diff --git a/src/trapping/cfrfme.cpp b/src/trapping/cfrfme.cpp
index 9415909a6e06d7f7481152cdf738346cd961871e..f7ccbaff69065bc3f1512d9c85bb99b0dbd201bb 100644
--- a/src/trapping/cfrfme.cpp
+++ b/src/trapping/cfrfme.cpp
@@ -105,6 +105,9 @@ void frfme(string data_file, string output_path) {
int wsum_size;
// End of vector size variables
if (jlmf != 1) {
+#ifdef USE_NVTX
+ nvtxRangePush("frfme() with jlmf != 1");
+#endif
int nxv, nyv, nzv;
if (tfrfme == NULL) tfrfme = TFRFME::from_binary(tfrfme_name, "HDF5");
if (tfrfme != NULL) {
@@ -147,7 +150,16 @@ void frfme(string data_file, string output_path) {
printf("ERROR: could not open TFRFME file.\n");
}
nks = nkv - 1;
- } else { // label 16
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
+ } else { // label 16; jlfm = 1
+#ifdef USE_NVTX
+ nvtxRangePush("frfme() with jlmf == 1");
+#endif
+#ifdef USE_NVTX
+ nvtxRangePush("Setup operations");
+#endif
int nksh, nrsh, nxsh, nysh, nzsh;
str_target = file_lines[last_read_line++];
for (int cli = 0; cli < 7; cli++) {
@@ -183,6 +195,9 @@ void frfme(string data_file, string output_path) {
}
str_target = file_lines[last_read_line++];
re = regex("[eEmM]");
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
if (regex_search(str_target, m, re)) {
more = m.str().at(0);
if (more == 'm' || more == 'M') {
@@ -200,6 +215,9 @@ void frfme(string data_file, string output_path) {
string tedf_name = output_path + "/" + namef + ".hd5";
ScattererConfiguration *tedf = ScattererConfiguration::from_binary(tedf_name, "HDF5");
if (tedf != NULL) {
+#ifdef USE_NVTX
+ nvtxRangePush("TEDF data import");
+#endif
int iduml, idum;
iduml = tedf->number_of_spheres;
idum = tedf->get_iog(iduml - 1);
@@ -223,6 +241,9 @@ void frfme(string data_file, string output_path) {
xi = xip;
}
// label 20
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
delete tedf;
double wn = wp / 3.0e8;
vk = xi * wn;
@@ -243,6 +264,9 @@ void frfme(string data_file, string output_path) {
fshmx = spd * (rir * (sqrt(uy - sthmx * sthmx) / sqrt(uy - sthlmx * sthlmx)) - uy);
}
// label 22
+#ifdef USE_NVTX
+ nvtxRangePush("Memory data loading");
+#endif
nlmmt = lm * (lm + 2) * 2;
nks = nksh * 2;
nkv = nks + 1;
@@ -286,6 +310,12 @@ void frfme(string data_file, string output_path) {
double *_yv = tfrfme->get_y();
double *_zv = tfrfme->get_z();
dcomplex **_wsum = tfrfme->get_matrix();
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
+#ifdef USE_NVTX
+ nvtxRangePush("Looped vector initialization");
+#endif
for (int i24 = nxshpo; i24 <= nxs; i24++) {
_xv[i24] = _xv[i24 - 1] + delxyz;
_xv[nxv - i24 - 1] = -_xv[i24];
@@ -304,7 +334,13 @@ void frfme(string data_file, string output_path) {
vkv[i28] = vkv[i28 - 1] + delk;
vkv[nkv - i28 - 1] = -vkv[i28];
} // i28 loop
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
if (tfrfme != NULL) {
+#ifdef USE_NVTX
+ nvtxRangePush("TFRFME initialization");
+#endif
tfrfme->set_param("vk", vk);
tfrfme->set_param("exri", exri);
tfrfme->set_param("an", an);
@@ -336,6 +372,12 @@ void frfme(string data_file, string output_path) {
tt2->set_param("nlmmt", 1.0 * nlmmt);
tt2->set_param("nrvc", 1.0 * nrvc);
tt2->write_binary(temp_name2, "HDF5");
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
+#ifdef USE_NVTX
+ nvtxRangePush("j80 loop");
+#endif
for (int j80 = jlmf; j80 <= jlml; j80++) {
dcomplex *tt1_wk = tt1->get_vector();
int wk_index = 0;
@@ -384,7 +426,13 @@ void frfme(string data_file, string output_path) {
} // iy70 loop
} // iz75 loop
} // j80 loop
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
// label 88
+#ifdef USE_NVTX
+ nvtxRangePush("Closing operations");
+#endif
tfrfme->write_binary(tfrfme_name, "HDF5");
string output_name = output_path + "/c_OFRFME";
FILE *output = fopen(output_name.c_str(), "w");
@@ -393,6 +441,9 @@ void frfme(string data_file, string output_path) {
if (spd > 0.0) fprintf(output, " FSHMX =%15.7lE\n", fshmx);
fprintf(output, " FRSH =%15.7lE\n", frsh);
fclose(output);
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
} else { // Should never happen.
printf("ERROR: could not open TFRFME file for output.\n");
}
@@ -405,8 +456,14 @@ void frfme(string data_file, string output_path) {
fprintf(output, " WRONG INPUT TAPE\n");
fclose(output);
}
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
}
// label 45
+#ifdef USE_NVTX
+ nvtxRangePush("frfme() memory clean");
+#endif
if (tfrfme != NULL) delete tfrfme;
delete[] file_lines;
if (tt2 != NULL) delete tt2;
@@ -416,6 +473,9 @@ void frfme(string data_file, string output_path) {
}
if (wk != NULL) delete[] wk;
if (tt1 != NULL) delete tt1;
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
printf("FRFME: Done.\n");
#ifdef USE_NVTX
nvtxRangePop();
diff --git a/src/trapping/clffft.cpp b/src/trapping/clffft.cpp
index 46828b2ddfb1d5de61249f6a570fa704dfc1f9e5..4d66d8bad883686c89c169c7afb90641735f1729 100644
--- a/src/trapping/clffft.cpp
+++ b/src/trapping/clffft.cpp
@@ -56,6 +56,10 @@
#include "../include/tra_subs.h"
#endif
+#ifdef USE_NVTX
+#include <nvtx3/nvToolsExt.h>
+#endif
+
using namespace std;
/*! \brief C++ implementation of LFFFT
@@ -64,6 +68,9 @@ using namespace std;
* \param output_path: `string` Directory to write the output files in.
*/
void lffft(string data_file, string output_path) {
+#ifdef USE_NVTX
+ nvtxRangePush("Running lffft()");
+#endif
const dcomplex uim = 0.0 + 1.0 * I;
const double sq2i = 1.0 / sqrt(2.0);
const dcomplex sq2iti = sq2i * uim;
@@ -476,4 +483,7 @@ void lffft(string data_file, string output_path) {
delete ccr;
delete[] file_lines;
printf("LFFT: Done.\n");
+#ifdef USE_NVTX
+ nvtxRangePop();
+#endif
}