These are chat archives for elemental/chat

21st
Aug 2016
Ryan H. Lewis
@rhl-
Aug 21 2016 00:03
@poulson: looking over it, i'm not sure why I needed the multMeta thing, you and I must have wanted to do something with DistGraph/DistSparseMatrix that I didn't complete yet.
Ryan H. Lewis
@rhl-
Aug 21 2016 00:54
holy crap, the OSX filesystem is not case sensitive
Ryan H. Lewis
@rhl-
Aug 21 2016 03:08
@poulson: ping?
Here is my proposed draft rewrite of Matrix.hpp: http://termbin.com/8sac
Ryan H. Lewis
@rhl-
Aug 21 2016 03:27
@poulson reading about the operator= for the various dist matrices
Ryan H. Lewis
@rhl-
Aug 21 2016 03:57
so, I propose that we begin to modernize the library for C++17. I believe it should be possible to signficantly reduce the complexity of the DistMatrix implementations.
Jack Poulson
@poulson
Aug 21 2016 14:23
@rhl- I can take a look at the PR now.
Jack Poulson
@poulson
Aug 21 2016 14:44
I'm not so sure that I like the idea of forward slash being the linear solve operator
for one thing, it's backwards from MATLAB/Octave
and there is no way for us to introduce a backslash
it might be easier to do a continuous deformation of the Matrix class through PR's
that way we can flesh out all of the consequences of each change
Ryan H. Lewis
@rhl-
Aug 21 2016 17:05
Yeah forward slash can go, backslash and ' not being usable as operators really blows.
Ryan H. Lewis
@rhl-
Aug 21 2016 17:28
@poulson: can we push the library to use c++14 ?
Then we can make use of: "[deprecated]
[deprecated("reason")]"
Jack Poulson
@poulson
Aug 21 2016 17:54
I'm okay with the minor bump from C++11 to C++14
AFAIK, all major compilers that support C++11 support C++14
Ryan H. Lewis
@rhl-
Aug 21 2016 17:59
[ 82%] Building CXX object CMakeFiles/tests-blas_like-Multiply.dir/tests/blas_like/Multiply.cpp.o
/Users/rhl/code/Elemental/tests/blas_like/Multiply.cpp:30:38: error: cannot throw object of incomplete type 'void'
if( nrm > limits::Epsilon<T>()){ throw RuntimeError("Sparse(I)x != Graph(I)x"); }
Jack Poulson
@poulson
Aug 21 2016 18:07
there is no need for the throw
RuntimeError is a routine that throws an exception
Ryan H. Lewis
@rhl-
Aug 21 2016 18:08
oh
I see
Jack Poulson
@poulson
Aug 21 2016 21:26
I'm committing a secular equation solver for the symmetric eigenvalue problem diag(d) + rho z z^T, where || z ||_2 = 1 in a few seconds
sequential and distributed D&C symmetric tridiagonal EVPs are imminent
Jack Poulson
@poulson
Aug 21 2016 22:07
elemental/Elemental@804d515
Ryan H. Lewis
@rhl-
Aug 21 2016 22:33
did you change GetLocal --> Get in DistSparseMatrix ?
I believe I have reverted my changes
the only odd things was that GetLocal was implemented, no signature, and Get was declared, no definition
so I just assume somehow I got out of sync with master
Jack Poulson
@poulson
Aug 21 2016 23:40
no, I didn't