Atlas C Operator Development
Develop operators using Atlas C
Operator Development
Design and implement foundational neural network units. It empowers developers to efficiently run innovative algorithms. Addressing operator development, CANN provides versatile approaches to streamline development, tuning, and deployment, reducing costs for AI developers:
- C/C++ support to align with developer habits
- CATLASS operator template library: faster matrix operator customization through premium code samples and modular components
- Triton support: empowering Python developers with simplified operator development
CATLASS Operator Development
Develop general matrix multiply (GEMM) and fused operators using the operator template library
Triton Operator Development
Tile-granular Python-based programming with automated performance optimization
'%20fill-opacity='1.000000'%20fill-rule='evenodd'/%3e%3c/g%3e%3c/svg%3e)
Get Resources
Download CANN
Install the CANN Toolkit to prepare your workspace for operator development.