peonist-ai / halogen-flash-server

The fastest way to run Qwen3.8-Flash-Next on Strix Halo (gfx1151)

DryhoppedIPA / bc250-gfx1013-fix

Kernel + Mesa/RADV patches that fix the ASRock BC-250's broken GPU compute queues: ~20% FPS from async compute