> Is it really impossible to use a standard Python dict instead of cfuhash?

I don't know if it is possible or not, I didn't try. Using PyDict has many requirements:

* the GIL must be held: I plan to hook also PyMem_Raw (which is called without the GIL being held)
* PyDict uses PyMem_Malloc() which would be a reentrant call
* the API uses PyObject* whereas tracemalloc uses C types and C structures
* PyDict can raises Python exceptions, which is not something expected from a memory allocator (especially PyMem_Malloc)
* etc.

The implementation is not done yet. I would like to be able to compute the size of memory allocated by the _tracemalloc module: size of hash tables.