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

[Peter Tury]

2006/11/2, martin rudalics <address@hidden>:
>  > 1. whitespace-buffer on slowtst.el takes ~9 seconds (=the same result
>  > what I got with your previous fix) IF point is at the beginning of the
>  > buffer. If point is somewhere down (e.g. at (point-max))
>  > whitespace-buffer finishes in ~2 seconds!?
>  >
>  > 2. newest whitespace-buffer on my "original most problematic" file
>  > needs ~ 55 seconds (before fix this was ~3(?) minutes and unexpected
>  > delay occured also at file opening and at file closing
>  > "automatically"; now file opening and closing is normal fast
>  > (=immediate)) IF point is at top of the buffer. If point is at about
>  > the middle, whitespace-buffer needs ~25 seconds to finish. If point is
>  > at (point-max), whitespace-buffer finishes in 3 seconds!
>  >
>  > So if you could find the reason of this strange difference related to
>  > (point), then there would be no performace problem with whitespace.
>
> Thanks for the valuable information.  Attached find a new patch which
> should handle this.

I tried this new whitespace.el. Results:

1. slowtst.el AND point is at top of buffer: whitespace-buffer takes
2(!)-5 seconds (= half of previous time)
2. slowtst.el AND point is at botom: 2-3 seconds (= the same as
previous, still faster than from top)
3. my original most problematic ("real") file AND point at top:
~2(!)-19 seconds!
4. original problematic AND bottom: ~2seconds

It seems that whitespace-buffer is lightning fast :-) at first run
(regardless of (point)), but subsequent runs are slower. If point is
at bottom, the slowing-down is not really problematic, but if point is
at top, slowing-down is significant.

I tried M-: (remove-overlays nil nil 'face 'whitespace-highlight) and
it took about the same time what was the increment between first and
other runs. Evaluating it when point is at bottom, very fast, but when
point is at top, it is slow.

I think you can easily test it by making slowtst.el a few times bigger
and then measuring the run times.

Anyway, thanks for your fixes: now whitespace is not blocking my
upgrade to newer Emacs.

Thanks,
P