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

vulkan: Support Q2_0 (#25430)

Browse Source 
* vulkan: Support Q2_0

The backend perf tests for mat-vec-mul weren't very good at first (worse than
q2_k), doubling the rows per workgroup made a big difference.

* reorder

* resolve merge conflict, adjust err threshold for f16->q2_0 set_rows
This commit is contained in:
Jeff Bolz
2026-07-17 08:42:59 +02:00
committed by GitHub
parent 0bd0ec6099
commit 788e07dc91
13 changed files with 266 additions and 13 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-vulkan/vulkan-shaders/dequant_funcs.glsl
+12 -1
View File
@@ -143,6 +143,17 @@ vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
}
#endif
#if defined(DATA_A_Q2_0)
vec2 dequantize(uint ib, uint iqs, uint a_offset) {
const uint bits = uint(data_a[a_offset + ib].qs[iqs / 4u]) >> (2u * (iqs % 4u));
return vec2(bits & 3u, (bits >> 2u) & 3u) - 1.0f;
}
vec4 dequantize4(uint ib, uint iqs, uint a_offset) {
const uint bits = uint(data_a[a_offset + ib].qs[iqs / 4u]);
return vec4(bits & 3u, (bits >> 2u) & 3u, (bits >> 4u) & 3u, bits >> 6u) - 1.0f;
}
#endif
#if defined(DATA_A_IQ1_S)
vec2 dequantize(uint ib, uint iqs, uint a_offset) {
const uint ib32 = iqs / 32;
@@ -547,7 +558,7 @@ vec2 get_dm(uint ib, uint a_offset) {
}
#endif
#if defined(DATA_A_Q4_0) || defined(DATA_A_Q5_0) || defined(DATA_A_Q8_0) || defined(DATA_A_IQ1_S) || defined(DATA_A_IQ2_XXS) || defined(DATA_A_IQ2_XS) || defined(DATA_A_IQ2_S) || defined(DATA_A_IQ3_XXS) || defined(DATA_A_IQ3_S) || defined(DATA_A_IQ4_XS) || defined(DATA_A_IQ4_NL)
#if defined(DATA_A_Q2_0) || defined(DATA_A_Q4_0) || defined(DATA_A_Q5_0) || defined(DATA_A_Q8_0) || defined(DATA_A_IQ1_S) || defined(DATA_A_IQ2_XXS) || defined(DATA_A_IQ2_XS) || defined(DATA_A_IQ2_S) || defined(DATA_A_IQ3_XXS) || defined(DATA_A_IQ3_S) || defined(DATA_A_IQ4_XS) || defined(DATA_A_IQ4_NL)
vec2 get_dm(uint ib, uint a_offset) {
return vec2(float(data_a[a_offset + ib].d), 0);
}