[mvapich-discuss] Forcing PSM on a machine with both PSM and PSM2 libs

[Hari Subramoni]

Hi Adam,

Right - my bad. Good catch. Will add this to the code. It should be
available with the next release.

Thx,
Hari.

On Tue, Aug 2, 2016 at 6:54 PM, Adam T. Moody <moody20 at llnl.gov> wrote:

> Hi Hari,
> The systems are down at the moment, so I can't test your suggestion.
> However, I suspect adding a path here won't help. You can see the configure
> output tested and found PSM, but it opted to use PSM2.
>
> checking psm.h usability... yes
> checking psm.h presence... yes
> checking for psm.h... yes
> checking psm2.h usability... yes
> checking psm2.h presence... yes
> checking for psm2.h... yes
> checking for psm2_init in -lpsm2... yes
>
>
> The subconfigure.m4 for the PSM channel is missing the case where someone
> specifies --with-psm explicitly.  For that case, it falls back to look for
> both PSM and PSM2 and then uses PSM2 if it finds it.  The patch here should
> handle the --with-psm case to force the build to use PSM on a system that
> has both installed.
> -Adam
>
>
> On 08/02/2016 03:27 PM, Hari Subramoni wrote:
>
>> Hi Adam,
>>
>> The "--with-psm" configure option takes am argument which is the path to
>> the psm library. Can you please mention the path and see if it works?
>>
>> Thx,
>> Hari.
>>
>> On Aug 2, 2016 5:50 PM, "Adam T. Moody" <moody20 at llnl.gov> wrote:
>>
>> Hello MVAPICH team,
>>>
>>> I have a machine that has both PSM and PSM2 installed.  I would like to
>>> create a build that links to PSM, but configure always picks PSM2 if it's
>>> available, even if I specify "--with-device=ch3:psm --with-psm".  I think
>>> the following patch would help, although I've not been able to test it
>>> yet.
>>>
>>>
>>> --- a/src/mpid/ch3/channels/psm/subconfigure.m4    2016-03-29
>>> 17:52:51.000000000 -0700
>>> +++ b/src/mpid/ch3/channels/psm/subconfigure.m4    2016-08-02
>>> 14:43:03.743895000 -0700
>>> @@ -60,6 +60,13 @@
>>>       AC_CHECK_LIB(psm2, psm2_init, , [
>>>       AC_MSG_ERROR(['psm2 library not found.  Did you specify
>>> --with-psm2=
>>> or --with-psm2-lib=?'])
>>>       ])
>>> +elif test "x$with_psm" != "x" ; then
>>> +    AC_CHECK_HEADER([psm.h], , [
>>> +     AC_MSG_ERROR(['psm.h not found.  Did you specify --with-psm= or
>>> --with-psm-include=?'])
>>> +    ])
>>> +    AC_CHECK_LIB(psm_infinipath, psm_init, , [
>>> +     AC_MSG_ERROR(['psm_infinipath library not found.  Did you specify
>>> --with-psm= or --with-psm-lib=?'])
>>> +    ])
>>>   else
>>>       AC_CHECK_HEADER([psm.h], [have_psm=yes], [have_psm=no])
>>>       AC_CHECK_HEADER([psm2.h], [have_psm2=yes], [have_psm2=no])
>>>
>>>
>>> Can you verify this?
>>>
>>> Thanks,
>>>
>>> -Adam
>>>
>>> _______________________________________________
>>> mvapich-discuss mailing list
>>> mvapich-discuss at cse.ohio-state.edu
>>> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>>>
>>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20160802/1a1d8008/attachment.html>