Commit b9f6783f authored by Giovanni La Mura's avatar Giovanni La Mura
Browse files

Implement an embarrasingly OMP parallel strategy for j80 loop

parent 43c640b0

src/trapping/cfrfme.cpp

+7 −6
Original line number Diff line number Diff line
@@ -374,14 +374,15 @@ void frfme(string data_file, string output_path) { 
#ifdef USE_NVTX

	  nvtxRangePop();

#endif

	  dcomplex *vec_w = new dcomplex[nkv * nkv]();

	  dcomplex **w = new dcomplex*[nkv];

	  for (int wi = 0; wi < nkv; wi++) w[wi] = vec_w + wi * nkv;

#ifdef USE_NVTX

	  nvtxRangePush("j80 loop");

#endif

#pragma omp parallel for

	  for (int j80 = jlmf; j80 <= jlml; j80++) {

	    int wk_index = 0;

	    dcomplex *vec_w = new dcomplex[nkv * nkv]();

	    dcomplex **w = new dcomplex*[nkv];

	    for (int wi = 0; wi < nkv; wi++) w[wi] = vec_w + wi * nkv;

	    int wk_index = (j80 - jlmf) * nkv * nkv;

	    for (int jy50 = 0; jy50 < nkv; jy50++) {

	      for (int jx50 = 0; jx50 < nkv; jx50++) {

		for (int wi = 0; wi < nlmmt; wi++) wk[wi] = tt1->vec_wk[wk_index++];
@@ -419,9 +420,9 @@ void frfme(string data_file, string output_path) { 
		} // ix65 loop

	      } // iy70 loop

	    } // iz75 loop

	  } // j80 loop

	    delete[] vec_w;

	    delete[] w;

	  } // j80 loop

#ifdef USE_NVTX

	  nvtxRangePop();

#endif