Commit 7289e202 authored by Giovanni La Mura's avatar Giovanni La Mura
Browse files

Use long long indices also in CPU parallel loops

parent a3f8befc

src/trapping/cfrfme.cpp

+13 −13
Original line number Diff line number Diff line
@@ -502,28 +502,28 @@ void frfme(string data_file, string output_path) { 
	  for (int j80 = jlmf - 1; j80 < jlml; j80++) {

	    dcomplex *vec_w = global_vec_w + nkvs * (j80 - jlmf + 1);

#pragma omp parallel for simd

	    for (int jxy50 = 0; jxy50 < nkvs; jxy50++) {

	      int wk_index = nlmmt * jxy50;

	    for (long long jxy50 = 0; jxy50 < nkvs; jxy50++) {

	      long long wk_index = nlmmt * jxy50;

	      dcomplex wk_value = vec_tt1_wk[wk_index + j80];

	      int jy50 = jxy50 / nkv;

	      int jx50 = jxy50 % nkv;

	      long long jy50 = jxy50 / nkv;

	      long long jx50 = jxy50 % nkv;

	      vec_w[(nkv * jx50) + jy50] = wk_value;

	    } // jxy50 loop

	    int nvxy = nxv * nyv;

	    long long nvxy = nxv * nyv;

#pragma omp parallel for

	    for (int ixyz = 0; ixyz < nrvc; ixyz++) {

	      int iz75 = ixyz / nvxy;

	      int iy70 = (ixyz % nvxy) / nxv;

	      int ix65 = ixyz % nxv;

	    for (long long ixyz = 0; ixyz < nrvc; ixyz++) {

	      long long iz75 = ixyz / nvxy;

	      long long iy70 = (ixyz % nvxy) / nxv;

	      long long ix65 = ixyz % nxv;

	      double z = _zv[iz75] + frsh;

	      double y = _yv[iy70];

	      double x = _xv[ix65];

	      dcomplex sumy = cc0;

#pragma omp parallel for simd reduction(+:sumy)

	      for (int jy60x55 = 0; jy60x55 < nkvs ; jy60x55++) {

		int jy60 = jy60x55 / nkv;

		int jx55 = jy60x55 % nkv;

		int w_index = (jx55 * nkv) + jy60;

	      for (long long jy60x55 = 0; jy60x55 < nkvs ; jy60x55++) {

		long long jy60 = jy60x55 / nkv;

		long long jx55 = jy60x55 % nkv;

		long long w_index = (jx55 * nkv) + jy60;

		double vky = vkv[jy60];

		if (jx55 == 0) {

		  // jx55 = 0: phasf