Refurbishment #2

[RuedigerMoeller]

• better stats output,
• separate VM for each serializer,
• moved to average measurement (median, quantile also implemented but just sysouted)
• categorized all serializers
• tested with cygwin, someone check linux
• preview of stats see https://github.com/RuedigerMoeller/fast-serialization/wiki/TestPage

[cakoose]

Why are the output files committed to the repo? Can you remove them?

[RuedigerMoeller]

The advantage is you can generate various stats without having to re-run
the bench (takes >1 hour to run). Do you still prefer to remove them ?

2014-03-13 2:03 GMT+01:00 Kannan Goundan [email protected]:

Why are the output files committed to the repo? Can you remove them?

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37489399
.

[RuedigerMoeller]

i could np

2014-03-13 2:06 GMT+01:00 Rüdiger Möller [email protected]:

The advantage is you can generate various stats without having to re-run
the bench (takes >1 hour to run). Do you still prefer to remove them ?

2014-03-13 2:03 GMT+01:00 Kannan Goundan [email protected]:

Why are the output files committed to the repo? Can you remove them?

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37489399
.

[cakoose]

I think it's valuable to save the result of test runs, but we should be more precise about it. For example, the file path should include some indication of the hardware/software configuration. For now, let's just remove it and expect people to run it at least once.

[RuedigerMoeller]

ok, i am at it. currently i denote hardware in the stats, which is flawed
as you can run the stats on a different computer

2014-03-13 2:10 GMT+01:00 Kannan Goundan [email protected]:

I think it's valuable to save the result of test runs, but we should be
more precise about it. For example, the file path should include some
indication of the hardware/software configuration. For now, let's just
remove it and expect people to run it at least once.

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37489711
.

[RuedigerMoeller]

done, afaik i don't need another pull request for the deletion ..

[cakoose]

Hey, sorry, but could you delete report.textile as well? Or if you're fine with me tweaking things, I can delete it when I merge. Either way is fine with me.

[RuedigerMoeller]

ok, should I paste the latest result to the wiki ? note i had trouble with
recurring CRLF on the scala compiler .. TT

2014-03-13 2:18 GMT+01:00 Kannan Goundan [email protected]:

Hey, sorry, but could you delete report.textile as well? Or if you're
fine with me tweaking things, I can delete it when I merge. Either way is
fine with me.

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37490121
.

[RuedigerMoeller]

done

2014-03-13 2:20 GMT+01:00 Rüdiger Möller [email protected]:

ok, should I paste the latest result to the wiki ? note i had trouble with
recurring CRLF on the scala compiler .. TT

2014-03-13 2:18 GMT+01:00 Kannan Goundan [email protected]:

Hey, sorry, but could you delete report.textile as well? Or if you're

fine with me tweaking things, I can delete it when I merge. Either way is
fine with me.

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37490121
.

[cakoose]

I'll look into the CRLF issue. It's happened before. I wonder if Git has that file in a bad state?

Let's wait on posting it to the wiki. We should give people more time to look at the changes and try running the benchmarks themselves. My guess is that one week is an acceptable delay.

[RuedigerMoeller]

np. I have tested it under Cygwin only ..

2014-03-13 2:27 GMT+01:00 Kannan Goundan [email protected]:

I'll look into the CRLF issue. It's happened before. I wonder if Git has
that file in a bad state?

Let's wait on posting it to the wiki. We should give people more time to
look at the changes and try running the benchmarks themselves. My guess is
that one week is an acceptable delay.

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37490658
.

[RuedigerMoeller]

Also people should review my classifications. I did not have time to review
any of the 80 serializers regarding their features. Just a quicj glance
onto their code

2014-03-13 2:29 GMT+01:00 Rüdiger Möller [email protected]:

np. I have tested it under Cygwin only ..

2014-03-13 2:27 GMT+01:00 Kannan Goundan [email protected]:

I'll look into the CRLF issue. It's happened before. I wonder if Git has

that file in a bad state?

Let's wait on posting it to the wiki. We should give people more time to
look at the changes and try running the benchmarks themselves. My guess is
that one week is an acceptable delay.

Reply to this email directly or view it on GitHubhttps://github.com//pull/44#issuecomment-37490658
.