David Poliakoff

The more I think about it, the more uncomfortable I am with supporting finalization without initialization. It feels like a program that does that is just _doing something wrong_. What's...

So, big change. I'm moving this to a new model, along the lines of what @Rombur was saying. This _does_ introduce breaking changes, but not into any user-facing code I...

@jrmadsen @masterleinad : I actually don't think you guys are too far apart. @jrmadsen, I'd recommend you just motivate why all this is in one place a bit for Daniel...

I'd just say move this to our internal documents edit: though I'll of course fix things where appropriate

Some V100 results: ``` [dzpolia@kokkos-dev-2 cuda-build]$ KOKKOS_PROFILE_LIBRARY=$HOME/src/apollo/build/src/libapollo.so ./perf_test/sparse/sparse_spmv --test kk -l 15000 -s 1000000 Initializing Apollo Tuning adapter == APOLLO: Looked for APOLLO_SINGLE_MODEL with getenv(), found nothing, using '0' (default)...

This is _profoundly, incredibly, extremely_ broken. The old version was right. I'll hack on this more to figure out why. Note that it wasn't broken if no tool was loaded,...

One thing to consider with these tuning problems is the use of Kokkos Autotuning. If the TeamPolicy constructor uses Kokkos::AUTO for team size (or vector length), autotuning will allow you...

I think 3.3 is the first release in which this is supported (edit: where "this" is team size tuning. Future releases add other kinds of tuning). Basically what happens is...

Right, it must run the parallel-for multiple times. Our most recent training material is [here](https://github.com/kokkos/kokkos-tutorials/wiki/Kokkos-Lecture-Series#kokkos-tools-lecture-series) under Autotuning, and we have some older stuff above, [here](https://www.youtube.com/watch?v=MH6zFYGw0HU&feature=youtu.be&t=3705)

Hey Tom, Thanks for the feedback, we're trying to schedule a group session to better document RAJA ASAP, I'll be sure to keep this in mind as one of our...