I couldn't figure out how to file a bug / submit a patch on this site so I will just post this here. Sorry for that.
The BitStringDecoder for ber is super slow since it decodes the stream bit by bit. I improved this by precomputing the bit lists and brought down the runtime for parsing the google.com cert to ~50% (total time, not just bitstream parsing time). I uploaded the patch of what I did here:
Maybe this helps...
Thanks for the patch!
I will see if I could improve BIT STRING handling performance beyond the precomputed values approach and report back here.
Current BIT STRING performance is just awful. %(
If you don't like the precomputed values approach, the most important problem to address in the current code is the accumulation of bits in a tuple. Tuples are imutable so this is essentially a quadratic algorithm.
If you just change this to be a list instead, it's already much faster. Not as fast as doing it byte wise with precomputation though :)
That's right, I'm looking into a byte wise design with some precomputation.
Original implementation was more of a proof of concept. %-\
Sign up for the SourceForge newsletter:
You seem to have CSS turned off.
Please don't fill out this field.