Julian Knodt

Let me see if adding a `cuda_sync` after managed makes it work for general sized vectors Edit: After doing `cuda_sync` I don't see any difference as it still segfaults on...

This will probably require a perf run, considering the size of consts increased, and if there's any alternative way to store consts more efficiently that would also be useful.

Since the size of the const is the same, I wonder if the perf changes are caused by an increased number of branches, causing it to switch from a jump...

@RalfJung I'll address your comments now! I just needed to get a working version first.

Mmmm I've not written anything for codegen at all, but if that makes more sense I can close this and move it to that, altho seeing a guide before it...

ah I thought reading from the array of sizes was handled around line 132, but I guess mapping discriminants to variant idxes is not guaranteed to be 1-1, is there...

I believe so? I don't remember the original issue.

Is it fine to remove this test for now? I assume it's checking for some optimization on enums, but since the coverage is now over all enum copies essentially, it...

Is this spurious? It compiles and runs on my machine™

It does panic on the variant with the explicit discriminant value, I might just disallow that case