NVIDIA introduced a tool for programming video cards in BASIC language

NVIDIA announced the cuTile BASIC project, designed to perform calculations on GPUs using the BASIC language.

Demonstration of the new CUDA Tile paradigm class=»notranslate»>__GTAG8__

The release is timed to coincide with April Fools' Day and is a demonstration of the capabilities of the new CUDA Tile paradigm introduced in version 13.1 of the CUDA software toolset. The developers have shown that thanks to the flexibility of intermediate code representation, hardware acceleration can be adapted to almost any existing syntax, notes xrust. Thus, NVIDIA engineers clearly illustrated how modern GPU technologies can work with syntax that was used long before the advent of multi-threaded programming.

Technical capabilities of the tool

In the published materials, the authors of the project described in detail the process of creating basic mathematical operations, as well as algorithms for working with large matrices. Users are given the opportunity to combine the processing power of modern GPUs with the syntactic simplicity of the BASIC language. Engineers have prepared fully working code examples, including vector addition operations, that can be run directly on real hardware through the appropriate software environment.

Hardware requirements

For this tool to work correctly, you need an NVIDIA video card with eighth generation architecture or later, as well as the latest drivers installed. This ensures that the joke compiler is compatible with the company's current hardware.

Project availability

The cuTile BASIC source code is published in an open repository on the GitHub platform. Enthusiasts and developers can freely download the materials and experience for themselves the unusual symbiosis of retro programming with advanced GPU technologies.

Ironic comment from the developers

In the final part of the publication, the authors of the project noted in a restrained ironic form that “the community should more accurately formulate its technical requests.” In addition, they promised to prepare a similar compiler for the COBOL language by April 1, 2027.

Context

The cuTile BASIC project was a direct demonstration of how flexible the new CUDA Tile paradigm can be in version 13.1. The company's engineers emphasized that the intermediate representation of code opens up the possibility of adapting hardware acceleration to various syntaxes, including the earliest programming languages ​​that predate the era of multi-threaded computing. This allows you to combine the high-performance capabilities of modern GPUs with the minimalistic approach to coding that was typical of BASIC.

Consequences

Although the initiative is humorous and dedicated to April Fools' Day, it highlights the technical flexibility of the CUDA platform. Developers and enthusiasts now have the opportunity to review the source code and test the tool on their own hardware. The promise to release a COBOL version by next April 1st indicates continued similar experiments in the future, which could bring additional attention to NVIDIA tools among the programming community.

By pages https://en.gamegpu

Xrust NVIDIA introduced a tool for programming video cards in the BASIC language