Skip to content

GPU race condition in MPI communication #98

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
diogosiebert merged 2 commits into from
Oct 15, 2025
Merged

GPU race condition in MPI communication #98

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1364d10
compilation fixes
jeremyfirst22 Oct 13, 2025
fe6f38a
bugfix: kernel packing of send buffers must complete before initiatin…
jeremyfirst22 Oct 14, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions IO/Reader.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
#include "IO/IOHelpers.h"
#include "IO/Mesh.h"
#include "IO/MeshDatabase.h"
#include "IO/silo.h"
#include "common/Utilities.h"

#include <ProfilerApp.h>
Expand Down
4 changes: 4 additions & 0 deletions cmake/macros.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -428,6 +428,10 @@ MACRO( IDENTIFY_COMPILER )
SET(USING_XL TRUE)
ADD_DEFINITIONS( -DUSING_XL )
MESSAGE("Using XL")
ELSEIF( CMAKE_CXX_COMPILER_ID MATCHES "NVHPC")
SET(USING_NVHPC TRUE)
ADD_DEFINITIONS( -DUSING_NVHPC )
MESSAGE("Using NVHPC")
ELSE()
MESSAGE( "CMAKE_C_COMPILER=${CMAKE_C_COMPILER}")
MESSAGE( "CMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}")
Expand Down
6 changes: 6 additions & 0 deletions common/ScaLBL.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2402,6 +2402,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {
ScaLBL_D3Q19_Pack(2, dvcSendList_x, sendCount_x, sendCount_x, sendbuf_x, Bq,
N);

ScaLBL_DeviceBarrier();
req1[0] =
MPI_COMM_SCALBL.Isend(sendbuf_x, 2 * sendCount_x, rank_x, sendtag + 0);
req2[0] =
Expand All @@ -2412,6 +2413,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {
ScaLBL_D3Q19_Pack(1, dvcSendList_X, sendCount_X, sendCount_X, sendbuf_X, Bq,
N);

ScaLBL_DeviceBarrier();
req1[1] =
MPI_COMM_SCALBL.Isend(sendbuf_X, 2 * sendCount_X, rank_X, sendtag + 1);
req2[1] =
Expand All @@ -2422,6 +2424,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {
ScaLBL_D3Q19_Pack(4, dvcSendList_y, sendCount_y, sendCount_y, sendbuf_y, Bq,
N);

ScaLBL_DeviceBarrier();
req1[2] =
MPI_COMM_SCALBL.Isend(sendbuf_y, 2 * sendCount_y, rank_y, sendtag + 2);
req2[2] =
Expand All @@ -2432,6 +2435,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {
ScaLBL_D3Q19_Pack(3, dvcSendList_Y, sendCount_Y, sendCount_Y, sendbuf_Y, Bq,
N);

ScaLBL_DeviceBarrier();
req1[3] =
MPI_COMM_SCALBL.Isend(sendbuf_Y, 2 * sendCount_Y, rank_Y, sendtag + 3);
req2[3] =
Expand All @@ -2442,6 +2446,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {
ScaLBL_D3Q19_Pack(6, dvcSendList_z, sendCount_z, sendCount_z, sendbuf_z, Bq,
N);

ScaLBL_DeviceBarrier();
req1[4] =
MPI_COMM_SCALBL.Isend(sendbuf_z, 2 * sendCount_z, rank_z, sendtag + 4);
req2[4] =
Expand All @@ -2454,6 +2459,7 @@ void ScaLBL_Communicator::BiSendD3Q7AA(double *Aq, double *Bq) {

//...................................................................................
// Send all the distributions
ScaLBL_DeviceBarrier();
req1[5] =
MPI_COMM_SCALBL.Isend(sendbuf_Z, 2 * sendCount_Z, rank_Z, sendtag + 5);
req2[5] =
Expand Down
Loading