Thanks for the benchmarking function. Here are the results I get running it three times on a fresh start of Emacs on src/xdisp.c:
with smart kerning:
GCs: 2 Elapsed time: 14.546864 seconds
GCs: 1 Elapsed time: 9.899920 seconds
GCs: 1 Elapsed time: 9.880759 seconds
without smart kerning
GCs: 2 Elapsed time: 12.188979 seconds
GCs: 0 Elapsed time: 7.264108 seconds
GCs: 1 Elapsed time: 7.346031 seconds
So it’s about a third slower. The difference is noticable.
> Btw, why aren't digit characters 0-9 in the regexp you use in sm-safe-table? More generally, why not all of the ASCII characters are there? isn't smart kerning performed for them as well? Come to think about that, what about Latin non-ASCII characters like á etc.?
The font I’m using, Commit Mono, only does smart kerning on a limited set of characters:
 | |
On Mar 15, 2024, at 2:46 AM, Eli Zaretskii <eliz@gnu.org> wrote:From: Stephen Molitor <stephen.molitor@icloud.com>
Date: Thu, 14 Mar 2024 17:10:37 -0500
Not that I’ve noticed.
I mean maybe it does, but I haven’t noticed. I have a fast machine. Is there an easy way to measure?
Yes, of course. Here's one example:
(defun scroll-up-benchmark ()
(interactive)
(let ((oldgc gcs-done)
(oldtime (float-time)))
(condition-case nil (while t (scroll-up) (redisplay))
(error (message "GCs: %d Elapsed time: %f seconds"
(- gcs-done oldgc) (- (float-time) oldtime))))))
Run this on a large file (I use src/xdisp.c) with and without your
set-char-table-range customization, and compare the results.
Btw, why aren't digit characters 0-9 in the regexp you use in
sm-safe-table? More generally, why not all of the ASCII characters
are there? isn't smart kerning performed for them as well? Come to
think about that, what about Latin non-ASCII characters like á etc.?