EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 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/mul_mat_vec_iq2_s.comp
View File

@@ -19,8 +19,8 @@ void calc_superblock(const uint a_offset, const uint b_offset, const uint itid,
const float db = d * (0.5 + scale) * 0.25;
const uint qh = data_a[ibi].qh[ib32];
const u8vec2 qs16 = unpack8(data_a_packed16[ibi].qs[itid]);
const u8vec2 sign16 = unpack8(data_a_packed16[ibi].qs[QUANT_K / 16 + itid]);
const u8vec2 qs16 = unpack8(uint32_t(data_a_packed16[ibi].qs[itid])).xy; // vec4 used due to #12147
const u8vec2 sign16 = unpack8(uint32_t(data_a_packed16[ibi].qs[QUANT_K / 16 + itid])).xy;
[[unroll]] for (uint l = 0; l < 2; ++l) {
const uint8_t sign = sign16[l];
const uint qs = qs16[l] | ((qh << (8 - nibble_shift - 2 * l)) & 0x300);