CS348Project/llama.cpp
5
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-11-13 10:57:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

vulkan: workaround for AMD Windows driver 16 bit unpack8 bug (#12472)

Browse Source
This commit is contained in:
Eve
2025-03-21 19:27:47 +00:00
committed by GitHub
parent af04481e6b
commit 30c42ef5cb
4 changed files with 12 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.comp
View File

@@ -82,8 +82,8 @@ vec2 dequantize(uint ib, uint iqs, uint a_offset) {
return vec2(int(data_a[a_offset + ib].qs[iqs]), int(data_a[a_offset + ib].qs[iqs + 1]));
}
vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const i8vec2 v0 = unpack8(data_a_packed16[a_offset + ib].qs[iqs/2]);
const i8vec2 v1 = unpack8(data_a_packed16[a_offset + ib].qs[iqs/2 + 1]);
const i8vec2 v0 = unpack8(int32_t(data_a_packed16[a_offset + ib].qs[iqs/2])).xy; // vec4 used due to #12147
const i8vec2 v1 = unpack8(int32_t(data_a_packed16[a_offset + ib].qs[iqs/2 + 1])).xy;
return vec4(v0.x, v0.y, v1.x, v1.y);
}
#endif