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

metal : clean up

Browse source
This commit is contained in:
Georgi Gerganov 2024-01-25 13:29:45 +02:00
parent f6416d4493
commit 2bf91c5306
No known key found for this signature in database
GPG key ID: BF970631944C16B7
2 changed files with 41 additions and 47 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-metal.m
View file

@ -2253,9 +2253,9 @@ static bool ggml_metal_graph_compute(
[encoder setBytes:&ne3 length:sizeof( int64_t) atIndex:26];
[encoder setBytes:&scale length:sizeof( float) atIndex:27];
const int64_t nsg = 2; // simdgroups per threadgroup (a.k.a. warps)
const int64_t nsg = 4; // simdgroups per threadgroup (a.k.a. warps)
const int64_t nqptg = 8; // queries per threadgroup !! sync with kernel template arguments !!
const int64_t ncpsg = 32;
const int64_t ncpsg = 32; // cache values per simdgroup (does not work for other values)
//const size_t smem = nqptg*(nhptg*ne00 + nsg*(nhptg*ne00 + 256))*(sizeof(float)/2);
const size_t smem = nqptg*(ne00 + nsg*(ne00 + 1*ncpsg))*(sizeof(float)/2);