bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

When I press ‘C-x C-=’ to temporarily increase font size Emacs is
getting considerably slower. For example, if I hold ‘C-n’ to move
several lines down the movement is not smooth. I don't think that
increasing of font size should normally result in such “effects”.



In GNU Emacs 25.0.50.19 (x86_64-unknown-linux-gnu, GTK+ Version 3.16.6)
 of 2015-08-31 on arch
Repository revision: 62661fe759dfc826f31b032e49f0f15b40a207d5
Windowing system distributor `The X.Org Foundation', version 11.0.11702000
Configured features:
XPM JPEG TIFF GIF PNG RSVG IMAGEMAGICK SOUND GPM DBUS GCONF GSETTINGS
NOTIFY ACL GNUTLS LIBXML2 FREETYPE M17N_FLT LIBOTF XFT ZLIB
TOOLKIT_SCROLL_BARS GTK3 X11

Important settings:
  value of $LANG: en_US.UTF-8
  locale-coding-system: utf-8-unix

Major mode: M

Minor modes in effect:
  global-git-commit-mode: t
  async-bytecomp-package-mode: t
  shell-dirtrack-mode: t
  diff-auto-refine-mode: t
  whitespace-mode: t
  flyspell-mode: t
  text-scale-mode: t
  yas-global-mode: t
  yas-minor-mode: t
  vimish-fold-global-mode: t
  vimish-fold-mode: t
  show-paren-mode: t
  minibuffer-electric-default-mode: t
  ido-vertical-mode: t
  ido-ubiquitous-mode: t
  ido-everywhere: t
  highlight-line-mode: t
  global-auto-revert-mode: t
  global-aggressive-indent-mode: t
  display-time-mode: t
  delete-selection-mode: t
  ace-popup-menu-mode: t
  smartparens-global-mode: t
  smartparens-mode: t
  tooltip-mode: t
  global-eldoc-mode: t
  file-name-shadow-mode: t
  global-font-lock-mode: t
  font-lock-mode: t
  auto-composition-mode: t
  auto-encryption-mode: t
  auto-compression-mode: t
  column-number-mode: t
  line-number-mode: t
  auto-fill-function: do-auto-fill
  transient-mark-mode: t
  auto-fill-mode: 1

Recent messages:
Undo!
Mark set
Replaced 13 occurrences
<menu> / is undefined
Undo!
Mark set (rectangle mode)
Undo!
Mark set (rectangle mode) [3 times]
Saving file /home/mark/projects/programs/modalka/README.md...
Wrote /home/mark/projects/programs/modalka/README.md

Load-path shadows:
None found.

