template<typename T, int BLOCK_SIZE, int NUM_PER_TH, int STOCHASTIC> __global__ void kQuantizeBlockwise(float * code, T * __restrict__ const A, float *absmax, unsigned char *out, float * __restrict__ const rand, const int rand_offset, const int n);
template<typename T, int BLOCK_SIZE, int THREADS, int NUM_PER_TH> __global__ void kDequantizeBlockwise(float *code, unsigned char * __restrict__ const A, float * __restrict__ const absmax, T *out, const int n);
template<typename T, int OPTIMIZER, int BLOCK_SIZE, int NUM_VALS>
template <typename T, int SPMM_ITEMS, int BITS> __global__ void kspmm_coo_very_sparse_naive(int *max_count, int *max_idx, int *offset_rowidx, int *rowidx, int *colidx, half *values, T *B, half *out, float *dequant_stats, int nnz, int rowsA, int rowsB, int colsB);
template <int ITEMS_PER_THREAD, int SUBTILE_ROWS, int THREADS>__global__ void kdequant_mm_int32_fp16(
int *__restrict__ const A, float *__restrict__ const rowStats, float *__restrict__ const colStats,
half *out, float* newRowStats, float* newcolStats, const int numRows, const int numCols, const int tileCols, const int n);
template<typename T, int THREADS, int ITEMS_PER_THREAD, int TILE_ROWS, int TILE_COLS, int SPARSE_DECOMP> __global__ void kgetColRowStats(T * __restrict__ A, float *rowStats, float *colStats, int * nnz_count_row, float nnz_threshold, int rows, int cols, int tiledRows, int tiledCols);
template <int THREADS, int ITEMS_PER_THREAD, int TILE_ROWS, int TILE_COLS, int SPARSE_DECOMP> __global__ void kDoubleRowColQuant(half *__restrict__ const A, float *__restrict__ const rowStats, float * __restrict__ const colStats, char *out_col_normed, char *out_row_normed, int *rowidx, int *colidx, half *val, int * __restrict__ nnz_block_ptr, float threshold, int rows, int cols, int tiledCols);
template <int THREADS, int ITEMS_PER_THREAD, int TILE_ROWS, int TILE_COLS, int TRANSPOSE, int FORMAT> __global__ void kTransformRowToFormat(char *__restrict__ const A, char *out, int rows, int cols, int tiledCols, int outRows, int outCols);