I found the reason of this: patch 7.3.1203:
набор изм-й: 4916:ba328b4a990e
метка: v7-3-1203
автор: Bram Moolenaar <[email protected]>
дата: Sat Jun 15 23:00:30 2013 +0200
файлы: src/screen.c src/version.c
описание:
updated for version 7.3.1203
Problem: Matches from matchadd() might be highlighted incorrectly when
they
are at a fixed position and inserting lines. (John Szakmeister)
Solution: Redraw all lines below a change if there are highlighted
matches.
(idea by Christian Brabandt)
This was due to Syntastic used matchaddpos() with static positions (what
matchaddpos() was designed to implement). See
https://groups.google.com/forum/#!topic/vim_dev/FuSXVJXySD8 for details.
Removing this patch increases speed of redrawing matching parens
dramatically! Syntastic developers may be encouraged to use new
matchaddpos() for static highlights: this will help whole vim to work
faster. I have a video that shows how faster it gets, but to show delays in
details i had to chose 60fps and this made the video size 24Mb: i am not
sure if it is ok to post such a big attachment here.
The patch that removes 7.3.1203 is attached.
UPDATE: Here is the video on in 30fps: http://vimeo.com/99325753, it does
not show delays very well, but they are still visible. The cases shown:
1. Slow vim. Heavy loaded by parens line on the top: slow.
2. Slow vim. Same line on the bottom: fast.
3. Fast vim. Same line on the top: fast exactly as in case 2 and 4.
4. Fast vim. Same line on the bottom: fast.
Cheers, Alexey.
2014-06-26 21:55 GMT+04:00 Alexey Radkov <[email protected]>:
> Meanwhile i found an obvious reproducible effect: horizontal navigation
> (e.g. to the left and to the right) along a heavily loaded by parentheses
> line differs in speed which depends on whether this line at the top of the
> window (the slowest), mid of the window (faster) and bottom of the window
> (fastest). I suspected that this was due to redrawing area lasted from the
> top-parenthesis line to the bottom window line, though the bottom line of
> redrawing area should be N + 1. I checked if b_mod_top and b_mod_bot is N
> and N + 1 respectively when the line is being drawn in win_line(): it
> appeared true. This means that something else forces screen to redraw to
> the bottom line, but i cannot still find what. I tried also to set rtype in
> match_add() and match_delete() to SOME_VALID: this should make redrawing
> line slowest at any window position: indeed this worked as i expected: the
> line began to redraw at the slowest speed and this should prove that the
> effect is really due to wrong/unexpected redrawing area (i.e. its bottom
> line lasts to the bottom of the window/screen).
>
> Cheers, Alexey.
>
>
>
> 2014-06-25 13:08 GMT+04:00 Alexey Radkov <[email protected]>:
>
> Ok, I'll do this during this or the next week.
>>
>> Cheers, Alexey.
>>
>>
>> 2014-06-23 23:11 GMT+04:00 Bram Moolenaar <[email protected]>:
>>
>>
>>> Alexey Radkov wrote:
>>>
>>> > Here is some small improvements for the patch:
>>> >
>>> > src/window.c, match_add():
>>> > - removed unnecessary (after alloc() to alloc_clear() transformation)
>>> > zero-initializations of m->pos fields,
>>> > - checking that li == NULL was moved inside for-loop body
>>> > - decrement i when the pos_list element is a number and has value 0 to
>>> be
>>> > symmetric with sublist counterpart
>>> >
>>> > src/screen.c:
>>> > - simple code alignment
>>> >
>>> > runtime/doc/eval.txt, matchaddpos():
>>> > this is a bit more important: actually column number and match length
>>> must
>>> > correspond to bytes as col() returns, not screen columns as virtcol()
>>> > returns! It does matter for Unicode texts like Cyrillics, i changed
>>> doc to
>>> > reflect this.
>>>
>>> Thanks.
>>>
>>> Now that we have one working test, perhaps you can add a few more tests?
>>> Would be good if we cover the alternatives in the code.
>>>
>>>
>>> --
>>> Proverb: A nightingale that forgets the lyrics is a hummingbird.
>>>
>>> /// Bram Moolenaar -- [email protected] -- http://www.Moolenaar.net
>>> \\\
>>> /// sponsor Vim, vote for features -- http://www.Vim.org/sponsor/
>>> \\\
>>> \\\ an exciting new programming language -- http://www.Zimbu.org
>>> ///
>>> \\\ help me help AIDS victims -- http://ICCF-Holland.org
>>> ///
>>>
>>> --
>>> --
>>> You received this message from the "vim_dev" maillist.
>>> Do not top-post! Type your reply below the text you are replying to.
>>> For more information, visit http://www.vim.org/maillist.php
>>>
>>> ---
>>> You received this message because you are subscribed to the Google
>>> Groups "vim_dev" group.
>>> To unsubscribe from this group and stop receiving emails from it, send
>>> an email to [email protected].
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>
>
--
--
You received this message from the "vim_dev" maillist.
Do not top-post! Type your reply below the text you are replying to.
For more information, visit http://www.vim.org/maillist.php
---
You received this message because you are subscribed to the Google Groups
"vim_dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
diff -r dd88acaa1a57 src/screen.c
--- a/src/screen.c Thu Jun 26 22:33:51 2014 +0200
+++ b/src/screen.c Fri Jun 27 15:30:49 2014 +0400
@@ -1768,10 +1768,6 @@
# endif
syntax_check_changed(lnum)))
#endif
-#ifdef FEAT_SEARCH_EXTRA
- /* match in fixed position might need redraw */
- || wp->w_match_head != NULL
-#endif
)))))
{
#ifdef FEAT_SEARCH_EXTRA
