kpriv/freebsd-ports
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

net/mpich: unbreak optimized runtime after 88e134883d

Browse source 
Runtime may fail without a L0 driver like intel-compute-runtime e.g.,

$ mpivars
Abort(268484367) on node 0 (rank 0 in comm 0): Fatal error in PMPI_Init_thread: Other MPI error, error stack:
MPIR_Init_thread(153):  gpu_init failed
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=268484367
:
system msg for write_line failure : Bad file descriptor
Attempting to use an MPI routine before initializing MPICH

$ MPIR_CVAR_ENABLE_GPU=0 mpivars
Abort(2139535) on node 0 (rank 0 in comm 0): Fatal error in PMPI_Init_thread: Other MPI error, error stack:
MPIR_Init_thread(159)......:
MPID_Init(591).............:
MPIDI_SHM_mpi_init_hook(22):
MPIDI_IPC_mpi_init_hook(36):
MPIDI_GPU_mpi_init_hook(79):  gpu_get_dev_count failed
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=2139535
:
system msg for write_line failure : Bad file descriptor
Abort(2139535) on node 0 (rank 0 in comm 0): Fatal error in PMPI_Init_thread: Other MPI error, error stack:
MPIR_Init_thread(159)......:
MPID_Init(591).............:
MPIDI_SHM_mpi_init_hook(22):
MPIDI_IPC_mpi_init_hook(36):
MPIDI_GPU_mpi_init_hook(79):  gpu_get_dev_count failed
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=2139535
:
system msg for write_line failure : Bad file descriptor
Segmentation fault

PR:		256244 (for tracking)
This commit is contained in:
Henrik Gulbrandsen 2021-08-12 16:35:20 +02:00 committed by Jan Beich
parent ea525437cd
commit b5815e7648
2 changed files with 45 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
net/mpich/Makefile
View file

@ -1,6 +1,6 @@
PORTNAME= mpich PORTNAME= mpich
PORTVERSION= 3.4.2 PORTVERSION= 3.4.2
PORTREVISION= 2 PORTREVISION= 3
CATEGORIES= net parallel CATEGORIES= net parallel
MASTER_SITES= https://www.mpich.org/static/downloads/${DISTVERSION}/ MASTER_SITES= https://www.mpich.org/static/downloads/${DISTVERSION}/

44
net/mpich/files/patch-l0-fallback Normal file
View file

@ -0,0 +1,44 @@
$ pkg delete intel-compute-runtime
$ mpivars
PCI: Failed to initialize libpciaccess with pci_system_init(): 6 (Permission denied)
Abort(268484367) on node 0 (rank 0 in comm 0): Fatal error in PMPI_Init_thread: Other MPI error, error stack:
MPIR_Init_thread(153): gpu_init failed
[unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=268484367
:
system msg for write_line failure : Bad file descriptor
Attempting to use an MPI routine before initializing MPICH
--- src/mpi/init/initthread.c.orig 2021-05-25 17:37:05 UTC
+++ src/mpi/init/initthread.c
@@ -150,7 +150,9 @@ int MPIR_Init_thread(int *argc, char ***argv, int user
* inside MPID_Init */
if (MPIR_CVAR_ENABLE_GPU) {
int mpl_errno = MPL_gpu_init();
- MPIR_ERR_CHKANDJUMP(mpl_errno != MPL_SUCCESS, mpi_errno, MPI_ERR_OTHER, "**gpu_init");
+ MPIR_ERR_CHKANDJUMP(
+ mpl_errno != MPL_SUCCESS && mpl_errno != MPL_ERR_GPU_INTERNAL,
+ mpi_errno, MPI_ERR_OTHER, "**gpu_init");
}
MPL_atomic_store_int(&MPIR_Process.mpich_state, MPICH_MPI_STATE__IN_INIT);
--- src/mpid/ch4/netmod/ofi/ofi_init.c.orig 2021-05-25 17:37:05 UTC
+++ src/mpid/ch4/netmod/ofi/ofi_init.c
@@ -731,7 +731,6 @@ int MPIDI_OFI_mpi_init_hook(int rank, int size, int ap
MPL_gpu_malloc_host(&(MPIDI_OFI_global.am_bufs[i]), MPIDI_OFI_AM_BUFF_SZ);
MPIDI_OFI_global.am_reqs[i].event_id = MPIDI_OFI_EVENT_AM_RECV;
MPIDI_OFI_global.am_reqs[i].index = i;
- MPIR_Assert(MPIDI_OFI_global.am_bufs[i]);
MPIDI_OFI_global.am_iov[i].iov_base = MPIDI_OFI_global.am_bufs[i];
MPIDI_OFI_global.am_iov[i].iov_len = MPIDI_OFI_AM_BUFF_SZ;
MPIDI_OFI_global.am_msg[i].msg_iov = &MPIDI_OFI_global.am_iov[i];
--- src/mpl/src/gpu/mpl_gpu_ze.c.orig 2021-05-25 17:37:05 UTC
+++ src/mpl/src/gpu/mpl_gpu_ze.c
@@ -33,7 +33,7 @@ int MPL_gpu_get_dev_count(int *dev_cnt, int *dev_id)
{
int ret = MPL_SUCCESS;
if (!gpu_initialized) {
- ret = MPL_gpu_init();
+ MPL_gpu_init();
}
*dev_cnt = device_count;