Features:
(shadow mail-extr emacsbug sendmail zygospore smartparens-html sgml-mode
markdown-toc smex esh-var esh-io esh-cmd esh-opt esh-ext esh-proc
esh-arg esh-groups eshell esh-module esh-mode esh-util dired-x
cider-debug cider-mode cider-resolve warnings cider-browse-ns
cider-inspector cider-repl cider-eldoc apropos cider-interaction spinner
arc-mode archive-mode cider-overlays cider-doc org-table cider-test
cider-stacktrace cider-client cider-util clojure-mode imenu nrepl-client
queue ewoc etags xref project names edebug autoload tar-mode cus-edit
cus-start cus-load mm-archive url-handlers markdown-mode bookmark pp
gnus-async qp gnus-ml disp-table calc-alg calc-ext calc-menu calc
calc-loaddefs calc-macs browse-url ace-window lisp-mnt url-http url-gw
url-cache url-auth url url-proxy url-privacy url-expand url-methods
url-history url-cookie url-domsuf url-util url-parse url-vars linum
magit-blame magit-stash magit-bisect magit-remote magit-commit
magit-sequence magit magit-apply magit-wip magit-log magit-diff
smerge-mode magit-core magit-process magit-popup magit-mode magit-git
crm magit-section magit-utils git-commit log-edit pcvs-util add-log
with-editor async-bytecomp async tramp-sh tramp tramp-compat
tramp-loaddefs trampver shell multiple-cursors
mc-hide-unmatched-lines-mode mc-separate-operations
rectangular-region-mode mc-mark-pop mc-mark-more mc-cycle-cursors
mc-edit-lines multiple-cursors-core rect org org-macro org-footnote
org-pcomplete pcomplete org-list org-faces org-entities noutline outline
org-version ob-emacs-lisp ob ob-tangle ob-ref ob-lob ob-table ob-exp
org-src ob-keys ob-comint ob-core ob-eval org-compat org-macs
org-loaddefs cal-menu calendar cal-loaddefs compile comint ansi-color
sort misearch multi-isearch vc vc-dispatcher vc-git diff-mode hl-line
nndraft nnmh nnfolder utf-7 epa-file epa derived epg gnutls
network-stream nsm auth-source eieio eieio-core starttls gnus-agent
gnus-srvr gnus-score score-mode nnvirtual gnus-msg gnus-art mm-uu
mml2015 mm-view mml-smime smime password-cache dig mailcap nntp
gnus-cache gnus-sum gnus-group gnus-undo gnus-start gnus-cloud nnimap
nnmail mail-source tls utf7 netrc nnoo parse-time gnus-spec gnus-int
gnus-range message dired format-spec rfc822 mml mml-sec mm-decode
mm-bodies mm-encode mail-parse rfc2231 rfc2047 rfc2045 ietf-drums
mailabbrev gmm-utils mailheader gnus-win gnus gnus-ems nnheader
gnus-util mail-utils mm-util help-fns mail-prsvr wid-edit
rainbow-delimiters hl-todo whitespace flyspell ispell
flycheck-color-mode-line face-remap flycheck-haskell haskell-cabal
haskell-utils haskell-customize flycheck find-func rx mk-visual
smart-mode-line rich-minority solarized-dark-theme solarized mk-texinfo
mk-tex mk-python mk-prolog mk-org mk-markdown mk-man mk-magit mk-lisp
mk-ibuffer mk-haskell mk-gnus mk-eshell mk-erc mk-elisp mk-dired mk-diff
mk-clojure kill-or-bury-alive mk-calendar mk-c mk-bookmark
mk-minor-modes common-lisp-snippets yasnippet cl whole-line-or-region
vimish-fold paren minibuf-eldef ido-vertical-mode ido-ubiquitous
ido-completing-read+ cl-macs gv ido highlight-line autorevert filenotify
cl-seq aggressive-indent easy-mmode time delsel ace-popup-menu
smartparens-config smartparens byte-opt thingatpt control-mode mk-global
edmacro kmacro ace-link avy ring misc server mk-utils subr-x f dash s
ucs-normalize finder-inf tex-site fix-word advice slime-autoloads info
bytecomp byte-compile cl-extra help-mode seq cl-loaddefs pcase cl-lib
cconv package easymenu epg-config time-date mule-util tooltip eldoc
electric uniquify ediff-hook vc-hooks lisp-float-type mwheel x-win
term/common-win x-dnd tool-bar dnd fontset image regexp-opt fringe
tabulated-list newcomment elisp-mode lisp-mode prog-mode register page
menu-bar rfn-eshadow timer select scroll-bar mouse jit-lock font-lock
syntax facemenu font-core frame cl-generic cham georgian utf-8-lang
misc-lang vietnamese tibetan thai tai-viet lao korean japanese eucjp-ms
cp51932 hebrew greek romanian slovak czech european ethiopic indian
cyrillic chinese charscript case-table epa-hook jka-cmpr-hook help
simple abbrev minibuffer cl-preloaded nadvice loaddefs button faces
cus-face macroexp files text-properties overlay sha1 md5 base64 format
env code-pages mule custom widget hashtable-print-readable backquote
dbusbind inotify dynamic-setting system-font-setting font-render-setting
move-toolbar gtk x-toolkit x multi-tty make-network-process emacs)

Memory information:
((conses 16 1301109 278239)
 (symbols 48 59662 0)
 (miscs 40 239 1733)
 (strings 32 181294 49935)
 (string-bytes 1 6370748)
 (vectors 16 89977)
 (vector-slots 8 2094007 64278)
 (floats 8 748 2372)
 (intervals 56 87834 1327)
 (buffers 976 36)
 (heap 1024 114703 68010))

bug#21453: When font size is temporarily increased Emacs is slow

[Mark Karpov]

Also, when point reaches edge of screen Emacs instantly shows you next
portion of text so point gets in the middle of the screen, instead of
just showing you one more line.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Date: Thu, 10 Sep 2015 15:56:46 +0600
> 
> When I press ‘C-x C-=’ to temporarily increase font size Emacs is
> getting considerably slower. For example, if I hold ‘C-n’ to move
> several lines down the movement is not smooth. I don't think that
> increasing of font size should normally result in such “effects”.

