[ofw] RE: OpenSM 3.3.3 available for interested parties.

[Leonid Keller]

> -----Original Message-----
> From: Smith, Stan [mailto:stan.smith at intel.com] 
> Sent: Thursday, November 05, 2009 6:49 PM
> To: Leonid Keller
> Cc: ofw at lists.openfabrics.org
> Subject: RE: OpenSM 3.3.3 available for interested parties.
> 
> Hello,
> 
> >> I think, to put new Opensm under trunk\tools\opensm is a good idea,
> > But I wouldn't delete the current branch till we distribute the old 
> > executables.
> 
> How about we add the new opensm as 
> trunk\ulp\opensm\user_3-3-3\ instead of adding it in tools?
> Keeping related items in one place makes things less complicated.

Agree.
But Opensm is application, not a protocol.
It's true place is under \tools.
And as far as we are any way going to remove the old opensm, it can stay
for the time being in old place.
[BTW, it won't by something unprecedent. We have got two versions of
perftests, living in two folders of our tree.  :( ]

> 
> > I'd suggest to change the target destination so as old Opensm 
> > executables would be placed somewhere under ulp\opensm\bin and with 
> > the same names as Sean suggest.
> > Then on the installation phase they can be copied under new 
> > opensm_3_0_0 folder.
> 
> OK, I can live with a trunk\opensm\user\bin\ target along 
> with changes to makebin.bat & WIX code.
> 
> > I don't think we may ask customers to check out old Opensm and to 
> > build on their own.
> > I don't think we can test that complicate program as opensm 
> enough as 
> > to be sure that it will work not worse than the previous one in all 
> > situations.
> 
> I strongly agree!
> 
> > I'd like first to see the new Opensm, working in real big clusters.
> 
> Would a 53 node HPC cluster be big enough? Certainly a good 
> step forward.
> 
Of course, it's a step forward.
But I meant clusters of hundreds or even thousands nodes.
One of them is already present in Top500 list.
> 
>