Aditya
Aditya
Points considered for the meeting on multivector/dense: - Conceptually it makes sense to separate multivector and dense, eg.: (i) column-wise dot products don't make sense for Dense matrices in general,...
> I've realized that there are issues if the jacobi preconditioner is used for matrices with negative diagonal elements. I need to further investigate this, so I will put the...
This is somewhat relevant here: the operation S1-1 A S2-1 S2 x = S1-1 b is needed for "batched scaled" batch solvers. Usually, the left matrix would be applied to...
@tcojean How does StarPU manage different tasks? Does it spawn threads? I feel like the internal complication arising from either options 2 or especially 3 needs good justification. In general,...
@tcojean I see. Is there a concept of ownership of data among threads that StarPU itself manages? Is that the case in general for tasking systems? If that is the...
In the second case you mentioned, perhaps cached data for the matrix, preconditioner etc. needs to be considered while setting task dependencies.
Yes, that's why I propose to keep the original pointer in a struct/class and delete using it once we are done. For now, I'm thinking of a static scope-based object...
Wait, actually, `aligned_alloc` [is there](https://en.cppreference.com/w/c/memory/aligned_alloc) in the C11 standard, and C++ 17 just "inherits" from that. Maybe we can just use the C11 version while still compiling with C++14. I...
The SWIG-Fortran talk from ORNL was quite interesting. - We of course need explicit instantiations for everything, and as far as I could see, we give each explicit instantiation a...