[solved] mve for fc painfully slow with low view-deviation settings

Here's the place for discussion related to coding in FreeCAD, C++ or Python. Design, interfaces and structures.
Forum rules
Be nice to others! Respect the FreeCAD code of conduct!
Post Reply
heda
Veteran
Posts: 1348
Joined: Sat Dec 12, 2015 5:49 pm

[solved] mve for fc painfully slow with low view-deviation settings

Post by heda »

recently i have given up on importing "real life" step models.
it has just been too slow and memory hungry (read throttling a decent cpu - all cores, and going to hd-swap on 32gb ram) - on step files that are not huge

do not remember this from earlier...

so far i have thought that it has been occt step importer, or regressions in versions somehow, and left it at that.

today i imported a model that i am pretty sure i imported last year without taking any significant notice of any slowness.
this time around i never got it imported, it just hogged down the whole system and i killed the imports, since you then know that even if the import finishes, fc is usually unusable.

in desperation i downloaded versions from 0.17 to 0.20 (all portable), they all behaved the same (pretty sure this was not the case earlier).
then started to break down the .step in one of the proprietary systems.
geo-checker in that gave some degenerate components.
those imported fine as step in fc

next i took a simple revolve with a few decorations.
that timed in at close to 6 minutes for importing into fc, peaked at +6Gb in memory, and this for a 26kb (text-readable) step...

still thinking this is something bogus with the step importer, tried the same with iges, which were as sluggish as the step - so no difference there.

next, set out to show how good fc is, so just recreated the revolve profile in fc.
to my disappointment, it was as slow as the import.

a step forward (no pun intended), we can rule out the importer...

then took the profile edge by edge to revolve a sheet, heureka - found it!
revolving an arc is what eats cpu and memory like there is no tomorrow!
tried with a clean file, same behaviour, even worse if one uses a sketched arc instead of a draft arc.

this was all on a win-machine

however, now i am at a complete loss
was going to make a mve on a linux-box for the forum,
and low and behold - it made the arc-revolve without breaking a sweat.
the infinite lower spec linux-box just completely knocks the socks off a highish spec win-box when it comes to an arc revolve.

what is going on here?
if it can be fixed/worked around on win-boxes, that would be superb - it is really unusable for anything related to this use case (which one supposes is a pretty common one)

how to check
from the file, just do a part-revolve of the line, should be instant
when doing the same revolve with the arc, you know immediately if it is slow, it will not be instant, it takes a minute and it will put the fan on overspeed.

do others see the same behaviour?

at last, starting to wonder if gpu has something to do with it...
Attachments
mve-slowfc-on-win.FCStd
(5.59 KiB) Downloaded 12 times
Last edited by heda on Tue Jul 19, 2022 7:14 pm, edited 1 time in total.
GeneFC
Veteran
Posts: 5373
Joined: Sat Mar 19, 2016 3:36 pm
Location: Punta Gorda, FL

Re: mve for fc painfully slow on windows

Post by GeneFC »

I did not see a problem. The arc revolve was slightly longer than the line, but it was still only about half of a second.

I do not have a blazing fast computer by any means. Intel i5 with onboard graphics. Lots of memory, but otherwise nothing special at all.

Gene
user1234
Veteran
Posts: 3340
Joined: Mon Jul 11, 2016 5:08 pm

Re: mve for fc painfully slow on windows

Post by user1234 »

heda wrote: Tue Jul 19, 2022 5:39 pm recently i have given up on importing "real life" step models.
it has just been too slow and memory hungry (read throttling a decent cpu - all cores, and going to hd-swap on 32gb ram) - on step files that are not huge
Edit --> Preferences --> Display --> Colors --> Enable preselection highlighting --> uncheck. Then assemblies are fast in handeling. Importing have the same speed as before.


heda wrote: Tue Jul 19, 2022 5:39 pm how to check
from the file, just do a part-revolve of the line, should be instant
when doing the same revolve with the arc, you know immediately if it is slow, it will not be instant, it takes a minute and it will put the fan on overspeed.

do others see the same behaviour?
No, works instant. I am to 90% sure, that your settings about Angular Deflection and Deviation is not proper adjusted.


Greetings
user1234
heda
Veteran
Posts: 1348
Joined: Sat Dec 12, 2015 5:49 pm

Re: mve for fc painfully slow on windows

Post by heda »

user1234 wrote: Tue Jul 19, 2022 6:22 pm 90% sure, that your settings about Angular Deflection and Deviation
make that a 100%, that was it, must be a reminiscence from fiddling with some dxf-files way back and having forgotten to turn that knob up again...

:D should have caught that myself, what a rookie mistake.

anyhow, thanks a whole bunch - things are back to normal now!
User avatar
adrianinsaval
Veteran
Posts: 5541
Joined: Thu Apr 05, 2018 5:15 pm

Re: [solved] mve for fc painfully slow with low view-deviation settings

Post by adrianinsaval »

when you encounter weird errors specific to one system always check if it also happens with a clean config
GeneFC
Veteran
Posts: 5373
Joined: Sat Mar 19, 2016 3:36 pm
Location: Punta Gorda, FL

Re: [solved] mve for fc painfully slow with low view-deviation settings

Post by GeneFC »

I always keep the Deviation cranked down as small as it will go, 0.01%. Angular does not do much, so I leave that one alone.

My results above were quoted based on my standard minimum Deviation.

In order to test I have again tried with the Angular set down to 1 degree. Still works almost instantly.

Gene
heda
Veteran
Posts: 1348
Joined: Sat Dec 12, 2015 5:49 pm

Re: [solved] mve for fc painfully slow with low view-deviation settings

Post by heda »

had 0.01 for bb and 0.5 for ad,
will check some different settings to see if there is a threshold to be found somewhere.

threshold for ad seems to be at 2.5, this is where there is a hint of delay, going to 2.0 the delay is for sure there,
and at 1.5 it is really noticeable (>5 sec for a simple shape)

as pointed out by genefc, bb setting does not seem to matter much for speed (for the simple test case at least).

this appears to be similar with lower and higher spec laptops
Post Reply