Skip to content

shared umpire device pool and mfem GPU-aware MPI #55

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
vladotomov merged 5 commits into from
Nov 25, 2025
Merged

shared umpire device pool and mfem GPU-aware MPI #55

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
2843170
Add shared umpire pool for hypre and mfem and add option to enable GP…
Nov 15, 2025
49bf9d8
Add GPU aware MPI option
Nov 15, 2025
b7047c8
fix endif
Nov 16, 2025
c251938
Add input option for device pool size
Nov 17, 2025
30bb37d
Print device setup after setting GPU aware MPI
rfhaque Nov 22, 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
31 changes: 31 additions & 0 deletions remhos.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,6 +43,12 @@
#include "fem/qinterp/grad.hpp"
#include "fem/integ/bilininteg_mass_kernels.hpp"

#if (defined(HYPRE_USING_UMPIRE) || defined(MFEM_USE_UMPIRE)) && (defined(MFEM_USE_CUDA) || defined(MFEM_USE_HIP))
#define REMHOS_USE_DEVICE_UMPIRE
#include <umpire/Umpire.hpp>
#include <umpire/strategy/QuickPool.hpp>
#endif

#ifdef USE_CALIPER
#include <caliper/cali.h>
#include <caliper/cali-manager.h>
Expand Down Expand Up @@ -202,6 +208,8 @@ MFEM_EXPORT int remhos(int argc, char *argv[], double &final_mass_u)
bool product_sync = false;
int vis_steps = 100;
const char *device_config = "cpu";
bool gpu_aware_mpi = false;
int dev_pool_size = 4;

int precision = 8;
cout.precision(precision);
Expand Down Expand Up @@ -257,6 +265,8 @@ MFEM_EXPORT int remhos(int argc, char *argv[], double &final_mass_u)
"Enable or disable next gen full assembly for the HO solution.");
args.AddOption(&device_config, "-d", "--device",
"Device configuration string, see Device::Configure().");
args.AddOption(&gpu_aware_mpi, "-gam", "--gpu-aware-mpi", "-no-gam",
"--no-gpu-aware-mpi", "Enable GPU aware MPI communications.");
args.AddOption(&smth_ind_type, "-si", "--smth_ind",
"Smoothness indicator: 0 - no smoothness indicator,\n\t"
" 1 - approx_quadratic,\n\t"
Expand Down Expand Up @@ -287,6 +297,8 @@ MFEM_EXPORT int remhos(int argc, char *argv[], double &final_mass_u)
"Enable remap of synchronized product fields.");
args.AddOption(&vis_steps, "-vs", "--visualization-steps",
"Visualize every n-th timestep.");
args.AddOption(&dev_pool_size, "-pool", "--dev-pool-size",
"Size (in GB) for the umpire device pool");
args.Parse();
if (!args.Good())
{
Expand All @@ -295,6 +307,24 @@ MFEM_EXPORT int remhos(int argc, char *argv[], double &final_mass_u)
}
if (myid == 0) { args.PrintOptions(cout); }

#ifdef REMHOS_USE_DEVICE_UMPIRE
auto &rm = umpire::ResourceManager::getInstance();
const char * allocator_name = "remhos_device_alloc";
size_t umpire_dev_pool_size = ((size_t) dev_pool_size) * 1024 * 1024 * 1024;
size_t umpire_dev_block_size = 1024 * 1024;
rm.makeAllocator<umpire::strategy::QuickPool>(allocator_name, rm.getAllocator("DEVICE"), umpire_dev_pool_size, umpire_dev_block_size);

#ifdef HYPRE_USING_UMPIRE
HYPRE_SetUmpireDevicePoolName(allocator_name);
#endif // HYPRE_USING_UMPIRE

#ifdef MFEM_USE_UMPIRE
MemoryManager::SetUmpireDeviceAllocatorName(allocator_name);
// the umpire host memory type is slow compared to the native host memory type
Device::SetMemoryTypes(MemoryType::HOST, MemoryType::DEVICE_UMPIRE);
#endif // MFEM_USING_UMPIRE
#endif // REMHOS_USE_DEVICE_UMPIRE

//setup caliper config manager
#ifdef USE_CALIPER
setupCaliper();
Expand All @@ -317,6 +347,7 @@ MFEM_EXPORT int remhos(int argc, char *argv[], double &final_mass_u)
// Enable hardware devices such as GPUs, and programming models such as
// CUDA, OCCA, RAJA and OpenMP based on command line options.
Device device(device_config);
device.SetGPUAwareMPI(gpu_aware_mpi);
if (myid == 0) { device.Print(); }

if (myid == 0) { KernelReporter::Enable(); }
Expand Down
Loading