avx512_bignum is a fast, infinite precision library built around the avx512
extensions to the x86 instruction set. 

Prerequisites
-------------
a) Hardware that supports avx512 running a recent linux distribution.
b) g++ version >= 13.2.1
c) python version >= 3.12.3

b) and c) are what was used to develop this library. Older versions may work.

Building the library.
---------------------
Clone the software to your machine. 
Run make debug|release|fast
debug runs the compiler with the -ggdb flag
release runs the compiler with the -Os flag
fast runs the compiler with the -Ofast flag.
The above will build the library b32.a and all the test binaries.

API
---
The library implements 2^32 base arithmetic in the form of a b32 class.
The header is incl/b32.h. The four arithmetic operations are
a) add_to
b) subtract_from
c) multiply_with
d) divide_by
For a better understanding of the API, see the cpp files in the tests directory

Tests
-----
Regression tests.
-----------------
operator_drv will run an operation for a particular pair of operands.
run_all.py will run regression tests on any or all of the operations for a set of randomly generated operands.

Performance tests.
------------------
file_ops will run and time an operation on a pair of numbers contained in files.
perf_test.py will generate operands of a given size and run and time an operation on those numbers.
All commands when run without arguments will print usage instructions.

Current Status
----------------
References to the __m512i datatype need to be removed for reasons of performance and clarity.
All operators will eventually be implemented in assembly.
The staging directory reflects the current status of this effort.
Changing away from the __m512i datatype will mean rewriting the whole library.
Furthur changes will mean a new repository.