[mpich-discuss] Implementation of MPI blocking Send
Darius Buntinas
buntinas at mcs.anl.gov
Fri Mar 25 12:30:59 CDT 2011
Sorry, there isn't, but you should be able to step through it with either a debugger, or using something like etags in emacs (you can find a wiki page on using etags with emacs here: http://wiki.mcs.anl.gov/mpich2/index.php/Etags ).
-d
On Mar 25, 2011, at 12:00 PM, vaibhav dutt wrote:
> Hi,
>
> Thanks for the information. Is there any document that describes a simple blocking send routine going through different layers from a high level view?
>
> Thanks
>
> On Fri, Mar 25, 2011 at 11:12 AM, Darius Buntinas <buntinas at mcs.anl.gov> wrote:
> MPICH2 has several interfaces to support derivative implementations. One interface is the ADI3 or "device" interface. You can find the device implementations in src/mpid/ . The base MPICH2 distribution contains two device implementations: CH3 and dcmfd. The dcmfd device is specifically for BlueGene, so you can ignore that. The MPID_ functions are implemented by the devices, so you can find them in src/mpid/ch3/src .
>
> The CH3 device also provides an "channel" interface for derivative implementations. The base MPICH2 distribution has a sock channel and a nemesis channel. You can find these in src/mpid/ch3/channels/ . Currently nemesis is the default channel. Nemesis uses shared memory for intranode communication, and uses network modules for internode communication. You can find the network modules in src/mpid/ch3/channels/nemesis/nemesis/netmod . Information on the netmod API can be found here:
> http://wiki.mcs.anl.gov/mpich2/index.php/Nemesis_Network_Module_API
>
> I hope this helps!
>
> -d
>
> On Mar 25, 2011, at 9:51 AM, vaibhav dutt wrote:
>
> > Thanks for the reply. Actually I am trying to understand the inter-node communication procedure in MPICH.
> > Are there any documents/papers on it as I was not able to find one. Does MPICH use RDMA for inter node communication?
> >
> > On Thu, Mar 24, 2011 at 5:28 PM, Dave Goodell <goodell at mcs.anl.gov> wrote:
> > MPICH2 is a large, complicated code base. Numerous tools have been developed over the years for navigating large, complicated codes; try grep, cscope, or ctags.
> >
> > -Dave
> >
> > On Mar 24, 2011, at 5:21 PM CDT, vaibhav dutt wrote:
> >
> > > Hi all,
> > >
> > > I was trying to find out the implementation of a simple blocking send in MPICH for my research and it is given in a
> > > file send.c. In that file, a routine MPID_Send is being used to send the message. But I was not able to find the implementation of
> > > MPID_Send. Is the code for MPID_Send routine given in one of C++ files?
> > >
> > > Thanks
> > > _______________________________________________
> > > mpich-discuss mailing list
> > > mpich-discuss at mcs.anl.gov
> > > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> >
> > _______________________________________________
> > mpich-discuss mailing list
> > mpich-discuss at mcs.anl.gov
> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
> >
> > _______________________________________________
> > mpich-discuss mailing list
> > mpich-discuss at mcs.anl.gov
> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
More information about the mpich-discuss
mailing list