docs · qxmx · Tangled

A single-model (Bonsai-27B) inference engine for Intel Arc GPUs

qxmx docs

2 files

· hoist loop-invariant address math in gemv + FFN GEMM kernels (decode +4.5%, prefill +6.7%) · 3mo ago

fused_fa_problem.md

M1.7c: fused FlashAttention via sycl::joint_matrix -- 741 tok/s (+20%)

gemv_dpas8_source_map.md

hoist loop-invariant address math in gemv + FFN GEMM kernels (decode +4.5%, prefill +6.7%)