There is no doubt you will lose performance, but remember that shared memory is
-only- used between procs on the same node. So the performance penalty for
multi-node applications is considerably less.
Ultimately, the magnitude of the performance loss depends strongly on the
design of the applic
Hi Ralph,
I wonder if I disable the btl-sm component ,the performance of message
passing will degrade.I need to test how much performance are lost. The code
quality now is far from to be integrated into trunk, I will try to clean up
my code before sending these changes to you .
在 2010年8月13日 下午1
On Aug 12, 2010, at 11:35 PM, 张晶 wrote:
> I think there are still many test needed to be launching .Maybe I'd
> better to look into MTT.
If you want to start using MTT, have a look at these wiki pages:
https://svn.open-mpi.org/trac/mtt/wiki/MTTOverview
https://svn.open-mpi.org/tr
Note that for submitting anything more than trivial patches, we will need a
signed Open MPI 3rd party contribution agreement on file. Please see:
http://www.open-mpi.org/community/contribute/
On Aug 13, 2010, at 12:00 AM, Ralph Castain wrote:
> This sounds like excellent progress! Jeff a
This sounds like excellent progress! Jeff and others know much more about MTT
than I do, so I'll leave that question to them.
You have two approaches to the mmap issue. Easiest for now would be to simply
disable the shared memory component - you can either turn it off at run-time
with -mca btl
Hi Ralph,Jeff and all
It is a good news that I can almost run the openmpi on the vxworks
,but there are also still some bugs.The final test which has passed is:
Rank 0 process calls mpi_send running on the host 0,rank 1 process
calls mpi_recv running on the host 1. It works well .For
Thank you ,Squyres , it is really useful !
在 2010年7月7日 下午7:22，Jeff Squyres 写道：
> On Jul 6, 2010, at 10:48 PM, 张晶 wrote:
>
> > 1.If I write a rlogin component ,
>
> Is the command line of rlogin that much different than that of rsh/ssh?
> For example, can you just s/rsh/rlogin/ on the overall com
On Jul 6, 2010, at 10:48 PM, 张晶 wrote:
> 1.If I write a rlogin component ,
Is the command line of rlogin that much different than that of rsh/ssh? For
example, can you just s/rsh/rlogin/ on the overall command line and have it
just work?
If so, I suspect that tweaking the rsh plm might be far
Hi all and ralph,
Going through the rsh component's launch job function , the daemon here
means the ssh session connecting to the nodes , I think this daemon is orted
, and in the orte_plm_base_launch_apps ,it use the selected odls component
to send command to the orted to spawn process. May what I
Hi all ,
It is really easy to create a new component , copy the origin one and modify
names ,run autogen.sh .thanks to the mca .But it is really hard to get out
what the role of the plm component play.I simpily know plm is used to launch
jobs. I tried to read the rsh component to figure out the me
According to the vxworks docs, they created a notion of "process" statin with
the 6.0 release. Use rtpSpawn to launch the process and you should be okay as
that is the equivalent of fork/exec.
On Jul 5, 2010, at 8:12 PM, 张晶 wrote:
> Hi Ralph ,
>
> It is really a bad news that vxworks even doe
The wiki(https://svn.open-mpi.org/trac/ompi/wiki) has some useful
information for developers. For creating a component -
https://svn.open-mpi.org/trac/ompi/wiki/devel/CreateComponent
Regards
--Nysal
2010/7/6 张晶
> Hi Ralph ,
>
> It is really a bad news that vxworks even doesn't include rsh serv
Hi Ralph ,
It is really a bad news that vxworks even doesn't include rsh server. I have
to write a plm component using rlogin. After going through the openmpi
website ,I can only find the some useful slides,However it is really not
enough . I wonder if there is a document like "how to write a
com
The rsh component checks for the existence of "fork", which may not be present
in vxworks. Is it? I'm afraid that component uses fork, so you may have to
create a vxworks component that doesn't, or at least uses the equivalent.
On Jul 1, 2010, at 9:32 PM, 张晶 wrote:
> Hi Ralph,
>
> I just copy
Hi Ralph,
I just copy your code in the trunk ,it seems working well. But the
"mpirun ls " still stuck at the orte_plm_select ,after using ompi_info
,there is no available component for plm. I think my configure option
is not the problem, for I use the same configure option to build the
program und
Would you mind filing them when you get a chance?
Thanks!
On Jun 30, 2010, at 10:03 AM, Ralph Castain wrote:
> Believe so - just a minor tweak to paffinity select.
>
> On Jun 30, 2010, at 5:03 AM, Jeff Squyres wrote:
>
> > Ralph -- are these easy CMRs?
> >
> > On Jun 30, 2010, at 12:03 AM, Ra
Believe so - just a minor tweak to paffinity select.
On Jun 30, 2010, at 5:03 AM, Jeff Squyres wrote:
> Ralph -- are these easy CMRs?
>
> On Jun 30, 2010, at 12:03 AM, Ralph Castain wrote:
>
>> You may be working off of an old version of OMPI - I updated opal_paffinity
>> awhile ago to no long
Ralph -- are these easy CMRs?
On Jun 30, 2010, at 12:03 AM, Ralph Castain wrote:
> You may be working off of an old version of OMPI - I updated opal_paffinity
> awhile ago to no longer require that a component be selected. Then you can
> no-build all paffinity components if you like and the sys
Hi Ralph ,
I used v1.4 and I checked out the 1.5.0 rc and the trunk . It seems
your code have been commited in the trunk but not the v1.5 under
branch.
Because it is not a easy work for me to switch the workcopy to trunk ,
I think it is a good idea to just conform the
opal_paffinity_base_select un
You may be working off of an old version of OMPI - I updated opal_paffinity
awhile ago to no longer require that a component be selected. Then you can
no-build all paffinity components if you like and the system will still startup
okay.
I don't believe this was moved over to the 1.4 release bra
Hi all ,
I tried to run the openmpi on vxworks. Now I can run program ompi_info
etc. But it failed running "mpirun ls", the error is :
--
It looks like opal_init failed for some reason; your parallel process is
likely to abor
21 matches
Mail list logo
