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

sycl : support nvfp4 type in mul_mat (#21227)

Browse Source
This commit is contained in:
Neo Zhang
2026-04-01 13:54:15 +03:00
committed by GitHub
parent 84f82e846c
commit 6b949d1078
6 changed files with 232 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-sycl/mmvq.cpp
+21 -1
View File
@@ -613,6 +613,23 @@ static void mul_mat_vec_mxfp4_q8_1_sycl(const void * vx, const void * vy, float
}
}
static void mul_mat_vec_nvfp4_q8_1_sycl(const void * vx, const void * vy, float * dst, const int ncols, const int nrows,
dpct::queue_ptr stream) {
GGML_ASSERT(ncols % QK_NVFP4 == 0);
const int block_num_y = (nrows + GGML_SYCL_MMV_Y - 1) / GGML_SYCL_MMV_Y;
const sycl::range<3> block_nums(1, 1, block_num_y);
const sycl::range<3> block_dims(1, GGML_SYCL_MMV_Y, WARP_SIZE);
{
stream->submit([&](sycl::handler & cgh) {
cgh.parallel_for(sycl::nd_range<3>(block_nums * block_dims, block_dims),
[=](sycl::nd_item<3> item_ct1) [[sycl::reqd_sub_group_size(WARP_SIZE)]] {
mul_mat_vec_q<QK_NVFP4, QI_NVFP4, block_nvfp4, VDR_NVFP4_Q8_1_MMVQ, vec_dot_nvfp4_q8_1>(
vx, vy, dst, ncols, nrows, item_ct1);
});
});
}
}
static void mul_mat_vec_q5_0_q8_1_sycl(const void *vx, const void *vy,
float *dst, const int ncols,
@@ -1145,8 +1162,11 @@ void ggml_sycl_op_mul_mat_vec_q(ggml_backend_sycl_context & ctx, const ggml_tens
case GGML_TYPE_MXFP4:
mul_mat_vec_mxfp4_q8_1_sycl(src0_dd_i, src1_ddq_i_bs, dst_dd_i_bs, ne00, row_diff, stream);
break;
case GGML_TYPE_NVFP4:
mul_mat_vec_nvfp4_q8_1_sycl(src0_dd_i, src1_ddq_i_bs, dst_dd_i_bs, ne00, row_diff, stream);
break;
default:
GGML_ABORT("fatal error");
GGML_ABORT("fatal error: unsupport data type=%s\n", ggml_type_name(src0->type));
}
}
GGML_UNUSED(src1);