Commit 19e36857 authored by Mulas, Giacomo's avatar Mulas, Giacomo
Browse files

fix omp parallel loop in scr0()

parent e5339261

src/libnptm/clu_subs.cpp

+25 −9
Original line number Diff line number Diff line
@@ -2146,24 +2146,19 @@ void scr0(double vk, double exri, C1 *c1, C1_AddOns *c1ao, C3 *c3, C4 * c4) { 
  double cccs = ccs / exdc;

  dcomplex csam = -(ccs / (exri * vk)) * 0.5 * I;

  //double scs = 0.0, ecs = 0.0, acs = 0.0;

  double scs = 0.0;

  double ecs = 0.0;

  double acs = 0.0;

  dcomplex tfsas = cc0;

  dcomplex *vec_rmi = c1->rmi[0];

  dcomplex *vec_rei = c1->rei[0];

#ifdef USE_NVTX

  nvtxRangePush("scr0 outer loop");

  nvtxRangePush("scr0 outer loop 1");

#endif



  //#pragma omp parallel for reduction(+:scs, ecs, acs, tfsas)

#pragma omp parallel for

  for (int i14 = 1; i14 <= c4->nsph; i14++) {

    int iogi = c1->iog[i14 - 1];

    if (iogi >= i14) {

      double sums = 0.0;

      dcomplex sum21 = cc0;

#ifdef USE_NVTX

      nvtxRangePush("scr0 inner loop");

      nvtxRangePush("scr0 inner loop 1");

#endif

#pragma omp target teams distribute parallel for simd reduction(+:sums, sum21)

      for (int l10 = 1; l10 <= c4->li; l10++) {
@@ -2177,6 +2172,9 @@ void scr0(double vk, double exri, C1 *c1, C1_AddOns *c1ao, C3 *c3, C4 * c4) { 
	sums += rvalue;

	sum21 += ((rm + re) * fl);

      } // l10 loop

#ifdef USE_NVTX

      nvtxRangePop();

#endif

      sum21 *= -1.0;

      double scasec = cccs * sums;

      double extsec = -cccs * real(sum21);
@@ -2191,11 +2189,29 @@ void scr0(double vk, double exri, C1 *c1, C1_AddOns *c1ao, C3 *c3, C4 * c4) { 
      c1->fsas[i14 - 1] = sum21 * csam;

    }

    // label 12

    // scs += c1->sscs[iogi - 1];

    // ecs += c1->sexs[iogi - 1];

    // acs += c1->sabs[iogi - 1];

    // tfsas += c1->fsas[iogi - 1];

  } // i14 loop

#ifdef USE_NVTX

  nvtxRangePop();

#endif

  double scs = 0.0;

  double ecs = 0.0;

  double acs = 0.0;

  dcomplex tfsas = cc0;

#ifdef USE_NVTX

      nvtxRangePush("scr0 loop 2");

#endif

#pragma omp target teams distribute parallel for simd reduction(+:scs, ecs, acs, tfsas)

  for (int i14 = 1; i14 <= c4->nsph; i14++) {

    int iogi = c1->iog[i14 - 1];

    scs += c1->sscs[iogi - 1];

    ecs += c1->sexs[iogi - 1];

    acs += c1->sabs[iogi - 1];

    tfsas += c1->fsas[iogi - 1];

  } // i14 loop

  }

  c3->scs = scs;

  c3->ecs = ecs;

  c3->acs = acs;