#include <assert.h>
#include "data_cm.h"
#include "meuk.h"

#define REPETITIONS 10000000
int main(int argc, char **argv) {

  assert(argc == 3);
  char *version = argv[1];
  char *number_of_updates = argv[2];
  const uint64_t Dim = 21;
  const uint64_t LDS = 24;
  // const double breakdown = 1e-3;
  const double breakdown = 1e-9; // this might be too small and cause NIs
  uint32_t rc;

  const uint64_t *N_updates;
  const double *Updates;
  const uint64_t *Updates_index;
  double *Slater, *Slater_invT;
  double determinant;
  if (number_of_updates[0] == '2') { // 2 Updates
    N_updates = &N_updates2;
    Updates = &Updates2[0];
    Updates_index = &Updates_index2[0];
    Slater = &Slater2[0];
    Slater_invT = &Slater_invT2[0]; // Slater_inv in QMC=Chem is actually its transpose
    determinant = determinant2;
  } else if (number_of_updates[0] == '3') { // 3 Updates
    N_updates = &N_updates3;
    Updates = &Updates3[0];
    Updates_index = &Updates_index3[0];
    Slater = &Slater3[0];
    Slater_invT = &Slater_invT3[0];
    determinant = determinant3;
  } else if (number_of_updates[0] == '5') { // 5 Updates
    N_updates = &N_updates5;
    Updates = &Updates5[0];
    Updates_index = &Updates_index5[0];
    Slater = &Slater5[0];
    Slater_invT = &Slater_invT5[0];
    determinant = determinant5;
  } else { // Exit
    printf("Incorrect number of updates given\n");
    return 1;
  }
  
  rc = check_residual(LDS, Dim, Slater_invT, Slater);
  assert(rc == 0 && "check_residual()");
  rc = test_kernel(version, LDS, Dim, *N_updates, Updates, Updates_index,
                   breakdown, Slater, Slater_invT, &determinant);
  assert(rc == 0 && "test_kernel()");

  // EVERYTHING WORKS UP UNTILL HERE

  uint64_t before = rdtsc();
  if (version[0] == 'a') { // Anthony
    for (int i = 0; i < REPETITIONS; i++) {
      const double* Upds;
      const uint64_t* Ui;
      for (int j = 0; j < *N_updates; j++) {
        Upds = &Updates[j*LDS];
        Ui = &Updates_index[j];
        detupd(Dim, LDS, Upds, Ui, Slater_invT, &determinant);
      }
    }
  } else if (version[0] == 'n') { // Naive
    for (int i = 0; i < REPETITIONS; i++) {
      rc = qmckl_sherman_morrison(LDS, Dim, *N_updates, Updates,
            Updates_index, breakdown, Slater_invT, &determinant);
      if (rc != 0) printf("qmckl_sherman_morrison failed\n");
    }
  } else if (version[0] == 's') { // Splitting
    for (int i = 0; i < REPETITIONS; i++) {
      rc = qmckl_sherman_morrison_splitting(LDS, Dim, *N_updates, Updates,
            Updates_index, breakdown, Slater_invT, &determinant);
      if (rc != 0) printf("qmckl_sherman_morrison_splitting failed\n");
    }
  } else if (version[0] == 'b') { // Blocked
    for (int i = 0; i < REPETITIONS; i++) {
      // rc = qmckl_woodbury_2(LDS, Dim, Updates, Updates_index,
      //                        breakdown, Slater_inv, &determinant);
      // rc = qmckl_woodbury_3(LDS, Dim, Updates, Updates_index,
      //                        breakdown, Slater_inv, &determinant);
      rc = qmckl_sherman_morrison_smw32s(LDS, Dim, *N_updates, Updates,
            Updates_index, breakdown, Slater_invT, &determinant);
      if (rc != 0) printf("qmckl_sherman_morrison_smw32s failed\n");
    }
  }
  uint64_t after = rdtsc();
  printf("cycles = %f\n", ((double)(after - before) / (double) REPETITIONS));
}