//
// Arquiteturas de Alto Desempenho 2025/2026
//
// DETI Coin Miner - CUDA implementation with histograms
//

#include <time.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <signal.h>
#include "aad_data_types.h"
#include "aad_utilities.h"
#include "aad_sha1_cpu.h"
#include "aad_cuda_utilities.h"
#include "aad_vault.h"

#define COINS_STORAGE_SIZE  1024u
#define MAX_HISTOGRAM_BINS  100

static volatile int keep_running = 1;

void signal_handler(int signum)
{
  (void)signum;
  keep_running = 0;
}


// Coin reconstruction from stored data
static void reconstruct_coin(u32_t *stored_data, u32_t coin[14])
{
  // Simply copy the complete coin data from storage
  for(int i = 0; i < 14; i++)
    coin[i] = stored_data[i];
}

//
// Mine DETI coins using CUDA
//
static void mine_coins_cuda(u64_t max_attempts, int use_scan_kernel)
{
  cuda_data_t cd;
  u32_t *host_storage;
  u64_t attempts = 0;
  u32_t coins_found = 0;
  u32_t kernel_runs = 0;

  // Initialize CUDA
  cd.device_number = 0;
  cd.cubin_file_name = "coin_miner_cuda_kernel.cubin";
  cd.kernel_name = use_scan_kernel ? "mine_deti_coins_scan_kernel" : "mine_deti_coins_kernel";
  cd.data_size[0] = COINS_STORAGE_SIZE * sizeof(u32_t);
  cd.data_size[1] = 0;

  initialize_cuda(&cd);

  host_storage = (u32_t *)cd.host_data[0];

  // Kernel configuration
  cd.block_dim_x = RECOMMENDED_CUDA_BLOCK_SIZE;
  cd.grid_dim_x = 4096;  // Large grid for maximum GPU utilization

  u32_t n_threads = cd.grid_dim_x * cd.block_dim_x;

  printf("Mining DETI coins using CUDA...\n");
  printf("Grid: %u blocks × %u threads = %u total threads\n",
         cd.grid_dim_x, cd.block_dim_x, n_threads);
  printf("Kernel: %s\n", cd.kernel_name);
  printf("Press Ctrl+C to stop\n\n");

  u32_t param1 = (u32_t)time(NULL);
  u32_t param2 = 0x12345678u;
  int scan_pos = 12;

  time_measurement();
  // double start_time = wall_time_delta();

  while(keep_running && (max_attempts == 0 || attempts < max_attempts))
  {
    // Initialize storage area
    host_storage[0] = 1u; // First unused index

    // Copy to device
    host_to_device_copy(&cd, 0);

    // Set kernel arguments
    cd.n_kernel_arguments = use_scan_kernel ? 4 : 3;
    cd.arg[0] = &cd.device_data[0];
    cd.arg[1] = &param1;
    cd.arg[2] = &param2;
    if(use_scan_kernel)
      cd.arg[3] = &scan_pos;

    // Launch the CUDA kernel
    launch_kernel(&cd);

    // Copy results back
    device_to_host_copy(&cd, 0);

    // Process found coins
    u32_t n_coins_this_kernel = 0;
    u32_t n_stored = (host_storage[0] - 1) / 14;

    if(n_stored > 0 && host_storage[0] < COINS_STORAGE_SIZE)
    {
      printf("DEBUG: host_storage[0] = %u, n_stored = %u\n", host_storage[0], n_stored);

      for(u32_t i = 0; i < n_stored; i++)
      {
        u32_t coin[14];
        reconstruct_coin(&host_storage[1 + i * 14], coin);

        // Verify it's actually a valid coin
        u32_t hash[5];
        sha1(coin, hash);

        printf("DEBUG: Coin %u - hash[0] = 0x%08X (expected 0xAAD20250)\n", i, hash[0]);

        // Print the coin as string
        if(i == 0) {
          printf("DEBUG: First coin content: ");
          u08_t *bytes = (u08_t *)coin;
          for(int j = 0; j < 55; j++) {
            char c = bytes[j ^ 3];
            if(c >= 32 && c <= 126)
              printf("%c", c);
            else
              printf("[0x%02X]", (u08_t)c);
          }
          printf("\n");
        }

        if(hash[0] == 0xAAD20250u)
        {
          coins_found++;
          n_coins_this_kernel++;
          printf("COIN FOUND! (kernel %u, coin %u in this kernel)\n",
                 kernel_runs, n_coins_this_kernel);
          save_coin(coin);
        }
      }
    }

    // Update counters
    kernel_runs++;
    if(use_scan_kernel)
      attempts += n_threads * 256; // Each thread tries 256 values
    else
      attempts += n_threads;

    // Update parameters for next iteration
    param1++;
    param2 = param2 ^ 0x9E3779B9u;
    if(use_scan_kernel)
      scan_pos = (scan_pos + 1) % 42 + 12; // Cycle through positions 12-53


  }

  time_measurement();
  double total_time = cpu_time_delta();

  printf("\n=== Mining Statistics ===\n");
  printf("Total attempts: %llu\n", (unsigned long long)attempts);
  printf("Total time: %.2f seconds\n", total_time);
  printf("Average rate: %.2f attempts/second\n", attempts / total_time);
  printf("Coins found: %u\n", coins_found);
  printf("Kernel launches: %u\n", kernel_runs);


  // Save any remaining coins
  save_coin(NULL);

  terminate_cuda(&cd);
}

int main(int argc, char *argv[])
{
  u64_t max_attempts = 0;
  int use_scan_kernel = 0;

  signal(SIGINT, signal_handler);

  if(argc > 1)
    max_attempts = strtoull(argv[1], NULL, 10);

  if(argc > 2 && strcmp(argv[2], "scan") == 0)
  {
    use_scan_kernel = 1;
    printf("Using scan kernel (tries 256 values per thread)\n");
  }

  mine_coins_cuda(max_attempts, use_scan_kernel);

  return 0;
}