Take the jpeg2000 release that stops holding a page in the colour transform - #96
Merged
Merged
Conversation
…nsform v0.11.1 runs the inverse ICT a band of 64 rows at a time. Its six pooled SIMD images were the size of the page -- six float64 planes, 48 bytes a pixel -- and InverseICT is pixel-wise, so the rows held at a time were free to choose all along. On a 4559 by 6942 colour scan, decoded alone, peak RSS falls 2416 -> 990 MB and the live heap 1208 -> 741 MB. 45% of the measured corpus is three-component, so this is not a corner. Nothing here changes: all 3215 documents of the measured corpus render byte-identically, rendered by exactly this pair.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
go-images/jpeg2000v0.11.1 runs the inverse ICT a band of 64 rows at atime. Its six pooled SIMD images were the size of the page — six
float64planes is 48 bytes a pixel — and
InverseICTis pixel-wise, so the rows held ata time were free to choose all along.
Decoding one 4559×6942 three-component scan and nothing else, so the figures are
the decoder's:
45% of the measured corpus is three-component — 54 130 of 119 164 JPEG 2000
streams — so this is not a corner. A colour page cost six times a grey one per
pixel before this.
Nothing in this repository changes. All 3215 documents of the measured corpus
render byte-identically (0 different, 66 drawn by neither, population floor
3000), rendered by exactly this pair, and the module in the cache was diffed
against the tree that was proved.
Speed is not claimed: the machine was carrying two system daemons at 132% and
86% of a core, and a timing taken under that describes the daemons.
🤖 Generated with Claude Code