<div dir="ltr">Hi Folks,<div><br></div><div>On cori/nersc, I have a patched criterion at </div><div><br></div><div>/global/homes/h/hpp/criterion_install</div><div><br></div><div>that can be used to run the GNI provider unit tests.</div><div><br></div><div>See</div><div><br></div><div><a href="https://github.com/ofi-cray/libfabric-cray/wiki/Building-and-running-the-unit-tests-(gnitest)">https://github.com/ofi-cray/libfabric-cray/wiki/Building-and-running-the-unit-tests-(gnitest)</a><br></div><div><br></div><div>for more info.</div><div><br></div><div>Howard</div><div><br></div>
</div><div class="gmail_extra"><br><div class="gmail_quote">2016-03-18 17:27 GMT-06:00 Sung-Eun Choi <span dir="ltr"><<a href="mailto:sungeun@cray.com" target="_blank">sungeun@cray.com</a>></span>:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">I left off one important detail about our unit tests run by<br>
run_gnitest: You'll need to download and build the unit test<br>
infrastructure (Criterion). See directions at<br>
<a href="https://github.com/ofi-cray/libfabric-cray/wiki/Building-and-running-the-unit-tests-(gnitest)" rel="noreferrer" target="_blank">https://github.com/ofi-cray/libfabric-cray/wiki/Building-and-running-the-unit-tests-(gnitest)</a>.<br>
<br>
It's possible that Howard Pritchard has installed a version of it on<br>
Cori for others to use. I'll let him speak up.<br>
<span class="HOEnZb"><font color="#888888"><br>
-- Sung<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
On Fri, Mar 18, 2016 at 04:20:39PM -0700, Sung-Eun Choi wrote:<br>
> Hi Greg,<br>
><br>
> Thanks for you mail. The launching mechanism used on Cori and other<br>
> XCs is not compatible with the way most of the fabtests need to be<br>
> started. That said, we're working on getting the gni provider working<br>
> in Cray Cluster Mode, which should make it easier to run the fabtests<br>
> (hopefully almost a no op).<br>
><br>
> In the meantime, there are a couple of options. We have a large<br>
> number of unit tests in our libfabric directory under prov/gni/test.<br>
> There's a script there called run_gnitest that will do the right thing<br>
> for slurm and aprun, though you have to get yourself a single node<br>
> allocation. Note that these tests are not distributed in the release.<br>
> As another option, you can also use our forked version of fabtests<br>
> (<a href="https://github.com/ofi-cray/fabtests-cray" rel="noreferrer" target="_blank">https://github.com/ofi-cray/fabtests-cray</a>), which includes a script<br>
> called cray_runall.sh. This runs as many of the original fabtests as<br>
> possible (using srun or aprun) and, if you build with PMI, a few<br>
> extras. There are also a couple files in the scripts directory that<br>
> list the expected and intermittent failures<br>
> (cray_runall_expected_failures and cray_runall_intermitten_failures).<br>
> Again, you'll have to obtain your own allocation for this script.<br>
><br>
> While we're on the topic, how could we have saved you the trouble of<br>
> trying all this stuff? Documentation in the man page? Info on our<br>
> wiki? Something else?<br>
><br>
> Thanks.<br>
><br>
> -- Sung<br>
><br>
> On Fri, Mar 18, 2016 at 10:52:38PM +0000, Eisenhauer, Greg S wrote:<br>
> > Hi Folks,<br>
> ><br>
> > I was trying to work with libfabric on Cori with the gni provider and have run into an early roadblock just trying to exercise runfabtests.sh in order to make sure everything was working. Have tried both v1.2 and current GIT HEAD for libfabric and fabtests. Am I missing something basic? A requirement to set SERVER, CLIENT or GOOD_ADDR? Output from fi_info, trying to run an individual test manually (other tests show similar output), and output from ‘runfabtests.sh gni’ are included below. I may be doing something stupid, but browsing various wikis, email lists and other resources for a while hasn’t produced any answers, so I thought I’d ask here.<br>
> ><br>
> > thanks,<br>
> > greg<br>
> ><br>
> ><br>
> ><br>
