docs: republish the macOS benchmarks from a pre-release run at 5c8b9769 #467

Merged

natsukium merged 1 commit from docs/republish-macos-bench-5c8b9769 into main 2026-09-26 08:57:44 +09:00

Conversation 0 Commits 1 Files changed 12 +73 -74

natsukium commented 2026-09-26 01:13:35 +09:00

Owner

Copy link

The published macOS figures came from 9d1d842f, which predates the parse, render and startup work that will ship in the release (#396 through #465). This republishes the macOS section of docs/reference/benchmarks.md, its charts and README's headline table from a fresh full run at 5c8b9769 (nix build .#felis). The Linux section is untouched.

Run. crossterm.py run over all eight suites and the full field, ROUNDS=3 with rotated order, at the published page's font pin (Menlo 14 pt). paneru was stopped for the whole run. Every leg completed in every round: no refused, unfinished or retried legs.

What moved.

• The grid is 120x35 at the same 960x600 window, 13% more cells than the published 106x35 pin (felis at 9d1d842f also opens at 120x35 under this config, so the difference is in the pin, not in felis). DOOM-fire fell by 8-12% for every terminal, not only felis.
• kittenbench: felis now leads all four rows, ASCII and Unicode included (599 and 710 MB/s against ghostty-tip's 542 and 574), with ranges apart.
• termbench many_lines 260 MB/s (was 212) and sgr_fg_lines 317 (was 237); felis leads every row.
• Startup: cold 99 ms, warm 91 ms (was 126 and 144); warm is now below cold in every round.
• Idle footprint rose from 33.2 to 50.1 MB (client 46.1 MB, daemon 4.0 MB), consistent across rounds (50.0-50.2), while every other terminal's idle figure moved by at most 2 MB. felis falls behind wezterm (46.8) and alacritty (47.9) on this row, and README's "under 33 MB" claim is corrected to what was measured. The cause is being investigated separately.
• Mean latency 24.6 ms against alacritty's 24.3, with overlapping ranges; no winner is claimed.

Kept on the page. Some legs ran more than 1.0 above the starting load average of 2.2, mostly felis's and ghostty-tip's in rounds 2 and 3; each chart names them and the three rounds agree.

Verified: every figure came from the harness's report.md (per-round artifacts checked for the prose claims); just prose-check clean; treefmt applied.
Assisted-by: Claude Code

The published macOS figures came from `9d1d842f`, which predates the parse, render and startup work that will ship in the release (#396 through #465). This republishes the macOS section of `docs/reference/benchmarks.md`, its charts and README's headline table from a fresh full run at `5c8b9769` (`nix build .#felis`). The Linux section is untouched. **Run.** `crossterm.py run` over all eight suites and the full field, `ROUNDS=3` with rotated order, at the published page's font pin (Menlo 14 pt). paneru was stopped for the whole run. Every leg completed in every round: no refused, unfinished or retried legs. **What moved.** - The grid is 120x35 at the same 960x600 window, 13% more cells than the published 106x35 pin (felis at `9d1d842f` also opens at 120x35 under this config, so the difference is in the pin, not in felis). DOOM-fire fell by 8-12% for every terminal, not only felis. - kittenbench: felis now leads all four rows, ASCII and Unicode included (599 and 710 MB/s against ghostty-tip's 542 and 574), with ranges apart. - termbench `many_lines` 260 MB/s (was 212) and `sgr_fg_lines` 317 (was 237); felis leads every row. - Startup: cold 99 ms, warm 91 ms (was 126 and 144); warm is now below cold in every round. - Idle footprint rose from 33.2 to 50.1 MB (client 46.1 MB, daemon 4.0 MB), consistent across rounds (50.0-50.2), while every other terminal's idle figure moved by at most 2 MB. felis falls behind wezterm (46.8) and alacritty (47.9) on this row, and README's "under 33 MB" claim is corrected to what was measured. The cause is being investigated separately. - Mean latency 24.6 ms against alacritty's 24.3, with overlapping ranges; no winner is claimed. **Kept on the page.** Some legs ran more than 1.0 above the starting load average of 2.2, mostly felis's and ghostty-tip's in rounds 2 and 3; each chart names them and the three rounds agree. Verified: every figure came from the harness's `report.md` (per-round artifacts checked for the prose claims); `just prose-check` clean; treefmt applied. Assisted-by: Claude Code

natsukium added 1 commit 2026-09-26 01:13:35 +09:00

docs: republish the macOS benchmarks from a pre-release run at 5c8b9769 ... 9e05e474cb

