GPI-2

GPI-2 (Global address space Programming Interface) is a threadsafe PGAS API for Infiniband,ROCE,ETHERNET,GEMINI and ARIES networks. GPI-2 aims at high performance, delivering wire-speed from the interconnect. It relies on one-sided and asynchronous communication that allow a perfect overlap between computation and communication. All GPI2 methods provide a timeout functionality for fault tolerant operation. With that in place, GPI-2 offers mechanisms that allow applications to react to failures and continue its execution.

Developer: Fraunhofer ITWM Platforms: Hawk (HPE Apollo) Category: Communication libraries License: commercial Website: GPI homepage

Using GPI-2 on Hawk

Load the necessary module. For example:

module load gpi2

Example

File: hello_gpi2.c

#include <stdio.h>
#include <stdlib.h>
#include <GASPI.h>

int main(int argc,char *argv[]){

gaspi_rank_t rank,tnc;
gaspi_float vers,vers_gni;
char mtype[16];

  if( gaspi_proc_init(GASPI_BLOCK) != GASPI_SUCCESS ){
    printf("gaspi_init failed !\n");
    exit(-1);
  }

  gaspi_version(&vers);
  gaspi_proc_rank(&rank);
  gaspi_proc_num(&tnc);

  printf("rank: %d tnc: %d (gpi2: %.2f ugni: %.2f)\n",rank,tnc,vers,vers_gni);

  if( gaspi_barrier(GASPI_GROUP_ALL,GASPI_BLOCK) != GASPI_SUCCESS ){
    printf("gaspi_barrier failed !\n");
    exit(-1);
  }

  gaspi_proc_term(GASPI_BLOCK);
  return 0;
}

Compilation example

gcc -O2 hello_gpi2.c -D_GNU_SOURCE -lpmi -lugni -lrca -lGPI2 -o hello_gpi2.bin

Example to run the program

GPI-2 Applications should be started with one process per NUMA Socket. Use threads to exploit SMP parallelism on each NUMA Socket (e.g. mctp3 for best performance). Ex.:

# 2 nodes, 128 cores (ht is enabled), interactive
qsub -I -l select=2:node_type=rome:mpiprocs=128,walltime=00:05:00

# 4 procs on 2 nodes,socket affinity mask, ht on
mpirun -np 8 ./hello_gpi2.bin

See also

• Software Development Tools, Compilers & Libraries

External links

• GPI homepage