[mcclim-devel] Tablature editor: incremental redisplay performance

[Christophe Rhodes]

Hi,

As some people here know, we've been working on a CLIM- and
Climacs-based editor for lute tablature.  We're at the stage now where
the performance characteristics are getting beyond our abilities to
manage, so we thought we would try to open up the problem to
specialists.

The issue seems to be that (with the patch I sent a week or so ago, to
use the bounding-rectangle of a region union rather than the precise
union itself), essentially all of the time spent in moving around a
Tabcode editor buffer is spent in COMPUTE-DIFFERENCE SET.  (Without
the patch I sent, essentially all the time is spent in computing
region unions: and it takes about 100 times as long doing that as it
does computing a difference set...).  Holding down a cursor key, on
one-year-old hardware, leads to perceptible lag and continued cursor
motion after the key is released.

We'd be happy to provide more details, or to explain other things
about our application and our aims for it; any advice or suggestions
for improvements to redisplay performance would be greatly
appreciated.

The code itself is at
<http://www.doc.gold.ac.uk/~mas01cr/tabcode-20050708.tar.gz>; papers
discussing the application are available from
<http://www.doc.gold.ac.uk/~mas01cr/papers/>.

Any contributions and observations welcome (though please try not to
laugh too much :-)

Cheers,

Christophe