22.2 slower than 21.3

22.2 slower than 21.3

[Robert]

Hi,

I just installed 22.2 (pre-built binary for Windows XP), but find it
slower than 21.3. I notice when opening C-files. I went straight back
to 21.3 now because of that.

I have not been following this group regularly, is this a well known
issue? Will this get faster in upcoming versions? The problem could
perhaps also be that the binary for Windows was built differently.
Have others seen the same as me?

-Robert

Re: 22.2 slower than 21.3

[Martin Fischer]

Hi,

  Robert writes:

  > Hi,
  >
  > I just installed 22.2 (pre-built binary for Windows XP), but find it
  > slower than 21.3. I notice when opening C-files. I went straight back
  > to 21.3 now because of that.

How did you measure, that it is slower.

  >
  > I have not been following this group regularly, is this a well known
  > issue? Will this get faster in upcoming versions? The problem could
  > perhaps also be that the binary for Windows was built differently.
  > Have others seen the same as me?

No, but you could try the 22.3 pretest binary from

http://alpha.gnu.org/gnu/emacs/pretest/windows/emacs-22.2.91-bin-i386.zip

and send a bug report if necessary.

Martin

Re: 22.2 slower than 21.3

[Robert]

On Aug 29, 1:22 pm, Martin Fischer <paroz...@web.de> wrote:
> Hi,
>
>   Robert writes:
>
>   > Hi,
>   >
>   > I just installed 22.2 (pre-built binary for Windows XP), but find it
>   > slower than 21.3. I notice when opening C-files. I went straight back
>   > to 21.3 now because of that.
>
> How did you measure, that it is slower.

I just noticed when opening a C source file I am working on. Counted
seconds now, see below.

>
>   >
>   > I have not been following this group regularly, is this a well known
>   > issue? Will this get faster in upcoming versions? The problem could
>   > perhaps also be that the binary for Windows was built differently.
>   > Have others seen the same as me?
>
> No, but you could try the 22.3 pretest binary from
>
> http://alpha.gnu.org/gnu/emacs/pretest/windows/emacs-22.2.91-bin-i386...

Thanks for the link, I tried that version and it is virtually
indistinguishable from the 21.3 version.
I tried to count now, and the file opens within one second with 21.3
and 22.2.91. With 22.2 it takes 4 seconds. Even when I kill the buffer
and reopen the file again from dired it takes 4 seconds. I am happy to
see that 22.2.91 is just as fast as 21.3. I won't bother filing any
bug then, I'll just skip the 22.2 version.

Thanks again.

-Robert

>
> and send a bug report if necessary.
>
> Martin

Re: 22.2 slower than 21.3

[Alan Mackenzie]

Hi, Robert!

On Fri, Aug 29, 2008 at 01:06:52AM -0700, Robert wrote:
> Hi,

> I just installed 22.2 (pre-built binary for Windows XP), but find it
> slower than 21.3. I notice when opening C-files. I went straight back
> to 21.3 now because of that.

There was a bug in older CC Modes which made it "necessary" to scan
entire buffers when loading them.  It was this:

#########################################################################
put the following into
a C file, somewhere near the beginning:

#warning for isn't a keyword here.

This will foul up the font locking in the rest of the file.  Even more
dramatically, insert this:

#warning for isn't a keyword here. (
//#warning for isnt a keyword here.
void foo (bar)
{
#error Brace yourself! }
    printf ("Hello, world!\n") ; /* the famous one liner! */
}

Now the paren matching functions are fouled up.
#########################################################################

However, the fix for this this shouldn't make Emacs that much slower to
load.  On my 1.2 GHz Athlon, it added around 1 second per Mbyte of C
buffer.

> I have not been following this group regularly, is this a well known
> issue?

No.  At least, not yet.  :-)

> Will this get faster in upcoming versions? The problem could perhaps
> also be that the binary for Windows was built differently.  Have
> others seen the same as me?

Could you give me some figures, please?  _HOW_ much slower is the new CC
mode?  Could you please do:
(i) C-x C-f for a large C file please, one stored on your local hard
  drive.  Then kill the buffer with C-x k, thus ensuring the file is in
  the file system cache, if any.
(ii) Do C-x C-f for this file again, and say how long it takes.
(iii) Do C-x C-f for the same file in Emacs 21.x.  How long does this
  take?
(iv) State how large the file is, and what make and speed of CPU you
  have.

Thanks!

If there is a real problem here, like files taking 1 minute to load,
it'll get fixed.  If it's a case of the loading feeling a bit sluggish
(an extra 1 or 2 seconds to load), it may be possible to optimise things
a bit.

Look forward to hearing from you!

[P.S. It would be helpful if you could put "C Mode" or "a C file" or
something like that in the subject line for things like this.  That way,
I'm less likely to miss it.]

