TornadoVM lets Java programs execute on accelerators. llm.c is a plain C implementation of OpenAI‘s GPT-2, the LLM that powered the 1st ChatGPT. Released in fall ‘22, it sparked an AI hype that still lasts. Both are not a perfect fit at first glance, but a Java version of llm.c could make them friends, so I tried to bring them together. Although there was already a Java port of llm.c, I made my own to get (back) into the groove of Java.

A few years ago I wrote a Java app that creates star maps (example). It does this by projecting the coordinates of celestial bodies onto a flat canvas. One of the features is to map images of artistic representations of certain star constellations onto the maps. The approach I took to perform the required calculations turned out to be quite slow. When I heard about CUDA I was excited by the idea of doing computations on graphics cards.

An infographic about my first approach to parallelizing Java code in 2017. It worked for me then and probably still does, but now there are tools that are much easier to use and are also much more flexible. One is TornadoVM, which essentially allows the programmer to mark up the code to be parallelized and does all the heavy lifting for execution on popular accelerators (AMD, Intel, NVIDIA) and multiple CPU cores.