Experimental - Initial commit for Kerngraph

[christopherngutierrez]

Proposed changes

Experimental
Kerngraph.py is a command line tool that takes output from kerngen.py through stdin and transform the Pisa for optimization.

This PR is the starting point.

• Parsing is currently supported for kerngen.py
• Currently supports the following high ops: Add, Sub, Mul, Muli, and Copy. Future PR will extend to composite high ops such as Relin and Mod.
• Transformation: loop interchange for RNS and Part are currently supported. Future PR will add loop interchange for Unit support.
• No graph representation is present at this time. Will add in a future PR.

Types of changes

What types of changes does your code introduce to the Encrypted Computing SDK project?
Put an x in the boxes that apply

• Bugfix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update (if none of the other choices apply)

Checklist

Put an x in the boxes that apply. You can also fill these out after creating
the PR. If you are unsure about any of them, do not hesitate to ask. We are
here to help! This is simply a reminder of what we are going to look for before
merging your code.

• I have read the CONTRIBUTING agreement
• Current formatting and unit tests / base functionality passes locally with my changes
• I have added tests that prove my fix is effective or that my feature works (if appropriate)
• I have added necessary documentation (if appropriate)
• Any dependent changes have been merged and published in downstream modules

Further comments

As an illustrate example, say that the contents of the file bgv.add.high contains:

CONTEXT BGV 16384 3 2

DATA c 2
DATA a 2
DATA b 2

ADD c a b

which produces the following output from kerngen:

### Context: scheme='BGV' poly_order=16384 key_rns=3 current_rns=2 num_digits=None max_rns=2 ###
### Kernel (0): Add(context=KernelContext(scheme='BGV', poly_order=16384, key_rns=3, current_rns=2, num_digits=None, max_rns=2, label='0'), output=Polys(name=c, parts=2, rns=2), input0=Polys(name=a, parts=2, rns=2), input1=Polys(name=b, parts=2, rns=2)) ###
0, add, c_0_0_0, a_0_0_0, b_0_0_0, 0
0, add, c_0_0_1, a_0_0_1, b_0_0_1, 0
0, add, c_1_0_0, a_1_0_0, b_1_0_0, 0
0, add, c_1_0_1, a_1_0_1, b_1_0_1, 0
0, add, c_0_1_0, a_0_1_0, b_0_1_0, 1
0, add, c_0_1_1, a_0_1_1, b_0_1_1, 1
0, add, c_1_1_0, a_1_1_0, b_1_1_0, 1
0, add, c_1_1_1, a_1_1_1, b_1_1_1, 1

With kerngraph, we can interchange the loop order to produce the following output:

# cat bgv.add.high | ./kerngen.py  | ./kerngraph.py -t add --primary part

0, add, c_0_0_0, a_0_0_0, b_0_0_0, 0
0, add, c_0_0_1, a_0_0_1, b_0_0_1, 0
0, add, c_0_1_0, a_0_1_0, b_0_1_0, 1
0, add, c_0_1_1, a_0_1_1, b_0_1_1, 1
0, add, c_1_0_0, a_1_0_0, b_1_0_0, 0
0, add, c_1_0_1, a_1_0_1, b_1_0_1, 0
0, add, c_1_1_0, a_1_1_0, b_1_1_0, 1
0, add, c_1_1_1, a_1_1_1, b_1_1_1, 1

Note the ordering of the generate pisa commands produced with kerngraph.

[faberga]

LGTM as an experimental feature that we want in the codebase so we can start discussing optimizations and showing people the potential of the various intermediate manipulations/optimizations that can be done.

[faberga]

Hi, @christopherngutierrez
That looks a good start. after this gets in as an experiment feature, let's have a chat on the semantics of the command line.

[kylanerace]

Looks good from an initial scan. Will attempt with wider variety next week, but doesn't seem breaking.

[kylanerace]

Only change I see would be to match the current file hierarchy on main (e.g. kerngen inside of p-isa_tools, same for functional_modeler, etc.). Just needs to be rebased with main

[faberga]

Hi @christopherngutierrez and @kylanerace
I agree with Kylan, it does need rebasing to fix the conflicts with the directory hierarchy.

[kylanerace]

LGTM