After upgrading to newly unmasked x11-base/xorg-server-1.5.3-r5 (plus all dependencies), text rendering on konsole became noticably slow. Nothing else was changed in the meantime. Reproducible: Always Steps to Reproduce: 1. Upgrade xorg-server to 1.5.3-r5 with all dependencies 2. start X and open a fullscreen konsole 3. start midnight commander and hold ctrl-o for a while - you see how the screen is painted Actual Results: Konsole contents are painted as they were sent over 19.2kbps serial line. Seriously. Expected Results: Screen blinks rapidly like a strobo lamp (this is how it worked with x11-base/xorg-server-1.5.3-r5, more or less). I can't provide you with any metrics - I have no idea how to measure it. I could do some profiling, but would need help with that. I am using terminus font (bitmap), but it looks silimar with other fonts. I didn't notice this problem with other KDE apps. Xterm and rxvt also work just fine. I had the same problem with ~version of xorg-server 1.5.3, which I tried in the past (Switched back to 1.3 because of that).
Created attachment 187476 [details] emerge_info.txt 'emerge --info' output
Created attachment 187478 [details] Xorg.0.log Last Xorg startup log. I am using amd/ati fglrx driver.
Created attachment 187480 [details] xorg-upgrade.txt Packages pulled in with the upgrade
I've just tried x11-drivers/xf86-video-ati-6.12.1-r1 - results are the same as with fglrx driver.
$ konsole --version Qt: 3.3.8b KDE: 3.5.10 Konsole: 1.6.6
Additional information available on this Forum thread: http://forums.gentoo.org/viewtopic-t-706046-start-0-postdays-0-postorder-asc-highlight-xorg+slow+font.html Many thanks for all your help!
I am adding that, as i mention here: http://forums.gentoo.org/viewtopic-p-5632373.html#5632373 the problem seems to be realated to anti-aliasing, as it disappears when anti-aliasing is disabled.
I just tested various fonts and font settings. Conclusions: 1) Bitmap fonts (Terminus, Terminal, Fixed) are slow. Scalable mono fonts (Bitstream Vera, DejaVu, Courier) are as fast as bitmap fonts were before the upgrade (just a visual comparison). 2) It doesn't matter if I run konsole with --noxft or set antialiasing to on/off system-wide - results from point 1) are still the same. I also: a) Did kernel upgrade to sys-kernel/gentoo-sources-2.6.29-r1 just in case. No luck. b) Downgraded libXft, libXrender and libX11 to pre-xorg-1.5 version. No luck. c) Run a small test: script showing two different, color blocks of text one after another in a loop. Timings show that on xorg 1.5 it takes 75% more wall time to execute. But if I run it with callgrind it shows no apparent difference in costs (why?!). Most of the time is spent in TEWidget::setImage (but that would be rather obvious).
One more test: compiled (by hand) qt3 with -thread -no-xft -no-xrandr -no-xinerama configure options and run konsole with this crafted qt preloaded. Works fine (terminal updates are fast). This would suggest that problem is with Xft, but in Comment #8 point b) I've tried to downgrade libXft already and it didn't help, so it's not libXft alone (or not libXft at all :-)). Anyway, this workaround looks similar to the orionbelt's one in Comment #7, although I really don't know why simple switching antialiasing off didn't help in my case.
Maybe this will be of some use: http://lists.freedesktop.org/pipermail/xorg/2009-April/045061.html "If Gentoo hasn't backported xserver commit f07f18231a921d3ae9dd9b75881c9e58e9e2e235 ('EXA: Allow using exaCompositeRects also when we can't use a mask in exaGlyphs.') yet, that could be worth a shot. (Note that 265d20068af5434489752b6dba0bf0065b3cc3ec is also required to fix a little bug in that)" -- Michel Dänzer
This may be worth nothing here too: I was testing something completly different and replaced my Radeon 9600 with GeForce FX 5900XT for a while (both are considered to be similar in GPU power). Gues what. There was no (visible) issue with font rendering.
For me this seems to be fixed as of lately. Last relevant update was to qt-* 4.5.1. I am using a nvidia card, kde 4.2.2 and konsole has been slow since I tried the 4.x brach of KDE the first time (all ~amd64). I'm using the terminus font and considered konsole being slow a _major_ annoyance. x11-base/xorg-server-1.5.3-r5. Do you need any other version numbers? Just wanted to give positive feedback, don't know what fixed it, but I'm glad it's usable again. Regards, Jan
I don't think any recent update fixed this for kde 3.5.10 konsole on qt3. But I did run some test on KDE 4.2.2 + QT 4.5.1 too. Here are my observations (although this is probably different issue than the original one): 1) With compositioning disabled in xorg.conf - konsole works flawlesly. 2) With compositioning enabled in xorg.conf but disabled in KDE - konsole updates faster than in kde 3.5.10, but still jumpy 3) With compositioning enabled in xorg.conf and enabled in KDE - same as 2) 4) With compositioning disabled in xorg.conf and enabled in... wait, it doesn't make any sense, does it? ;-) This is all on GeForce 6600. I guess we will have to leave compositioning for demonstration purposes only, at least for now. ;-) I wonder if fix mentioned in Comment #10 helps. Anybody tried it? Or do we already have it with xorg-server-1.5.3-r5?
