<div>Sundeep,</div> <div> </div> <div>I am sorry. I did not make myself very clear in my questions. </div> <div> </div> <div>I have got the iWARP running on my set-up for the Ammasso and Chelsio cards.</div> <div> </div> <div>To do this, I downloaded the iWARP code from the gen2 branch.</div> <div> </div> <div>Now I was trying to run the OSC MPI tools MVAPICH2 there to run the iWARP traffic in my set-up.</div> <div> </div> <div>When I am trying to build the MVAPICH2 0.9.8 tool, it has some mandatory variable called, OPEN_IB_HOME = /usr/local/ofed in the makefile make.mvapich2.iwarp, the build file which is to be run for the iWARP code. </div> <div> </div> <div> </div> <div>My question is why should I have this reference to /usr/local/ofed there if I do not need to download the OFED distribution code to run the iWARP.</div> <div> </div> <div>Is it possible to add more information in your MVAPICH2 0.9.8 User Guide
describing how to build this and what are the dependencies?</div> <div><BR>David<BR><B><I>Sundeep Narravula <narravul@cse.ohio-state.edu></I></B> wrote:</div> <BLOCKQUOTE class=replbq style="PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #1010ff 2px solid">Hi David,<BR><BR>> iWARP is actually a part of the Open Fabrics SVN. It is available from<BR>> a different branch.<BR>><BR>> I am cc'ing this note to my group. One of the students (Sundeep) will<BR>> send you the detailed instructions on which branch of OF to download<BR>> and use.<BR><BR>The instructions for setting up iwarp on OpenFrabics is avalable on the<BR>openIB wiki at<BR>https://openib.org/tiki/tiki-index.php?page=Install+OpenIB+for+Chelsio+T3<BR>https://openib.org/tiki/tiki-index.php?page=Install+OpenIB+for+Ammasso1100<BR><BR>Further, the branch you can download from the svn is<BR>https://openib.org/svn/gen2/branches/iwarp/tags/iwarp-2.6.17-stable<BR><BR>Please let us know if you
have any further questions.<BR><BR>Regards,<BR>--Sundeep.<BR><BR><BR>> > Is there any document describing the build process of the MVAPICH2<BR>> > tool? I am going through the MVAPICH2 0.9.8 users guide and that<BR>> > does not seem to be giving me the detailed information.<BR>><BR>> We will add this additional information on our user guide.<BR>><BR>> > Can you please provide README files for the iWARP which describes<BR>> > TODO steps?<BR>><BR>> Sundeep's information will help. If you have any additional questions,<BR>> please feel free to ask us.<BR>><BR>> Thanks,<BR>><BR>> DK<BR>><BR>> > Thanks,<BR>> > David<BR>> ><BR>> > Dhabaleswar Panda <PANDA@CSE.OHIO-STATE.EDU>wrote:<BR>> > The MVAPICH team is pleased to announce the availability of MVAPICH2<BR>> > 0.9.8 with the following NEW features:<BR>> ><BR>> > - Checkpoint/Restart support for application transparent
systems-level<BR>> > fault tolerance. BLCR-based support using native InfiniBand Gen2<BR>> > interface is provided. Flexible interface to work with different<BR>> > file systems. Tested with ext3 (local disk), NFS and PVFS2.<BR>> ><BR>> > Performance of sample applications with checkpoint-restart using<BR>> > PVFS2 and Lustre can be found here:<BR>> ><BR>> > http://nowlab.cse.ohio-state.edu/projects/mpi-iba/performance/mvapich2/application/MVAPICH2-ckpt.html<BR>> ><BR>> > - iWARP support: Incorporates the support for OpenFabrics/Gen2-iWARP.<BR>> > Tested with Chelsio T3 (10GigE) and Ammasso iWARP adapters and<BR>> > drivers.<BR>> ><BR>> > - RDMA CM-based Connection management support<BR>> ><BR>> > - Shared memory optimizations for collective communication operations.<BR>> > Efficient algorithms and optimizations for barrier, reduce and<BR>> > all-reduce operations.
