Mark Hoemmen

Implement the suggestion here: https://github.com/ORNL/cpp-proposals-pub/issues/288

Document where integer overflow - related preconditions get checked

2

A user recently asked what parts of mdspan impose preconditions relating to integer overflow. One can deduce this from the proposals, but it would make sense to restate the preconditions...

@johan-overbye [reports that](https://github.com/kokkos/mdspan/issues/166#issuecomment-1201197622): > `mdarray`s with all static extents seem to require an extra 4 bytes on top of the container. (Latest MSVC for x86-64 with `/std:c++latest`.) The reference implementation...

Tpetra::Map::isCompatible could use Directory to avoid all-reduce if Map is contiguous

4

@trilinos/tpetra Tpetra::Map::isCompatible currently requires an all-reduce for contiguous but nonuniform Maps (that is, Maps created by the [Map constructor that takes a global and a local number of indices](https://github.com/trilinos/Trilinos/blob/f685733187b4faaa40174e0d14705f49024ccd65/packages/tpetra/core/src/Tpetra_Map_decl.hpp#L391)): https://github.com/trilinos/Trilinos/blob/f685733187b4faaa40174e0d14705f49024ccd65/packages/tpetra/core/src/Tpetra_Map_def.hpp#L1232...

Galeri::Xpetra::BuildProblem has redundant template parameters

3

@trilinos/galeri @trilinos/xpetra Galeri::Xpetra::BuildProblem takes six template parameters: Scalar, LocalOrdinal, GlobalOrdinal, Map, Matrix, and MultiVector. The latter three are redundant or very nearly so. With Tpetra, they are actually redundant. With...

@youyu3 discovered that the `layout_stride::mapping(const extents_type&, span)` constructor doesn't compile. Here is an example: https://godbolt.org/z/Tf57xdrGq ```c++ using extents_type = stdex::dextents; extents_type ex{2, 2}; // this one works fine stdex::layout_stride::mapping map0{ex,...

Add layout_padded example

4

I added a `layout_padded` example. This is a strided layout that ensures a given overalignment of the stride-1 extent (either the leftmost, for column-major mode, or the rightmost, for row-major...

Investigate making the implementation simpler, and benchmark the effects

5

@crtrott @dalg24 @nliber The following discussion * https://github.com/NVIDIA/libcudacxx/pull/299#issuecomment-1252094052 * https://github.com/NVIDIA/libcudacxx/pull/299#issuecomment-1252566731 * https://github.com/NVIDIA/libcudacxx/pull/299#issuecomment-1252626640 suggests that the implementation is more complex than needed, because it does not fully take advantage of the...

Due to a recent build issue in a different project, I accidentally constructed an `extents` type where the first template parameter `IndexType` was an `extents` specialization, instead of an integral...

@youyu3 @crtrott ## New issue I revised this issue on 2022/09/06, based on offline discussion with @crtrott . We do not want to give users the impression that they are...