[maemo-developers] Re: testing pango with 770
ext Federico Mena Quintero wrote: >This is interesting, since pango-1.11.1 has our recent optimizations in >it. > >You mentioned that you were using gettimeofday(). Can you please test >the "pango-profile" module from CVS? It has a slightly more >sophisticated timer, using times(). I've found it to be quite reliable, >and it gives reproducible results. > > > I've been running pango-language-profile now aswell, also produced new results with fs-tests BUT then when I made tests with x86 I found out with callgrind&kcachegrind that quite much of the time was spent on typecasts and checks ... so actually I've been running stuff with debug turned on (instead of 'minimal' in stable release) :-/ With x86 this does not seem to make much difference but with arm I'd expect results with new pango to be better. I'm sorry for my misconclusions, I'll run the tests again and post results! // Tapani ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Hi, > Note: In Maemo UI user cannot resize windows and applications > themselves do that very rarely for dialogs, so this is fairly > moot point for Maemo. > Try popping up the vkb. Opera takes ~4 seconds to resize the first time. I doubt this is related to Gtk as VKB pops up just fine for other applications. If you attach to the Browser process with strace or ltrace (ltrace -S -p PID), you can see what it's doing. - Eero ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
ext Koen Kooi wrote: > Eero Tamminen wrote: > > >Note: In Maemo UI user cannot resize windows and applications themselves > >do that very rarely for dialogs, so this is fairly moot point for Maemo. > > > Try popping up the vkb. Opera takes ~4 seconds to resize the first time. > I think this is a special case which does not happen with other applications, browser is doing something more here. > regards, > > Koen // Tapani ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Koen Kooi kirjoitti 29.12.2005 kello 23.54: Eero Tamminen wrote: Note: In Maemo UI user cannot resize windows and applications themselves do that very rarely for dialogs, so this is fairly moot point for Maemo. Try popping up the vkb. Opera takes ~4 seconds to resize the first time. Hi, Isn't the Opera engine written using QT?-) Seriously, I don't know what widget set they use in the 770, but the browser widget/window (unlike the rest of the UI) is obviously not gtk+. Compare e.g. the scrollbars to those of native gtk+ apps. Other than that, switching to/from full screen mode is one common case where the applications resize themselves. But I don't think that's unusually slow in any of the default applications at least... BR, Lassi ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Eero Tamminen wrote: > Note: In Maemo UI user cannot resize windows and applications themselves > do that very rarely for dialogs, so this is fairly moot point for Maemo. Try popping up the vkb. Opera takes ~4 seconds to resize the first time. regards, Koen -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (Darwin) iD8DBQFDtFsJMkyGM64RGpERAsekAKC5UhvLKHVxLL8SHFq8dT2vePFYAQCfeFZ9 7EAK01TeKkiFyDG230Yew+0= =Bw+t -END PGP SIGNATURE- ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Hi, > These are just some impressions I had on my systems: > * Konqueror has a faster UI than Nautilus > * Opera has a faster UI than Firefox (whereas Mozilla GTK-1.2 > builds were pretty snappy till they switched to GTK-2) > * The Motif and Fox-Ports of Eclipse are MUCH faster then Eclipse/GTK2 > * Layouting works usually more smooth (=higher update frequency) > when resizing something with QT than GTK. > * Even smaller UIs like that one from GFTP feel a bit slow if you > resize the views. Could you give names of the applications you've used for the comparison? E.g. was this Kword vs. Abiword etc (and not Kate vs. Abiword :))? I think it would then be pretty straightforward to get some hard numbers with xresponse. Note: In Maemo UI user cannot resize windows and applications themselves do that very rarely for dialogs, so this is fairly moot point for Maemo. - Eero ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Hi Eric, > This is so interesting since GTK+ applications feels so much faster on > all machines I've been using whereas QT once are somewhat slow.. Are > you running them