Re: Speed tips, again, for extremely large scores?
On Mon, Feb 01, 2010 at 12:37:42PM -, Trevor Daniels wrote: > > Mats Bengtsson wrote Monday, February 01, 2010 11:52 AM >> >> Martin Tarenskeen wrote: >>> >>> On Mon, 1 Feb 2010, Michael Kappler wrote: >>> I'm also still very interested if there are possibilities to increase LilyPond performance further. My machine is very slow, though and I cannot speak for many people when raising performance issues. >>> >>> Would it be an idea to create a "Lilypond Benchmark" webpage, small, >>> interesting, and useless ;-) > > It would be both interesting and a useful check on > whether code additions to new releases have had an > effect on processing speed. Although for this we > would have to establish one or maybe several standard > configurations so the tests are directly comparable. ARGH! Bloody mao! (not directed at you, Trevor) Like **so many** things in lilypond, this was done ages ago. For the past few YEARS, the regtests have included this information. The problem is that it isn't documented anywhere, the output is hard to read, and I have a sneaky suspicion that part of it is broken at the moment but would only require a 10-minute bugfix if the person knew what he was doing. You want to improve the situation? There's two options: 1) start reading the sources to figure out how this benchmarking works. Figure out what the output means, whether or not it's currently working, look at old versions to see what it looked like back then, etc. 2) start helping me with easy stuff (like writing plain text for the new website), so that *I* can work on #1. I have the technical knowledge and persistance required to solve #1, but given the lack of other people on more critical things, I doubt I'll start working on it for at least 3 months. - Graham ___ lilypond-user mailing list lilypond-user@gnu.org http://lists.gnu.org/mailman/listinfo/lilypond-user
Re: Speed tips, again, for extremely large scores?
Mats Bengtsson wrote Monday, February 01, 2010 11:52 AM Martin Tarenskeen wrote: On Mon, 1 Feb 2010, Michael Kappler wrote: I'm also still very interested if there are possibilities to increase LilyPond performance further. My machine is very slow, though and I cannot speak for many people when raising performance issues. Would it be an idea to create a "Lilypond Benchmark" webpage, small, interesting, and useless ;-) This webpage would show a list with the following info: 1. Hardware / Processor type 2. Platform / OS 3. Lilypond version 4. The benchmark result: I suggest to give te time needed to "make" the (at least in this mailinglist) already famous "Reubke Psalm 94" score. It would be both interesting and a useful check on whether code additions to new releases have had an effect on processing speed. Although for this we would have to establish one or maybe several standard configurations so the tests are directly comparable. Don't forget that RAM size is a major factor when it comes to processing time. Probably the major factor along with cpu speed, unless you already have enough RAM. There are several other parameters which will affect the speed but which are harder to determine. These vary depending on the particular computer model, so the results might still be very variable. These include Processor cache size and speed Memory transfer rates Hard disk seek/search times (affects page rates if RAM is insufficient) Whether page file is fragmented (ditto) Also the effect of other processes would need to be excluded by doing the test in a newly-booted clean system, and running the test several times, retaining just the fastest time. Trevor ___ lilypond-user mailing list lilypond-user@gnu.org http://lists.gnu.org/mailman/listinfo/lilypond-user
Re: Speed tips, again, for extremely large scores?
Martin Tarenskeen wrote: On Mon, 1 Feb 2010, Michael Kappler wrote: I'm also still very interested if there are possibilities to increase LilyPond performance further. My machine is very slow, though and I cannot speak for many people when raising performance issues. Would it be an idea to create a "Lilypond Benchmark" webpage, small, interesting, and useless ;-) This webpage would show a list with the following info: 1. Hardware / Processor type 2. Platform / OS 3. Lilypond version 4. The benchmark result: I suggest to give te time needed to "make" the (at least in this mailinglist) already famous "Reubke Psalm 94" score. You can find some examples of such benchmarking in the mailing list archives. Don't forget that RAM size is a major factor when it comes to processing time. /Mats ___ lilypond-user mailing list lilypond-user@gnu.org http://lists.gnu.org/mailman/listinfo/lilypond-user
Re: Speed tips, again, for extremely large scores?
On Mon, 1 Feb 2010, Michael Kappler wrote: I'm also still very interested if there are possibilities to increase LilyPond performance further. My machine is very slow, though and I cannot speak for many people when raising performance issues. Would it be an idea to create a "Lilypond Benchmark" webpage, small, interesting, and useless ;-) This webpage would show a list with the following info: 1. Hardware / Processor type 2. Platform / OS 3. Lilypond version 4. The benchmark result: I suggest to give te time needed to "make" the (at least in this mailinglist) already famous "Reubke Psalm 94" score. The website should have a simple interface where Lilypond users can enter their results and show a list sorted by results. On linux it is easy to measure the time using the "time" command. Would be interesting to see results with relatively old computers as well as the newest i7 speeddevils ... Lilypond does benifit a lot from this fast new processor technology. Speed is not Lilypond's strongest point. Other, less powerful but still quite good, text-to-musicscore commandline programs like mup (shareware), abcm2ps, and pmw are MUCH faster. Even on my ancient Atari computer running at 32 MHz, I could run these programs with reasonable speed. But there are several other reasons why I prefer and use mostly Lilypond now. If one day computers will be too fast to give reliable results, we can always change the benchmark test by repeating the same test 10 or 100 times :-) There have been some individual postings on this subject, but a small dedicated website would be nice. An extra page on the normal Lilypond homepage maybe ? I'm not a very capable webpage designer. Any volunteers ? -- Martin ___ lilypond-user mailing list lilypond-user@gnu.org http://lists.gnu.org/mailman/listinfo/lilypond-user