Feat/signextend in asm

[amkCha]

Handles the case of instruction SIGNEXTEND in BIN module

Adds 1000 traces for INST=11 extracted from bin.all.bz2 in bin.accepts

Note : at first, I thought I needed the index of the word's first non null byte, which I don't. Hence the util function first_non_null_byte - I kept it as it could be useful, but if you'd rather keep the utils lightweight, let me know 👌

---

Note

Adds SIGNEXTEND handling to bin and introduces signextend and first_nonzero_byte utilities with extensive test vectors.

• ASM BIN module:
  • Handle SIGNEXTEND (INST=11) in testdata/asm/bench/bin.zkasm: add dispatch to signextend_call, return unchanged for oversized index, and include ../util/signextend.zkasm.
• Utilities:
  • Add testdata/asm/util/signextend.zkasm implementing signextend(size, word) using byte256 and fill_bytes_between.
  • Add testdata/asm/util/first_byte.zkasm implementing first_nonzero_byte helpers for u256 down to u16.
• Test vectors/bench data:
  • Append large set of INST=11 cases to testdata/asm/bench/bin.accepts.
  • Add accepts for signextend in testdata/asm/util/signextend.accepts and for first_nonzero_byte in testdata/asm/util/first_byte.accepts.
• Go tests:
  • Add Test_AsmUtil_FirstByte and Test_AsmUtil_Signextend in pkg/test/assembly_util_test.go.

^{Written by Cursor Bugbot for commit 55c808a. This will update automatically on new commits. Configure here.}

[DavePearce]

Yup, this looks good. I didn't find any issues. The only tiny thing I could find would be that we could increase the number of specific tests for util/signextend.zkasm. But, I don't think its necessary as we already have a lot coming from the tests for the bin module.

[amkCha]

Yup, this looks good. I didn't find any issues. The only tiny thing I could find would be that we could increase the number of specific tests for util/signextend.zkasm. But, I don't think its necessary as we already have a lot coming from the tests for the bin module.

Indeed, it was a little light, so I ended up added 250 randomly picked traces extracted from the same source - they should be slightly different than the traces from the bin module