I am moving code from a single-threaded to multi-threaded process. One section runs much more slowly than expected, which I suspect has to do with deepcopy.
The datastructures in question are quite complex, but immutable in the parallelized code. Hence I have tried to make use of "ptr". Cleanup of resulting memory issues caused the slowdown. I suspect some substructure is getting copied unnecessarily... however there is quite a bit of code to read, so hard to locate "manually".
Any suggestions for spitting out diagnostics on deepcopy or even causing program to crash so I can zero in on the problem?