Performance 21.3.50 W2K

Performance 21.3.50 W2K

[Dhruva Krishnamurthy]

Hi,
 I tried opening a 58Mb file (Yes, I know it was a stupid thing to do) in
 Emacs. It just froze for couple of minutes (2 to 3 mins) and opened the
 file. What surprises me is, the same file when I open in XEmacs (latest
 beta), it took less than 5 seconds. I see a very big performance
 difference in this regard.

with regards,
dhruva
-- 
Dhruva Krishnamurthy
Home: http://www.geocities.com/gnued/

Re: Performance 21.3.50 W2K

[Eli Zaretskii]

> Date: Wed, 30 Oct 2002 09:32:13 UT
> From: "Dhruva Krishnamurthy" <seagull@fastmail.fm>
> 
>  I tried opening a 58Mb file (Yes, I know it was a stupid thing to do) in
>  Emacs. It just froze for couple of minutes (2 to 3 mins) and opened the
>  file. What surprises me is, the same file when I open in XEmacs (latest
>  beta), it took less than 5 seconds. I see a very big performance
>  difference in this regard.

Perhaps your XEmacs was built without Mule.  In that case, "M-x
find-file-literally" instead of "C-x C-f" should be as fast as XEmacs.

The reason for this performance hit is that "C-x C-f" by default tries
to detect whether the file has encoded non-ASCII characters, and if
so, decodes them into internal representation used for non-ASCII
characters within Emacs buffers.  XEmacs compiled without Mule doesn't
do that, so visiting a large file is faster.

Re: Performance 21.3.50 W2K

[Richard Stallman]

    >  I tried opening a 58Mb file (Yes, I know it was a stupid thing to do) in
    >  Emacs. It just froze for couple of minutes (2 to 3 mins) and opened the
    >  file.

If you do find-file-literally, how fast is that?

Re: Performance 21.3.50 W2K

[Dhruva Krishnamurthy]

On Thu, 31 Oct 2002 12:26:57 -0500, "Richard Stallman" <rms@gnu.org>
said:
>     >  I tried opening a 58Mb file (Yes, I know it was a stupid thing to do) in
>     >  Emacs. It just froze for couple of minutes (2 to 3 mins) and opened the
>     >  file.
> 
> If you do find-file-literally, how fast is that?

It takes less than 3 to 4 seconds. Emacs and XEmacs have almost identical
performance. Well, since I am on this list, I could get clarifications.
For a user, I feel we should come up with a slightly different behaviour.

1. If the file size is greater than a given size (user customizable) and
if it has non-ascii chars, we should prompt whether to use
'find-file-literally instead of 'find-file.

or

1. Come up with buffered reading for large files. Read a portion of the
file, display it and repeat this till the file is completely read. An
assumption can be made, user will not want to go to the end of the file
immediately.

with regards,
dhruva
-- 
Dhruva Krishnamurthy
Home: http://www.geocities.com/gnued/

Re: Performance 21.3.50 W2K

[Richard Stallman]

    1. If the file size is greater than a given size (user customizable) and
    if it has non-ascii chars, we should prompt whether to use
    'find-file-literally instead of 'find-file.

This might be a good idea for very large files.  We could add a
variable to specify the threshold for this, release Emacs with that
threshold set as very large, then ask users to try smaller values
and say if they prefer those.

    1. Come up with buffered reading for large files. Read a portion of the
    file, display it and repeat this till the file is completely read. 

This would be a good feature, but doing it in a general way would be
very difficult.

The simplest variant of this idea would be to call redisplay in the
middle of inserting the file contents.  Emacs would not be able to
execute any commands until it finishes reading the file, but you could
at least see the beginning of the file.  That might not be too hard.
Of course, it would only help to a limited extent.