Exploits the multi-core optimized shared<BR>> > memory point-to-point communication support introduced in MVAPICH2<BR>> > 0.9.6.<BR>> ><BR>> > Performance of sample collective operations with this new feature<BR>> > can be found here:<BR>> ><BR>> > http://nowlab.cse.ohio-state.edu/projects/mpi-iba/perf-coll.html<BR>> ><BR>> > - uDAPL support for NetEffect 10GigE adapter. Tested with<BR>> > NetEffect NE010 adapter.<BR>> ><BR>> > More details on all features and supported platforms can be obtained<BR>> > by visiting the following URL:<BR>> ><BR>> > http://nowlab.cse.ohio-state.edu/projects/mpi-iba/mvapich2_features.html<BR>> ><BR>> > MVAPICH2 0.9.8 release is tested with the latest OFED 1.1 stack. It<BR>> > continues to deliver excellent performance. Sample performance<BR>> > numbers include:<BR>> ><BR>> > - OpenFabrics/Gen2 on EM64T dual-core with
PCI-Ex and IBA-DDR:<BR>> > Two-sided operations:<BR>> > - 2.81 microsec one-way latency (4 bytes)<BR>> > - 1561 MB/sec unidirectional bandwidth<BR>> > - 2935 MB/sec bidirectional bandwidth<BR>> ><BR>> > One-sided operations:<BR>> > - 4.92 microsec Put latency<BR>> > - 1569 MB/sec unidirectional Put bandwidth<BR>> > - 2935 MB/sec bidirectional Put bandwidth<BR>> ><BR>> > - OpenFabrics/Gen2 on EM64T dual-core with PCI-Ex and IBA-DDR (Dual-rail):<BR>> > Two-sided operations:<BR>> > - 2.81 microsec one-way latency (4 bytes)<BR>> > - 3127 MB/sec unidirectional bandwidth<BR>> > - 5917 MB/sec bidirectional bandwidth<BR>> ><BR>> > One-sided operations:<BR>> > - 4.37 microsec Put latency<BR>> > - 3137 MB/sec unidirectional Put bandwidth<BR>> > - 5917 MB/sec bidirectional Put bandwidth<BR>> ><BR>> > - OpenFabrics/Gen2 on Opteron single-core with PCI-Ex
and IBA-DDR:<BR>> > Two-sided operations:<BR>> > - 3.01 microsec one-way latency (4 bytes)<BR>> > - 1402 MB/sec unidirectional bandwidth<BR>> > - 2238 MB/sec bidirectional bandwidth<BR>> ><BR>> > One-sided operations:<BR>> > - 4.65 microsec Put latency<BR>> > - 1402 MB/sec unidirectional Put bandwidth<BR>> > - 2238 MB/sec bidirectional Put bandwidth<BR>> ><BR>> > Performance numbers for all other platforms, system configurations and<BR>> > operations can be viewed by visiting `Performance' section of the<BR>> > project's web page.<BR>> ><BR>> > With the ADI-3-level design, MVAPICH2 0.9.8 delivers similar<BR>> > performance for two-sided operations compared to MVAPICH 0.9.8.<BR>> > Organizations and users interested in getting the best performance for<BR>> > both two-sided and one-sided operations and also want to exploit<BR>> > advanced features (such as fault
tolerance with checkpoint/restart,<BR>> > iWARP, RDMA CM connection management, multi-threading, integrated<BR>> > multi-rail, multi-core optimization, memory hook support and optimized<BR>> > collectives) may migrate from MVAPICH code base to MVAPICH2 code base.<BR>> ><BR>> > For downloading MVAPICH2 0.9.8 package and accessing the anonymous<BR>> > SVN, please visit the following URL:<BR>> ><BR>> > http://nowlab.cse.ohio-state.edu/projects/mpi-iba/<BR>> ><BR>> > A stripped down version of this release is also available at the<BR>> > OpenFabrics SVN.<BR>> ><BR>> > All feedbacks, including bug reports and hints for performance tuning,<BR>> > are welcome. Please post it to the mvapich-discuss mailing list.<BR>> ><BR>> > Thanks,<BR>> ><BR>> > MVAPICH Team at OSU/NBCL<BR>> ><BR>> > ======================================================================<BR>>