I cannot reproduce this in "emacs -Q" with today's master.  Do you
have any customizations related to scrolling?  This further input from
you:

> Also, when point reaches edge of screen Emacs instantly shows you next
> portion of text so point gets in the middle of the screen, instead of
> just showing you one more line.                            ^^^^^^^^^^
  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

seems to indicate that you have customized scrolling, because by
default Emacs does not scroll just one line.  Perhaps you have
scroll-conservatively set to a large value; if so, I'd indeed expect
some slowdown when the font size is changed, because supporting
scroll-conservatively is sometimes costly, especially when the last
displayed line is only partially visible.

So please try this in "emacs -Q" first.

Thanks.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

Eli Zaretskii <eliz@gnu.org> writes:

> seems to indicate that you have customized scrolling, because by
> default Emacs does not scroll just one line.  Perhaps you have
> scroll-conservatively set to a large value; if so, I'd indeed expect
> some slowdown when the font size is changed, because supporting
> scroll-conservatively is sometimes costly, especially when the last
> displayed line is only partially visible.

I indeed have some customization of scrolling, but it's not
‘scroll-conservatively’, its value is 0.

I can reproduce it with ‘emacs -Q’, although my Emacs build is a bit
older then current master. I'll try with master.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

I've built recent master and the problem is still there even in ‘emacs
-Q’. Maybe it's not Emacs.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Fri, 11 Sep 2015 18:16:24 +0600
> 
> 
> I've built recent master and the problem is still there even in ‘emacs
> -Q’. Maybe it's not Emacs.

How much slower it is?

Also, what text is in the buffer that you are using in your recipe?
Does it use more than one font?

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

Eli Zaretskii <eliz@gnu.org> writes:

> How much slower it is?

Hard to tell. I don't know how to measure this properly. From subjective
perception it's slow enough to be noticed when I hold ‘C-n’.

> Also, what text is in the buffer that you are using in your recipe?
> Does it use more than one font?

No it doesn't. I tried it with normal text, something from my
configuration. I can only tell that it happens even when text doesn't
have long lines (beyond 80 characters). Just checked this out.

I think the text itself is not that important.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Fri, 11 Sep 2015 18:46:33 +0600
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > How much slower it is?
> 
> Hard to tell. I don't know how to measure this properly. From subjective
> perception it's slow enough to be noticed when I hold ‘C-n’.

One way to measure it is to write a simple Lisp function that scrolls
with next-line to the end of a long buffer, and measures the time it
takes to do that.  Then run it and show the results.  (Some slow-down
is expected because a larger font with the same window dimensions
means more scrolls, but you can alleviate that effect to some extent
by enlarging the window to have the same number of visible lines.)

> > Also, what text is in the buffer that you are using in your recipe?
> > Does it use more than one font?
> 
> No it doesn't. I tried it with normal text, something from my
> configuration. I can only tell that it happens even when text doesn't
> have long lines (beyond 80 characters). Just checked this out.
> 
> I think the text itself is not that important.

Does the text have faces other than the default face?  What modes were
active in that buffer?  Do you see the same slow-down in a buffer
that's in Fundamental mode?

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

Eli Zaretskii <eliz@gnu.org> writes:

> Does the text have faces other than the default face?  What modes were
> active in that buffer?  Do you see the same slow-down in a buffer
> that's in Fundamental mode?

Well yes, since that test file is part of my Emacs configuration it has
some syntax highlighting (everything is by default since it's run with
‘-Q’ flag).

Modes? You can check yourself, I just started Emacs without config, then
opened one of my init files and pressed ‘C-x C-=’, then scrolled. That's
all.

Yes I can reproduce it in Fundamental mode too.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Fri, 11 Sep 2015 22:08:21 +0600
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Does the text have faces other than the default face?  What modes were
> > active in that buffer?  Do you see the same slow-down in a buffer
> > that's in Fundamental mode?
> 
> Well yes, since that test file is part of my Emacs configuration it has
> some syntax highlighting (everything is by default since it's run with
> ‘-Q’ flag).
> 
> Modes? You can check yourself, I just started Emacs without config, then
> opened one of my init files and pressed ‘C-x C-=’, then scrolled. That's
> all.
> 
> Yes I can reproduce it in Fundamental mode too.

