Re: macOS metal rendering engine in mac port

[Eli Zaretskii]

> From: Aaron Jensen <aaronjensen@gmail.com>
> Date: Mon, 24 May 2021 13:21:48 -0700
> Cc: Alan Third <alan@idiocy.org>, YAMAMOTO Mitsuharu 
> <mituharu@math.s.chiba-u.ac.jp>, emacs-devel@gnu.org
> 
> The mystery to me is how your computer has optimized those extra
> calls.

No mystery: face merging is just a small fraction of what redisplay
does, so the twofold increase in face merging shouldn't slow down
redisplay too much.

> What's the result of this for you:
> 
> (let ((idx 0))
>   (catch 'nth-elt
>     (dolist (f face-new-frame-defaults)
>       (when (equal 'line-number (car f)) (throw 'nth-elt idx))
>       (setq idx (1+ idx)))
>     nil))
> 
> I get 115.

95 here.

> Also, it just occurred to me that I'm using native-compilation. Are
> you?

No.

> I just tested without native comp and on emacs -Q I got:
> 
> 7.3s w/o line numbers
> 8.2s w/ line numbers

So the effect is still significant, not what I see here.

> > What about "emacs -Q -nw": do you see a similar slowdown when you turn
> > on line numbers?
> 
> Similar, yes: 1.53s w/o line numbers and 2.12s with line numbers (this
> was with native comp).

Sigh.