Skip to content

[WIP] Basic ALP Encoder (not ready for review)#4

Draft
alamb wants to merge 4 commits into
sdf-jkl:alpfrom
alamb:alamb/begin_alp_encoder
Draft

[WIP] Basic ALP Encoder (not ready for review)#4
alamb wants to merge 4 commits into
sdf-jkl:alpfrom
alamb:alamb/begin_alp_encoder

Conversation

@alamb
Copy link
Copy Markdown

@alamb alamb commented May 8, 2026

This is part of the ALP encoding work

This PR puts in place a basic ALP encoder (it is as small as I can make it).

Currrently the encoder encodes all values as excpetions, which is allowed in the spec, but has worse compression than plain encoding 😆

I basically plan at least two more PRs after this one:

  1. implement the FOR encoding
  2. Implement sampling to determine factor/scale

Todos:

  • Implement variable Vector sizes (currently uses hard coded 2>>11)
  • Implement sampling, etc for encoding parameters

@alamb alamb changed the title [WIP] Basic ALP Encoder [WIP] Basic ALP Encoder (not ready for review) May 8, 2026
@alamb alamb force-pushed the alamb/begin_alp_encoder branch from 44c3fc7 to 66e3034 Compare May 9, 2026 10:48
@alamb alamb force-pushed the alamb/begin_alp_encoder branch from 66e3034 to 606bdc7 Compare May 15, 2026 17:36
@alamb
Copy link
Copy Markdown
Author

alamb commented May 15, 2026

An update:

  1. I wired in the ALP decoder instantiation so I can make it for either f32 or f64 (can be its own PR)
  2. I an currently stuck in trait hell trying to make the code compile, but will get it soon

@alamb
Copy link
Copy Markdown
Author

alamb commented May 15, 2026

I hit a snag while trying to be all fancy with a streaming encodr -- which is you don't know up front how many vectors we'll have. I have some ideas about this that I will work on

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

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant