nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 27 Packages Projects Releases Wiki Activity

vulkan: fix MMQ shader push constants and multi-dispatch (#19732)

Browse Source
This commit is contained in:
Ruben Ortlam
2026-02-19 14:59:16 +01:00
committed by GitHub
parent da348c9dfb
commit abb9f3c42b
1 changed files with 5 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq.comp
+5 -3
View File
@@ -57,6 +57,8 @@ layout (push_constant) uniform parameter
uint nbi1; uint nbi1;
uint ne11; uint ne11;
#else #else
uint base_work_group_z;
uint num_batches;
uint k_split; uint k_split;
uint ne02; uint ne02;
uint ne12; uint ne12;
@@ -108,7 +110,7 @@ void main() {
const uint ic = gl_WorkGroupID.y; const uint ic = gl_WorkGroupID.y;
#ifdef MUL_MAT_ID #ifdef MUL_MAT_ID
const uint expert_idx = gl_GlobalInvocationID.z; const uint expert_idx = gl_WorkGroupID.z;
if (ic * BN >= data_expert_count[expert_idx]) { if (ic * BN >= data_expert_count[expert_idx]) {
return; return;
} }
@@ -118,7 +120,7 @@ void main() {
#endif #endif
#ifndef MUL_MAT_ID #ifndef MUL_MAT_ID
const uint batch_idx = gl_GlobalInvocationID.z; const uint batch_idx = gl_WorkGroupID.z + p.base_work_group_z;
const uint i13 = batch_idx / p.ne12; const uint i13 = batch_idx / p.ne12;
const uint i12 = batch_idx % p.ne12; const uint i12 = batch_idx % p.ne12;
@@ -276,7 +278,7 @@ void main() {
const uint dc = ic * BN + warp_c * WN; const uint dc = ic * BN + warp_c * WN;
#ifndef MUL_MAT_ID #ifndef MUL_MAT_ID
const uint offsets = batch_idx * p.batch_stride_d + ik * p.batch_stride_d * gl_NumWorkGroups.z; const uint offsets = batch_idx * p.batch_stride_d + ik * p.batch_stride_d * p.num_batches;
#endif #endif
[[unroll]] for (uint wsic = 0; wsic < WNITER; wsic++) { [[unroll]] for (uint wsic = 0; wsic < WNITER; wsic++) {