[mpich-discuss] [mvapich-discuss] mvapich-2.2rc2 build failure with psm/per-object thread-cs.

Huiwei Lu huiweilu at mcs.anl.gov
Fri Mar 27 22:19:09 CDT 2015


Hi Jian and Abhishek,

The build problem should have been fixed in the latest mpich master branch.

You can find the change here:
http://trac.mpich.org/projects/mpich/changeset/983a60e8fd6490daf15301f5f092795de941ed8d/

Please let me know if it works.

Thanks,
--
Huiwei Lu
Postdoc Appointee
Mathematics and Computer Science Division
Argonne National Laboratory

On Thu, Mar 26, 2015 at 7:56 PM, Huiwei Lu <huiweilu at mcs.anl.gov> wrote:

> Hi Jian and Abhishek,
>
> Thanks for reporting this.
>
> When configured with --enable-thread-cs=per-object, MPID_cc_t will be
> defined as OPA_int_t, which can not be directly compared with 0 using
> "==".
>
> I will create a patch for this.
>
> Best regards,
>
> --
> Huiwei Lu
> Postdoc Appointee
> Mathematics and Computer Science Division
> Argonne National Laboratory
>
> On Thu, Mar 26, 2015 at 4:32 PM, Jian Lin <lin.2180 at osu.edu> wrote:
>
>> Hi, Abhishek,
>>
>> Thanks for your note. By using the full configure options you provided,
>> we are able to reproduce this behavior. However, MVAPICH2 2.1rc2 is
>> based on MPICH 3.1.4. We found that the error also exists with the
>> MPICH 3.1.4/3.2b1 even if using ch3:nemesis device. It will be good if
>> the MPICH team resolves this. I am cc'ing this note to the MPICH team.
>>
>> Hi, MPICH community,
>>
>> The forwarded compilation error reported by the user is related to
>> unmatched MPID_cc_t typedefs if some thread-related configure options
>> are used. The default build without these thread-related options does
>> not have any issue. Will you please take a look. Thanks!
>>
>>
>> On Mon, 23 Mar 2015 20:29:12 -0400
>> Abhishek Kulkarni <adkulkar at indiana.edu> wrote:
>>
>> > Jonathan,
>> >
>> > I am using the following configure flags:
>> >
>> > ./configure --enable-threads=multiple --enable-async-progress
>> > --with-device=ch3:psm --enable-thread-cs=per-object CFLAGS=-O3
>> > --enable-refcount=lock-free --enable-handle-allocation=tls
>> > --with-atomic-primitives
>> >
>> > The distribution is Ubuntu 12.04.5 LTS and the compiler version is gcc
>> > (Ubuntu/Linaro 4.6.4-1ubuntu1~12.04) 4.6.4.
>> >
>> > Abhishek
>> >
>> >
>> > On Tue, Mar 17, 2015 at 5:06 PM, Jonathan Perkins <
>> > perkinjo at cse.ohio-state.edu> wrote:
>> >
>> > > Thank you for your report.  When trying to reproduce this problem
>> > > locally I am running into a similar but different build error.  Can
>> > > you provide us more information on how you're building?  Please let
>> > > us know the Linux distribution and version and vendor of the
>> > > compiler that you're using.
>> > >
>> > > On Tue, Mar 17, 2015 at 2:02 PM Abhishek Kulkarni
>> > > <adkulkar at indiana.edu> wrote:
>> > >
>> > >> Hello,
>> > >>
>> > >> While trying to build mvapich 2.2rc2, I run into a build failure
>> > >> when I use the following two flags together: --with-device=ch3:psm
>> > >> --enable-thread-cs=per-object
>> > >>
>> > >> Is this a known issue? Since I expect to mostly run in a hybrid
>> > >> (OpenMP+MPI) configuration, I am assuming that the granularity of
>> > >> critical sections might be critical for good performance.
>> > >>
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_rma_acc_ops.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_rma_ops.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_rma_reqops.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_rma_sync.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_rndv.lo
>> > >> src/mpid/ch3/src/ch3u_handle_recv_req.c: In function
>> > >> ‘MPIDI_CH3_ReqHandler_FOPComplete’:
>> > >> src/mpid/ch3/src/ch3u_handle_recv_req.c:963:9: warning: passing
>> > >> argument 4 of ‘MPIDI_CH3_iStartMsgv’ from incompatible pointer
>> > >> type [enabled by default]
>> > >> ./src/mpid/ch3/include/mpidimpl.h:1580:5: note: expected ‘struct
>> > >> MPID_Request **’ but argument is of type ‘struct MPID_Request *’
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_eager.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_eagersync.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-ch3u_win_fns.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-mpid_abort.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-mpid_cancel_recv.lo
>> > >>   CC       src/mpid/ch3/src/lib_libmpi_la-mpid_cancel_send.lo
>> > >> src/mpid/ch3/src/ch3u_rndv.c: In function
>> > >> ‘MPIDI_CH3_PktHandler_RndvReqToSend’:
>> > >> src/mpid/ch3/src/ch3u_rndv.c:210:28: error: invalid operands to
>> > >> binary == (have ‘MPID_cc_t’ and ‘int’)
>> > >> make[2]: *** [src/mpid/ch3/src/lib_libmpi_la-ch3u_rndv.lo] Error 1
>> > >> make[2]: *** Waiting for unfinished jobs....
>> > >> src/mpid/ch3/src/ch3u_eager.c: In function
>> > >> ‘MPIDI_CH3_PktHandler_EagerSend’:
>> > >> src/mpid/ch3/src/ch3u_eager.c:880:28: error: invalid operands to
>> > >> binary == (have ‘MPID_cc_t’ and ‘int’)
>> > >> src/mpid/ch3/src/ch3u_eager.c: In function
>> > >> ‘MPIDI_CH3_PktHandler_ReadySend’:
>> > >> src/mpid/ch3/src/ch3u_eager.c:1022:28: error: invalid operands to
>> > >> binary == (have ‘MPID_cc_t’ and ‘int’)
>> > >> make[2]: *** [src/mpid/ch3/src/lib_libmpi_la-ch3u_eager.lo] Error 1
>> > >> src/mpid/ch3/src/ch3u_eagersync.c: In function
>> > >> ‘MPIDI_CH3_PktHandler_EagerSyncSend’:
>> > >> src/mpid/ch3/src/ch3u_eagersync.c:259:28: error: invalid operands
>> > >> to binary == (have ‘MPID_cc_t’ and ‘int’)
>> > >> make[2]: *** [src/mpid/ch3/src/lib_libmpi_la-ch3u_eagersync.lo]
>> > >> Error 1
>> > >>
>> > >> Thanks,
>> > >> Abhishek
>> > >> _______________________________________________
>> > >> mvapich-discuss mailing list
>> > >> mvapich-discuss at cse.ohio-state.edu
>> > >> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>> > >>
>> > >
>>
>>
>>
>> --
>> Jian Lin
>> http://linjian.org
>> _______________________________________________
>> discuss mailing list     discuss at mpich.org
>> To manage subscription options or unsubscribe:
>> https://lists.mpich.org/mailman/listinfo/discuss
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20150327/ac2fc1dc/attachment.html>
-------------- next part --------------
_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss


More information about the discuss mailing list