all in KDE? I'm running everything in Gnome and it's > fast enough even on my 800MHZ machine at home. I have to admit that I also use KDE but I am not a fan of one or the other I just use it because I like its filemanager more than gnome's. For pure GTK apps it should not make any difference wether they run on KDE or Gnome except initial loading time which is not what I am talking about. All the stuff I talked about is from the user point of view, it is not my intention to flame arround nore I am a fan of QT or GTK apps which I would like to highlight. I am just a bit dissappointed by the performance of some GTK2 apps, thats all. These are just some impressions I had on my systems: * Konqueror has a faster UI than Nautilus * Opera has a faster UI than Firefox (whereas Mozilla GTK-1.2 builds where pretty snappy till they switched to GTK-2) * The Motif an Fox-Potrs of Eclipse are MUCH faster then Eclipse/GTK2 * Layouting works usually more smooth (=higher update frequency) wen resizing something with QT than GTK. * Even smaller UIs like that one from GFTP feel a bit slow if you resize the views. However that are just impressions, I did not make any time-tables to document behaviours nore do I say QT is better than GTK. I would be just happy if GTK would receive some intensiv profiling and tuning versions instead of new features to be on the line with other toolkits from the performance point of view. It could also be the case that GTK apps suffer a bit form the nvidia-drivers since I use nvidia cards on all of my 3 systems. Thanks for listening, sorry for bothering, lg Clemens ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Hi, Nils Faerber wrote: > Have you ever tried recent GPE on a 200MHz iPaq? > I find that very "snappy" too and it uses GTK2... > > The latest GTK release uses Cairo for rendering - which is nice in > general but really a performance hog. But this is only in the very > latest release. I hope and guess that this problem has already been > recognised by the GTK developers and they hopefully work on it. i just tried two GPE images on an iPAQ h5550 yesterday. One using GTK 2.6.x and one using GTK 2.8.x. Comparing the behaviour without any measurement 2.8 just feels slower. Like discussed before memory usage is another issue... > I would like to have Cairo as a compile time option and not a fixed > dependency. For small devices like the 770 or other PDAs you could then > go without it (or slower devices). This would be a good solution, but this would lead to the situation that you can have incompatible builds of the same GTK version. I guess this is the main reason why the GTK developers only provide a very few compile time options. For example disabling deprecated bits at build time would be such an option too. It would save some space and make sure developers do not use deprecated features anymore. Greetings Florian -- The dream of yesterday Florian Boor is the hope of todayTel: 0271-771091-14 and the reality of tomorrow.Fax: 0271-771091-19 [Robert Hutchings Goddard, 1904][EMAIL PROTECTED] 6C 44 30 4C 43 20 6B 61 16 07 0F AA E6 97 70 A8 ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
> - Even large QT apps are snappy and very responsive whereas many large > GTK-2 apps tend to be unuseable on slower systems (like my Dorun-800). > 800mhz to render some buttons and layout texts?? > Maybe I am just too old to understand whats going on :( This is so interesting since GTK+ applications feels so much faster on all machines I've been using whereas QT once are somewhat slow.. Are you running them all in KDE? I'm running everything in Gnome and it's fast enough even on my 800MHZ machine at home. I read people saying what you are saying all the time, but I just figure they are KDE fans who wants to throw trash as Gnome/GTK (note, I'm not saying you are, which is why I respond... normally I just ignore them :) ) Regards, Erik ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
