ClaudeBot/llama.cpp
0
0
Fork 1
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-17 20:31:47 +02:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: make MUL_MAT_ID BN/2 tail unconditional (#28923)

Browse Source 
Use BN/2 as the default for BNover2 and as the disabled fallback for BNover4, and remove the enable gate from the MUL_MAT_ID BN/2 branch. The BN/4 branch remains gated by enable_smaller_matrices, while the p.N path is unchanged.
This commit is contained in:
SG-Amadeus
2026-09-16 08:45:44 +02:00
committed by GitHub
parent 0a8b29a607
commit d4365d9554
1 changed files with 3 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/mul_mm_cm2.comp
+3 -3
View File
@@ -41,8 +41,8 @@ layout (constant_id = 2) const uint BN = 64;
layout (constant_id = 3) const uint BK = 16; // Assumed to be 32 if working with a quant
layout (constant_id = 4) const bool enable_smaller_matrices = false;
const uint BNover2 = enable_smaller_matrices ? (BN / 2) : BN;
const uint BNover4 = enable_smaller_matrices ? (BN / 4) : BN;
const uint BNover2 = BN / 2;
const uint BNover4 = enable_smaller_matrices ? (BN / 4) : (BN / 2);
layout (constant_id = 5) const uint ALIGNED = 0;
layout (constant_id = 6) const uint subgroup_size = 32;
@@ -677,7 +677,7 @@ void main() {
coopMatPerElementNV(mat_d, mat_d, perElemOpD, ir, ic);
return;
}
if (enable_smaller_matrices && ic * BN + BNover2 >= _ne1) {
if (ic * BN + BNover2 >= _ne1) {
coopmat<ACC_TYPE, gl_ScopeWorkgroup, BM, BNover2, gl_MatrixUseAccumulator> sum;
sum = coopmat<ACC_TYPE, gl_ScopeWorkgroup, BM, BNover2, gl_MatrixUseAccumulator>(0.0);