llama.cpp
ggml-cpu: implement MXFP4 SIMD for s390x
#16193
Merged

ggml-cpu: implement MXFP4 SIMD for s390x #16193

taronaeo
taronaeo ggml-cpu: impl mxfp4 s390x
618ef467
taronaeo ggml-cpu: missing s = sumf
6549e3bc
taronaeo ggml-cpu: fix incorrect kval_mxfp4 type
377d0fca
taronaeo ggml-cpu: rework mxfp4
35389304
taronaeo ggml-cpu: missing delta calc
cf927d8d
taronaeo ggml-cpu: fix typo
ae718c7d
taronaeo ggml-cpu: fix typo for vec_splats
f7e7539c
taronaeo ggml-cpu: expand to 2 blocks per loop
5fb1bb99
taronaeo ggml-cpu: add unroll to boost perf
4f85c33d
taronaeo ggml-cpu: back to 1 block per loop to test perf
1fe55724
taronaeo Revert "ggml-cpu: back to 1 block per loop to test perf"
1f99e511
taronaeo taronaeo requested a review from ggerganov ggerganov 358 days ago
taronaeo taronaeo requested a review from slaren slaren 358 days ago
github-actions github-actions added ggml
slaren
slaren approved these changes on 2025-09-24
taronaeo ggml-cpu: rm unroll from single block
96cba331
taronaeo
ggerganov
ggerganov approved these changes on 2025-09-26
ggerganov ggerganov merged 9b265118 into master 355 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone