Creating nightly hwloc snapshot git tarball was a success.
Snapshot: hwloc dev-574-gc6fa7b7
Start time: Wed May 20 21:01:02 EDT 2015
End time: Wed May 20 21:02:48 EDT 2015
Your friendly daemon,
Cyrador
George,
first i'd like to amend my initial message.
i previously wrote the same algo is used to parse rules per communicator
size and per message size.
this is true, but i missed the part where it is mandatory to define a
rule for zero size message.
consequently, a given message is either in
A few observations.
1. The smcuda btl is only built when --with-cuda is part of the configure line
so folks who do not do this will not even have this btl and will never run into
this issue.
2. The priority of the smcuda btl has been higher since Open MPI 1.7.5 (March
2014). The idea is that
Rolf - this doesn’t sound right to me. I assume that smcuda is only supposed to
build if cuda support was found/requested, but if there are no cuda adapters,
then I would have thought it should disqualify itself.
Can we do something about this for 1.8.6?
> On May 20, 2015, at 11:14 AM,
I was making basic performance measurements on our machine after installing
1.8.5, the performance were looking bad. It turns out that the smcuda btl has a
higher exclusivity than both vader and sm, even on machines with no nvidia
adapters. Is there a strong reason why the default exclusivity
Hi Nathan,
Not entirely sure it is related but I'm also seeing a hang at the end of
Put_all_local with -mca pml ob1 -mca btl tcp,sm,self.
It seems to have finished the test but doesn't proceed to the next one. When
run alone, Put_all_local finishes fine.
Also, I verified with master and I see
Each rule define an interval with the previous rule, and everything in an
interval will be bound the the rule with the next message size. You cannot
define a rule for a specific amount. Thus, the fact that the rules must be
ordered by message size was done by design.
Returning a NULL rule as
I could be wrong (George, please feel free to correct me), but I *think*
this was the designed behavior. If you read Jelena's paper,
http://www.open-mpi.org/papers/euro-pvmmpi-2006-collective-alg-selection/euro-pvmmpi-2006-collective-alg-selection.pdf
you basically construct a new decision map
Guys, you are way off-base here. This is why Jeff asked that we table this
conversation until the devel meeting. As he and I discussed at length on the
phone, your starting premise is incorrect.
This entire thread stems from Jeff’s recent attempt to do a bisect search on
the master. He hit
Howard,
i made PR 593 https://github.com/open-mpi/ompi/pull/593 in order to fix
this.
George,
could you please review this ?
Cheers,
Gilles
On 5/20/2015 12:57 PM, Howard Pritchard wrote:
HI Gilles,
First a disclaimer - I do not know what the intended design was nor
where the design
On Tue, May 19, 2015 at 9:37 PM, Howard Pritchard
wrote:
> Pretty soon the developer will get trained to use the PR process, unless
> they are that engineer I've yet to meet who always writes flawless code.
I've never met that developer, either.
However, I have met one
On 20/05/15 14:37, Howard Pritchard wrote:
> It would also be easy to trap the I-want-to-bypass-PR-because-I
> know-what-I'm-doing-developer with a second level of protection. Just
> set up a jenkins project that does a smoke test after ever commit to
> master. If the smoke test fails, send a
Hi Dave,
> The other way to solve this issue would be to stop treating the master as
> a general dumping ground for potentially unstable code where anyone can
> just push any time they want. If we switched to using PRs for
> (essentially) all code that goes into master as well, then we
HI Gilles,
First a disclaimer - I do not know what the intended design was nor where
the design document
for this feature is located.
However, I would certainly prefer that if the communicator size wasn't
specifically specified
in the rule file, a fall back do-no-harm algorithm would be
14 matches
Mail list logo