Andzelmas

I haven't looked at your code too closely, but when multithreading be careful of data races - if different threads access the same memory at the same time undefined behavior occurs.
In the e0 and e1 arrays if each thread goes through the members and does something with them (even without changing the values in the arrays) some bad stuff might happen.
If you need to share values between threads ring buffers are the way to go.
If you already took care of all that, sorry - like i said didn't have time to take a closer look at the code.
Interesting project anyway: ) good luck!

INFO FOR OTHERS WITH SIMILAR PROBLEM
Os: arch linux 6.7.6-arch1-1
Hardware: lenovo laptop with amd 7945hx and nvidia 4070
Checked journalctl dump, saw lines similar to: #1 0x00007101df2002f3 je_tcache_get (libjemalloc.so.1 +
On my system there is only libjemalloc.so.2 and libjemalloc.so.1 is only in the houdini directory
So added LD_PRELOAD="/usr/lib/libjemalloc.so.2" before launching houdini and all is well.
Launches with opengl on integrated gpu or nvidia.

Turns out all i had to do was set a HCUSTOM_CFLAGS variable to -DOPENEXR_DLL -DHALF_EXPORT
Now the dll compiles.