Arsène Pérard-Gayot

To expand a little bit on that, you don't need to closure-convert `loop`, it is enough to just lift it, because it is not passed to another function or stored...

Thank you for this detailed bug report. I will have a look at it as soon as I can.

I cannot reproduce your issue with the following code: ```cpp #include #include #include #include #include #include #include #include #include #include using Scalar = float; using Vec3 = bvh::v2::Vec; using BBox...

I am running Fedora 38 on a Ryzen 2700U laptop. Does the same code work if you omit the thread pool?

> Turning the number of threads down to 1 does not seem to help. Could you try using the version of the function that does not take a thread pool...

Perhaps this isn't linked to threading at all, and is just a consequence of the fact that the order in which the reinsertion candidates are selected is different. Could you...

Ah, right, sorry. It's still possible to add it to the default builder itself, in `default_builder.h`, line 43, right before the call to the reinsertion optimizer.

It appears your BVH has an incorrect structure. I'm not sure why, but node 2207 should be a left child (it has an odd index), but it appears it is...

Ah, of course, this would have created invalid BVHs! That does indeed explain why it was seemingly random and not happening on my machine. Nice work figuring this one out....

It's unclear how to pass existing allocations/BVHs/buffers to the BVH builders. For instance, the mini tree builder needs to create mini-trees that require their own allocations of impredictable size, and...