[tahoe-dev] pycryptopp benchmarks

Zooko Wilcox-O'Hearn zooko at zooko.com
Wed Mar 21 18:00:08 UTC 2012


 I wrote:

> 5. Unless there's something wrong with my benchmarks, encryption isn't quite as fast as I'd like on  François's little low-power ARM box.

There was indeed something wrong with my benchmarks of ciphers and
hashes (but not digital signatures). Something so bone-headedly wrong
that I would prefer if you didn't look at this patch that fixes it:
https://github.com/tahoe-lafs/pycryptopp/commit/e855f9b477bafe19bfe3a551f5f0e383bb224e2d

Everything I wrote about digital signatures in the previous letter was
correct. As for ciphers, the performance is more than 1000 times
better than I thought.

On 64-bit servers, the it costs something around 10 *nanoseconds* per
byte, and on François's little low-power ARM box, it costs about 200
nanoseconds per byte.

Regards,

Zooko



More information about the tahoe-dev mailing list