Poulpy Emitter

[martonmoro]

Step 3 (Implement Code Generation) from #3096.

Adds lib/Target/Poulpy/{PoulpyEmitter.h,.cpp,BUILD,PoulpyTemplates.h}
and registers the --emit-poulpy translation in heir-translate.

• ModuleOp printer: Rust prelude, backend detection from
  !poulpy.module argument types, type BE = ...; and derived type
  aliases (Ct/Tsk/Akm)
• func::FuncOp / func::ReturnOp printers: typed, mutability-aware
  argument list, Result return, zero/one-operand return handling
• convertType: !poulpy.module, !poulpy.scratch, !poulpy.tensor_key,
  !poulpy.automorphism_key_map, and rank-0
  memref<!poulpy.ciphertext> (&Ct/&mut Ct/Ct depending on position
  and mutation)
• Argument mutability analysis: decides &Ct vs &mut Ct from which
  values are written by a compute op in the function body
• Lazy memref.alloc materialization: a fresh ciphertext buffer's
  Rust allocation is synthesized at the first op that writes into
  it, deriving its layout at runtime from a sibling operand
• Rejects an _assign op whose dst is still unmaterialized, since
  that would silently read uninitialized memory
• Op printers: add, add_assign, sub, sub_assign, mul, mul_assign,
  rotate, rotate_assign

Adds tests/Emitter/Poulpy/{BUILD,emit_poulpy.mlir,
emit_poulpy_invalid_assign.mlir}. Output has also been verified with
cargo check against the real poulpy 0.7.0 crates, not just
pattern-matched.

[github-actions]

Welcome to HEIR!

Thank you for opening your first pull request! We appreciate you joining our community.

Please note that our contributing policy) requires unrecognized users to meet a maintainer in a non-textual format before their first PR will be reviewed.

If you have already met a maintainer, or you're sure the maintainers know who you are, you can safely ignore this message. Otherwise, a maintainer will likely close this PR shortly.

Don't take it personally! Instead, we welcome you to attend one of our open meetings on our community calendar, schedule a dedicated call with a maintainer, or reach out in the #heir channel on Discord.

We look forward to connecting with you!

[j2kun]

Looks pretty good! I think it would be worth sharing a sample of the generated code with @Pro7ech to get a validation check on the API usage and the prelude.

[martonmoro]

Looks pretty good! I think it would be worth sharing a sample of the generated code with @Pro7ech to get a validation check on the API usage and the prelude.

will do!

[j2kun]

Please rebase and squash your changes and we can optimistically get it merged pending feedback from Jean-Philippe

[Pro7ech]

I went over the example, the emitted code looks good, assuming that hard-coded randomness was intentional. A future optimization would be to use the <op>_tmp_bytes(...) and <struct>.bytes_of(...) API to assess the scratch space required instead of hard-coding it's size, and borrow the tmp values from it instead of doing heap allocations.

Also, just to avoid ambiguity, in Poulpy Galoiskeys are keyed by their signed Galois element.

[martonmoro]

Please rebase and squash your changes and we can optimistically get it merged pending feedback from Jean-Philippe

I've squashed the commits. Also i left the randomness as is for now and added a comment so the emitter prints code that can be compiled at this stage.