Incompatibility between the Intel® MPI Library and certain versions of the Mellanox* OFED drivers

Cause

There is an incompatibility between the Intel® MPI Library and some versions of the Mellanox* OFED drivers when using the DAPL* fabric.  This has been reported with Mellanox* OFED 2.1-1.0.0, and could exist in other versions.

Symptoms

You will likely see Segmentation Faults in MPI_Init when running.  A stack trace could show:

#0 0x00007f222c9ad85c in I_MPI_dat_extension_op (handle=0x1e5e770, ext_op=6) at ./../../include/I_MPI_wrap_dat.h:352
#1 0x00007f222c9b0b56 in dapl_rc_init_infostructure_20 (proc=0x7f222d02e040, param=0x7fffe3b5d6c8, rc_proc=0x7f222d02db00, rc_param=0x7fffe3b5d650, rc_secondary_proc_params=0x7f222d02dfe0, p_ia_attr=0x7fffe3b5d068, p_provider_attr=0x7fffe3b5d318) at ../../dapl_init_rc.c:948
#2 0x00007f222c9aea39 in MPID_nem_dapl_rc_init_20 (params=0x7fffe3b5d6c8, rc_params=0x7fffe3b5d650, pg_p=0x1e41480, pg_rank=1, fallback_device=0, provider_found=0x7fffe3b5d64c) at ../../dapl_init_rc.c:207
#3 0x00007f222ca10b0c in MPID_nem_dapl_init (pg_p=0x1e41480, pg_rank=1, bc_val_p=0x7fffe3b5d980, val_max_sz_p=0x7fffe3b5da98) at ../../dapls_module_init.c:437
#4 0x00007f222caf41a7 in MPID_nem_impi_netmod_init (pg_rank=1, pg_p=0x1e41480, ckpt_restart=0, bc_val=0x1e41bb0 "", pval_max_remaining=0x7fffe3b5da98, num_processes=2) at ../../mpid_nem_init.c:146
#5 0x00007f222caf7482 in MPID_nem_init_ckpt (pg_rank=1, pg_p=0x1e41480, ckpt_restart=0, has_parent=0) at ../../mpid_nem_init.c:859
#6 0x00007f222caf563e in MPID_nem_init (pg_rank=1, pg_p=0x1e41480, has_parent=0) at ../../mpid_nem_init.c:490
#7 0x00007f222c8f6a2a in MPIDI_CH3_Init (has_parent=0, pg_p=0x1e41480, pg_rank=1) at ../../ch3_init.c:64
#8 0x00007f222cad5244 in MPID_Init (argc=0x7fffe3b5e8a8, argv=0x7fffe3b5e8b0, requested=0, provided=0x7fffe3b5e4c0, has_args=0x7fffe3b5e4c4, has_env=0x7fffe3b5e4c8) at ../../mpid_init.c:193
#9 0x00007f222caa3019 in MPIR_Init_thread (argc=0x7fffe3b5e8a8, argv=0x7fffe3b5e8b0, required=0, provided=0x7fffe3b5e6bc) at ../../initthread.c:539
#10 0x00007f222ca90190 in PMPI_Init (argc=0x7fffe3b5e8a8, argv=0x7fffe3b5e8b0) at ../../init.c:195
#11 0x0000000000400a81 in main (argc=1, argv=0x7fffe3b5e9c8) at hello.c:26

Resolution

To correct this, upgrade the Mellanox* OFED driver to at least version 2.1-1.0.6.

Para obtener información más completa sobre las optimizaciones del compilador, consulte nuestro Aviso de optimización.