Re: macOS metal rendering engine in mac port

[Aaron Jensen]

On Sat, May 29, 2021 at 2:40 PM Aaron Jensen <aaronjensen@gmail.com> wrote:
>
> Here's another benchmark:
>
> (defun type-benchmark ()
>        (interactive)
>        (let ((oldgc gcs-done)
>              (oldtime (float-time)))
>          (cl-loop for x from 1 to 300 do
>                   (insert-char ?j) (redisplay))
>          (message "GCs: %d Elapsed time: %f seconds"
>                     (- gcs-done oldgc) (- (float-time) oldtime))))
>
> All with Surface stuff branch and -O2.
>
> With Native Comp, Hash, and line numbers: 1630ms
> With Native Comp, Hash, and no line numbers: 874ms
>
> With No Native Comp, Hash, and line numbers: 1712ms
> With No Native Comp, Hash, and no line numbers: 900ms
>
> With No Native Comp, alist, and line numbers: 2368ms
> With No Native Comp, alist, and no line numbers: 1062ms
>
> With Native Comp, alist, and line numbers: 3321ms
> With Native Comp, alist, and no line numbers: 1333ms
>
>
> So according to this, native comp is slower for me in this typing test
> but only if alists are used? That's surprising right?

Okay, thankfully I can no longer reproduce the difference between
native and not with alists. I must have had something else built that
I ran that on. No idea what. But I built both fresh and they're both
similar to the native comp numbers I posted. So, please disregard the
"No Native Comp, alist" numbers. Native comp doesn't appear to make
much of a difference whether or not I'm using hash or alist.