Skip to content

Add scatterplots in-terminal to the comparison#10

Open
msooseth wants to merge 6 commits into
mainfrom
fix-by-mate2
Open

Add scatterplots in-terminal to the comparison#10
msooseth wants to merge 6 commits into
mainfrom
fix-by-mate2

Conversation

@msooseth

@msooseth msooseth commented May 19, 2026

Copy link
Copy Markdown
Contributor

They make more sense now that we have more things being compiled.

Also removing the printing to the millionth of a second and nonsense like that. We can't measure to that precision with any reliability. I am reducing precision printed to 1/100th of a second. Noise level is way higher than that. Same with memory, reducing to 0.01MB precision, nobody cares about anything more.

msooseth and others added 6 commits May 19, 2026 10:57
In cross-version mode, `compare` now draws a comparison scatter: one
dot per benchmark, x = baseline cpu_time, y = target cpu_time, with the
y=x diagonal. Dots on the line are unchanged; below it the target is
faster.

The plot is rendered with gnuplot and always saved to
solc-bench-compare.png so it can be reused. When stdout is a terminal
that speaks the kitty graphics protocol or sixel (detected by querying
the terminal), it is also drawn inline. gnuplot's kitty/sixel terminals
home the cursor and clear the screen, so those escapes are stripped to
keep the image flowing inline below the table.

gnuplot is bundled onto the wrapped binary's PATH via the Nix flake.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The compare command now draws baseline-vs-target scatters for cpu_time,
cycles, and peak_rss instead of cpu_time alone. Each panel is a square
384x384 plot (down from 640x640), saved as compare_<metric>.png, and the
panels render side by side inline via gnuplot multiplot.

Each dot is one benchmark, pairing the baseline and target medians.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Trim the verbose comments and docstrings to the essentials; no behavior
change.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Base automatically changed from fix-by-mate to main May 19, 2026 11:59
@msooseth msooseth requested a review from r0qs May 19, 2026 15:25
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