Re: some more performance benchmarks

2011-04-14 Thread Greg Stein
On Thu, Apr 7, 2011 at 10:42, Neels Hofmeyr wrote: >... > Any ideas/opinions about adding to tools/dev/ would be appreciated. Commited in r1092569 as tools/dev/benchmarks/suite1. Having them in source control is better than worrying too much about the naming. We can now evolve these for other pl

Re: some more performance benchmarks

2011-04-08 Thread Daniel Shahaf
Neels Hofmeyr wrote on Thu, Apr 07, 2011 at 16:42:43 +0200: > [[[ > COMPARE 5x5_1.6 to 5x5_trunk > 1.23|+0.45 means factor=1.23, difference in seconds = 0.45 > factor < 1 or difference < 0 means '5x5_trunk' is faster than > '5x5_1.6' > min max avg operat

Re: some more performance benchmarks

2011-04-07 Thread Daniel Shahaf
Neels Hofmeyr wrote on Thu, Apr 07, 2011 at 16:42:43 +0200: > I have no idea about other loads on our vm. Maybe we could use the > user/sys output of 'time -p' instead of the real-world time measurements > it is taking right now, although my intention was to stick to the actual > real world time. O

Re: some more performance benchmarks

2011-04-07 Thread Neels Hofmeyr
On Thu, 2011-04-07 at 08:03 -0500, Hyrum K Wright wrote: > This is great work, though as you admit the results are bit > incomprehensible at this stage. Would it make sense to somehow Hmm, I think stsp was a bit early to post, I'm still in the sanity checking phase ;) E.g.,: [[[ Timings for 5x

Re: some more performance benchmarks

2011-04-07 Thread C. Michael Pilato
On 04/07/2011 07:29 AM, Stefan Sperling wrote: > Neels has set up his performance benchmarks to run on the svn VM > we have at apache.org. [...] > Below are the results of the latest run. > They're probably a bit hard to decipher as is. The plan is to make the > scripts produce more easily digest

Re: some more performance benchmarks

2011-04-07 Thread Hyrum K Wright
On Thu, Apr 7, 2011 at 6:29 AM, Stefan Sperling wrote: > Neels has set up his performance benchmarks to run on the svn VM > we have at apache.org. > > The tests only use ra_Local and are supposed to measure wc-ng > performance, not httpv2. They focus on edge cases and are thus a > good addition to

some more performance benchmarks

2011-04-07 Thread Stefan Sperling
Neels has set up his performance benchmarks to run on the svn VM we have at apache.org. The tests only use ra_Local and are supposed to measure wc-ng performance, not httpv2. They focus on edge cases and are thus a good addition to Mark Phippard's test which focus more on real world usage. Below