Repository navigation
Optimize efficiency of core algorithms (conditional entropy, etc.) #32
Description
Profile the codebase to identify inefficiencies, especially in the conditional entropy calculation (see Katz et al. 2021 for prior art). Optimize, rewrite, or replace inefficient algorithms and document performance improvements.
Activity
johnh2o2 commentedon Sep 28, 2026
Release triage for PR #69 (cuvarbase 1.0.1, source
6b6c48945160).This issue is implemented in PR #69 and will close automatically when that PR merges into master. Keeping it open until the merge.
The release includes device-sized CE launch grids, removal of unnecessary global histograms on the fast path, and BLS/LS/PDM host and allocation optimizations. CE documents single-precision ratios of 1.2x, 1.9x and 8x on one shared A40, with slower/break-even double-precision cases retained. These are qualified within-session measurements, not a claim of superiority over Katz et al. or all competitors. CE remains in maintenance mode; broader comparative benchmark coverage stays tracked in #19.
Evidence: CE occupancy and memory optimization, with measured ratios · CE grid sizing implementation · BLS/LS/PDM/CE performance changes and qualifications