[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Processing speed in large C++ raw strings.

From: Alan Mackenzie
Subject: Re: Processing speed in large C++ raw strings.
Date: Sat, 27 Apr 2019 11:37:34 +0000
User-agent: Mutt/1.10.1 (2018-07-13)

Hello, Tadeus.

On Fri, Apr 26, 2019 at 22:15:28 +0200, Tadeus Prastowo wrote:
> Hi Alan,

> On Thu, Apr 25, 2019 at 8:41 PM Alan Mackenzie <address@hidden> wrote:
> > I have now optimized CC Mode, such that each such self-insert-command
> > now takes around 0.07s-0.1s.  I used two strategies for this: (i) breaking 
> > out
> > of (expensive) fontification routines early, when the region being
> > fontified is nothing but a string; (ii) Not fontifying the entire raw
> > string for a change to only a small part of it.

> I have been working with C++ template metaprograms for two years, and
> since quite early, I have experienced CC Mode being very sluggish that
> I often switch back-and-forth between cc-mode and text-mode.

> I have not made any report since I don't feel like sparing my time to
> collect some data for the report now that switching to text-mode does
> the trick.  So, since you are already at it, perhaps you may want to
> find out why even scrolling the buffer in cc-mode is very slow when
> visiting the file at
> http://git.savannah.nongnu.org/cgit/tice.git/plain/tests/v1/test-v1_internals_program.cpp?

You've submitted a bug report now.  :-)

> Specifically, I built master at commit
> 82fc8ca6ba179596928db614ab903e44968ef44e and fired the result with
> `emacs -Q' to visit the file locally.  Then, to experience the
> sluggishness, I pressed M-> followed by M-v several times.  To further
> experience it, I started to edit some template parameters, such as
> changing `node1' to node `node5' in line 680.

Thanks for drawing my attention to this problem with lots of templates.
The bug is that, in a certain low-level place, CC Mode was failing to
mark the < and > "as parentheses" (with syntax-table text properties),
hence a lot of unneeded buffer scanning was being done.

The following patch should speed up the scrolling operation markedly.
It does not become instantaneous, but takes a fraction of a second,
compared with the several seconds it has taken up to now.  Would you
please try it out.  

diff -r 1940f96b9799 cc-engine.el
--- a/cc-engine.el      Fri Apr 26 01:01:07 2019 +0000
+++ b/cc-engine.el      Sat Apr 27 11:30:06 2019 +0000
@@ -5539,7 +5539,9 @@
                      (prog1 (looking-at "\\s(")
-         (if (c-forward-<>-arglist nil) ; Should always work.
+         (if (let ((c-parse-and-markup-<>-arglists t)
+                   (c-restricted-<>-arglists t))
+               (c-forward-<>-arglist nil)) ; Should always work.
              (when (> (point) to)
                (setq bound-<> (point)))

> Thank you very much for your work.

> --
> Best regards,
> Tadeus

Alan Mackenzie (Nuremberg, Germany).

reply via email to

[Prev in Thread] Current Thread [Next in Thread]