Polygon over 24 sides hangs = must kill Alibre in TaskMngr

[DingBEN]

Alibre Design Expert V27
WIN 10, DirectX 12
9th gen Intel Core 5 i5-9400F Processor (2b) 64bit/(32bit) - 2.90GHz
Chip type: NVIDIA GeForce GTX 1660 Ti
full hardware specs attached...

All optimizations per the instructions on the Alibre website Performance section.
System was mostly already set correctly, except I had to point the GPU to the \Program Files\alibre.exe in the OS settings.
Note: the GPU is in 32bit mode(in attached), I don't know if I can/should change that?
...is there something in the BIOS that would need optimizing?

First install... just practicing features/sketching figures thus nothing on the screen, one or 2 polygons and a few lines.
Polygon 24 sides - press Apply - all good... but takes 8-15secs to draw the sides-nodes and release operation
I may have succeeded with a 28 or 30 sided attempt, not sure... a 32 froze. I tried many attempts at various settings.
I need a 48... system just freezes, have waited 1 1/2 hour to see if it would eventually complete... each time, I had to kill Alibre in TaskMngr
Played with resolution, DPI... zero improvement.
No other issues of snagging, slowness, delays, hick ups, ...everything responds well.
I can create a circular pattern of 12 count feature, without issue.

Thank you for any help you may provide.

 

[simonb65]

Does the same in v26. 24 sides takes a bout 6-10 seconds on my desktop (spec in signature) ... and if you hover the mouse over the sketch during that 6-10 seconds you get a crash ...



Tried 30 sides about 5 minutes ago, still hasn't returned after hitting apply! Response time looks exponential based on number of sides (typical classic symptom of linear processing that restarts as it iterates through the number of data points!).

Edit: 30 sides took 6 minutes 14 seconds!

Should be very easy for Support to repeat this one! You need to raise a ticket unless someone from Alibre picks this topic up.

 

[stepalibre]

Memory usage spikes when using that tool. It must be calculating each side + angles in an inefficient way or something. Turning off constraint symbols and other sketch options sped up compute time on my machine.

 

[DingBEN]

Memory usage spikes when using that tool. It must be calculating each side + angles in an inefficient way or something. Turning off constraint symbols and other sketch options sped up compute time on my machine.

Does the same in v26. 24 sides takes a bout 6-10 seconds on my desktop (spec in signature) ... and if you hover the mouse over the sketch during that 6-10 seconds you get a crash ...

View attachment 40018

Tried 30 sides about 5 minutes ago, still hasn't returned after hitting apply! Response time looks exponential based on number of sides (typical classic symptom of linear processing that restarts as it iterates through the number of data points!).

Edit: 30 sides took 6 minutes 14 seconds!

Should be very easy for Support to repeat this one! You need to raise a ticket unless someone from Alibre picks this topic up.

Simon: Thank you so much for chiming in. This is a match. Everything you stated I suspected, especially the exponential factor. Will save me to test it in V26 and I see that you have 4 times the memory yet it does not help! Thus adding memory will not be helpful.
It may rely strictly on the GPU mem.
This is disappointing, especially knowing V26 has the same issue. Perhaps there is a setting that needs to be tweaked. I just can't see they would let the software out knowing the issue. Could be an even easier fix.

Stepalibre: Can you report a time benchmark on a 48 sided polygon, and what you actually turned off besides the constraints symbols?

I am putting in a tech ticket, pointing to this thread.

Thank you everyone.

 

[stepalibre]

48 ~5 min
30 - ~1.5min
I normally work with all sketch options toggled off, but sketch options should not affect the tool in this way.

 

[Ex Machina]

This is definitely something for support to look over. Especially since my parametric polygons calculate significantly faster than this.

In the meantime check this part out this part and this method of creating polygons. In my systems it goes from 6 sides to 32 sides in under a second.
I'm also attaching the video I made on it:

P.S. If this works this much faster, you can save the sketch as a catalog feature and keep adding it when you need to make polygons. It'll also bring it's own parameters in and now you can control the number of sides even after making the polygon. I have another video on how to use Catalog Features. It's one of the top 3 most powerful tools in Alibre and most people don't know about it.

 

[JimCad]

Thanks Konstantinos,

I tried 48 sides and my system hung.
I'll watch your video tonight and try it out.
Jim

 

[Ex Machina]

Hmmm, I'll try 48 too and see what happens. But it did 32 so easily.

 

[Ex Machina]

So, here's me changing the sides count of that file. It seems to do it pretty quickly:

And this is our company's backup system, so:

Ryzen 7 4800h
16GB RAM
RTX3050 4GB VRAM

I haven't tried it on the fixed inscribed circle polygon cut, but I don't expect it to be much slower...

 

[simonb65]

It has to be an issue in the way the constraints are added/applied/calculated, because the math to calculate a number of vertices/nodes based on the centre point x,y,(z) and a radius is very simple!

