Repository navigation

Rework README around benchmark evidence - #1

Merged

ponythewhite merged 33 commits into

mainfrom

benchmark-readme-1.0.5

May 4, 2026

Commits on Apr 25, 2026
- Add hardening benchmark harness and initial results
  ponythewhite
  committed
  77c3f4a
- Document Clostera quality improvement benchmark plan
  ponythewhite
  committed
  d19f884
- Add dense ADC and hybrid PQ k-means paths
  ponythewhite
  committed
  5066822
- Expose Clostera quality modes in Python API
  ponythewhite
  committed
  166c39a
- Fix Clostera variant benchmark class-count inference
  ponythewhite
  committed
  9c72246
- Add szymon3 Clostera variant sweep results
  ponythewhite
  committed
  c3809fb
- Activate Clostera clustering knobs and speed wins
  ponythewhite
  committed
  8b212b3
- Add SIMD dispatch and frontier benchmark schedule
  ponythewhite
  committed
  8afe9c1
- Add packed PQ4 assignment benchmark lane
  ponythewhite
  committed
  1e219b0
- Use packed PQ4 for hybrid shortlist
  ponythewhite
  committed
  ca1bfa6
- Add quantized PQ4 FastScan kernels
  ponythewhite
  committed
  b1adf17
- Reduce dense update cache contention
  ponythewhite
  committed
  0d2e2d2
- Add PQ4 frontier benchmark results
  ponythewhite
  committed
  9f35f67
- Reduce PQKMeans cache and allocation churn
  ponythewhite
  committed
  3a19c2a
- Schedule cache optimized frontier sweep
  ponythewhite
  committed
  0589e5a
- Reduce PQ training accumulator churn
  ponythewhite
  committed
  b05a4c4
- Reuse PQKMeans hot-path scratch buffers
  ponythewhite
  committed
  a11d40f
- Add cache frontier benchmark results
  ponythewhite
  committed
  99ac234
- Schedule scratch-buffer frontier sweep
  ponythewhite
  committed
  01c8b42
- Chunk encoder and auto-k parallel writes
  ponythewhite
  committed
  ed295d3
- Schedule chunked frontier sweep
  ponythewhite
  committed
  5c55051
- Incorporate research supplement roadmap
  ponythewhite
  committed
  56447e7
- Add PDX exact path and spherical cosine centers
  ponythewhite
  committed
  301bc99
Commits on Apr 26, 2026
- Add optimized clustering paths and resweep schedule
  ponythewhite
  committed
  d275431
- Optimize dense exact assignment dispatch
  ponythewhite
  committed
  dcd5fd8
- Add fixed 32 and 64 K grid points
  ponythewhite
  committed
  336376b
- Enforce cached sweep timeouts per row
  ponythewhite
  committed
  6289eee
Commits on May 4, 2026
- Implement benchmark-driven auto algorithm API
  ponythewhite
  committed
  5386960
- Rework README around benchmark evidence
  ponythewhite
  committed
  6eaaa4c
- Strengthen README benchmark positioning
  ponythewhite
  committed
  22d8fc8
- Tighten public README benchmark presentation
  ponythewhite
  committed
  1b1935a
- Use Clostera README hero banner
  ponythewhite
  committed
  06170aa
- Let README hero banner lead
  ponythewhite
  committed
  3539394

Navigation Menu

Repository navigation

Rework README around benchmark evidence - #1

Rework README around benchmark evidence#1
ponythewhite merged 33 commits into
mainfrom
benchmark-readme-1.0.5

Commits on Apr 25, 2026

Add hardening benchmark harness and initial results

Document Clostera quality improvement benchmark plan

Add dense ADC and hybrid PQ k-means paths

Expose Clostera quality modes in Python API

Fix Clostera variant benchmark class-count inference

Add szymon3 Clostera variant sweep results

Activate Clostera clustering knobs and speed wins

Add SIMD dispatch and frontier benchmark schedule

Add packed PQ4 assignment benchmark lane

Use packed PQ4 for hybrid shortlist

Add quantized PQ4 FastScan kernels

Reduce dense update cache contention

Add PQ4 frontier benchmark results

Reduce PQKMeans cache and allocation churn

Schedule cache optimized frontier sweep

Reduce PQ training accumulator churn

Reuse PQKMeans hot-path scratch buffers

Add cache frontier benchmark results

Schedule scratch-buffer frontier sweep

Chunk encoder and auto-k parallel writes

Schedule chunked frontier sweep

Incorporate research supplement roadmap

Add PDX exact path and spherical cosine centers

Commits on Apr 26, 2026

Add optimized clustering paths and resweep schedule

Optimize dense exact assignment dispatch

Add fixed 32 and 64 K grid points

Enforce cached sweep timeouts per row

Commits on May 4, 2026

Implement benchmark-driven auto algorithm API

Rework README around benchmark evidence

Strengthen README benchmark positioning

Tighten public README benchmark presentation

Use Clostera README hero banner

Let README hero banner lead