[Trilinos-Users] Tpetra::CrsMatrix caching related to C= A^T B and D = A*C
csiefer at sandia.gov
Mon Sep 24 10:48:48 EDT 2018
My apologies! My email has been inaccessible for the last week (the
wonders of modern technology, eh?). Replies below inline.
On 09/24/2018 07:46 AM, Denis Davydov wrote:
> Hi Chris,
> Sorry to bother, but can you please reply to comment on the questions
> (1) and (3) below?
> You said I can run C=A^T B with C being fillComplete() but as I
> mention there is an exception thrown.
> I am also still curious what’s the distribution of C=A^T B when both A
> and B and distributed row-wise.
> Perhaps based on our discussion the documentation
> of Tpetra::MatrixMatrix::Multiply() could be improved.
>> On 14 Sep 2018, at 11:48, Denis Davydov <davydden at gmail.com
>> <mailto:davydden at gmail.com>> wrote:
>> Hi Chris,
>>> On 12 Sep 2018, at 16:00, Siefert, Christopher <csiefer at sandia.gov
>>> <mailto:csiefer at sandia.gov>> wrote:
>>> 1) The way this is usually used is row-wise, matching A.
>> but I am doing A^T B where both A and B are distributed in MPI
>> row-wise only.
>> So what’s the distribution of C = A^T B in this case?
It should match the DomainMap of A, IIRC.
>>> 3) Running with fillComplete'd matrices for is going to be much,
>>> much faster than running w/o fillComplete. You do *not* need to do
>>> what you're doing. Could you try removing the resumeFill and
>>> redoing the timing?
>> looks like I have to do this, otherwise I get:
>> Exception on processing:
>> Throw number = 1
>> Throw test that evaluated to true: C.isFillActive() == false
>> the check is here:
>> So to me it looks like it’s not possible to call C = A^T B with C
>> being fillComplete().
Huh. That's weird. I need to look into that.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Trilinos-Users