tqcq/llama.cpp
0
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2025-08-01 15:09:32 -04:00
Code Issues Packages Projects Releases Wiki Activity

CUDA: revise q8_1 data layout for mul_mat_q (#7824)

Browse Source
This commit is contained in:
Johannes Gäßler
2024-06-09 09:42:25 +02:00
committed by GitHub
parent 2decf57bc6
commit 42b53d192f
5 changed files with 282 additions and 151 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
ggml-cuda/mmq.cu
View File

@@ -11,6 +11,7 @@ void ggml_cuda_op_mul_mat_q(
const int64_t nb01 = src0->nb[1];
const int64_t ne10 = src1->ne[0];
const int64_t ne11 = src1->ne[1];
GGML_ASSERT(ne10 % QK8_1 == 0);
const int64_t ne0 = dst->ne[0];
@@ -25,7 +26,7 @@ void ggml_cuda_op_mul_mat_q(
// nrows_dst == nrows of the matrix that the kernel writes into
const int64_t nrows_dst = id == ctx.device ? ne0 : row_diff;
const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, nrows_dst};
const mmq_args args = {src0_dd_i, src1_ddq_i, dst_dd_i, ne00, row_diff, stride00, src1_padded_row_size, src1_ncols, ne11, nrows_dst};
switch (src0->type) {
case GGML_TYPE_Q4_0: