Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] [OMPI svn-full] svn:open-mpi r27739 - in trunk: ompi/mca/btl/sm ompi/mca/common/sm ompi/mca/mpool/sm opal/mca/shmem opal/mca/shmem/mmap opal/mca/shmem/posix opal/mca/shmem/sysv opal/mca/shmem/windows
From: Ralph Castain (rhc_at_[hidden])
Date: 2013-01-04 16:16:39


On Jan 4, 2013, at 12:57 PM, George Bosilca <bosilca_at_[hidden]> wrote:

> Sam,
>
> This is a major change and would have deserved an RFC, as it impose a drastic/major non-scalable change (up to now the backend file creation was centralized, not in addition we exchange the data through the modex).

I guess that's subject to interpretation as to how drastic and non-scalable this is - but this was discussed at length at last year's developer meeting and in subsequent calls.

> A quick look highlight the fact that quite a lot of new modex entries have appeared after this patch. On a 4 proc (2x2) we got more than 20 entries each one of them up to 32 bytes (he list is attached at the end of this email).

It looks to me like an optimization is missing - I suspect because Sam was doing this in two phases. When completed, we should see only the local-rank=0 proc on each node emit three keys (hoping to condense that to only two as I believe one may no longer be needed). All other procs will simply consume them.

>
> Clearly this new approach is significantly less scalable compared with the old one.

It shouldn't be, when completed.

> In the past we had issues adding one single integer per process, I fail to understand how our standards changed so much that now few hundreds bytes per process become acceptable. Moreover, what is the benefit this change provides in exchange of this loss of scalability?

We need to remove the RML from the shared memory startup. In the past, we used the RML to pass the rendezvous info between the procs on the node. Once the BTLs are moved, this will no longer be possible. So the only other option for now is to use the modex.

As has been discussed at the meetings, future plans (once the BTLs have moved) will remove this info from the modex except for direct-launch cases. For cases where we have orteds, the orteds will open the backing files and pass the info down to the local procs, thus eliminating the entire rendezvous protocol. Hopefully, we'll get there in the not-too-distant future.

>
> George.
>
> PS: The exhaustive list of new SM-related modex entries:
> [dancer01:01049] [[50563,1],0] db:hash:store: storing key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer02:01720] [[50563,1],1] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],0]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],0]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer01:01049] [[50563,1],0] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer02:01721] [[50563,1],3] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-0-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-0-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-1-0[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-1-1[OPAL_STRING] for proc [[50563,1],1]
> [dancer01:01050] [[50563,1],2] db:hash:store: storing pointer of key btl.sm.1.9-2[OPAL_BYTE_OBJECT] for proc [[50563,1],1]
>
>
> On Jan 3, 2013, at 22:52 , svn-commit-mailer_at_[hidden] wrote:
>
>> Author: samuel (Samuel K. Gutierrez)
>> Date: 2013-01-03 16:52:20 EST (Thu, 03 Jan 2013)
>> New Revision: 27739
>> URL: https://svn.open-mpi.org/trac/ompi/changeset/27739
>>
>> Log:
>> sm BTL initialization via modex, as discussed at last year's meeting.
>>
>> Text files modified:
>> trunk/ompi/mca/btl/sm/btl_sm.c | 337 +++++++++++++++++++++--------
>> trunk/ompi/mca/btl/sm/btl_sm.h | 60 +++++
>> trunk/ompi/mca/btl/sm/btl_sm_component.c | 444 ++++++++++++++++++++++++++++++++++++++-
>> trunk/ompi/mca/btl/sm/help-mpi-btl-sm.txt | 6
>> trunk/ompi/mca/common/sm/common_sm.c | 92 +++++--
>> trunk/ompi/mca/common/sm/common_sm.h | 45 +++
>> trunk/ompi/mca/mpool/sm/mpool_sm.h | 17
>> trunk/ompi/mca/mpool/sm/mpool_sm_component.c | 111 ++++-----
>> trunk/opal/mca/shmem/mmap/shmem_mmap_module.c | 7
>> trunk/opal/mca/shmem/posix/shmem_posix_module.c | 9
>> trunk/opal/mca/shmem/shmem_types.h | 36 ++
>> trunk/opal/mca/shmem/sysv/shmem_sysv_module.c | 11
>> trunk/opal/mca/shmem/windows/shmem_windows_module.c | 7
>> 13 files changed, 933 insertions(+), 249 deletions(-)
>
>
> _______________________________________________
> devel mailing list
> devel_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/devel