Phil Miller - NOAA comments

Results 319 comments of


                                            Phil Miller - NOAA

Nightly build failures with intel/19, "missing return statement at end of non-void function"

Issue was develop-only as noted, so closing.

Warnings in cuda builds: Kokkos_ViewCtor.hpp(304): warning: missing return statement at end of non-void function

Issue was develop-only as noted, so closing.

Synchronization of sort(View) with exec space argument

4.0 released with this.

openmp.team_broadcast_long unit test segfaults with gcc/10.2.0

Transcribing my debugging process suggestions from a Slack conversation while I was away with only my phone: > What I'd suggest is to look for variables in the tsan logs...

Optimized behavior of resize/realloc of views with aliasing mirrors is surprising

Some possible responses: * Leave things as they are, and document the current behavior * Deprecate/remove `Kokkos::realloc` and `Kokkos::resize`, and just have applications explicitly write the exact behavior they want...

Optimized behavior of resize/realloc of views with aliasing mirrors is surprising

Seeing that these functions are called in a number of places throughout Trilinos, we may be able to get some testing of whether usage with aliased input is expected, and...

Optimized behavior of resize/realloc of views with aliasing mirrors is surprising

This goes back to #904 and #916 by @mhoemmen and amended in #4207

Optimized behavior of resize/realloc of views with aliasing mirrors is surprising

@crtrott is drafting documentation that will serve as a base for discussion.

Kokkos + Legion interop

> Just an update, I have a non-LANL Legion user who also would like to be able to use Kokkos for implementing GPU tasks, but definitely needs to run on...

Kokkos + Legion interop

For the broader issue of Kokkos-Legion interoperability, there's initialization testing introduced in #5206, and improvements for HIP in #5603. At this point, I think we can close this issue, and...