After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.
Bug 131461 - gnome-terminal is slow!
gnome-terminal is slow!
Status: RESOLVED DUPLICATE of bug 137864
Product: gnome-terminal
Classification: Core
Component: general
2.4.x
Other Windows
: Normal normal
: ---
Assigned To: GNOME Terminal Maintainers
GNOME Terminal Maintainers
Depends on: 137864
Blocks:
 
 
Reported: 2004-01-14 16:56 UTC by Cristian Adam
Modified: 2004-12-22 21:47 UTC
See Also:
GNOME target: ---
GNOME version: 2.3/2.4



Description Cristian Adam 2004-01-14 16:56:29 UTC
I have run ls -lR on my home directory with xterm, konsole and
gnome-terminal (maximized window and without dircolors), here are my results:

konsole
  first  run: 0m7.442s
  second run: 0m8.538s
  third  run: 0m7.738s
  --------------------
  average   : 0m7.905s
 
xterm
  first  run: 0m6.547s
  second run: 0m5.423s
  thrid  run: 0m5.385s
  --------------------
  average   : 0m5.784s
 
gnome-terminal
  first  run: 0m35.169s
  second run: 0m37.131s
  third  run: 0m37.086s
  ---------------------
  average   : 0m36.461 

The conclusion
xterm ........... 100 %
konsole ......... 136 %
gnome-terminal .. 630 % 

The results speak for themselfs! The test was run on a Slackware 9.1 with
Dropline GNOME 2.4.1.
Comment 1 Mariano Suárez-Alvarez 2004-01-23 23:48:14 UTC
*** Bug 132285 has been marked as a duplicate of this bug. ***
Comment 2 Mariano Suárez-Alvarez 2004-01-23 23:48:51 UTC
Changing the summary so that we can use this as a general “g-t is
slow” bug
Comment 3 Gabor Farkas 2004-02-05 08:36:08 UTC
hi. i'll add my "benchmarks" also:

gentoo-linux, gnome-hmm...newest devel packages (around 2.5.3?),
gnome-terminal-2.5.1
vte-0.11.10

laptop, centrino-1.4ghz, 512MB ram, ati-9000m graphics card.

there are 2 possible drivers: a binary one from ati, and the "radeon"
driver from xfree.

i use 1400x1050, the font is fixed-misc.

as a benchmark i used the "time cat mplayer.c" command 
(mplayer.c has around 3700 lines), 
and repeated it 3-5 times, tohave the file cached in memory. then i
calculated the average.
 
when i use the ati-binary drivers:
konsole: 0.32 seconds
gnome-term: 1.7seconds

when i use the xfree ati drivers:
konsole: 0.2 seconds
gnome-term: 0.4 seconds

(enabling/disabling transparency doesn't change anything)

so, there are 2 "problems":
1. the ati-binary drivers are slower, 
2. gnome-term is ALWAYS slower than konsole
Comment 4 Gabor Farkas 2004-02-05 08:44:25 UTC
so it would be good to know on what hardware+drivers did drac@gmx.net
(Cristian Adam) make his tests (nvidia binary drivers? ati drivers?
what graphics card?). 

(btw. on my desktop computer (nvidia-binary drivers), i also get
terrible gnome-term performance, while konsole performance seems to be ok)
Comment 5 Cristian Adam 2004-02-06 00:16:36 UTC
I've rerun the tests (this time without konsole, because I've removed
KDE from my system). 

My configuration: Intel Celeron 1000MHz, Gnome Terminal 2.4.2, NVIDIA
GeForce 2 MX 400 with NVIDIA 5336 binary drivers and 1152x864 resolution.

1. Gnome Terminal (with dircolors)
time ls -lR on my home
1m02.105s
0m59.148s
1m00.783s

2. XTerm (with dircolors) and the following configuration (~/.XResources)
xterm*background : black
xterm*foreground: white
xterm*faceName : Andale Mono
xterm*faceSize : 10

                                                                     
                              to have the same true type font that
Gnome Terminal uses
0m11.251s
0m11.545s
0m10.195s

Xterm looks the same as gnome terminal and is 6 times faster.

