runmat

I'm on the RunMat team. Prior releases of RunMat focused on accelerating large math w/ auto offload to GPUs via transpiling to WGPU shaders, and correct interpretation of the MATLAB language semantics (we're 99%+ on grammar / semantics now). We left a placeholder for a Cranelift based JIT, and when we went back to fill it in, we though -- why not take it a bit further. Since we were putting in the logic for branches of the CFG to be compiled to native code, we figured why not fundamentally build it such that you could run "runmat compile simulation.m -o simulation" and get a standalone binary target. I.e. compile a MATLAB input project into to a standalone native binary target. We've detailed more in the blog. There's a bit of work we're finishing still, but a build with the core native compile pieces is now out (we still have to propagat

Hi, I’m Nabeel. In August I released RunMat as an open-source runtime for MATLAB code that was already much faster than GNU Octave on the workloads I tried. https://news.ycombinator.com/item?id=44972919 Since then, I’ve taken it further with RunMat Accelerate: the runtime now automatically fuses operations and routes work between CPU and GPU. You write MATLAB-style code, and RunMat runs your computation across CPUs and GPUs for speed. No CUDA, no kernel code. Under the hood, it builds a graph of your array math, fuses long chains into a few kernels, keeps data on the GPU when that helps, and falls back to CPU JIT / BLAS for small cases. On an Apple M2 Max (32 GB), here are some current benchmarks (median of several runs): * 5M-path Monte Carlo * RunMat ≈ 0.61 s * PyTorch ≈ 1.70 s * NumPy ≈ 79.9 s → ~2.8× faster than P