Skip to content

Commit 3edd87c

Browse files
shawngu-quiclhez
andauthored
opencl: optimize mxfp4 kernels (#16037)
- flatten mxfp4 and packed fp4->fp16 bit-wise convert function (replace lut) - MoE kernel optimizations --------- Co-authored-by: Li He <[email protected]>
1 parent c0b4509 commit 3edd87c

File tree

5 files changed

+701
-2
lines changed

5 files changed

+701
-2
lines changed

ggml/src/ggml-opencl/CMakeLists.txt

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -83,8 +83,10 @@ set(GGML_OPENCL_KERNELS
8383
mul_mv_q4_0_f32_1d_16x_flat
8484
mul_mv_q6_k
8585
mul_mv_mxfp4_f32
86+
mul_mv_mxfp4_f32_flat
8687
mul_mv_id_q4_0_f32_8x_flat
8788
mul_mv_id_mxfp4_f32
89+
mul_mv_id_mxfp4_f32_flat
8890
mul_mm_f32_f32_l4_lm
8991
mul_mm_f16_f32_l4_lm
9092
mul

0 commit comments

Comments
 (0)