Commit 5e782a5

committed

et-backend: Q8_0 matrix-engine GEMM kernel for prefill

Adds a Q8_0 matrix-engine mul_mat kernel with static-reuse dequant across N-tiles and K-splitting, plus fused bias support for MUL_MAT+ADD. Routes N >= 47 to the matrix engine and N <= 47 to vecdot. Registers the new kernel in the KERNELS list, which it was missing from, so the runtime kernel lookup was silently failing and leaving the destination buffer unwritten. Co-authored-by: Rehan Qasim <rehan.qasim@10xengineers.ai>

1 parent 7cbd610 commit 5e782a5Copy full SHA for 5e782a5

4 files changed

ggml/src/ggml-et
- CMakeLists.txt
- et-kernels/src
  - mul_mat_Q8_0_matrix_engine.c
  - platform.h
- ggml-et-ops.cpp

`‎ggml/src/ggml-et/CMakeLists.txt‎`

-Original file line number
+Diff line change
     mul_mat_id_Q4_0
     mul_mat_id_Q8_0
     mul_mat_Q8_0
 +    mul_mat_Q8_0_matrix_engine
     mul_mat_Q4_0
     mul_mat_Q4_0_matrix_engine
     mul_mat_f16

Comments

(0)