#pragma once

#include <hip/hip_runtime.h>

#include "ck/ck.hpp"
#include "ck/stream_config.hpp"
#include "ck/device_utility/hip_check_error.hpp"

template <typename... Args, typename F>
float launch_and_time_kernel(const StreamConfig& stream_config,
                             F kernel,
                             dim3 grid_dim,
                             dim3 block_dim,
                             std::size_t lds_byte,
                             Args... args)
{
#if CK_TIME_KERNEL
    if(stream_config.time_kernel_)
    {
        printf("%s: grid_dim {%d, %d, %d}, block_dim {%d, %d, %d} \n",
               __func__,
               grid_dim.x,
               grid_dim.y,
               grid_dim.z,
               block_dim.x,
               block_dim.y,
               block_dim.z);

        const int nrepeat = 10;

        printf("Warm up 1 time\n");

        // warm up
        kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);

        printf("Start running %d times...\n", nrepeat);

        hipEvent_t start, stop;

        hip_check_error(hipEventCreate(&start));
        hip_check_error(hipEventCreate(&stop));

        hip_check_error(hipDeviceSynchronize());
        hip_check_error(hipEventRecord(start, stream_config.stream_id_));

        for(int i = 0; i < nrepeat; ++i)
        {
            kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);
        }

        hip_check_error(hipEventRecord(stop, stream_config.stream_id_));
        hip_check_error(hipEventSynchronize(stop));

        float total_time = 0;

        hip_check_error(hipEventElapsedTime(&total_time, start, stop));

        return total_time / nrepeat;
    }
    else
    {
        kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);

        return 0;
    }
#else
    kernel<<<grid_dim, block_dim, lds_byte, stream_config.stream_id_>>>(args...);

    return 0;
#endif
}