Re: penalty for double buffering

[Steven Jenkins <steven jenkins ieee org>]

Owen Taylor wrote:
> Well, glancing at the code briefly, certainly there is no reason
> to use double buffering if you are already saving using a backing
> pixmap, as you are in your code.

OK, good. That answers a question and solves a problem.

> But I wouldn't expect any significant performance impact assuming
> that you have sufficient video memory. The extra copy inside the
> video card will basically be free.
>
> If you want to investigate further what is going on, I'd suggest
> that the first step would be to create a small test case independent
> of your application that does essentially the same thing.

I may try that over the holidays. My motivation is diminished now that
the problem is solved :-).

> If I had to make a random guess, my guess would be that there is
> some pathology lurking in the way you are handling refreshes by
> recursing the main loop (which is... err... odd) and that small
> differences in timing are resulting in big differences in the
> amount of drawing you are doing.

Just to clarify: there's no recursion. There is a top-level loop that
interleaves input/output processing, drawing and
gdk_window_invalidate_rect() calls, and gtk_main_iteration() calls.
GTK+-based graphical output is a compile-time option, so I can't use 
gtk_main() at the top level.

The test case is playback of recorded data from a file. The timing is 
deterministic and the area exposed is identical for every input data point.

Steve