Re: testing pango with 770

On 12/21/05, Tapani Pälli <tapani palli nokia com> wrote:
> I understand that this is a kind of 'raw speed' testing, I am interested
> to try other tests aswell. Before profiling something we should have
> specific testcases, also note that profiling is not so trivial with arm
> as with x86. Consider this as a 'rough' figure what is the performance
> difference before going further.

I have two suggestions for testing pango performance:

For more "pure" figures on pango there is the pango-profile:

It creates a layout for translation databases for different languages
so data is varied (more real-life-like), lays the stings out but
doesn't draw them so no graphics performance is involved. Don't know
how much memory it chomps though.

For more real use case, I "happen" to know that ScwView available from:

uses heavily pango layouts for it's drawing. It also has an
unoptimized horizontal resize which lays the text out for every line.
So, put a thousand lines in a ScwView (IIRC the device choked after
1500 lines when I appended stuff indefinetly to a view ;), vary the
adjustment randomly and tell it to resize vertically to get a bunch of
laying out mixed with some drawing.

There's even a stress test included that inserts and removes stuff so
it too could be used for testing, although those are not as taxing
(*some* parts of it are at least optimized).

Kalle Vahlman, zuh iki fi
Powered by
Interesting stuff at

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]