Inkscape very slow at high-level zoom

Bug #484151 reported by EricG on 2009-11-17
56
This bug affects 11 people
Affects Status Importance Assigned to Milestone
Inkscape
Medium
Unassigned

Bug Description

Inkscape very slow at high-level zoom

Vista32, all recent Inkscape versions up to 0.47pre4-1

I'm used to create maps from GIS vector data which have usually 20,000 or more vertices and at a re-sized dimension up to 4,000 px. When editing these contours with F2 tool at a zoom level of 2,000 % or higher (but it happens also even at 50 %), editing becomes frequently very slow, meaning Inkscape may have difficulties to handle all this amount of information.

I noticed also that after navigating in the image through the scroll bars or wheel, the reactivity of Inkscape drops down drastically, selecting a single object (even very simple) with F1 can take several seconds, this when the whole image is a bit “complicated” with several thousands of points. Only zooming in (or out) and back brings the things in a normal state of reactivity again.

My configuration: Vista32, Intel Dual Core E8400, RAM 2 Go

su_v (suv-lp) on 2009-11-17
tags: added: performance renderer
ubby (kostas-sytske) wrote :

I also noticed it, I am using Inkscape 0.47pre4 r22446, built Oct 15 2009 on Ubuntu 9.10 64 bit.

LucaDC (dicappello) wrote :

Grouped objects also have impact on this: when at high zoom level, if you have some objects grouped and experience the slow down, ungrouping them gives a boost.

Sometimes the slow down is so heavy that I can't double click on objects anymore, as the two clicks (regardless of how fast I can do them, and I'm fast!) are always managed as two single clicks. Zooming out decreases the slow-down and double click restarts working.

Windows XP SP3.

jazzynico (jazzynico) wrote :

Hi Eric,

I rarely use such a big file, and thus can't confirm this issue.
Could you please attach a test file so that we can try to reproduce it?
Thanks.

Changed in inkscape:
importance: Undecided → Medium
status: New → Incomplete
tags: added: zoom
Marcel Partap (empee584) wrote :
Download full text (4.4 KiB)

Ok i _think_ i might be onto something. Today i got myself familar with using oprofile, and i gave it a shot (trying to debug starcraft II demo stalls under wine, actually ^^)...
The attached file is from an oprofile run i started when a complex document was fully loaded already. Nothing complex in the current viewport - only some text. Panning works satisfactory. However, zooming in and out takes FAR longer than it should - half a minute for a single canvas update...
This oprofile report is just from zooming in and out at a rather uncomplex position in a moderately complex text collage document. Seems to narrow down to the bounds_exact_transformed() function!?
And seems to be from this backtrace...
Can someone verify this?

Breakpoint 1, bounds_exact_transformed (pv=std::vector of length 2, capacity 2 = {...}, t=...) at helper/geom.cpp:159
#0 bounds_exact_transformed (pv=std::vector of length 2, capacity 2 = {...}, t=...) at helper/geom.cpp:159
#1 0x000000000058e9c5 in nr_arena_glyphs_update (item=0x50a1dc0, gc=0x7fffe8bbf130) at display/nr-arena-glyphs.cpp:120
#2 0x0000000000593071 in nr_arena_item_invoke_update (item=0x50a1dc0, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#3 0x000000000058ffb3 in nr_arena_group_update (item=0x509eb40, area=0x0, gc=0x7fffe8bbf2d0, state=510, reset=510) at display/nr-arena-group.cpp:176
#4 0x0000000000593071 in nr_arena_item_invoke_update (item=0x509eb40, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#5 0x000000000058ffb3 in nr_arena_group_update (item=0x509fb00, area=0x0, gc=0x7fffe8bbf470, state=510, reset=510) at display/nr-arena-group.cpp:176
#6 0x0000000000593071 in nr_arena_item_invoke_update (item=0x509fb00, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#7 0x000000000058ffb3 in nr_arena_group_update (item=0x5070dc0, area=0x0, gc=0x7fffe8bbf610, state=510, reset=510) at display/nr-arena-group.cpp:176
#8 0x0000000000593071 in nr_arena_item_invoke_update (item=0x5070dc0, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#9 0x000000000058ffb3 in nr_arena_group_update (item=0x1bde9a0, area=0x0, gc=0x7fffe8bbf7b0, state=510, reset=510) at display/nr-arena-group.cpp:176
#10 0x0000000000593071 in nr_arena_item_invoke_update (item=0x1bde9a0, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#11 0x000000000058ffb3 in nr_arena_group_update (item=0x1bdedc0, area=0x0, gc=0x7fffe8bbf950, state=510, reset=510) at display/nr-arena-group.cpp:176
#12 0x0000000000593071 in nr_arena_item_invoke_update (item=0x1bdedc0, area=0x0, gc=<optimized out>, state=510, reset=510) at display/nr-arena-item.cpp:266
#13 0x00000000009709c5 in sp_canvas_arena_update (item=0x32ed080, affine=..., flags=<optimized out>) at display/canvas-arena.cpp:164
#14 0x00000000005ab276 in sp_canvas_item_invoke_update (item=0x32ed080, affine=<optimized out>, flags=2) at display/sp-canvas.cpp:292
#15 0x00000000005ad2af in sp_canvas_group_update (item=0x32b71b0, affine=..., flags=2) at display/sp-canvas.cpp:785
#16 0x00000000005ab276 in sp_canvas_item_invoke_up...

Read more...

Marcel Partap (empee584) on 2011-08-18
Changed in inkscape:
status: Incomplete → New
Marcel Partap (empee584) wrote :

Seems this is caused by commit
rev10451 Merge SPCanvasArena caching layer work
Further analysis shows that bounds_exact_transformed() is not actually all the problem, pixman_op has the biggest chunk of CPU cycles. Here's a call graph, recorded for a 'zoom on empty space in otherwise complex document' event. Inkscape 0.48.2 doesn't have the problem so for now, i have rolled back.

Krzysztof (tongue tumbler alert, lol) please look at this again ;)

I have these problem to...

mint14 Mate edition

i have a core i7 and optimos Geforce gt540

jazzynico (jazzynico) on 2013-07-09
Changed in inkscape:
status: New → Triaged
Beluga (buovjaga) wrote :

Is this still happening with 0.91 or 0.92? Example file would be cool.

Christian (zwixx) wrote :

Yes, with a complex picture it's not really working. 20 Seconds on a Intel i7 6790K if I try to zoom in.

To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Duplicates of this bug

Other bug subscribers