| 21 | // cpu function of Ddot |
| 22 | template <typename T> |
| 23 | void DnMatDnMatDivide( |
| 24 | DeviceDnTen<T>& dnmat3, DeviceDnTen<T>& dnmat1, const DeviceDnTen<T>& dnmat2, const size_s power, |
| 25 | const cudaStream_t& stream = (cudaStream_t) 0, int block_size = 1024 |
| 26 | ) { |
| 27 | int num_block = (dnmat1.total_size + block_size - 1) / block_size; |
| 28 | dnmat_divide_kernel<<<num_block, block_size, 0, stream>>>( |
| 29 | dnmat3.vals, dnmat1.vals, dnmat2.vals, dnmat1.total_size, power |
| 30 | ); |
| 31 | return; |
| 32 | } |
| 33 | |
| 34 | #endif // DENSE_MATDOT_H |