> -Robert

-- 
Alan Mackenzie (Nuremberg, Germany).

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/

Re: 22.2 slower than 21.3

[Robert]

Hi Alan!

On Aug 29, 4:27 pm, Alan Mackenzie <a...@muc.de> wrote:
> Hi, Robert!
>
> On Fri, Aug 29, 2008 at 01:06:52AM -0700, Robert wrote:
> > Hi,
> > I just installed 22.2 (pre-built binary for Windows XP), but find it
> > slower than 21.3. I notice when opening C-files. I went straight back
> > to 21.3 now because of that.
>
> There was a bug in older CC Modes which made it "necessary" to scan
> entire buffers when loading them.  It was this:

OK, didn't know there was a new CC-mode.

> > Will this get faster in upcoming versions? The problem could perhaps
> > also be that the binary for Windows was built differently.  Have
> > others seen the same as me?
>
> Could you give me some figures, please?  _HOW_ much slower is the new CC
> mode?  Could you please do:
> (i) C-x C-f for a large C file please, one stored on your local hard
>   drive.  Then kill the buffer with C-x k, thus ensuring the file is in
>   the file system cache, if any.
> (ii) Do C-x C-f for this file again, and say how long it takes.
> (iii) Do C-x C-f for the same file in Emacs 21.x.  How long does this
>   take?
> (iv) State how large the file is, and what make and speed of CPU you
>   have.

I only loaded a file of 518k, a C-file of about 10000 lines. I tried
now and it takes about 5 seconds to load in 22.2, actually 6 seconds
if I don't use reverse video (which I normally use). The 22.2.91
version takes less than 1 second.

> If there is a real problem here, like files taking 1 minute to load,
> it'll get fixed.  If it's a case of the loading feeling a bit sluggish
> (an extra 1 or 2 seconds to load), it may be possible to optimise things
> a bit.

OK, so this is only a few seconds difference. Now I concatenated the
file I had to get a larger C-file of 2 MB:

2.22:       17 seconds
2.22.91:   0.8 seconds

I loaded the file first, then killed the buffer and then I timed the
loading.

With 2.22.91 I get another problem though, with SLIME:

---
An error has occurred while loading `e:/home/rfo/.emacs':

Wrong number of arguments: (lambda (&optional contribs) Setup Emacs so
that lisp-mode buffers always use SLIME.
CONTRIBS is a list of contrib packages to load. (when (member (quote
lisp-mode) slime-lisp-modes) (add-hook (quote lisp-mode-hook) (quote
slime-lisp-mode-hook))) (when contribs (add-to-list (quote load-path)
(expand-file-name contrib slime-path)) (dolist (c contribs) (require
c) (let ((init (intern (format %s-init c)))) (when (fboundp init)
(funcall init)))))), 2
---

This does not happen in 21.3 or 22.2.

Regards,
-Robert

Re: 22.2 slower than 21.3

[Robert]

Correction: The problem below with SLIME is not there when I revert to
an older version of SLIME, so this is not an issue for this group.

> With 2.22.91 I get another problem though, with SLIME:
>
> ---
> An error has occurred while loading `e:/home/rfo/.emacs':
>
> Wrong number of arguments: (lambda (&optional contribs) Setup Emacs so
> that lisp-mode buffers always use SLIME.
> CONTRIBS is a list of contrib packages to load. (when (member (quote
> lisp-mode) slime-lisp-modes) (add-hook (quote lisp-mode-hook) (quote
> slime-lisp-mode-hook))) (when contribs (add-to-list (quote load-path)
> (expand-file-name contrib slime-path)) (dolist (c contribs) (require
> c) (let ((init (intern (format %s-init c)))) (when (fboundp init)
> (funcall init)))))), 2
> ---
>
> This does not happen in 21.3 or 22.2.

Yes it does, I needed an older version of SLIME.

-Robert

Re: 22.2 slower than 21.3

[David Combs]

In article <mailman.17893.1220019825.18990.help-gnu-emacs@gnu.org>,
Alan Mackenzie  <bug-cc-mode@gnu.org> wrote:
>Hi, Robert!
>
>On Fri, Aug 29, 2008 at 01:06:52AM -0700, Robert wrote:
>> Hi,
>
>> I just installed 22.2 (pre-built binary for Windows XP), but find it
>> slower than 21.3. I notice when opening C-files. I went straight back
>> to 21.3 now because of that.
>
>There was a bug in older CC Modes which made it "necessary" to scan
>entire buffers when loading them.  It was this:
>
>#########################################################################
>put the following into
>a C file, somewhere near the beginning:
>
>#warning for isn't a keyword here.

Sure looks like a C-comment to me -- so by what means did
emacs "see into" that comment?  And why would it do anything
like that?


Thanks,

David