Re: X performance suffers under emacs 21.1.1

[Kim F. Storm]

address@hidden (Gerd Moellmann) writes:

> Richard Stallman <address@hidden> writes:
> 
> > One idea is to record the time at which each image was last displayed,
> > and get rid of any image that has not been displayed in 10 minutes.
> > I don't know if this is really feasible in the code as it is written.
> > Gerd, what do you think?  Any advice?
> 
> Incidentally, what you describe is already implemented and exactly the
> way you proposed: clear_image_cache is called from redisplay,
> indirectly, when it clears the face cache, and clears images that
> haven't been displayed for 30 minutes.
> 
> Maybe it suffices if that time is simply made shorter.  Or maybe one
> could implement another idea: record where the image was displayed,
> and remove images from the cache when the buffer they were displayed
> in is killed (which wouldn't capture the case when a buffer is reused;
> I'm not sure what Gnus does, perhaps someone else knows.).

IMO, the image cache should have an upper bound to the number of
cached entries, and some form of functionality to clean up the cache
in case there are no (or very few) free entries.

However, this only makes sense if the display engine doesn't require
that all images to be shown for a buffer are cached.  Otherwise, there
would be cases where the cache might be too small to display all the
necessary images.

Afaics, the display engine uses an image number which is an index into
the cache for its internal representation of images to be displayed -
but it is hard for me to see for how long these indexes need to be
valid.

Gerd,
can you say something about for how long an image must stay in the
cache for the display engine to work?

++kfs