Xfce 4.4: The best lightweight desktop environment (Linux.com)

Posted Feb 22, 2007 8:02 UTC (Thu) by drag (subscriber, #31333) [Link]

I am using PowerPC 1.2ghz G4 with ATI 9200 DRI/XAA drivers.

The
time python -c "print 'a'*10000000"
microbenchmark

With Gnome-terminal:
real 0m19.088s
user 0m0.059s
sys 0m0.382s

With Xterm:
real 0m24.315s
user 0m0.052s
sys 0m0.364

I tried the tables thing by sticking:
xterm*multiScroll: on

in .Xresources file. But it didn't seem to make any difference.

Now with remote X it was pretty bizzare.

This is going over ssh on a internet connection with a 45KB/s connection between both hosts.
ssh -X remote-host gnome-terminal

I ran:
time python -c "print 'a'*10000000"
real 0m5.861s
user 0m0.016s
sys 0m0.068s

!!! that's bizzare.

Then I tried my first microbenchmark which was run find on my home directory
find ./

First time I ran it I got 1minute 23 seconds. But second time I ran it I got:
real 0m28.653s
user 0m0.584s
sys 0m1.480s

And the 3rd time was exactly the same.

Just so you know...
time find ./ |wc
133705 151271 10996476

real 0m19.248s
user 0m1.444s
sys 0m0.812s

(which I suppose is why the python command is much better, eh?)

Now with:

user 0m0.052s
sys 0m0.364

I tried the tables thing by sticking:
xterm*multiScroll: on

in .Xresources file. But it didn't seem to make any difference.

Now with remote X it was pretty bizzare.

This is going over ssh on a internet connection with a 45KB/s connection between both hosts.
ssh -X remote-host gnome-terminal

I ran:
time python -c "print 'a'*10000000"
real 0m5.861s
user 0m0.016s
sys 0m0.068s

!!! that's bizzare.

Then I tried my first microbenchmark which was run find on my home directory
find ./

First time I ran it I got 1minute 23 seconds. But second time I ran it I got:
real 0m28.653s
user 0m0.584s
sys 0m1.480s

And the 3rd time was exactly the same.

Just so you know...
time find ./ |wc
133705 151271 10996476

real 0m19.248s
user 0m1.444s
sys 0m0.812s

(which I suppose is why the python command is much better, eh?)

Now with:
ssh -X remote-host xterm

real 9m3.364s
user 0m0.020s
sys 0m0.048

!!!

That's over 9 minutes for xterm versus 5 _seconds_ for gnome-terminal.

And I was watching the bandwidth usage. It got a solid 42-46 KB/s the entire time. There was only a disconnect for about 6 seconds.

time find ./
real 14m37.836s
user 0m0.464s
sys 0m1.100s

That's 15 minutes for Xterm vs 25 seconds for Gnome-terminal.

That's bizzare. The Gnome folks are nailing some sort of optimization in their libraries for this. Are they incorporating NX optimizations or something?

I am sorry, but this is hilarious. Completely unexpected.
Comparing Gnome-terminal vs Xterm over a slow network connection is like watch Gozilla vs Methuselah (in his later days) in a no-hold-barred cage fight.

To be fair I tried
ssh -XC remote-host xterm

The python command gave me this:
real 0m22.725s
user 0m0.020s
sys 0m0.088s

So compression helps a lot.

The find ./ game me this
real 1m30.950s
user 0m0.464s
sys 0m1.140s

So I tried it with gnome-terminal and compression:

The python command:
real 0m5.912s
user 0m0.016s
sys 0m0.088s

the find command:
real 0m24.609s
user 0m0.512s
sys 0m1.460s

So not that much of a difference, for gnome-teminal.

So I figure I don't want any disk I/O limits to limit performance.
So I mounted tmpfs to /mnt
and went:
find ./ > /mnt/cow

That created a 11meg file.

So with:
ssh -XC remote-host gnome-terminal

Then:
time cat /mnt/cow
real 0m7.314s
user 0m0.000s
sys 0m0.168s

With
ssh -XC remote-host xterm
real 1m18.927s
user 0m0.008s
sys 0m0.128s

Then for comparision the amount of time it takes to transfer the file.

time scp -C remote-host:/mnt/cow ./
cow 100% 10MB 488.1KB/s 00:22

real 0m27.538s
user 0m0.483s
sys 0m0.265s

Keep in mind that there was about 3-4 seconds before the command launched.

So Gnome-terminal is cheating. Cheating heavily. I don't dissaprove.

Is there a better benchmark? Or is this going to reflect real-world performance? I don't have any problem using either over the connection.