(BTW on a real 80x25 console, ls -lR is executed in 2 seconds. but
it's not 142x45 like the xterm and gnome-terminal)
Comment 6 Kjartan Maraas 2004-02-06 17:41:10 UTC
I experimented a bit with this now and I can see that making changes
to .XResources does nothing on my fedora core 1 system. I changed
gnome-terminal to use the same font as xterm - "fixed 10 pt" and now I
see xterm taking between 0.7 - 1.4 secons to do ls -l /usr/lib.
gnome-terminal does the same at a consistent 1.6 seconds. This makes
it seem to me that the difference isn't that huge.

I've got a 1.2 MHz PIII laptop using the XFree radeon driver.
Comment 7 Cristian Adam 2004-02-06 19:42:58 UTC
There was a typo there the file is not .XResources but .Xresources
Comment 8 Kjartan Maraas 2004-02-06 20:57:10 UTC
Actually, that's what I used out of habit, but after not seeing any
change I tried .XResources too, with no luck of course. :-?
Comment 9 Cristian Adam 2004-02-06 22:19:53 UTC
Now I've completed the tests:

1. xfree86 drivers (nv)

  gnome-terminal       xterm
  0m48.787s            0m44.556s 

2. nvidia drivers (nvidia with "RenderAccel" "off")
  
  gnome-terminal       xterm
  1m6.616s             0m43.969s

3. nvidia drivers (nvidia with "RenderAccel" "on")

  gnome-terminal       xterm
  0m56.487s            0m9.997s

Option "RenderAccel" "boolean"
         Enable or disable hardware acceleration of the RENDER
         extension.  THIS OPTION IS EXPERIMENTAL.  ENABLE IT AT YOUR
         OWN RISK.  There is no correctness test suite for the
         RENDER extension so NVIDIA can not verify that RENDER
         acceleration works correctly.

The conclusion is:
gnome-terminal doesn't use the hardware acceleration of the RENDER
extension, while xterm and konsole do.
Comment 10 Duraid Madina 2004-05-01 12:00:27 UTC
Added dependency on #137864 - take a look for some more detailed profiling.
Comment 11 Vladimir Vukicevic 2004-05-02 18:55:20 UTC
More benchmarks... the test machine was a 1.3ghz centrino laptop with a radeon
mobility 7500, with the stock xorg drivers.  The test was a simple

time perl -e 'foreach $i (1 .. 10000) { print "This is line $i This is line $i T
his is line $i\n"; }'

repeated a few times, picking an average-looking run from each set.  I don't
trust the cpu % numbers in the time output though; top reported far different
percentages.  The X load was gleaned from just watching top on a separate
computer via ssh.  Not a very scientific test, but interesting... a
gnome-rxvt-unicode looks interesting :)  Terminal emulators tried were g-t,
xterm, original rxvt, and rxvt-unicode (adds unicode and Xft support).

linux: (xorg ati drivers, g-t 2.6.0)
g-t (vera sans mono): perl -e   0.06s user 0.03s system 1% cpu 7.893 total; 65%
X load
rxvt-unicode (vera sans mono, xft): perl -e   0.06s user 0.04s system 1% cpu 9.2
08 total; 90% X load

g-t (fixed): perl -e   0.09s user 0.09s system 2% cpu 6.517 total; 25% X load
xterm (fixed): perl -e   0.10s user 0.11s system 4% cpu 4.675 total; 30% X load
rxvt (fixed): perl -e   0.10s user 0.08s system 29% cpu 0.600 total; negligible
X load
rxvt-unicode (fixed): perl -e   0.07s user 0.07s system 24% cpu 0.583 total; neg
ligible X load

Comment 12 Mariano Suárez-Alvarez 2004-05-07 11:17:40 UTC

*** This bug has been marked as a duplicate of 137864 ***