root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-25 02:44:36 +00:00
Code Issues Actions 9 Packages Projects Releases Wiki Activity

vulkan: increase small tile size for NV_coopmat2

Browse Source
This commit is contained in:
Jeff Bolz 2024-12-21 22:36:56 -06:00
parent 26252831ac
commit e52a0f28e7
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-vulkan/ggml-vulkan.cpp
View File

@ -1404,10 +1404,10 @@ static void ggml_vk_load_shaders(vk_device& device) {
// spec constants and tile sizes for non-quant matmul/matmul_id // spec constants and tile sizes for non-quant matmul/matmul_id
l_warptile = { 256, 128, 256, 64 }; l_warptile = { 256, 128, 256, 64 };
m_warptile = { 256, 128, 128, 64 }; m_warptile = { 256, 128, 128, 64 };
s_warptile = { 128, 32, 16, 64 }; s_warptile = { 128, 64, 64, 64 };
l_wg_denoms = {128, 256, 1 }; l_wg_denoms = {128, 256, 1 };
m_wg_denoms = {128, 128, 1 }; m_wg_denoms = {128, 128, 1 };
s_wg_denoms = { 32, 16, 1 }; s_wg_denoms = { 64, 64, 1 };
// spec constants and tile sizes for quant matmul (non-Qi_K) // spec constants and tile sizes for quant matmul (non-Qi_K)
l_warptile_mmq = { 256, 128, 256, 64 }; l_warptile_mmq = { 256, 128, 256, 64 };