Lumpiasty/llama.cpp
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vulkan: Fuse mul_mat_id+add_id+mul and mul_mat+add+add. (#17287)

Browse Source 
These both show up in gpt-oss. Also, cleanup the mul_mat_vec fusion code a bit.
This commit is contained in:
Jeff Bolz
2025-11-15 19:54:23 +01:00
committed by GitHub
parent 4dca015b7e
commit 24dc769f1b
6 changed files with 296 additions and 147 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp
+8 -12
View File
@@ -10,14 +10,7 @@
layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
layout (binding = 1) readonly buffer B {B_TYPE data_b[];};
layout (binding = 2) writeonly buffer D {D_TYPE dst[];};
layout (binding = 0) readonly buffer AV4 {A_TYPE_VEC4 data_a_v4[];};
layout (binding = 1) readonly buffer BV4 {B_TYPE_VEC4 data_b_v4[];};
layout (binding = 3) readonly buffer Bias {D_TYPE data_bias[];};
#include "mul_mat_vec_iface.glsl"
layout(constant_id = 0) const int BLOCK_SIZE = 32;
// gqa_ratio is in the range [1,8]
@@ -31,7 +24,7 @@ layout (push_constant) uniform parameter
uint nchannels_y;
uint b_offset;
uint d_offset;
uint enable_bias;
uint fusion_flags;
} p;
#if !USE_SUBGROUP_ADD
@@ -151,10 +144,13 @@ void main() {
[[unroll]] for (uint c = 0; c < gqa_ratio; ++c) {
// dst is not transposed and not permuted
const uint idst = (channel + c)*nrows_dst + row_dst;
if (p.enable_bias != 0) {
temp[c] += FLOAT_TYPE(data_bias[idst]);
if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
temp[c] += FLOAT_TYPE(data_fuse0[idst]);
}
dst[idst] = temp[c];
if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
temp[c] += FLOAT_TYPE(data_fuse1[idst]);
}
data_d[idst] = temp[c];
}
}
}