> MVAPICH/MVAPICH2 project is currently supported with funding from<BR>> > U.S. National Science Foundation, U.S. DOE Office of Science,<BR>> > Mellanox, Intel, Cisco Systems, Sun Microsystems and Linux Networx;<BR>> > and with equipment support from Advanced Clustering, AMD, Apple,<BR>> > Appro, Dell, IBM, Intel, Mellanox, Microway, PathScale, SilverStorm<BR>> > and Sun Microsystems. Other technology partner includes Etnus.<BR>> > ======================================================================<BR>> ><BR>> > _______________________________________________<BR>> > openfabrics-ewg mailing list<BR>> > openfabrics-ewg@openib.org<BR>> > http://openib.org/mailman/listinfo/openfabrics-ewg<BR>> ><BR>> ><BR>> ><BR>> ><BR>> > ---------------------------------<BR>> > Sponsored Link<BR>> ><BR>> > Try Netflix today! With plans starting at only $5.99 a month what
are you waiting for?<BR>> > --0-119921422-1163376356=:19890<BR>> > Content-Type: text/html; charset=iso-8859-1<BR>> > Content-Transfer-Encoding: 8bit<BR>> ><BR>> > <DIV>I am trying to use the OSC MPI tool for the iWARP and quite new to the open fabrics tools.</DIV> <DIV> </DIV> <DIV>If I know it correctly, iWARP is not yet part of the OFED release. </DIV> <DIV>But the iWARP makefile has reference to ofed code. Is it really required.</DIV> <DIV> </DIV> <DIV>Is there any document describing the build process of the MVAPICH2 tool? I am going through the MVAPICH2 0.9.8 users guide and that does not seem to be giving me the detailed information.</DIV> <DIV> </DIV> <DIV>Can you please provide README files for the iWARP which describes TODO steps?</DIV> <DIV> </DIV> <DIV>Thanks,</DIV> <DIV>David<BR><BR><B><I>Dhabaleswar Panda <panda@cse.ohio-state.edu></I></B> wrote:</DIV> <BLOCKQUOTE class=replbq
style="PADDING-LEFT: 5px; MARGIN-LEFT: 5px; BORDER-LEFT: #1010ff 2px solid">The MVAPICH team is pleased to announce the availability of MVAPICH2<BR>0.9.8 with the following NEW features:<BR><BR>-<BR>> > Checkpoint/Restart support for application transparent systems-level<BR>fault tolerance. BLCR-based support using native InfiniBand Gen2<BR>interface is provided. Flexible interface to work with different<BR>file systems. Tested with ext3 (local disk), NFS and PVFS2. <BR><BR>Performance of sample applications with checkpoint-restart using <BR>PVFS2 and Lustre can be found here:<BR><BR>http://nowlab.cse.ohio-state.edu/projects/mpi-iba/performance/mvapich2/application/MVAPICH2-ckpt.html<BR><BR>- iWARP support: Incorporates the support for OpenFabrics/Gen2-iWARP.<BR>Tested with Chelsio T3 (10GigE) and Ammasso iWARP adapters and<BR>drivers.<BR><BR>- RDMA CM-based Connection management support<BR><BR>- Shared memory optimizations for collective communication operations.
<BR>Efficient algorithms and optimizations for barrier, reduce and<BR>all-reduce operations. Exploits the multi-core optimized shared<BR>memory point-to-point communication support introduced in<BR>> > MVAPICH2<BR>0.9.6.<BR><BR>Performance of sample collective operations with this new feature<BR>can be found here:<BR><BR>http://nowlab.cse.ohio-state.edu/projects/mpi-iba/perf-coll.html <BR><BR>- uDAPL support for NetEffect 10GigE adapter. Tested with <BR>NetEffect NE010 adapter.<BR><BR>More details on all features and supported platforms can be obtained<BR>by visiting the following URL:<BR><BR>http://nowlab.cse.ohio-state.edu/projects/mpi-iba/mvapich2_features.html<BR><BR>MVAPICH2 0.9.8 release is tested with the latest OFED 1.1 stack. It<BR>continues to deliver excellent performance. Sample performance<BR>numbers include:<BR><BR>- OpenFabrics/Gen2 on EM64T dual-core with PCI-Ex and IBA-DDR:<BR>Two-sided operations: <BR>- 2.81 microsec one-way latency (4 bytes)<BR>-
