Add TornadoVM CUDA backend support and tensor-core (MMA) accelerated batch prefill (FP16 & Q8_0, Llama & Qwen3)#127

Merged

stratika merged 22 commits into

mainfrom

Jul 21, 2026

Commits on Jun 22, 2026

Commits on Jun 23, 2026

Revert specific workflow runner labels
orionpapadakis
committed

Commits on Jul 2, 2026

Convert QVK, Wo and W2 prefill projections to use MMA GEMM
mairooni
committed
Fuse QKV/GateUp MMA GEMMs, pipeline gemmMMA loads, and fix flash attention accumulation in FP16 batch prefill
mairooni
committed

Commits on Jul 7, 2026

Commits on Jul 10, 2026

Commits on Jul 19, 2026

Merge branch 'main' into feat/mma_cuda
mairooni
committed

Commits on Jul 21, 2026