From a9a646a4b8445faa878d008390579aa1e73f5673 Mon Sep 17 00:00:00 2001
From: Berenger Bramas <bbramas@mpcdf.mpg.de>
Date: Thu, 20 Apr 2017 12:41:23 +0200
Subject: [PATCH] Remove possible race condition in distr particles computation
---
.../particles/abstract_particles_distr.hpp | 50 +++++++++----------
1 file changed, 23 insertions(+), 27 deletions(-)
diff --git a/bfps/cpp/particles/abstract_particles_distr.hpp b/bfps/cpp/particles/abstract_particles_distr.hpp
index 65c770c6..28837b5c 100644
--- a/bfps/cpp/particles/abstract_particles_distr.hpp
+++ b/bfps/cpp/particles/abstract_particles_distr.hpp
@@ -381,10 +381,9 @@ public:
//////////////////////////////////////////////////////////////////////
/// Merge
//////////////////////////////////////////////////////////////////////
- if(releasedAction.first == MERGE_PARTICLES){
+ if(releasedAction.first == MERGE_PARTICLES && more_than_one_thread == false){
NeighborDescriptor& descriptor = neigDescriptors[releasedAction.second];
- // We can merge safely on first and last partitions since no one is working on it
if(descriptor.isLower){
TIMEZONE("reduce");
assert(descriptor.toRecvAndMerge != nullptr);
@@ -400,36 +399,13 @@ public:
}
}
}
-
- if(more_than_one_thread){
- TIMEZONE_OMP_INIT_PRETASK(timeZoneTaskKey)
- #pragma omp taskgroup
- {
- // Do for first and last partition (or the single partition if current_partition_size is one)
- for(int idxPartition = 0 ; idxPartition < current_partition_size ; idxPartition = std::max(idxPartition+1,current_partition_size-1)){
- for(int idxPart = current_offset_particles_for_partition[idxPartition] ;
- idxPart < current_offset_particles_for_partition[idxPartition+1] ; idxPart += 300){
-
- const int sizeToDo = std::min(300, current_offset_particles_for_partition[idxPartition+1]-idxPart);
-
- #pragma omp task default(shared) firstprivate(idxPart, sizeToDo) priority(0) TIMEZONE_OMP_PRAGMA_TASK_KEY(timeZoneTaskKey)
- {
- TIMEZONE_OMP_TASK("apply_computation_border", timeZoneTaskKey);
- apply_computation(&particles_positions[idxPart*size_particle_positions],
- &particles_current_rhs[idxPart*size_particle_rhs],
- sizeToDo);
- }
- }
- }
- }
- }
}
if(more_than_one_thread && omp_get_thread_num() == 1){
TIMEZONE_OMP_INIT_PRETASK(timeZoneTaskKey)
#pragma omp taskgroup
{
// Do for all partitions except the first and last one
- for(int idxPartition = 1 ; idxPartition < current_partition_size-1 ; ++idxPartition){
+ for(int idxPartition = 0 ; idxPartition < current_partition_size ; ++idxPartition){
for(int idxPart = current_offset_particles_for_partition[idxPartition] ;
idxPart < current_offset_particles_for_partition[idxPartition+1] ; idxPart += 300){
@@ -449,8 +425,28 @@ public:
}
}
+ if(more_than_one_thread == true){
+ for(int idxDescr = 0 ; idxDescr < int(neigDescriptors.size()) ; ++idxDescr){
+ NeighborDescriptor& descriptor = neigDescriptors[idxDescr];
+ if(descriptor.nbParticlesToSend){
+ if(descriptor.isLower){
+ TIMEZONE("reduce_later");
+ assert(descriptor.toRecvAndMerge != nullptr);
+ reduce_particles_rhs(&particles_current_rhs[0], descriptor.toRecvAndMerge.get(), descriptor.nbParticlesToSend);
+ descriptor.toRecvAndMerge.release();
+ }
+ else {
+ TIMEZONE("reduce_later");
+ assert(descriptor.toRecvAndMerge != nullptr);
+ reduce_particles_rhs(&particles_current_rhs[(current_offset_particles_for_partition[current_partition_size]-descriptor.nbParticlesToSend)*size_particle_rhs],
+ descriptor.toRecvAndMerge.get(), descriptor.nbParticlesToSend);
+ descriptor.toRecvAndMerge.release();
+ }
+ }
+ }
+ }
- // Do my own computation
+ // Do my own computation if not threaded
if(more_than_one_thread == false){
TIMEZONE("compute-my_compute");
// Compute my particles
--
GitLab