Speedup charts of bbtree++ and bbtree (Rust) against std::map, a row per
bench cell: each dot sits at std::map's time divided by the map's, on a
log axis, so right of 1× is faster.
| file | data |
|---|---|
integer_keys.csv, integer_keys.png |
COMPARISON.md, "Integer keys" |
string_keys.csv, string_keys.png |
COMPARISON.md, "String keys": bbtree++ with the compact string fastest on each cell, named in the CSV |
The main README shows the 1M-entry cells only (integer_keys_1m.png,
string_keys_1m.png).
The CSVs hold the tables' figures, nanoseconds per operation (iterate and many maps: per entry). Regenerate the charts with uv, which fetches matplotlib for the script:
T=bbtree/bench/tools/speedup_chart.py C=bbtree/bench/charts
uv run $T $C/integer_keys.csv --title "Integer keys: bbtree++ and bbtree (Rust) against std::map"
uv run $T $C/string_keys.csv --title "String keys: bbtree++ and bbtree (Rust) against std::map"
uv run $T $C/integer_keys.csv --1m --title "Integer keys, 1M entries: bbtree++ and bbtree (Rust) against std::map"
uv run $T $C/string_keys.csv --1m --title "String keys, 1M entries: bbtree++ and bbtree (Rust) against std::map"PNGs are 2000 pixels wide; --svg writes SVG instead.
--1m keeps only the cells of one map of 1M entries and writes
<csv name>_1m.png; --baseline picks another column to compare with.

