[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Timur  changed:

   What|Removed |Added

   Severity|major   |normal

--- Comment #19 from Timur  ---
That 10 sec looked Major, now 2 sec seems High-Normal.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Xisco Faulí  changed:

   What|Removed |Added

   See Also||https://bugs.documentfounda
   ||tion.org/show_bug.cgi?id=15
   ||9131

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #18 from ady  ---
An aside comment (but maybe not _that_ aside)...

FWIW, there was some performance issue that received some kind of patch and was
later reverted, and applied again with some modification, only to be later
reverted again [1]. This back and forth is around versions 7.4 to 7.6 and
master (during 24.2 dev), so the _specific_ version/build that is tested for
this report tdf#158167 might be relevant (with or without that patch
applied/reverted).

The only reason I mention this here is because we are talking about performance
under certain circumstances. The original reasons that triggered each of the
lags/crashes and the respective patches might not be related to each other at
all, but the possible delays (or improvements) in each test case might be
wrongly attributed to some assumed proposed solution while in fact the
resulting behavior might be a consequence of something else, unrelated.

IOW, seeing delays/lags in one version but not in the other, or in both (7.6.x
vs 24.2.dev) while testing the reported behavior might be related to diverse
factors, including the inclusion of a patch that was later reverted (twice)
during the relevant development cycle.

Whether the (reverted) patch is actually of any influence for the reported
behavior in this tdf#158167, I have no idea.

[1]: 

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #17 from Matt K  ---
I also repro about a 2 sec delay on Windows with the fix on a debug build,
which is better than the 10+ sec delay that was there before.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #16 from Timur  ---
This was very slow for me in GTK and now it less slow, like 2 secs, which is
accepratable for debug build, but I am not sure if good for a normal use,
should be checked in a release.
That bisected commit is from 7.5 but with installed LO 7.5.9 I see no delay,
even those 2 secs.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Buovjaga  changed:

   What|Removed |Added

 CC||ilmari.lauhakangas@libreoff
   ||ice.org

--- Comment #15 from Buovjaga  ---
(In reply to Matt K from comment #14)
> Created attachment 192006 [details]
> Windows Performance Analyzer output
> 
> I took a perf trace under Windows using the Windows Performance Recorder UI
> tool with CPU Usage collection selected.  I've attached the output which
> shows the code is spending at least 15 seconds (3 intervals of almost 5
> seconds each) doing sclo.dll!ScColContainer::resize which just calls
> "reserve" on a std::vector container.  I don't know why moving a row has
> anything to do with resizing columns, but maybe someone more familiar with
> the code can determine the problem.

That sounds exactly like the bibisect result I got for bug 159131, so can Timur
or others still repro after Noel's fix? Noel does say in the fixing commit:
"I'm not sure why the above commit causes trouble on Windows, but not Linux"

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Matt K  changed:

   What|Removed |Added

 CC||matt...@gmail.com

--- Comment #14 from Matt K  ---
Created attachment 192006
  --> https://bugs.documentfoundation.org/attachment.cgi?id=192006=edit
Windows Performance Analyzer output

I took a perf trace under Windows using the Windows Performance Recorder UI
tool with CPU Usage collection selected.  I've attached the output which shows
the code is spending at least 15 seconds (3 intervals of almost 5 seconds each)
doing sclo.dll!ScColContainer::resize which just calls "reserve" on a
std::vector container.  I don't know why moving a row has anything to do with
resizing columns, but maybe someone more familiar with the code can determine
the problem.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Aron Budea  changed:

   What|Removed |Added

   Keywords|bibisectRequest |notBibisectable

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Libreoffice-bugs] [Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #13 from ady  ---
(In reply to Timur from comment #12)
> I see no difference with 16k option or without 16k: 1st time I move row
> after loading, it lags 5-6 secs, but later I cannot move it at all.

My intention was for someone to test with older versions, before and after the
official/default introduction of 16k rows in LO 7.4.

I wasn't aware that the 2nd (consecutive) move was behaving differently/worse
than the 1st; maybe I missed that detail from prior comments.

Now my test...


Since the introduction of 16384 columns in LO 7.4, there seems to be some
(minor?) additional lag when moving a row (in comparison to LO 7.3), but the
lag is _much_ more notable in LO 24.2.

Additionally, until LO 7.6, while moving the selected row/area (dragging the
mouse but before releasing the button) we can see an edge/border indicating the
destination area of the move. In contrast, in LO 24.2, such border over the
to-be-destination area (either an entire row or any other selected area/cell)
is not seen (although my 24.2 alpha is several weeks old ATM, so YMMV).

Version: 24.2.0.0.alpha0+ (X86_64) / LibreOffice Community
Build ID: c0c8cffd3541e3cd616c96791b04e7ebf2b2ed03
CPU threads: 4; OS: Windows 10.0 Build 19045; UI render: Skia/Raster; VCL: win
Locale: en-US (es_AR); UI: en-US
Calc: CL threaded

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Libreoffice-bugs] [Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #12 from Timur  ---
I see no difference with 16k option or without 16k: 1st time I move row after
loading, it lags 5-6 secs, but later I cannot move it at all.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Libreoffice-bugs] [Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #11 from ady  ---
(In reply to Aron Budea from comment #10)
> I'd think this is related to switching to 16k columns

If that's the case, the behavior should be already noticed in LO 7.4, when the
16384 columns were introduced. ATM this is reported against LO 24.2. Timur?

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Libreoffice-bugs] [Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

--- Comment #10 from Aron Budea  ---
I'd think this is related to switching to 16k columns, so the amount of data to
move is 16x as it used to be before, and a debug build is obviously not
optimized for performance.

I can't tell whether this needs improvement or could be further improved.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[Libreoffice-bugs] [Bug 158167] Calc: Move selected row and scroll sheet extremely slow with CPU in local debug build

[bugzilla-daemon]

https://bugs.documentfoundation.org/show_bug.cgi?id=158167

Timur  changed:

   What|Removed |Added

Summary|Calc: cannot move selected  |Calc: Move selected row and
   |row in local debug build|scroll sheet extremely slow
   ||with CPU in local debug
   ||build

-- 
You are receiving this mail because:
You are the assignee for the bug.