ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

metal : use FA-vec kernel up to batch size 20 (#13496)

Browse source 
* batched-bench : fix pp batch contents

* metal : optimize multi-sequence FA vec kernel

ggml-ci

* metal : use FA-vec kernel up to batch size 20

ggml-ci
This commit is contained in:
Georgi Gerganov 2025-05-13 18:04:39 +03:00 • committed by GitHub
parent c252e0c409
commit f0995d28ce
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml/src/ggml-metal/ggml-metal.m
View file

@ -4358,7 +4358,7 @@ static bool ggml_metal_encode_node(
// TODO: add vec kernels for (ne00%64 == 0) and maybe also for (ne00%32 == 0) // TODO: add vec kernels for (ne00%64 == 0) and maybe also for (ne00%32 == 0)
// for now avoiding mainly to keep the number of templates/kernels a bit lower // for now avoiding mainly to keep the number of templates/kernels a bit lower
// these are now trivial to add after: https://github.com/ggml-org/llama.cpp/pull/12612 // these are now trivial to add after: https://github.com/ggml-org/llama.cpp/pull/12612
if (ne01 >= 4 || (ne00%128 != 0 && ne00 != 96 && ne00 != 192 && ne00 != 576)) { if (ne01 >= 20 || (ne00%128 != 0 && ne00 != 96 && ne00 != 192 && ne00 != 576)) {
switch (src1->type) { switch (src1->type) {
case GGML_TYPE_F16: case GGML_TYPE_F16:
{ {