Open MPI logo

Open MPI Development Mailing List Archives

  |   Home   |   Support   |   FAQ   |   all Development mailing list

Subject: Re: [OMPI devel] 0.9.1rc2 is available
From: Brice Goglin (Brice.Goglin_at_[hidden])
Date: 2009-10-22 05:05:11

Ashley Pittman wrote:
>> [csamuel_at_tango069 ~]$ ~/local/hwloc/0.9.1rc2/bin/lstopo
>> System(31GB)
>> Node#0(15GB) + Socket#0 + L3(6144KB) + L2(512KB) + L1(64KB) + Core#0 + P#0
>> Node#1(16GB) + Socket#1 + L3(6144KB)
>> L2(512KB) + L1(64KB) + Core#0 + P#4
>> L2(512KB) + L1(64KB) + Core#1 + P#5
>> L2(512KB) + L1(64KB) + Core#2 + P#6
>> L2(512KB) + L1(64KB) + Core#3 + P#7
> Does this imply the default is to report on processes in the current
> cpuset rather than the entire system? Does anyone else feel that
> violates the principal of least surprise?

Yes, by default, it's the current cpuset. Maybe lstopo should report the
whole system (it does if you pass --whole-system), or display a clear
message saying that's it's only showing the current cpuset. Apart from
lstopo, for real applications, I feel like using the current cpuset only
is better.

By the way, we'll probably add another behavior in 1.0: only report the
topology inside the current process binding mask. But for sure this
behavior won't be enabled by default.