> This thread is pretty funny. Nokia tests 1 version of gtk with different > versions of pango and suddenly people say that *different* versions of > gtk were tested and jump to conclusions Reading-Knowledge is an advantage ;) But I never expect some of the gtk-diehards to tlisten any criticism since GTK is that wonderful and if something is not ok everybody is able to fix it themself. Furthermore if its really that slow anybody else would have fixed it before. Who said that different versios of GTK were *tested*. there was just sereral mails which claimed that their autors are not happy with GTK performance and since I am too disappointed by GTK-2.x speed I wrote my opinion down. By the way GTK-1 was not pango powered so these tests would be pretty useless ;) lg Clemens ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
ext Koen Kooi wrote: > Clemens Eisserer wrote: > > >>It's disappointing that with every release gtk gets slower... > > > >I absolutly agree - GTK-1.2 has been a nice and fast toolkit but since > >the jump to 2.0 GTK is just one of those "too-bloated-to-be-tuneable" > >pieces of software no one wants to touch since it could break > >something. > > > This thread is pretty funny. Nokia tests 1 version of gtk with different > versions of pango and suddenly people say that *different* versions of > gtk were tested and jump to conclusions > Yep. My intention for posting the results was to have a bit of discussion what could be wrong. I will run some other tests aswell and try to investigate. Let's not begin another endless flamewar about toolkits :-) > Koen // Tapani ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Clemens Eisserer wrote: >>It's disappointing that with every release gtk gets slower... > > > I absolutly agree - GTK-1.2 has been a nice and fast toolkit but since > the jump to 2.0 GTK is just one of those "too-bloated-to-be-tuneable" > pieces of software no one wants to touch since it could break > something. This thread is pretty funny. Nokia tests 1 version of gtk with different versions of pango and suddenly people say that *different* versions of gtk were tested and jump to conclusions Koen -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (Darwin) iD8DBQFDsoXtMkyGM64RGpERArTqAJ9mRUbj319IHPfzaThsxgTEmhrbigCgq+H+ z7vbZmBclp2XCeKnkr+Nvu8= =gn8L -END PGP SIGNATURE- ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
Hi again, > Have you ever tried recent GPE on a 200MHz iPaq? > I find that very "snappy" too and it uses GTK2... I have to admit that I do not have a lot of knowledge how qualified different toolkits are for handheld use - my observations just showed that GTK2 performs quite poorly on the 770 as it does for larger desktop applications. (like Eclipse/GTK2 (Eclips/Motif or Eclipse/FOX perform well), Evolution, ...) - the promise to tune it is arround for years now but I haven't seen major progress in this area. Its not possible to scroll widgets/tables withought articafts and window-repaints of simple windows with just a few widgets take 200-500ms. Skinned FLTK apps are fast even on that underpowered device... > I would like to have Cairo as a compile time option and not a fixed > dependency. For small devices like the 770 or other PDAs you could then > go without it (or slower devices). But since the plan is to expose the cairo-api this would be quite difficult ... just another layer somewhere in the stack. One size simlpy does not fit all. lg Clemens ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Clemens Eisserer schrieb: >>It's disappointing that with every release gtk gets slower... > > I absolutly agree - GTK-1.2 has been a nice and fast toolkit but since > the jump to 2.0 GTK is just one of those "too-bloated-to-be-tuneable" > pieces of software no one wants to touch since it could break > something. > > And do not argue that big, feature richt toolkits are slow in general > - Even large QT apps are snappy and very responsive whereas many large > GTK-2 apps tend to be unuseable on slower systems (like my Dorun-800). > 800mhz to render some buttons and layout texts?? > Maybe I am just too old to understand whats going on :( > > A very favourite argument is that skinned toolkits are slow or X is > slow, but how do other, much faster toolkits face with that. Have you ever tried recent GPE on a 200MHz iPaq? I find that very "snappy" too and it uses GTK2... The latest GTK release uses Cairo for rendering - which is nice in general but really a performance hog. But this is only in the very latest release. I hope and guess that this problem has already been recognised by the GTK developers and they hopefully work on it. I would like to have Cairo as a compile time option and not a fixed dependency. For small devices like the 770 or other PDAs you could then go without it (or slower devices). > lg Clemens Cheers nils faerber - -- kernel concepts Tel: +49-271-771091-12 Dreisbachstr. 24 Fax: +49-271-771091-19 D-57250 Netphen Mob: +49-176-21024535 - -- -BEGIN PGP SIGNATURE- Version: GnuPG v1.4.1 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFDsnsQJXeIURG1qHgRAiofAJ9MNGan2jE43nnM0tkz7uKy19P9gwCghrvR P/NKx8hENuZStXB36YJ3NjE= =Ue2C -END PGP SIGNATURE- ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
Re: [maemo-developers] Re: testing pango with 770
> It's disappointing that with every release gtk gets slower... I absolutly agree - GTK-1.2 has been a nice and fast toolkit but since the jump to 2.0 GTK is just one of those "too-bloated-to-be-tuneable" pieces of software no one wants to touch since it could break something. And do not argue that big, feature richt toolkits are slow in general - Even large QT apps are snappy and very responsive whereas many large GTK-2 apps tend to be unuseable on slower systems (like my Dorun-800). 800mhz to render some buttons and layout texts?? Maybe I am just too old to understand whats going on :( A very favourite argument is that skinned toolkits are slow or X is slow, but how do other, much faster toolkits face with that. lg Clemens ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
[maemo-developers] Re: testing pango with 770
Note that many of the optimizations that went into Pango 1.11 only affect the cairo backend. I cannot remember any change that slows down pango. behdad On Wed, 21 Dec 2005, [ISO-8859-1] Tapani Pälli wrote: > Hello, > Here are test results for Pango performance on 770 : > > === > TEST : Pango performance on 770 > === > > Test programs are made by OpenedHand, all tests are measuring fullscreen > write speed. > Timestamps are taken using gettimeofday() after rendering and issuing > XSync(). > > Short description on tests (api, rendering function) : >*test-pango: xlib,pango_xft_render >*test-pango-gdk: gtk/gdk, pango_xft_render >*test-pango-gtk: gtk, pango_xft_render >*test-pango-layout : gtk, gdk_draw_layout > > Text 'abcdefghijklmnopqrstuvwxyz' is rendered to 20 rows 100 times using > 'sans serif-18' font. > (Each test is renderering 52000 glyphs.) > > using 770 macroboard, latest sw-build. > > > -test-1-- > - glib 2.6.2 > - pango 1.8.1 > > test-pango: Total time 6254 ms, approx 8314 glyphs per second > test-pango-gdk: Total time 6292 ms, approx 8264 glyphs per second > test-pango-gtk: Total time 6313 ms, approx 8236 glyphs per second > test-pango-layout : Total time 7383 ms, approx 7043 glyphs per second > test-gtk-layout : Total time 7782 ms, approx 6682 glyphs per second > > -test-2-- > - glib 2.9.1 > - pango 1.10.2 > > test-pango: Total time 6787 ms, approx 7661 glyphs per second > test-pango-gdk: Total time 6610 ms, approx 7866 glyphs per second > test-pango-gtk: Total time 6569 ms, approx 7915 glyphs per second > test-pango-layout : Total time 8038 ms, approx 6469 glyphs per second > test-gtk-layout : Total time 8593 ms, approx 6051 glyphs per second > > -test-3-- > - glib 2.9.1 > - pango 1.11.1 > > test-pango: Total time 6979 ms, approx 7450 glyphs per second > test-pango-gdk: Total time 6899 ms, approx 7537 glyphs per second > test-pango-gtk: Total time 6893 ms, approx 7543 glyphs per second > test-pango-layout : Total time 8064 ms, approx 6448 glyphs per second > test-gtk-layout : Total time 8423 ms, approx 6173 glyphs per second > > > Notes : > > * Gtk was not built with new glib, I was using the same gtk > (2:2.6.4-1.osso76) all the time. There might be some speed differences > to tests using gtk then. > > * 'test-pango' is the most accurate figure here, since it's not using gtk. > > * Packages were made using upstream tarballs and applying maemo-patches > (glib 5, pango 1) on top > > * I will happily try running other tests aswell, don't hesitate to answer > > * What is causing slowness? > > // Tapani > > ___ > gtk-i18n-list mailing list > gtk-i18n-list@gnome.org > http://mail.gnome.org/mailman/listinfo/gtk-i18n-list > > --behdad http://behdad.org/ "Commandment Three says Do Not Kill, Amendment Two says Blood Will Spill" -- Dan Bern, "New American Language" ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
[maemo-developers] Re: testing pango with 770
