wylab/llama.cpp-rocm
1
0
Fork 0
forked from wylab/llama.cpp
Code Pull Requests Actions 29 Packages Activity

fixes

Browse Source
This commit is contained in:
Ruben Ortlam
2026-04-10 06:52:55 +02:00
parent 90005edeea
commit c2f4d7aedf
1 changed files with 9 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/ggml-vulkan.cpp
+9 -3
View File
@@ -2912,7 +2912,7 @@ static vk_fa_tuning_params get_fa_tuning_params_scalar(const vk_device& device,
result.d_split = std::min(std::min(result.subgroup_size, 8u), D_lsb / 4);
result.shmem_staging = (hsk < 256 && hsv < 256) ? 1 : 0;
result.shmem_staging = (device->vendor_id == VK_VENDOR_ID_NVIDIA && hsk < 256 && hsv < 256) ? 1 : 0;
if (!reduce_block_rows && !ggml_vk_flash_attn_scalar_shmem_support(device, result, hsk, hsv, f32acc)) {
result.block_rows /= 2;
@@ -3446,6 +3446,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
CREATE_FA(GGML_TYPE_F32, f32, FA_SCALAR, )
CREATE_FA(GGML_TYPE_F16, f16, FA_SCALAR, )
#if defined(GGML_VULKAN_INTEGER_DOT_GLSLC_SUPPORT)
if (device->integer_dot_product) {
CREATE_FA(GGML_TYPE_Q4_0, q4_0, FA_SCALAR, _int8)
CREATE_FA(GGML_TYPE_Q8_0, q8_0, FA_SCALAR, _int8)
@@ -3453,7 +3454,9 @@ static void ggml_vk_load_shaders(vk_device& device) {
CREATE_FA(GGML_TYPE_Q5_0, q5_0, FA_SCALAR, _int8)
CREATE_FA(GGML_TYPE_Q5_1, q5_1, FA_SCALAR, _int8)
CREATE_FA(GGML_TYPE_IQ4_NL, iq4_nl, FA_SCALAR, _int8)
} else {
} else
#endif
{
CREATE_FA(GGML_TYPE_Q4_0, q4_0, FA_SCALAR, )
CREATE_FA(GGML_TYPE_Q8_0, q8_0, FA_SCALAR, )
CREATE_FA(GGML_TYPE_Q4_1, q4_1, FA_SCALAR, )
@@ -3465,6 +3468,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
CREATE_FA(GGML_TYPE_F32, f32, FA_SCALAR, _fp32)
CREATE_FA(GGML_TYPE_F16, f16, FA_SCALAR, _fp32)
#if defined(GGML_VULKAN_INTEGER_DOT_GLSLC_SUPPORT)
if (device->integer_dot_product) {
CREATE_FA(GGML_TYPE_Q4_0, q4_0, FA_SCALAR, _fp32_int8)
CREATE_FA(GGML_TYPE_Q8_0, q8_0, FA_SCALAR, _fp32_int8)
@@ -3472,7 +3476,9 @@ static void ggml_vk_load_shaders(vk_device& device) {
CREATE_FA(GGML_TYPE_Q5_0, q5_0, FA_SCALAR, _fp32_int8)
CREATE_FA(GGML_TYPE_Q5_1, q5_1, FA_SCALAR, _fp32_int8)
CREATE_FA(GGML_TYPE_IQ4_NL, iq4_nl, FA_SCALAR, _fp32_int8)
} else {
} else
#endif
{
CREATE_FA(GGML_TYPE_Q4_0, q4_0, FA_SCALAR, _fp32)
CREATE_FA(GGML_TYPE_Q8_0, q8_0, FA_SCALAR, _fp32)
CREATE_FA(GGML_TYPE_Q4_1, q4_1, FA_SCALAR, _fp32)