CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-16 11:27:03 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: handle large sizes for get_rows (#15686)

Browse Source
This commit is contained in:
Jeff Bolz
2025-08-31 03:13:27 -05:00
committed by GitHub
parent c37052ab4d
commit bbbf5ecccb
3 changed files with 46 additions and 25 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
ggml/src/ggml-vulkan/vulkan-shaders/get_rows.comp
View File

@@ -7,27 +7,36 @@ layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
void main() {
const uint i00 = gl_GlobalInvocationID.x;
const uint i10 = gl_GlobalInvocationID.y;
const uint i11 = (gl_GlobalInvocationID.z)/p.ne12;
const uint i12 = (gl_GlobalInvocationID.z)%p.ne12;
if (i00 >= p.ne00) {
return;
}
const uint i01 = data_b[get_boffset() + i10*p.nb10 + i11*p.nb11 + i12*p.nb12];
uint gid_z = gl_GlobalInvocationID.z;
while (gid_z < p.ne11 * p.ne12) {
uint gid_y = gl_GlobalInvocationID.y;
while (gid_y < p.ne10) {
const uint i10 = gid_y;
const uint i11 = gid_z / p.ne12;
const uint i12 = gid_z % p.ne12;
const uint a_offset = get_aoffset() + i01*p.nb01 + i11*p.nb02 + i12*p.nb03;
const uint d_offset = get_doffset() + i10*p.nb21 + i11*p.nb22 + i12*p.nb23;
const uint i01 = data_b[get_boffset() + i10*p.nb10 + i11*p.nb11 + i12*p.nb12];
const uint a_offset = get_aoffset() + i01*p.nb01 + i11*p.nb02 + i12*p.nb03;
const uint d_offset = get_doffset() + i10*p.nb21 + i11*p.nb22 + i12*p.nb23;
#if defined(DATA_A_BF16)
FLOAT_TYPE v = FLOAT_TYPE(bf16_to_fp32(data_a[a_offset + i00]));
FLOAT_TYPE v = FLOAT_TYPE(bf16_to_fp32(data_a[a_offset + i00]));
#else
FLOAT_TYPE v = FLOAT_TYPE(data_a[a_offset + i00]);
FLOAT_TYPE v = FLOAT_TYPE(data_a[a_offset + i00]);
#endif
#ifndef OPTIMIZATION_ERROR_WORKAROUND
data_d[d_offset + i00] = D_TYPE(v);
data_d[d_offset + i00] = D_TYPE(v);
#else
data_d[d_offset + i00] = D_TYPE(v);
data_d[d_offset + i00] = D_TYPE(v);
#endif
gid_y += gl_WorkGroupSize.y * gl_NumWorkGroups.y;
}
gid_z += gl_WorkGroupSize.z * gl_NumWorkGroups.z;
}
}