On Wed, 2005-12-21 at 15:15 +0200, Tapani Pälli wrote: > Text 'abcdefghijklmnopqrstuvwxyz' is rendered to 20 rows 100 times using > 'sans serif-18' font. > (Each test is renderering 52000 glyphs.) This test could be more thorough. You aren't testing word breaking or the non-latin shapers, for example. Also, your test needs to run for a longer time. Otherwise, you'll get a lot of variation among runs. Make it so that each run takes a few minutes. > -test-2-- > - glib 2.9.1 > - pango 1.10.2 > > test-pango: Total time 6787 ms, approx 7661 glyphs per second > test-pango-gdk: Total time 6610 ms, approx 7866 glyphs per second > test-pango-gtk: Total time 6569 ms, approx 7915 glyphs per second > test-pango-layout : Total time 8038 ms, approx 6469 glyphs per second > test-gtk-layout : Total time 8593 ms, approx 6051 glyphs per second > > -test-3-- > - glib 2.9.1 > - pango 1.11.1 > > test-pango: Total time 6979 ms, approx 7450 glyphs per second > test-pango-gdk: Total time 6899 ms, approx 7537 glyphs per second > test-pango-gtk: Total time 6893 ms, approx 7543 glyphs per second > test-pango-layout : Total time 8064 ms, approx 6448 glyphs per second > test-gtk-layout : Total time 8423 ms, approx 6173 glyphs per second This is interesting, since pango-1.11.1 has our recent optimizations in it. You mentioned that you were using gettimeofday(). Can you please test the "pango-profile" module from CVS? It has a slightly more sophisticated timer, using times(). I've found it to be quite reliable, and it gives reproducible results. > * What is causing slowness? Ask a profiler :) For your test, I'd like to know the CPU usage for both Pango and the X server. On my machine, a "render a lot of text" benchmark spends 50% of the time in the X server (xorg without the MMX compositing stuff), about 30% in Pango, and the rest in other processes. Sysprof can tell you that very nicely. If you use the benchmark in pango-profile, beware that it just tests the Pango machinery, not the rendering machinery. You may want to stick a gdk_draw_layout() in the code. Does Sysprof run on the 770 yet? Federico ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers
[maemo-developers] Re: testing pango with 770
ext Federico Mena Quintero wrote: >On Wed, 2005-12-21 at 15:15 +0200, Tapani Pälli wrote: > > > >>Text 'abcdefghijklmnopqrstuvwxyz' is rendered to 20 rows 100 times using >>'sans serif-18' font. >>(Each test is renderering 52000 glyphs.) >> >> > >This test could be more thorough. You aren't testing word breaking or >the non-latin shapers, for example. > > > Yes, I will think of real-life usecases also. But writing lines of text is pretty much simplest usecase there is ... I wonder why the difference is that big, I will try to find that out. >Also, your test needs to run for a longer time. Otherwise, you'll get a >lot of variation among runs. Make it so that each run takes a few >minutes. > > > This is true, I will run this again with longer time. For gtk I think it will speed up but xlib test should stay pretty neutral. >You mentioned that you were using gettimeofday(). Can you please test >the "pango-profile" module from CVS? It has a slightly more >sophisticated timer, using times(). I've found it to be quite reliable, >and it gives reproducible results. > > > I will try pango-profile, thanks! >>* What is causing slowness? >> >> > >Ask a profiler :) > >For your test, I'd like to know the CPU usage for both Pango and the X >server. On my machine, a "render a lot of text" benchmark spends 50% of >the time in the X server (xorg without the MMX compositing stuff), about >30% in Pango, and the rest in other processes. Sysprof can tell you >that very nicely. > > > Yep the whole pipeline has to be inspected, the biggest bottleneck can be elsewhere (whether the actual blitting can be optimized more that's another question). but we haven't actually had any changes with our X for awhile and I was using same X build for all tests. >If you use the benchmark in pango-profile, beware that it just tests the >Pango machinery, not the rendering machinery. You may want to stick a >gdk_draw_layout() in the code. > > > Ok, this is good to know. >Does Sysprof run on the 770 yet? > > Nope :-/ // Tapani ___ maemo-developers mailing list maemo-developers@maemo.org https://maemo.org/mailman/listinfo/maemo-developers