Skip to content

landing: single-core NumPy figures replace the multicore claims - #90

Merged
tannevaled merged 3 commits into
mainfrom
landing/single-core-claims
Nov 14, 2026
Merged

tannevaled merged 3 commits into
mainfrom
landing/single-core-claims

Conversation

@tannevaled

Copy link
Copy Markdown
Contributor

The landing's speed claims were multicore figures: "2.3–4.5×" on Sum/Mean/Max on 16 cores, and "3.7–5.9×" over one-thread OpenBLAS with multicore Go. The first single-core measurement of the whole API against NumPy 2.5.4 (go-ndarray/ndarray v0.41.0, tracked in go-ndarray/ndarray#118 by cause) contradicts them:

  • Max takes 1.13× NumPy's time on one core;
  • MatMul at 128² takes 1.15×;
  • 188 of 256 public operations are slower, and 54 are faster.

This PR changes the stats, the lead paragraph and the closing band to state the single-core result:

  • the faster operations, with their ratios;
  • the slower count, with the main causes;
  • a link to the tracking issue.

The figures are copied from docs/perf-table.md of go-ndarray/ndarray#119. This should merge after that PR and the ndarray release that carries the table, so the "performance page" link shows the same numbers.

landingkit check public passes.

Refs go-ndarray/ndarray#118, Refs go-ndarray/ndarray#119

🤖 Generated with Claude Code

tannevaled and others added 3 commits November 14, 2026 22:14
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…against Go)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@tannevaled
tannevaled merged commit 33cb762 into main Nov 14, 2026
7 checks passed
@tannevaled
tannevaled deleted the landing/single-core-claims branch November 14, 2026 21:45
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant