Fwd: Serious performace problems on Windows XP with new(!) GNU Emacs v22 (both patched and unpatched EmacsW32 were tried)

[Peter Tury]

Hi,

additional info:

normally I start emacs with the "default EmacsW32 icon" (= gnuclientw.exe -sqf).

Now I've tried emacs.exe -Q; then opening the 1MB file: it's fine (1-2
seconds), but fundamental mode was used. In this running "pure" emacs
I loaded the files for the specific mode support for this source file,
and tried opening again: it still fast enough (2-3 secs). So that mode
itself seems to be OK, but something else slows down emacs.

How to figure out what? Is the only way what I can try to "run" my
.emacs line by line (from a running Emacs via eval-last-sexp) and try
opening that 1MB file each time?

I also tried emacs.exe (so no -Q was used) just to verify if
gnuclientw is wrong. Emacs.exe is similarly slow as gnuclientw.exe.

Let me emphasize again that the same .emacs with older EmacsW32 is
fast, so something is changed in these newer Emacses what results slow
down -- if I use my .emacs ...

---------- Forwarded message ----------
From: Peter Tury <address@hidden>
Date: 2006.10.11. 22:15
Subject: Serious performace problems on Windows XP with new(!) GNU
Emacs v22 (both patched and unpatched EmacsW32 were tried)
To: address@hidden


Hi,

I found that newer v22 Emacses (I tried:
- patched EmacsW32 vP060914 (=mid September);
- patched EmacsW32 vP060926 (= end of September) and
- unpatched(!) EmacsW32 vU061011(=current CVS))
are much slower than older ones (I tried for comparison patched
EmacsW32 vP060415 (=mid April)).

This is recognized with a special mode(?) and big (>200 kbytes) files.
In these cases even C-x k takes several (>5-6) seconds and opening
these files takes even more; it seems to be exponentially worse with
bigger files. Processor load is 100%, 99% is taken by emacs.exe. I
have 200 MB free memory during these steps.

With old Emacs (I only renamed the directories in Program Files, so
.emacs and every other settings are untouched) everything seems to be
OK.

My measurements: opening a ~1MB file took more than 5 minutes; killing
it took ~4
minutes! I realized that if I type something in the middle of this 5
minutes, it appears at the beginning of the buffer, even if I can see
it only ~2 minutes later (when the content becomes visible).

For comparison: opening the same 1MB file with patched EmacsW32
vP060415 (same .emacs and everything) took ~3 seconds; killing it took
~2 seconds.

Can you help what to check? What could went wrong?

How can I check which function takes the processor time?

Have you heard/found similar problems?

Br,
P