Skip to content

mercury: remaining fast-path eligibility gaps #434

Description

@boxerab

bail census over grok-test-data (421 files) after #423 closed. each still sends the stream to classic, so #426 cannot retire classic until they move.

  • HT and BYPASS code-block modes (Cmodes 0x40, 0x48, 0xc0, 0x01, 0x34): 22 files. mercury's packet parser reads one segment length per contribution and needs the part-15 / multi-segment length signalling, then the T1 shim needs the HT coder.
  • COD/COC in tile-part headers (0xff52, 0xff53): 18 files. needs a per-tile coding style in the plan; today the header is rejected in comb_tile_part.
  • Psot=0 (last tile-part runs to EOC): 6 files. the SOT walk needs the file length as the implied end.
  • derived quantization (Sqcd style 1): 2 files. one LL step size scaled per band, not wired in the band delta table.
  • PPM (packed packet headers in the main header): grok bails. mercury reads packet headers from the packet stream only.
  • RGN (ROI upshift, main or tile-part header): grok bails, mercury rejects the tile-part marker. needs the max-shift descaling in the block decode.
  • PPT and tile-part POC/RGN: rejected by mercury, classic keeps them.

also classic-only by decision in #423: component filter, asynchronous decode, transcode packet-length recording, part 2 transforms.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions