Tag: GPU Programming
-
NVIDIA’s Christmas Surprise: Open Sourcing CUDA Tile IR to Break the Moat
As a Christmas gift to the open-source community, NVIDIA has released the CUDA Tile IR intermediate representation as open source, abandoning its proprietary licensing model. Earlier in December, alongside the launch of CUDA 13.1, the company unveiled the CUDA Tile programming model, describing it as “the largest and most far-reaching update to the CUDA platform…
-
NVIDIA CUDA 13.1: ‘CUDA Tile’ Abstraction Simplifies High-Level GPU Programming
NVIDIA has announced the most significant update to the CUDA platform since its inception in 2006. With CUDA 13.1, the company is introducing CUDA Tile, a new paradigm for GPU programming built on a block-oriented, or tile, model. Its defining feature is that developers describe algorithms at a much higher level, without binding themselves to…