nicolas
/
ik_llama_opt
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Fix more Q8_0 repacking mess on AVX2 (#719) 

Co-authored-by: Iwan Kawrakow <iwan.kawrakow@gmail.com>
This commit is contained in:
Kawrakow 2025-08-23 09:04:51 +03:00 committed by GitHub
parent 866145b2b9
commit 8e23ecdd96
1 changed files with 2 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/iqk/iqk_mul_mat.cpp
View File

@ -502,10 +502,8 @@ extern "C" IQK_API bool iqk_mul_mat(long Nx, long Ny, long ne00,
MulMat mm;
auto etypeA = ggml_type(typeA);
if (auto dequant_type = MulMat::is_dequant_better(etypeA, Ny); dequant_type != etypeA) {
if (!MulMat::prepare(dequant_type, typeB, ne00, mm, Ny)) {
return false;
}
if (auto dequant_type = MulMat::is_dequant_better(etypeA, Ny);
dequant_type != etypeA && MulMat::prepare(dequant_type, typeB, ne00, mm, Ny)) {
constexpr int k_x_step = 32;