GPU Compiler Software Development

Project Description:

Working on GPU support for OpenAI/Triton — a language and compiler for writing highly efficient custom Deep-Learning primitives. Work with the open-source community to analyze, develop, test, and deploy performance improvements for neural networks implemented with Triton on GPUs with ROCm.

Responsibilities:

New features development, support and optimization of OpenAI/Triton project for GPUs. Communication with other developers, customers and project managers. Test implementation, project documentation and verification of the system with unit/component/functional tests

Mandatory Skills Description:

Strong C/C++ programming skills

Experience with compiler internals (llvm, gcc or any other)

Basic Python programming skills

Experience in performance analysis

Nice-to-Have Skills Description:

• Basic understanding of ML technologies

• Experience with GPGPU (General purpose GPU) computing (HIP, CUDA, OpenCL, etc.)

• Experience with PyTorch

• Experience with LLVM and MLIR compiler infrastructure, analysis or optimizations implementation

• Knowledge of ROCm infrastructure

• Experience in CMake, make/ninja build system

• GEMM performance fundamentals

• Experience with Docker