 

[DavidJ]

Konstantinos' approach has a lot less 'stuff' in the sketch compared to the polygon tool. That is probably significant.

 

[simonb65]

So, here's me changing the sides count of that file. It seems to do it pretty quickly:

I don't think modifying is the issue, it's the original geometry creation the the first instance and application of constraints that is the bottleneck!

 

[Ex Machina]

Konstantinos' approach has a lot less 'stuff' in the sketch compared to the polygon tool. That is probably significant.

But I wonder if you take that isosceles generator triangle, make it 2 reference line and one geometry one, and pattern it around in sketch, would it be any slower? I think it would be the same speed.

P.S. I tried doing it in sketch. Two problems:
1) The pattern count in sketch cannot be driven by a parameter
2) The resulting pattern is underdefined. The sides can be moved by dragging.

So, it's not a valid alternative. But it is lightning fast.

 

[DingBEN]

Is anyone wondering, that it takes a noob like me to unearth a major issue of this type?
Seems this should have been found out already, and fixed, or instructions provided to resolve the issue.
I have this crazy question mark spinning erratically above me occipital bone!

I will be back on this in a couple of days, and attempt to do creative shtuff like Konstantinos.
I had thought to do it via concentric circles with degreeeeed rays... pattern circle tool.
...and that means having to cutout unnecessary geometry, or creating reference material.

My pinball machine design is stagnating in the Alibre speed swamp for the time being.

 

[Ex Machina]

Ah, OK. By the way, why do you need a regular 48 sided polygon?

 

[HaroldL]

Is anyone wondering, that it takes a noob like me to unearth a major issue of this type?

Not really. These kinds of things come up once in a while.

Would you be surprised to know that when I was working with SolidWorks we would find instances where it could not complete the feature we needed? We had to rely on a workaround (different workflow) to get the job done. And look how "mature" SolidWorks is supposed to be.

 

[Ex Machina]

Yeah, Harold. That came across as a bit of a toxic and ignorant comment by Ding... But in written speach sometimes some of the content and intent is lost. And he is a self proclaimed noob, so it happens.

 

[simonb65]

Yeah, Harold. That came across as a bit of a toxic and ignorant comment by Ding... But in written speach sometimes some of the content and intent is lost. And he is a self proclaimed noob, so it happens.

As a software developer myself, I don't see it as toxic or ignorant ... there are, and have been plenty of cases in the past, of things being found by users that should be found by a) stress testing and b) QA. Some do slip the net, but any feature that has user defined parameters should have known limits and behaviour, even if those numbers may seem a little excessive. The way to control it in software is to find acceptable 'bounds' through development and stress test, then limit the number that can be entered as a parameter in order that the application behaves in a known and predictable way.

If it not a bug, but a limiting performance issue of the geometry engine then the 'sides' field should only allow values from 3 to 30 ... until a time that a) performance can be improved or b) a different method of implementing the feature in code can be achieved that allows the upper bound to be increased.

Maybe @DingBEN comment may have been interpreted in a different context, I personally agree with the sentiment! I wouldn't just put it down to him been a 'noob', as these are the people that intuitively expect software to work, without having to learn it's limitations by trial and error.

 

[stepalibre]

I agree with Simon, it should have an upper bound to account for the performance issues of the tool where the number of sides can hit unstable state.

Another issue is that the tool includes sketch relations/constraints. This adds overhead and is separate from the regular polygon code. I would guess the problem is there and not with the polygon code, since it is simple symmetrical geometry.

This code runs in ~1-2 seconds without sketch relations/constraints:

import math
Diameter = 100
Sides = 60
EDia = Diameter / math.cos(math.pi / Sides)
P = CurrentPart()
S = P.AddSketch('Regular_Polygon_60_Sides_Dia_100', P.GetPlane('XY-Plane'))
S.AddPolygon(0, 0, EDia, Sides, False)

 

[DingBEN]

...thank you Simon... you interpret my candid statement as intended.
...and that is the way everyone else should take it, no animosity, no malice, just fact expressed thru surprise.
...and to hear that some of the more mature software on the market suffers comparable issues, warns me
to not feel as bewildered as I will undoubtedly run into other instances.

As to the limit, I have tinkered with other trial software that allowed high limits and performed quite well.
Incidentally, the limit for Alibre Design Expert, I read somewhere in their docs, is 100.

Yes, I am a noob when it comes to additive/subtractive manufacturing technology[CNC/3D printing] and CAD
I have no shame in exposing my level explicitly.

As to why I need 48 sides... the feature in my design is a circular blind maze bunker that faces 24 separate and different 'respondors', dynamically friend or foe, all around it, thus a need to have an opening for each... the polygon seemed to be the more natural and matching geometry.

I hope this smooths out any emotional bumps.