You guys _might_ want to put the X11 herd into the loop if you want xorg-server patches... FWIW, 1.5 will stay as-is, I already put 70+ patches on it, I'm not going to add more. 1.6, on the other hand, is still not frozen upstream. I'm trying to get those patches ported to it, but don't hold your breath as those patches are not easy to backport. Cheers
Hello, I'm facing this problem as well. Do you know if there's any update about it? Also, I am having trouble regarding the fonts in Konsole: I'm using DejaVu Sans Mono, size 8, and the options "Regular" and "Bold" show exactly the same font. I will probably open a bug for it tomorrow, unless some maintainer ask me not to do it. Thank you.
X.org 7.4 (>= xorg-server-1.6) has been stable for while now. Is this still an issue with up-to-date system?
I just checked this again, and i confirm that problem is still there for me, as i described it above. I am using x11-base/xorg-server-1.6.3.901-r2 .
Could anyone provide an up-to-date Xorg.0.log after the upgrade to xorg-server 1.6? Thanks
Created attachment 206832 [details] Xorg.0.log for x11-base/xorg-server-1.6.3.901-r2 ..and my font in the konsole is terminus, too. Will my emerge --info be useful?
@Tomăś, Looks like DRI2 isn't set up here, could you take a look? Thanks
Created attachment 206840 [details] Xorg.0.log for x11-base/xorg-server-1.6.3.901-r2
And you're using nvidia-drivers... Erf... Could you guys make sure that you have USE=sse2 enabled on pixman? Could you guys try some of the EXA options described in 'man exa'? Please do try them only one at a time. If nothing works, we'll have to ping both X and KDE upstream devs to see if they've heard about this before. Thanks
[I] x11-libs/pixman Available versions: 0.10.0 0.12.0 0.14.0-r1 0.16.0 ~0.16.2 {altivec debug mmx sse sse2} Installed versions: 0.16.0(11:16:57 01.10.2009)(mmx sse2 -altivec -debug)
(In reply to comment #22) > And you're using nvidia-drivers... Erf... It's not by choice. I was using nv, but flash inside Firefox would bring the browser to a grinding halt until i switched to nvidia. I could hunt down the bug report, if you want. > Could you guys make sure that you have USE=sse2 enabled on pixman? % equery uses pixman - - altivec : Adds support for optimizations for G4 and G5/ppc970 processors - - debug : Enable extra debug codepaths, like asserts and extra output. If you want to get meaningful backtraces see http://www.gentoo.org/proj/en/qa/backtraces.xml + + mmx : Adds support for optimizations for Pentium MMX and Athlon class processors + + sse2 : faster floating point optimization for SSE2 capable chips > Could you guys try some of the EXA options described in 'man exa'? Please do > try them only one at a time. I will try it next time i log off my X session. > If nothing works, we'll have to ping both X and KDE upstream devs to see if > they've heard about this before. Thanks for your time!
I am using radeon driver from x11-drivers/xf86-video-ati-6.12.4, not nvidia-drivers (I suffer an aversion to nonfrees ;) Testing these EXA options will take some time. I should be able to have the results till tomorrow.
Ok, I have tested the EXA options. I have run the following command: sleep 1; time for x in {1..50} do; dmesg; done For comparision purposes I have run this command also in an inactive konsole tab (I switched to another tab, so the screen was not updated). The results were: inactive tab: 2 seconds Option "AccelMethod" "XAA" : 16 seconds Option "AccelMethod" "EXA" : 10 seconds EXA with various *Disable options turned on: ~13 seconds I have tested all the combinations of the options from "man exa", but the differences were minimal. The interesting thing is that in (freshly emerged) KDE 4.3.1 this test runs 2 seconds with actually drawing the text, and 0.7 of a second in inactive tab.
I'm running out of ideas... sorry. Please _each_ file a bug in FreeDesktop's bugzilla (since you are using different drivers) and paste the url here. I'll CC myself on both bugs to keep track of them. Thanks
A data point: I recently upgraded to KDE4 (kde-base/kdelibs-4.3.1-r2) and the bug disappeared...