Re: Motif support

[Stefan Monnier]

> The question that interests me is which part(s) of the Emacs display
> code you'd like to accelerate using this hardware.

Indeed, and maybe for that a first step would be to try and get actual
numbers about where the time is spent.

My assumption has always been that the times when redisplay isn't fast
aren't related to the complexity of actually drawing on the glass, but
instead it's due to the time spent updating the glyph matrices (and
other things before that such as running jit-lock).

How this could take advantage of GPUs is very much unclear to me.
Instead, I tried to speed this up by reducing the number of cases where
we just re-render all the windows.

Then again, maybe my assumption here is completely wrong.

My impression is that would be "easier" to speed things up by making use
of multiple CPUs rather than making use of GPUs.  True concurrency when
running ELisp is hard, but there are other ways to make use of multiple
CPUs:

- make the GC parallel.
- make the GC concurrent.
- use a separate thread per window-redisplay (so when redisplaying 20
  windows we can use 20 CPUs to update their glyph matrices).
- ...

I don't consider these examples as easy, but at least I have
a fairly clear idea of what it could look like.


        Stefan