NVIDIA/cuda-tile

CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure for CUDA kernel optimization, focusing on tile-based computation patterns and optimizations targeting NVIDIA tensor core units.

Owner: NVIDIA (organization)
Stars: Stars: 1,050
Forks: Forks: 91
Language: C++
License: None
Created: 2025
Tier: Beta (700 to 1,999 ★) What this tier means

View on GitHub

More repos by NVIDIA

All 269 repos by NVIDIA

NVIDIA/SkillSpectorStars: 23,897PythonLTS
NVIDIA/NemoClawStars: 22,915TypeScriptLTS
NVIDIA/OpenShellStars: 20,930RustLTS
NVIDIA/Megatron-LMStars: 18,312PythonStable
NVIDIA/nvidia-dockerStars: 17,607Stable
NVIDIA/open-gpu-kernel-modulesStars: 17,544CStable

Nearby in C++

All C++ repos

KhronosGroup/Vulkan-ValidationLayersStars: 1,050C++Beta
MOLAorg/molaStars: 1,050C++Beta
rl-tools/rl-toolsStars: 1,050C++Beta