EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

vulkan: Use push constant offset to handle misaligned descriptors (#10987)

Browse Source
This commit is contained in:
Jeff Bolz
2024-12-29 02:35:11 -06:00
committed by GitHub
parent f865ea149d
commit fdd2188912
19 changed files with 103 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
ggml/src/ggml-vulkan/vulkan-shaders/contig_copy.comp
View File

@@ -19,9 +19,9 @@ void main() {
if (idx + (num_iter-1)*num_threads < p.ne) {
[[unroll]] for (uint i = 0; i < num_iter; ++i) {
#ifndef OPTIMIZATION_ERROR_WORKAROUND
data_d[p.d_offset + idx] = D_TYPE(data_a[idx]);
data_d[get_doffset() + idx] = D_TYPE(data_a[get_aoffset() + idx]);
#else
data_d[p.d_offset + idx] = data_a[idx];
data_d[get_doffset() + idx] = data_a[get_aoffset() + idx];
#endif
idx += num_threads;
}
@@ -32,9 +32,9 @@ void main() {
}
#ifndef OPTIMIZATION_ERROR_WORKAROUND
data_d[p.d_offset + idx] = D_TYPE(data_a[idx]);
data_d[get_doffset() + idx] = D_TYPE(data_a[get_aoffset() + idx]);
#else
data_d[p.d_offset + idx] = data_a[idx];
data_d[get_doffset() + idx] = data_a[get_aoffset() + idx];
#endif
idx += num_threads;
}