mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 30 Packages Projects Releases Wiki Activity

vulkan: fix SSM_CONV PP scaling with large ubatch sizes (#20379)

Browse Source 
* vulkan: optimize SSM_CONV workgroup dispatch for large ubatch

Tile tokens into 2D workgroups (32x16) to reduce workgroup launch
overhead at large ubatch sizes. Add vec4 fast path for nc=4 (common
d_conv size). Fixes PP performance degradation with ubatch > 512.

Ref: ggml-org/llama.cpp#18725

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

* vulkan: remove unused shared memory declaration in SSM_CONV

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Progeny Alpha <ProgenyAlpha@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
ProgenyAlpha
2026-03-12 05:03:18 -04:00
committed by GitHub
parent de190154c8
commit 40c550d4f6
2 changed files with 17 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/ggml-vulkan.cpp
+1 -1
View File
@@ -4576,7 +4576,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
ggml_vk_create_pipeline(device, device->pipeline_ssm_scan_f32_d256, "ssm_scan_256_f32", ssm_scan_f32_len, ssm_scan_f32_data, "main", 8, sizeof(vk_op_ssm_scan_push_constants), {1, 1, 1}, {256, device->subgroup_size, 16}, 1, true, true);
}
ggml_vk_create_pipeline(device, device->pipeline_ssm_conv_f32, "ssm_conv_f32", ssm_conv_f32_len, ssm_conv_f32_data, "main", 3, sizeof(vk_op_ssm_conv_push_constants), {32, 1, 1}, {32}, 1);
ggml_vk_create_pipeline(device, device->pipeline_ssm_conv_f32, "ssm_conv_f32", ssm_conv_f32_len, ssm_conv_f32_data, "main", 3, sizeof(vk_op_ssm_conv_push_constants), {32, 16, 1}, {32, 16}, 1);
ggml_vk_create_pipeline(device, device->pipeline_opt_step_adamw_f32, "opt_step_adamw_f32", opt_step_adamw_f32_len, opt_step_adamw_f32_data, "main", 5, sizeof(vk_op_push_constants), {512, 1, 1}, {}, 1);