Alex Fikl

As discussed, it seems to make sense to transition from `DeviceDataRecord` to `arraycontext.dataclass_array_container`. The operations needed are the same * `DeviceDataRecord.get` to `arraycontext.to_numpy`. * `DeviceDataRecord.to_device` to `arraycontext.from_numpy`. * `DeviceDataRecord.with_queue` to...

@inducer @gaohao95 Thanks for the feedback on that! I'll try removing `ImmutableHostDeviceArray` support and bug you once this gets into a better state.

> Thanks for the PR - sorry for taking so long to get to this. I'm going to try and figure out why CI is failing here. No worries! Some...

This should be fixed by the buildsystem changes in #185 (at least it is for me on Arch as well). For the `CLANG_ROOT` error in particular, Arch just seems to...

> @kaushikcfd Is this worth pursuing given that compile times are suffering because of expression traversal complexity in pymbolic? For what it's worth, the `loopy` branch is up to date...

> However, I wonder how the loopy CI failures could be fixed? By creating a derived version of every pymbolic expression type in loopy so that they can use loopy-specific...

@inducer From your comment, it seemed like #303 is needed before anything useful can be done here. And adding events unconditionally could also be done in `boxtree` as a stopgap,...

> Sound fair? Yeah, that sounds like a good plan to me!

@alejandrogallo Managed to fix and/or skip all the tests that were causing problems so far. Let me know if any better ways to deal with some of them come to...

> I'll take a look this weekend and try to merge it, thank you for this great job!! That would be awesome! Thanks! :rocket: