chore: Add u1 ops to interpreter, convert Value panics to errors

[jfecher]

Description

Problem*

Summary*

Breaks out ssa interpreter changes from this PR: #8436

Additional Context

We apparently allow u1 + u1 in addition to other odd operations like u1 << u1

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[aakoshh]

👍

[aakoshh]

There are still some problems with this. I'm testing after merging into my branch, and getting a failure like this:

Internal(ConstantDoesNotFitInType { constant: 18446744073709551588, typ: Unsigned { bit_size: 8 } })

It's trying to cast i64 of -28 to u8. The problem is that interpret_instruction handling Cast first converts the NumericValue to Field, where we now use the signedness, but then treats that Field as if it was the target type.

[aakoshh]

Added an test to show the above error in 09e8403

[jfecher]

@aakoshh should be fixed now

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Compilation Time'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: b4dede8	Previous: 9c53936	Ratio
rollup-root	2.05 s	1.666 s	1.23

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[aakoshh]

When we cast signed integers most of the time we sign-extend but only when casting to Fields we zero-extend instead.

What exactly do we mean by "sign-extend" and "zero-extend"?

[jfecher]

What exactly do we mean by "sign-extend" and "zero-extend"?

It refers to the values in binary when you have -1 as an i8 it looks like 11111111 and you need to cast to e.g. a u16 XXXXXXXX 11111111. There's extra space so you need to decide what to fill the X's with.

zero-extend is to fill it with zeroes: 00000000 11111111, this is the standard for unsigned values in most languages.
sign-extend is to fill it with the same bit as the sign: 11111111 11111111. If the value was positive it'd be filled with zeroes just like zero-extending it. This is the standard for signed values in most languages.

zero-extending a value like -1 as i8 into a i32 would produce 255 : i32 while sign-extending would produce -1 : i32.
For Noir when we cast to Fields we essentially zero-extend since we produce 255 : Field instead of sign-extending and producing a very large field value.

Edit: This is also why given x: i8, x as u32 != x as u8 as u32 because the former sign-extends while the later casts to a u8 so it zero-extends.

[aakoshh]

This shall be our spec 📖 👨‍⚖️