The published macOS figures came from `9d1d842f`, which predates the parse, render and startup work that will ship in the release (#396 through #465). This republishes the macOS section of `docs/reference/benchmarks.md`, its charts and README's headline table from a fresh full run at `5c8b9769` (`nix build .#felis`). The Linux section is untouched.

**Run.** `crossterm.py run` over all eight suites and the full field, `ROUNDS=3` with rotated order, at the published page's font pin (Menlo 14 pt). paneru was stopped for the whole run. Every leg completed in every round: no refused, unfinished or retried legs.

**What moved.**

- The grid is 120x35 at the same 960x600 window: felis's cell is now 8 px wide instead of 9 (#411 rounds the cell width to the nearest pixel), so the pin holds 13% more cells. DOOM-fire fell by 8-12% for every terminal, not only felis.
- kittenbench: felis now leads all four rows, ASCII and Unicode included (599 and 710 MB/s against ghostty-tip's 542 and 574), with ranges apart.
- termbench `many_lines` 260 MB/s (was 212) and `sgr_fg_lines` 317 (was 237); felis leads every row.
- Startup: cold 99 ms, warm 91 ms (was 126 and 144); warm is now below cold in every round.
- Idle footprint rose from 33.2 to 50.1 MB (client 46.1 MB, daemon 4.0 MB), consistent across rounds (50.0-50.2), while every other terminal's idle figure moved by at most 2 MB. felis falls behind wezterm (46.8) and alacritty (47.9) on this row, and README's "under 33 MB" claim is corrected to what was measured. The cause is being investigated separately.
- Mean latency 24.6 ms against alacritty's 24.3, with overlapping ranges; no winner is claimed.

**Kept on the page.** Some legs ran more than 1.0 above the starting load average of 2.2, mostly felis's and ghostty-tip's in rounds 2 and 3; each chart names them and the three rounds agree.

Verified: every figure came from the harness's `report.md` (per-round artifacts checked for the prose claims); `just prose-check` clean; treefmt applied.
Assisted-by: Claude Code

natsukium force-pushed docs/republish-macos-bench-5c8b9769 from 9e05e474cb to 56cc08bfcc 2026-09-26 01:20:42 +09:00 Compare

natsukium merged commit e82f672cf2 into main 2026-09-26 08:57:44 +09:00

natsukium deleted branch docs/republish-macos-bench-5c8b9769 2026-09-26 08:57:44 +09:00

natsukium referenced this pull request from a commit 2026-09-26 08:57:46 +09:00

docs: republish the macOS benchmarks from a pre-release run at 5c8b9769 (#467)

Commenting is not possible because the repository is archived.

Reviewers

No reviewers

Labels

Clear labels   

priority/P0

Must be resolved before the v0.1.0 tag

  

priority/P1

Must be settled before the first public contract freezes

  

priority/P2

Explicitly deferred until after v0.1.0

  

release/v0.1.0

Work tracked for the first tagged release

  

status/blocked

Cannot start now: waits on an upstream release, another issue, or a recorded revisit trigger

  

status/planned

Triaged with a source-grounded plan and a passed pi review; ready to implement

  

type/bug

Shipped behavior that contradicts the docs or the code's own intent; needs a product code change

  

type/design

An open scope or trade-off question; needs a recorded decision in the owning explanation doc before code

  

type/test-gap

Documented or reviewed behavior that no test pins; usually split from a review finding, needs no product code change

  

type/tracker

Parent issue that tracks child work

No labels

priority/P0

priority/P1

priority/P2

release/v0.1.0

status/blocked

status/planned

type/bug

type/design

type/test-gap

type/tracker

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

natsukium/felis!467

No description provided.