Re: macOS metal rendering engine in mac port

[Aaron Jensen <aaronjensen@gmail.com>]

On Tue, May 25, 2021 at 11:26 PM Aaron Jensen <aaronjensen@gmail.com> wrote:
>
> On Tue, May 25, 2021 at 11:23 PM Alan Third <alan@idiocy.org> wrote:
> >
> > It probably could, I've seen something similar when I was testing
> > something or other. It will be bypassing the viewWillDraw code that
> > forces a redisplay, but I think in theory we shouldn't need it with
> > this code.
> >
> > Oh, I suppose that might break the live resizing. But we can work that
> > out if so.
>
> Live resizing? I don't see the flicker and if I drag to resize it
> remains painted usually. I did see it flash white once. And I think
> I've perhaps seen it blank one time today briefly, but I don't
> remember what I was doing.

Just did some latency measurements with emacs -Q (my previous ones
were w/ my config)

emacs27-drawing branch: 79.2 62.5 75.0 70.4 (71.775 avg)
emacs28 surface-stuff branch: 66.7 66.7 75.0 70.8 (69.8 avg)
emacs28 surface-stuff branch+last patch: 70.8 75.0 66.7 62.5 (68.75 avg)

Those are similar, if not slightly better than iterm2 latency on my
machine, so that's probably pretty good. The margin of error is going
to be pretty high on those, so I can't definitely say that 28 is
faster.

The scroll-up-benchmark w/o the patch is slightly better from my
limited testing (4.9s vs 5.2s).

Aaron