python

Background

I write respective Python wrapper functions for abstracting interaction with shared object, which is compiled using make genlib. Check wrapper definition in module ff_p.

Before you use Python module defined in this directory, make sure you've run make genlib and generated shared object.

Now you can import ff_p and perform basic prime field arithmetics on accelerator, but you'll probably never want to do this, as offloading single prime field computation into accelerator doesn't help anyhow.

Note, main purpose of creating this Python module is depicting that it can be done easily & testing my SYCL prime field arithmetics implementation against another module written in Python, named galois

Note, prime field modulus is 2**64 - 2**32 + 1

from ff_p import ff_p

ff = ff_p() # create object which will be
            # used for further interaction
            # with device, it encapsulates
            # SYCL queue where jobs will be
            # submitted to

assert 3 == ff.add(1, 2)
assert 18446744069414584320 == ff.sub(1, 2)
assert 19456 == ff.multiply(19, 1 << 10)
assert (1 << 10) == ff.exponentiate(2, 10)
assert 18302628881372282881 == ff.inverse(128)
assert 8198552919739815254 == ff.divide(2, 9)

Tests

There're some randomised testcases I've written, for strongly asserting results produced by ff_p module are indeed correct, using another Finite Field implementation galois module.

For running those test cases, I suggest you start with enabling virtualenv

# assuming you're in this directory

python3 -m pip install --user virtualenv # if `venv` not installed
python3 -m venv .
source bin/activate

# you're inside `venv` now

Now you can install required dependencies inside venv

python3 -m pip install -r requirements.txt

Finally run randomized test cases

python3 ff_p_test.py

When done, consider deactivating virtual environment

deactivate