[jacopoabramo]

• Compare the key with every stored slot on the first in, add, discard or remove after a read, and build the index on a later lookup
• Remove every stored copy of a key on that first discard or remove
• Compare 8-byte keys, floats included, in blocks written so the compiler can use SSE2 instructions
• Drop stored duplicates before a full set refuses an add
• Add benchbox ops rows that read a set, change one key and write it back
• Accept the TypeError a forged kind above 5 raises in the forged-bytes test

[codspeed]

Merging this PR will improve performance by ×3

⚡ 3 improved benchmarks
✅ 72 untouched benchmarks
⏩ 1 skipped benchmark¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
⚡	test_contains_bounded_set[bs64k-65536]	19.9 ms	3 ms	×6.6
⚡	test_contains_bounded_set[bs1k-1024]	353.2 µs	96.6 µs	×3.7
⚡	test_contains_bounded_set[bs16-16]	57.1 µs	51 µs	+11.94%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing perf/bounded-set-first-lookup (85df223) with main (ff0b2c4)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