Re: request for numbers and graphs for Release Notes

[Travis Reitter <treitter-dev netdrain com>]

I re-ran my gnome-terminal "large text file output" benchmark with vte
0.11.8 and gnome-terminal 2.13.91.

This new benchmark is on the same computer as before (1.8 GHz Pentium-M
laptop), but running Ubuntu Dapper Drake.

It's interesting to note that while the average (and even the worst
case) for the newest gnome-terminal and vte were a bit lower, there was
quite a range even on an otherwise-unloaded system. Lowest was 11.08
seconds to output the (cached) file, highest was 17.82 seconds. I'm not
sure why the results were so inconsistent, but it's certainly better
overall.

Including the benchmarks from before (I've confirmed that uxterm and
xterm perform the same as they did last time):

Gnome-terminal (GNOME 2.12.0)
  real  0m18.419s
  user  0m0.000s
  sys   0m0.191s

Gnome-terminal (GNOME 2.13.91)
  real  0m14.02s
  user  0m0.01s
  sys   0m0.21s

Uxterm
  real  3m56.133s
  user  0m0.003s
  sys   0m0.139s

Xterm
  real  0m20.574s
  user  0m0.001s
  sys   0m0.144s

In all cases, the file has been read immediately before benchmarks, so
any level of caching should be even. Also, the processor was running at
full clock speed (1.8 GHz) during all tests, and few other programs were
running (none intensely).

For more details, refer to the more thorough profiling document:

http://www.netdrain.com/performance/gnome-term-prof/

Nice job, Behdad and everyone else :)

-Travis


On Tue, 2006-02-14 at 11:11 +0800, Davyd Madeley wrote:
> Due to the number of fine speedups in GNOME 2.14, it would be really
> nice to put some attractive graphs in the release notes about things
> that are faster.
> 
> Did somebody want to get my some hard numbers (and perhaps even some
> graphs) that would be suitable for the release notes. Things that I was
> thinking of include:
>  - Pango rendering;
>  - vte; and
>  - g_new vs GMemChunk vs GSlice (for the developer section)
> 
> Please CC replies, not on list.
> 
> Thanks all,
> 
> --d
>