Dear community,
What are the current bit sizes possible with python numpy? Can’t find any reference in the docs…
Hello @benjmanable
The team was off, I think concrete-numpy supports up to 16 bits of precision but with those sizes it can be pretty slow, up to 8 bits should be faster I’ll let @umutsahin confirm
Cheers
It’s not in the doc because it depends on the circuit. For fully leveled circuits (i.e., no table lookups), concrete-numpy can go up to 57-bits for example.
These limitations are related to Concrete Compiler, so they might be documented in detail in the documentation of Concrete Compiler when it becomes public.
Let me know if this helps 