[Primitives] Fix slice primitives

[leonardt]

• Promote int-like values in get_slice/set_slice
• Update doc to refer to set_index instead of set_bit
• Support set_slice of length 1 for constant values

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 92.85714% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 85.03%. Comparing base (42b8850) to head (4edeeae).
⚠️ Report is 265 commits behind head on master.

Files with missing lines	Patch %	Lines
magma/primitives/slice.py	92.85%	1 Missing ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1139      +/-   ##
==========================================
+ Coverage   85.00%   85.03%   +0.02%     
==========================================
  Files         154      154              
  Lines       16368    16375       +7     
==========================================
+ Hits        13914    13924      +10     
+ Misses       2454     2451       -3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[rsetaluri]

I would be in favor of renaming the function (publicly, i.e. in the docs) to set_bit. We could deprecate set_index for now with warning (doubt anyone is using it anyway...)

[leonardt]

It was changed to set_index because the pattern is more general than just set_bit (i.e. you can use it on generic arrays, rather than just on Bits), so I think it makes more sense to keep it as set_index for generality

[rsetaluri]

I see. In that case, why do we need a separate function? I.e. why is it not covered by set slice with width 1?

[leonardt]

Simpler to use when you know you're only interested in one index. For the same reasons in Python we have x[1] versus having to do x[1:2].

[rsetaluri]

Makes sense -- but I would argue x[1] and x[1:2] is the same interface with different arguments (in fact it literally is the function __getitem__).

Maybe we don't need to do this in this change, but seems reasonable to refactor to have the following interface:

"""
* target: Array[N, T]
* value: Union[Array[M, T], T]
* start: UInt[W]

where M < N and W <= clog2(N)
"""
def set(target: Array, value: Union[Type, Array], start: UInt):
    ...

Now that I think about it there are few new concerns here:

• Should this function work with all Arrays and not just Bits? I kind of assumed it did, but looking at the code closer I'm not certain.
• Do we need a width argument? Can it just be inferred from value? If value is an array of the same dimensionality as target then it is a slice set; else if it is of dimensionality one less than value then it is an index set (of course assuming the contained types match up and all).
• In the case start being a constant value, should we optimize for the simple wiring (non-dynamic mux) case?

[rsetaluri]

Moving this conversation here:

I'm not sure if set_slice of length makes sense unless the start value is a constant 0. Otherwise, the smallest start value we could have is a Bit which could take on value 1 or 0. A start index of 1 would be invalid for setting a slice on length 1.

I think my note was unclear. I meant the case when width is 1, i.e. value is of length 1 -- not target being length 1. This seems like it should be uniform with any other width value. Right now it seems to be broken.