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

Fix small matrix multiplication errors in AMD GPUs on Windows or with amdvlk

Browse source 
Co-authored-by: Engininja2 <139037756+Engininja2@users.noreply.github.com>
This commit is contained in:
0cc4m 2024-01-31 06:26:32 +01:00
parent ffe841cdd9
commit 7c8cf29925
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
ggml-vulkan.cpp
View file

@ -817,7 +817,7 @@ static void ggml_vk_load_shaders() {
// mulmat
std::initializer_list<uint32_t> warptile_l = { 128, 128, 128, 16, vk_device.subgroup_size * 2, 64, 2, 4, 4, vk_device.subgroup_size };
std::initializer_list<uint32_t> warptile_m = { 128, 64, 64, 16, vk_device.subgroup_size, 32, 2, 4, 2, vk_device.subgroup_size };
std::initializer_list<uint32_t> warptile_s = { vk_device.subgroup_size, 32, 32, 8, 32, 32, 2, 2, 2, vk_device.subgroup_size };
std::initializer_list<uint32_t> warptile_s = { vk_device.subgroup_size, 32, 32, 16, 32, 32, 2, 2, 2, vk_device.subgroup_size };
std::array<uint32_t, 3> l_wg_denoms = {128, 128, 1 };
std::array<uint32_t, 3> m_wg_denoms = { 64, 64, 1 };