OK, then the timing information will be really welcome.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Fri, 11 Sep 2015 22:08:21 +0600
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Does the text have faces other than the default face?  What modes were
> > active in that buffer?  Do you see the same slow-down in a buffer
> > that's in Fundamental mode?
> 
> Well yes, since that test file is part of my Emacs configuration it has
> some syntax highlighting (everything is by default since it's run with
> ‘-Q’ flag).
> 
> Modes? You can check yourself, I just started Emacs without config, then
> opened one of my init files and pressed ‘C-x C-=’, then scrolled. That's
> all.
> 
> Yes I can reproduce it in Fundamental mode too.

Any news on this one?  Last time I asked for some time measurements,
to have a more quantitative information about this slow-down.  Could
you provide that, please?

Thanks.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

Eli Zaretskii <eliz@gnu.org> writes:

> Any news on this one?  Last time I asked for some time measurements,
> to have a more quantitative information about this slow-down.  Could
> you provide that, please?
>
> Thanks.

Sorry for the delay. I've been quite busy and forgot about it. I'll try
to measure it as soon as I can get to it.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Sat, 26 Sep 2015 18:38:36 +0600
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Any news on this one?  Last time I asked for some time measurements,
> > to have a more quantitative information about this slow-down.  Could
> > you provide that, please?
> >
> > Thanks.
> 
> Sorry for the delay. I've been quite busy and forgot about it. I'll try
> to measure it as soon as I can get to it.

OK, thanks.

bug#21453: When font size is temporarily increased Emacs is slow

[Mark Karpov]

Here I describe my experiment and its outcome.

Setting
=======

To measure how long it takes to scroll a long buffer from beginning to
end with ‘next-line’ command the following code was evaluated:

  (defmacro with-timer (title &rest forms)
    "Run the given FORMS, counting the elapsed time.
  A message including the given TITLE and the corresponding elapsed
  time is displayed."
    (declare (indent 1))
    (let ((nowvar (make-symbol "now"))
          (body   `(progn ,@forms)))
      `(let ((,nowvar (current-time)))
         (message "%s..." ,title)
         (prog1 ,body
           (let ((elapsed
                  (float-time (time-subtract (current-time) ,nowvar))))
             (message "%s... done (%.3fs)" ,title elapsed))))))
  
  (defun next-line-racer ()
    "Scroll till end of buffer with `next-line' and measure how long it takes.
  
  To make it realistic `next-line' is called interactively."
    (interactive)
    (with-timer 
        (while (not (eobp))
          (call-interactively #'next-line))))

The file used:

https://github.com/Fuco1/smartparens/blob/master/smartparens.el

which is 7992 lines long.

The Experiment
==============

The file is opened and cursor is at the beginning of the file. Font size
is either altered with ‘C-x C-=’ (increased, denoted in results as +1)
or ‘C-x C--’ (decreased, denoted in results as -1) or not altered at
all. No adjustments to compensate number of visible lines were performed
in this experiment because I think difference wouldn't be noticeable.

Then ‘M-x next-line-racer’ is called. Emacs seemingly freezes for some
time until it shows the point (cursor) at the end of the buffer and
message in mini-buffer displaying time it took to accomplish the
scrolling.

No demanding processes were run in the background during the experiment.

The Results
===========

Here I post measured times:

  Font Size       Time in Seconds
  normal           301.423
  +1              1055.157
  -1               487.935

Hope it helps.

bug#21453: 25.0.50; When font size is temporarily increased Emacs is slow

[Mark Karpov]

Oh my! It seems I've spoiled the whole thing. I didn't notice that
`with-timer' takes TITLE argument. This leads to the form evaluated
twice. (I copied this snippet from one site.)