> > eisen@nid00050:~/fabtests-1.2.0/scripts> fi_info<br>
> > gni: gni<br>
> > version: 1.0<br>
> > type: FI_EP_RDM<br>
> > protocol: FI_PROTO_GNI<br>
> > UDP: UDP-IP<br>
> > version: 1.0<br>
> > type: FI_EP_DGRAM<br>
> > protocol: FI_PROTO_UDP<br>
> > sockets: IP<br>
> > version: 1.0<br>
> > type: FI_EP_MSG<br>
> > protocol: FI_PROTO_SOCK_TCP<br>
> > sockets: IP<br>
> > version: 1.0<br>
> > type: FI_EP_DGRAM<br>
> > protocol: FI_PROTO_SOCK_TCP<br>
> > sockets: IP<br>
> > version: 1.0<br>
> > type: FI_EP_RDM<br>
> > protocol: FI_PROTO_SOCK_TCP<br>
> > eisen@nid00050:~/fabtests-1.2.0/scripts> fi_msg_rma -o write -I 5 -f gni<br>
> > fi_getinfo(): common/shared.c:340, ret=-61 (No data available)<br>
> > eisen@nid00050:~/fabtests-1.2.0/scripts> ./runfabtests.sh gni<br>
> > # Test Result<br>
> > # --------------------------------------------------------------<br>
> > fi_av_test -d 192.168.10.1 -n 1 -s 127.0.0.1 -f gni: Notrun<br>
> > fi_dom_test -n 2 -f gni: Pass<br>
> > fi_eq_test -f gni: Fail<br>
> > fi_size_left_test -f gni: Fail<br>
> > fi_cq_data -f gni: Notrun<br>
> > fi_dgram -f gni: Notrun<br>
> > fi_dgram_waitset -f gni: Notrun<br>
> > fi_msg -f gni: Notrun<br>
> > fi_msg_epoll -f gni: Notrun<br>
> > fi_msg_sockets -f gni: Notrun<br>
> > fi_poll -f gni: Notrun<br>
> > fi_rdm -f gni: Notrun<br>
> > fi_rdm_rma_simple -f gni: Notrun<br>
> > fi_rdm_rma_trigger -f gni: Notrun<br>
> > fi_rdm_shared_ctx -f gni: Notrun<br>
> > fi_rdm_tagged_peek -f gni: Notrun<br>
> > fi_scalable_ep -f gni: Notrun<br>
> > fi_cmatose -f gni: Notrun<br>
> > fi_msg_pingpong -I 5 -f gni: Notrun<br>
> > fi_msg_rma -o write -I 5 -f gni: Notrun<br>
> > fi_msg_rma -o read -I 5 -f gni: Notrun<br>
> > fi_msg_rma -o writedata -I 5 -f gni: Notrun<br>
> > fi_rdm_atomic -I 5 -o all -f gni: Notrun<br>
> > fi_rdm_cntr_pingpong -I 5 -f gni: Notrun<br>
> > fi_rdm_inject_pingpong -I 5 -f gni: Notrun<br>
> > fi_rdm_multi_recv -I 5 -f gni: Notrun<br>
> > fi_rdm_pingpong -I 5 -f gni: Notrun<br>
> > fi_rdm_rma -o write -I 5 -f gni: Notrun<br>
> > fi_rdm_rma -o read -I 5 -f gni: Notrun<br>
> > fi_rdm_rma -o writedata -I 5 -f gni: Notrun<br>
> > fi_rdm_tagged_pingpong -I 5 -f gni: Notrun<br>
> > fi_ud_pingpong -I 5 -f gni: Notrun<br>
> > fi_rc_pingpong -n 5 -f gni: Notrun<br>
> > fi_rc_pingpong -n 5 -e -f gni: Notrun<br>
> > # --------------------------------------------------------------<br>
> > # Total Pass 1<br>
> > # Total Notrun 31<br>
> > # Total Fail 2<br>
> > # Percentage of Pass 33<br>
> > # --------------------------------------------------------------<br>
> > eisen@nid00050:~/fabtests-1.2.0/scripts><br>
> ><br>
> ><br>
> > --------------------------<br>
> > Greg Eisenhauer<br>
> > <a href="mailto:eisen@cc.gatech.edu">eisen@cc.gatech.edu</a><mailto:<a href="mailto:eisen@cc.gatech.edu">eisen@cc.gatech.edu</a>><br>
> ><br>
><br>
> > _______________________________________________<br>
> > Libfabric-users mailing list<br>
> > <a href="mailto:Libfabric-users@lists.openfabrics.org">Libfabric-users@lists.openfabrics.org</a><br>
> > <a href="http://lists.openfabrics.org/mailman/listinfo/libfabric-users" rel="noreferrer" target="_blank">http://lists.openfabrics.org/mailman/listinfo/libfabric-users</a><br>
><br>
> _______________________________________________<br>
> Libfabric-users mailing list<br>
> <a href="mailto:Libfabric-users@lists.openfabrics.org">Libfabric-users@lists.openfabrics.org</a><br>
> <a href="http://lists.openfabrics.org/mailman/listinfo/libfabric-users" rel="noreferrer" target="_blank">http://lists.openfabrics.org/mailman/listinfo/libfabric-users</a><br>
_______________________________________________<br>
Libfabric-users mailing list<br>
<a href="mailto:Libfabric-users@lists.openfabrics.org">Libfabric-users@lists.openfabrics.org</a><br>
<a href="http://lists.openfabrics.org/mailman/listinfo/libfabric-users" rel="noreferrer" target="_blank">http://lists.openfabrics.org/mailman/listinfo/libfabric-users</a><br>
</div></div></blockquote></div><br></div>