From b9f6783fa6bdcad1824d5b3c75f8fc73d057588e Mon Sep 17 00:00:00 2001
From: Giovanni La Mura <giovanni.lamura@inaf.it>
Date: Tue, 3 Jun 2025 18:12:05 +0200
Subject: [PATCH] Implement an embarrasingly OMP parallel strategy for j80 loop
---
src/trapping/cfrfme.cpp | 13 +++++++------
1 file changed, 7 insertions(+), 6 deletions(-)
diff --git a/src/trapping/cfrfme.cpp b/src/trapping/cfrfme.cpp
index 5901be1..24a780c 100644
--- a/src/trapping/cfrfme.cpp
+++ b/src/trapping/cfrfme.cpp
@@ -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
+ delete[] vec_w;
+ delete[] w;
} // j80 loop
- delete[] vec_w;
- delete[] w;
#ifdef USE_NVTX
nvtxRangePop();
#endif
--
GitLab