Re: Early Gnome-terminal profiling

[John Rice <John Rice Sun COM>]

Interesting stuff Travis.

Would be useful to also put in a few tests on some 'real world' usage 
patterns of the terminal. For instance:
ls of /usr/bin
cat of a 1-2k file

With the ls /usr/bin, xterm seems to outperform gnome-terminal by an 
order of x2 on my test box [Acer Ferrari 3400 running Gnome 2.12 with 
Solaris nevada b22, 1 Gig Ram] [I was testing on 80x24 terminal, with 
matching fonts; GT: Bitstream Vera Sans Mono, 10; xterm: Bitstream Vera 
Sans Mono, 12 (changed using xfontsel)]

ls /usr/bin
xterm: 0.04 sec
gt: 0.08 sec

Note: ls /usr/bin | wc -l = 881

This is what I would have expected as the xterm is using non AA fonts 
while GT is using AA fonts, so has a lot more rendering work to do 
[confirm using xmag].

When I ran your test on the large_text.txt file I got these results:

GT:
real    0m14.525s
user    0m0.001s
sys     0m0.030s

xterm:
real    0m11.447s
user    0m0.001s
sys     0m0.031s

So interesting variation between the two systems.

I did notice the hotspot you mentioned with vte_terminal_find_charcell() 
[resulting in some 300k calls to lookup_type_check triggered in the 
g_return macro in this function - now sorted with Kjartan's patch].

JR

Travis Reitter wrote:

> I've done a little profiling of gnome-terminal, uxterm, and xterm, to
> see how they stack up to each other in a few different tests.
>
> Here's the rough draft (please don't link to it - the final release will
> be a blog post, once I've gotten a little review, and polished it up a
> little): http://www.netdrain.com/performance/gnome-term-prof/index.html
>
> I'd appreciate any feedback and suggestions (especially corrections).
>
> (I'll also clean up the formatting and give a little more summary
> information in the final post).
>
> -Travis
> _______________________________________________
> Performance-list mailing list
> Performance-list gnome org
> http://mail.gnome.org/mailman/listinfo/performance-list
>