Remove several TODOs related to C++17 support (#3574)

Co-authored-by: Paul Romano <[email protected]>

Author: GuySten

src/bank.cpp

@@ -79,16 +79,9 @@ void sort_fission_bank()
 
   // Perform exclusive scan summation to determine starting indices in fission
   // bank for each parent particle id
-  int64_t tmp = simulation::progeny_per_particle[0];
-  simulation::progeny_per_particle[0] = 0;
-  for (int64_t i = 1; i < simulation::progeny_per_particle.size(); i++) {
-    int64_t value = simulation::progeny_per_particle[i - 1] + tmp;
-    tmp = simulation::progeny_per_particle[i];
-    simulation::progeny_per_particle[i] = value;
-  }
-
-  // TODO: C++17 introduces the exclusive_scan() function which could be
-  // used to replace everything above this point in this function.
+  std::exclusive_scan(simulation::progeny_per_particle.begin(),
+    simulation::progeny_per_particle.end(),
+    simulation::progeny_per_particle.begin(), 0);
 
   // We need a scratch vector to make permutation of the fission bank into
   // sorted order easy. Under normal usage conditions, the fission bank is

src/event.cpp

@@ -1,4 +1,5 @@
 #include "openmc/event.h"
+
 #include "openmc/material.h"
 #include "openmc/simulation.h"
 #include "openmc/timer.h"
@@ -73,17 +74,17 @@ void process_calculate_xs_events(SharedArray<EventQueueItem>& queue)
 {
   simulation::time_event_calculate_xs.start();
 
-  // TODO: If using C++17, perform a parallel sort of the queue
-  // by particle type, material type, and then energy, in order to
-  // improve cache locality and reduce thread divergence on GPU. Prior
-  // to C++17, std::sort is a serial only operation, which in this case
-  // makes it too slow to be practical for most test problems.
+  // TODO: If using C++17, we could perform a parallel sort of the queue by
+  // particle type, material type, and then energy, in order to improve cache
+  // locality and reduce thread divergence on GPU. However, the parallel
+  // algorithms typically require linking against an additional library (Intel
+  // TBB). Prior to C++17, std::sort is a serial only operation, which in this
+  // case makes it too slow to be practical for most test problems.
   //
   // std::sort(std::execution::par_unseq, queue.data(), queue.data() +
   // queue.size());
 
   int64_t offset = simulation::advance_particle_queue.size();
-  ;
 
 #pragma omp parallel for schedule(runtime)
   for (int64_t i = 0; i < queue.size(); i++) {

src/hdf5_interface.cpp

@@ -225,8 +225,7 @@ void get_name(hid_t obj_id, std::string& name)
 {
   size_t size = 1 + H5Iget_name(obj_id, nullptr, 0);
   name.resize(size);
-  // TODO: switch to name.data() when using C++17
-  H5Iget_name(obj_id, &name[0], size);
+  H5Iget_name(obj_id, name.data(), size);
 }
 
 int get_num_datasets(hid_t group_id)