MostAwesomeDude: So I hear that fragprog compiler for r5xx is in demand. I'll work more on that, as time permits.
MostAwesomeDude: Also I finally moved over to xserver git. Took a good half-hour to get everything building, especially since I don't have working VTs.
MostAwesomeDude: Also synaptics doesn't work with xserver from git...
MostAwesomeDude: ...and I just got another no-warning X crash.
MostAwesomeDude: Wonder how many bisects there are between here and 1.3...
MostAwesomeDude: ...The answer, I guess, is "Not enough to be a problem."
MostAwesomeDude: I'm screened now, so I won't vanish again.
spreeuw: hi folks
spreeuw: current git exa has slow xterm
spreeuw: I use fontconfig fonts
spreeuw: and they scroll in a tearing screen refresh fashion
spreeuw: real 0m3.788s user 0m0.001s sys 0m0.002s
spreeuw: > time dmesg
spreeuw: thats insane ;p
arekm: so are you saying that few commits earlier git exa didn't have this issue?
spreeuw: it used to be way faster than 4seconds yes
spreeuw: a dmesg shouldnt take 4 seconds to display
spreeuw: it should be half a second or so
spreeuw: or less
arekm: so maybe do git bisect to find out bad commit
spreeuw: real 0m0.033s
spreeuw: turned of renderfonts in xterm for now
otaylor: spreeuw: Are fonts uniformly slow everywhere or just in xterm?
spreeuw: only xterm had it
spreeuw: with renderfonts on
spreeuw: xperf says still 30000
otaylor: spreeuw: for what test?
spreeuw: the one from the radeon blog
otaylor: spreeuw: which blog is this?
spreeuw: x11perf -aa10text
otaylor: And what hardware?
spreeuw: 160000 reps @ 0.0365 msec ( 27400.0/sec): Char in 80-char aa line (Charter 10)
spreeuw: X700 64bit 128MB
spreeuw: EXA git ati
otaylor: spreeuw: Eeew, that's slow
spreeuw: :((( ;)
otaylor: spreeuw: Though still, not as slow as you describe xterm
otaylor: spreeuw: Are you sure it's actually git head?
otaylor: spreeuw: that;s about the speed we were getting before the recent fixes landed
otaylor: (well, twice what I was seeing, but R420 vs. R300)
spreeuw: Date: Thu Apr 17 13:14:53 2008 +0200
spreeuw: commit c5d62fa0e8f52c3264ff9db3ff10cdf5a806bfc0
spreeuw: Author: Owen Taylor
spreeuw: my last
spreeuw: I'm not using any composite bullshit
spreeuw: does that matter?
otaylor: spreeuw: No
otaylor: spreeuw: And you are sure this is installed on your system and picked up by the X server
otaylor: sudo grep radeon_drv.so /proc/`/sbin/pidof Xorg`/maps .... then do a ls -l of the .so it is using
otaylor: And see if the timestamp is what you expect
spreeuw: b7b79000-b7bdb000 r-xp 00000000 03:02 113639444 /usr/local/lib/xorg/modules/drivers/radeon_drv.so
spreeuw: b7bdb000-b7be2000 rw-p 00062000 03:02 113639444 /usr/local/lib/xorg/modules/drivers/radeon_drv.so
spreeuw: -rwxr-xr-x 1 root root 464K 2008-04-19 10:03 /usr/local/lib/xorg/modules/drivers/radeon_drv.so
spreeuw: the first is my disabled northbridge
spreeuw: it has an r300 there too
spreeuw: which I havent tried in ages
spreeuw: X200 express thing
spreeuw: maybe I have to update my xserver?
otaylor: spreeuw: actually, they are just different maps of the same library, both presumabyl for your enbaled card ... note the r-xp vs. rw-p
spreeuw: oh, ok
otaylor: spreeuw: How old is your xserver ?
spreeuw: X.Org X Server 1.4.0
spreeuw: Release Date: 5 September 2007
otaylor: spreeuw: Oh, that's not going to be speedy at all :-)
spreeuw: ah,I'll chase down another then, thanks
spreeuw: saw there were some rc's
otaylor: spreeuw: Bascially you need the glyph-pixmaps work that landed shortly after 1.4.0 came out
spreeuw: oh nice
spreeuw: 188.8.131.521 is last rc?
otaylor: spreeuw: I'd go for git rather than a tarball if you are going to be building by hand anyways
spreeuw: this may just solve my web page crasherbug too
spreeuw: hmm this feels zippy indeed
spreeuw: and 1280x960 works again
spreeuw: only worked with xrandr before
spreeuw: Error: couldn't find RGB GLX visual
spreeuw: the rest looks normal
spreeuw: any obvious thing I still need to upgrade?
spreeuw: xorg log has everything loaded without error
otaylor: 5600000 reps @ 0.0010 msec (1020000.0/sec): Char in 80-char aa line (Charter 10) ... now if only about a quarter of my text wasn't mysteriously disappearing, I'd be really happy with that
Magnade: otaylor: just make the middle 1/4 of words disappear and your good ;P
tilman: any idea why i cannot access the screen pixmap in exaDriverInit()?
tilman: this results in a segfault, even though fbScreenInit should have been called already
tilman: that patch works on intel fwiw
otaylor: 4800000 reps @ 0.0012 msec (848000.0/sec): Char in 80-char aa line (Charter 10) ... OK it's a little slower now that it's drawing all the characters, but I'm not going to complain :-)
tilman: to be specific, the devPrivates on the screen pixmap are messed up (*privates is 0x18, not quite a valid pointer ;p)
otaylor: tilman: sorry, no immediate idea
tilman: too bad
tilman: on radeon, the CreateScreenResouce patch works (the one i sent to the list), but it breaks on intel
tilman: vice versa with the patch i pasted above
otaylor: have you tried putting some logging where the screen pixmap is created?
Magnade: otaylor: what were you getting on that bench before?
otaylor: Magnade: about 130k
otaylor: (and before that, about 19k)
otaylor: Magnade: I can't actually think what needs 850,000 characters drawn a second, but at least it removes it as a bottleneck
tilman: the screen pixmap is created in miCreateScreenResources
leio: 4000000 trep @ 0.0068 msec (147000.0/sec): Char in 80-char aa line (Charter 10)