zhangyiss/mlx
1
0
Fork 0
mirror of https://github.com/ml-explore/mlx.git synced 2025-12-16 01:49:05 +08:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

add cuda gemv (#2400)

Browse Source
This commit is contained in:
Awni Hannun
2025-07-22 08:24:13 -07:00
committed by GitHub
parent 1e496ddb82
commit d107d8d495
12 changed files with 198 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
mlx/backend/cuda/device/utils.cuh
View File

@@ -218,20 +218,8 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT, IdxT> elem_to_loc_nd(
return cuda::std::make_tuple(a_loc, b_loc, c_loc);
}
// Optimized version when ndim is larger than 4.
template <typename IdxT = int64_t>
inline __host__ __device__ IdxT
elem_to_loc_4d(IdxT elem, const int* shape, const int64_t* strides, int ndim) {
IdxT loc = 0;
for (int i = ndim - 1; i >= 0; --i) {
loc += (elem % shape[i]) * IdxT(strides[i]);
elem /= shape[i];
}
return loc;
}
template <typename IdxT = int64_t>
inline __host__ __device__ cuda::std::tuple<IdxT, IdxT> elem_to_loc_4d(
inline __host__ __device__ cuda::std::tuple<IdxT, IdxT> elem_to_loc(
IdxT elem,
const int* shape,
const int64_t* a_strides,
@@ -249,7 +237,7 @@ inline __host__ __device__ cuda::std::tuple<IdxT, IdxT> elem_to_loc_4d(
}
template <typename IdxT = int64_t>
inline __host__ __device__ cuda::std::tuple<IdxT, IdxT, IdxT> elem_to_loc_4d(
inline __host__ __device__ cuda::std::tuple<IdxT, IdxT, IdxT> elem_to_loc(
IdxT elem,
const int* shape,
const int64_t* a_strides,