like operations.

pypto-case-reduction-sum

Reduce 3D tensors along one axis while preserving original dimensions in Pypto.

pypto-case-norm-layernorm

Reshape batched tensors to 2D and apply per-row LayerNorm with PyPto.

pypto-case-matvec

Substitute matmul with elementwise multiplication and sum for large-K matrix-vector multiplication.

pypto-optimization

Allocate and tune PyPTO optimization rules for kernel performance on selected backends.

pypto-pitfalls

Identify and prevent common PyPTO generation pitfalls in ascend backend kernel development.

pypto-loop-view

Enforce compile-time constant shapes for pypto.view inside pypto.loop kernels.

pypto-api

Document PyPTO API signatures and constraints for Ascend kernel development.

pypto-basics

Design deterministic PyPTO kernels for elementwise, matmul, and normalization patterns.

kernel-agent-overview

Guide kernel generation workflows from user intent to validated results.

cpu-optimization-x64

Optimizes C++ x86_64 CPU code with SIMD vectorization and cache-friendly layouts.

cpu-optimization-arm

Optimize ARM CPU compute kernels with NEON SIMD and loop unrolling.

cpu-basics

Explain foundational CPU kernel concepts and patterns for AKG operators.

Frequently Asked Questions About MindSpore

FAQPage Schema

What specific hardware backends are supported for kernel generation?▼

MindSpore supports kernel generation and optimization for Ascend NPU (Atlas A2/A3) and standard GPU architectures. It provides specialized DSLs including Triton-Ascend, TileLang, and CUDA-C to target these specific hardware backends for maximum compute efficiency.

Which engineering personas benefit from these operator development capabilities?▼

These capabilities are designed for deep learning infrastructure engineers, compiler developers, and performance engineers. They provide the necessary primitives and design patterns to implement custom operators, optimize memory access, and manage distributed sharding strategies for large-scale model training.

What are the primary dependencies for running generated kernels?▼

Generated kernels require a compatible environment including the MindSpore framework, appropriate NPU or GPU drivers, and the specific backend runtime (e.g., Ascend CANN or CUDA). Users must ensure the environment is provisioned with the necessary compiler toolchains for the chosen DSL.