nikola/llama_cpp
1
0
Fork 0
Code Issues Pull Requests Actions 4 Packages Projects Releases Wiki Activity

opencl: q8_0 gemv precision improvement (#24923)

Browse Source
This commit is contained in:
Shawn Gu
2026-06-22 22:25:21 -07:00
committed by GitHub
parent dec5ca5577
commit 23ee8797e1
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
ggml/src/ggml-opencl/kernels/gemv_noshuffle_q8_0_f32.cl
View File

@@ -174,7 +174,7 @@ __kernel void kernel_gemv_noshuffle_q8_0_f32(
regA.s6 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 6)).x;
regA.s7 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 7)).x;
dequantizeBlockAccum_ns_sgbroadcast_1(totalSum, regA, regS, regB);
dequantizeBlockAccum_ns_sgbroadcast_1(totalSum, regA, convert_float(regS), regB);
}
// reduction in local memory, assumes #wave=4