#ifndef __AMSCUARRAY_DOPS_HPP__
#define __AMSCUARRAY_DOPS_HPP__

//Device Operations on Arrays
//

//Device Operations on Device Buffers
// dodb

namespace amscuda
{


    //sum
    template<typename T> T devcuarray_sum(cuarray<T> *devptr);

    template<typename T> T dbuff_sum(T *devbuffer, int N);


    struct dbuff_statstruct
    {
        public:
        float min;
        float max;
        float mean;
        float stdev;
        float sum;
    };
    
    //stats (min,max,mean,stdev)

    template<typename T> void dbuff_minmax(T *devbuffer, int N, T *min, T *max);

    template<typename T> dbuff_statstruct dbuff_stats(T *devbuffer, int N); //

    //sets all elements to setto
    template<typename T> void dbuff_setall(T *devbuffer, int N, T setto, int nblocks, int nthreads);

    //random device buffer functions
    void dbuff_rand_dpr32(float *devbuffer, int N, int32_t *rseedinout,  int nblocks, int nthreads); //
    void dbuff_rand_dpr32n(float *devbuffer, int N, int32_t *rseedinout,  int nblocks, int nthreads); //
    

    void dbuff_rand_dpr64(float *devbuffer, int N, int64_t *rseedinout,  int nblocks, int nthreads); //

    //Elementwise device-buffer vector binary operation
    //takes two input arrays ( , ) --> one output array
    template<typename T1, typename T2, typename T3> void dbuff_vectorbinop(T1 *dbuf_a, T2 *dbuf_b, T3 *dbuf_out, int N, T3 (*fpnt)(T1,T2), int nblocks, int nthreads);

    //Elementwise device-buffer vector two-parameter operation
    //takes one input array, and a constant paramter ( ) ---> one output array
    template<typename T1, typename T2, typename T3> void dbuff_vectorbinop(T1 *dbuf_a, T2 par_b, T3 *dbuf_out, int N, T3 (*fpnt)(T1,T2), int nblocks, int nthreads);


    //vector_add
    template<typename T> void dbuff_add(T *dbuff_a, T *dbuff_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_add(T *dbuff_a, T par_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_sub(T *dbuff_a, T *dbuff_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_sub(T *dbuff_a, T par_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_mult(T *dbuff_a, T *dbuff_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_mult(T *dbuff_a, T par_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_div(T *dbuff_a, T *dbuff_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_div(T *dbuff_a, T par_b, T *dbuff_out, int N, int nblocks, int nthreads); 
    template<typename T> void dbuff_div(T par_a, T *dbuff_b, T *dbuff_out, int N, int nblocks, int nthreads); 


    // Tests //

    void test_dbuff_rand_dpr32();

};

#include <amsculib2/amscuarray_dops_impl.hpp>

#endif