NVIDIA's April Fools' Day post announces cuTile BASIC, a tongue-in-cheek implementation of the CUDA Tile programming model in the BASIC language. Built on CUDA 13.1's CUDA Tile IR, it lets developers write tile-based GPU kernels using classic BASIC syntax with line numbers. The post includes working code examples for vector addition and matrix multiplication (GEMM), complete with installation instructions via pip and actual GitHub repository links. The joke highlights CUDA Tile's language-agnostic design — any language can target CUDA Tile IR — while poking fun at legacy programming nostalgia. An editor's note teases cuTile COBOL for April 1, 2027.
Table of contents
What is cuTile BASIC?Who is cuTile BASIC for?Get setupcuTile BASIC examplePutting it all togetherA BASIC matrix multiplicationHow developers can get cuTileGet startedCUDA Tile in any languageSort: