Skip to content
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

Jump to bottom

Fix MPI on Ubuntu 18.04 with CUDA #2271

Merged
merged 2 commits into from
Sep 18, 2018
Merged

Fix MPI on Ubuntu 18.04 with CUDA #2271

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cf1fe2a
Fix MPI on Ubuntu 18.04 with CUDA
mkuron Sep 18, 2018
43b49c3
Use setenv instad of MPI_T interface
mkuron Sep 18, 2018
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
26 changes: 26 additions & 0 deletions src/core/communication.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,8 +227,34 @@ int mpi_check_runtime_errors(void);
* procedures
**********************************************/

#if defined(OPEN_MPI)
/*! Workaround for "Read -1, expected XXXXXXX, errno = 14" that sometimes
appears when CUDA is used. This is a bug in OpenMPI 2.0-2.1.2 and 3.0.0
according to
https://www.mail-archive.com/users@lists.open-mpi.org/msg32357.html,
so we set btl_vader_single_copy_mechanism = none.
*/
static void openmpi_fix_vader() {
if (OMPI_MAJOR_VERSION < 2 || OMPI_MAJOR_VERSION > 3)
return;
if (OMPI_MAJOR_VERSION == 2 && OMPI_MINOR_VERSION == 1 &&
OMPI_RELEASE_VERSION >= 3)
return;
if (OMPI_MAJOR_VERSION == 3 &&
(OMPI_MINOR_VERSION > 0 || OMPI_RELEASE_VERSION > 0))
return;

std::string varname = "btl_vader_single_copy_mechanism";
std::string varval = "none";

setenv((std::string("OMPI_MCA_") + varname).c_str(), varval.c_str(), 0);
}
#endif

void mpi_init() {
#ifdef OPEN_MPI
openmpi_fix_vader();

void *handle = 0;
int mode = RTLD_NOW | RTLD_GLOBAL;
#ifdef RTLD_NOLOAD
Expand Down