1561 MB/sec unidirectional bandwidth <BR>- 2935 MB/sec bidirectional bandwidth <BR><BR>One-sided operations:<BR>- 4.92 microsec Put latency <BR>- 1569 MB/sec unidirectional Put bandwidth <BR>- 2935 MB/sec<BR>> > bidirectional Put bandwidth <BR><BR>- OpenFabrics/Gen2 on EM64T dual-core with PCI-Ex and IBA-DDR (Dual-rail):<BR>Two-sided operations:<BR>- 2.81 microsec one-way latency (4 bytes)<BR>- 3127 MB/sec unidirectional bandwidth<BR>- 5917 MB/sec bidirectional bandwidth<BR><BR>One-sided operations:<BR>- 4.37 microsec Put latency<BR>- 3137 MB/sec unidirectional Put bandwidth<BR>- 5917 MB/sec bidirectional Put bandwidth<BR><BR>- OpenFabrics/Gen2 on Opteron single-core with PCI-Ex and IBA-DDR: <BR>Two-sided operations:<BR>- 3.01 microsec one-way latency (4 bytes) <BR>- 1402 MB/sec unidirectional bandwidth<BR>- 2238 MB/sec bidirectional bandwidth <BR><BR>One-sided operations:<BR>- 4.65 microsec Put latency<BR>- 1402 MB/sec unidirectional Put bandwidth<BR>- 2238 MB/sec
bidirectional Put bandwidth<BR><BR>Performance numbers for all other platforms, system configurations and<BR>operations can be viewed by visiting `Performance' section of the<BR>project's web page.<BR><BR>With the<BR>> > ADI-3-level design, MVAPICH2 0.9.8 delivers similar<BR>performance for two-sided operations compared to MVAPICH 0.9.8.<BR>Organizations and users interested in getting the best performance for<BR>both two-sided and one-sided operations and also want to exploit<BR>advanced features (such as fault tolerance with checkpoint/restart,<BR>iWARP, RDMA CM connection management, multi-threading, integrated<BR>multi-rail, multi-core optimization, memory hook support and optimized<BR>collectives) may migrate from MVAPICH code base to MVAPICH2 code base.<BR><BR>For downloading MVAPICH2 0.9.8 package and accessing the anonymous<BR>SVN, please visit the following URL:<BR><BR>http://nowlab.cse.ohio-state.edu/projects/mpi-iba/<BR><BR>A stripped down version of this
release is also available at the<BR>OpenFabrics SVN.<BR><BR>All feedbacks, including bug reports and hints for performance tuning,<BR>are welcome. Please post it to the mvapich-discuss mailing list.<BR><BR>Thanks,<BR>> > <BR><BR>MVAPICH Team at OSU/NBCL <BR><BR>======================================================================<BR>MVAPICH/MVAPICH2 project is currently supported with funding from<BR>U.S. National Science Foundation, U.S. DOE Office of Science,<BR>Mellanox, Intel, Cisco Systems, Sun Microsystems and Linux Networx;<BR>and with equipment support from Advanced Clustering, AMD, Apple,<BR>Appro, Dell, IBM, Intel, Mellanox, Microway, PathScale, SilverStorm<BR>and Sun Microsystems. Other technology partner includes Etnus.<BR>======================================================================<BR><BR>_______________________________________________<BR>openfabrics-ewg mailing
list<BR>openfabrics-ewg@openib.org<BR>http://openib.org/mailman/listinfo/openfabrics-ewg<BR><BR></BLOCKQUOTE><BR> <div><BR>> ><BR>> ><BR>> > <HR SIZE=1> Sponsored Link<BR><BR><BR>> > <A href="http://clk.atdmt.com/NFX/go/yhxxxnfx0020000913nfx/direct/01/">Try Netflix today! With plans starting at only $5.99 a month what are you waiting for?</A><BR>> > --0-119921422-1163376356=:19890--<BR>> ><BR>><BR><BR></BLOCKQUOTE><BR><p>
<hr size=1>Sponsored Link<br><br>
$200,000 mortgage for $660/mo - 30/15 yr fixed, reduce debt, home equity - <a href="http://www.ratemarketplace.com/forms/form.jsp?ADEL=Os4fVJ19lSwzAOc/zMz/CYLqA5sBlJ/HICjGxkSe1FKtmWyM5TR8NmykLaxQaZnu&PI=y_mail_textlink">Click now for info</a>