TornadoVM 7 Enables GPU-Native Workloads from Java Applications
Comments
mikepapadim
Akseleu_omirbay
And what about TPU support? Is it compatible with Google TPUs or strictly limited to gpu backends like CUDA and Metal?
mikepapadim
no TPU support yet, just NVIDIA, Apple and any other HW thats supports OpenCL
zmmmmm
Seems like a nice win even if it just does the example they have:
void matmul(Matrix2DFloat A, Matrix2DFloat B, Matrix2DFloat C, int size) {
for (@Parallel int i=0; i<size; i++) {
for (@Parallel int j=0; j<size; j++) {
float sum = 0f;
for (int k=0; k<size; k++)
sum += A.get(i,k)*B.get(k,j);
C.set(i, j, sum);
}
}
}
And why confuse everybody by calling it a VM when it ships as a JAR file that runs in an existing VM?
pjmlp
I actually use the logging API, it is good enough, and maybe I am old enough not to like to cowboy install dependencies if I can avoid it.
russli1993
what about tiling? seems better to call into existing kernels, fused megakernels than rewrite kernel in java.
mikepapadim
You can do both. It supports nvidia's cuTile programming model through Tilecontext API or you can write megakernels in Java and compile then to CUDA or even add native libraries in the mix like cudnn. JITllm is a good example how all these can work together
zmmmmm
i might be overly optimistic, but I'm assuming it'll do the tiling by itself with the above. As long as there's not an algorithmic reason to do it a specific way, in theory you could just optimise it for the target hardware.
mikepapadim
For now it supports directly cuTile through what they call TileContext API. Is a 1-1 mapping of the cuTile C++ implementation.
the_arun
Does this mean we can use Java to build AI infrastructure?
vlowrian
Or an abstract AI infrastructure factory?
mikepapadim
https://github.com/beehive-lab/jitllm there is already an vllm-like inference engine that uses tornadovm as the GPU backend all written in java
keeda
That’s probably the next thing after Software Factories.
kitd
I'll be disappointed if this isn't named AbstractAiInfrastructureFactory
pjmlp
These kind of complaints get tiresome, here for some Objective-C fun,
mike_hearn
Yes (or any JVM lang). There is also this:
... which is a single JAR that does SIMD accelerated inferencing of local models, dependency free. Everything is written in Java from the tokenizer up. It can also use Metal on macOS. Small models can be surprisingly powerful these days and all you need is a fast CPU. It works not only for LLMs but also image models, video models and speech.
Remarkably, it is significantly faster than llama.cpp on some tasks like prefill (decode is memory bandwidth constrained so language hardly makes a difference). At least as long as you run on GraalVM (JVM or native image, both work). So you can actually beat C++ performance for AI in Java.
It's impressive stuff.
vips7L
Why GraalVM specifically? Is it something special with the Graal JIT?
mike_hearn
It optimizes more aggressively, yes.
amir734jj
Basically JAX for Java instead of Python
mikepapadim
Many components for a jax for java are already tested in https://github.com/beehive-lab/jitllm
mathisfun123
"your scientists were so preoccupied with whether they could they didn't stop to think if they should"
https://github.com/beehive-lab/TornadoVM this is main repo.
You can use it with any JDK21+ and Nvidia or Apple hardware and their Nvidia libs or MLX libs.