verifying encoding_comparison.html results
Thu Jan 31 22:41:01 CET 2002
I downloaded the Megaco measurement software from your website and ran it
locally on my machine which is an Intel Celeron 900MHz box running RedHat
Linux. The Erlang/OTP environment was compiled with "--enable-hipe"
option. While checking the results, the following questions arose:
1) Your charts at http://www.erlang.org/project/megaco/encode_time.jpg
refer to binary encoders as just "PER" and "BER". Exactly which variant
of your encoders are being referenced here - "original" or "native"? As
you can see in the attachment to this email, the tool that you provided
reports results for both your "original" and "native" implementations.
2) According to my measurements, on average the erl_dist encoder is 11.7
times faster than the BER native encoder and 17.3 times faster than the
ordinary (non-native) BER encoder. However, your charts at
show that the erl_dist encoder is "only" 8.8 times faster then BER. How do
you explain this difference where my results shows erl_dist is 17.3 times
faster than BER, while yours show only 8.8? Is it that the version of
your libraries that you provided has a more highly optimized version of
the erl_dist encoder/decoder than the one that you used in your own
reports? If that is not the case, do you have any idea why it performs
significantly better than is reported on the website?
You can find the full set of output from my test run in the attached file.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 2808 bytes
More information about the erlang-questions