Let me know if you think it should be tested again. Although current
results demonstrate that slow-down is considerable and I doubt it won't
change if we evaluate the form only once.

bug#21453: When font size is temporarily increased Emacs is slow

[Mark Karpov]

* won't → will

bug#21453: When font size is temporarily increased Emacs is slow

[Eli Zaretskii]

> From: Mark Karpov <markkarpov@openmailbox.org>
> Cc: 21453@debbugs.gnu.org
> Date: Sun, 27 Sep 2015 17:02:12 +0600
> 
> The Results
> ===========
> 
> Here I post measured times:
> 
>   Font Size       Time in Seconds
>   normal           301.423
>   +1              1055.157
>   -1               487.935

I see something very different on my system, so this could be specific
to X or even your font back-end (xft).

Here are my times:

  Font Size       Time in Seconds
                   ELisp  Fundamental
  normal           139.9  113.2
  +1               140.4  112.2
  -1               282.0  126.4
  -2                      148.8
  +2                      114.4

The second column of numbers is with the same file, but with font-lock
turned off, after rejecting the file-local variables, and in
Fundamental mode.  As you see, the differences are much smaller in
Fundamental mode.  This points to face remapping as the possible
reason for slower redisplay, at least in my case.  When you enlarge or
decrease the default font size, Emacs needs to look up the faces in
face-remapping-alist each time it needs to redisplay something, which
could explain some slowdown.

I don't know why the +1 case is so much slower on your system; perhaps
you could use some kind of profiler to see what takes most of the time
in that case.  You could start with the Emacs's built-in profiler,
although I doubt it will tell something revealing, because I think the
reason is below the level the built-in profiler can look.

In any case, I understand now why you perceive the slowdown as
significant: if it takes 1000 sec to scroll through a 8000-line file,
it means you get just 8 lines-per-second, which is slow enough to
annoy.  As you see, on my system, I get 5 times faster performance,
which is below human annoyance threshold.

bug#21453: When font size is temporarily increased Emacs is slow also on 25.2.1

[Gennady Uraltsev]

[-- Attachment #1.1: Type: text/plain, Size: 213 bytes --]

Sorry, found this bug only now. Did some profile measurements and
reported it here:

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26402

Let me know if I can be useful. Seems like quite an big issue.




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#21453: When font size is temporarily increased Emacs is slow also on 25.2.1

[Gennady Uraltsev]

[-- Attachment #1.1: Type: text/plain, Size: 213 bytes --]

Sorry, found this bug only now. Did some profile measurements and
reported it here:

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26402

Let me know if I can be useful. Seems like quite an big issue.




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#21453: When font size is temporarily increased Emacs is slow also on 25.2.1

[Gennady Uraltsev]

[-- Attachment #1.1: Type: text/plain, Size: 213 bytes --]

Sorry, found this bug only now. Did some profile measurements and
reported it here:

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26402

Let me know if I can be useful. Seems like quite an big issue.




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#21453: When font size is temporarily increased Emacs is slow also on 25.2.1

[Gennady Uraltsev]

Sorry, found this bug only now. Did some profile measurements and
reported it here:

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26402

Let me know if I can be useful. Seems like quite an big issue.

bug#21453: When font size is temporarily increased Emacs is slow also on 25.2.1

[Gennady Uraltsev]

[-- Attachment #1.1: Type: text/plain, Size: 209 bytes --]

Sorry, found this bug only now. Did some profile measurements and
reported it here:

https://debbugs.gnu.org/cgi/bugreport.cgi?bug=26402

Let me know if I can be useful. Seems like quite an big issue.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

bug#21453: When font size is temporarily increased Emacs is slow

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> Here I post measured times:
>> 
>>   Font Size       Time in Seconds
>>   normal           301.423
>>   +1              1055.157
>>   -1               487.935
>
> I see something very different on my system, so this could be specific
> to X or even your font back-end (xft).
>
> Here are my times:
>
>   Font Size       Time in Seconds
>                    ELisp  Fundamental
>   normal           139.9  113.2
>   +1               140.4  112.2
>   -1               282.0  126.4
>   -2                      148.8
>   +2                      114.4

I tried this on my system (with the test file provided).

The test case is:

(benchmark-progn
  (while (not (eobp))
    (call-interactively #'next-line)))

normal 25s
+1     31s
+2     25s

So basically, they're all more or less the same now, so I'm guessing
that whatever Mark was seeing in Emacs 25 has gone away now (as Emacs
has gotten a new font backend etc), so I'm closing this bug report.

If this is still an issue in Emacs 27, please reopen.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no