Hi,

While optimizing an implementation of levenshtein distance, I found that
(min ...) took up a significant part of the profile (16%), so I tried
replacing it with a simple implementation in Scheme. That worked out
surprisingly well (shaved off about 25% of the total runtime) and when I
benchmarked it directly it turned out to be between 1.5 and 3 times
faster than the min in numbers.c.

Would it be a good idea to move min and max from C to Scheme? And what
would be needed?

Attached is my implementation directly in arithmetic.bm

drop it into benchmark-suite/benchmarks/arithmetic.bm and run
./benchmark-guile arithmetic.bm

to see the results.

Best wishes,
Arne

Attachment: arithmetic.bm
Description: Binary data

-- 
Unpolitisch sein
heißt politisch sein,
ohne es zu merken.
draketo.de

Attachment: signature.asc
Description: PGP signature

Reply via email to