Another approach to assembly solver (A2plus)

[kbwbe]

I'm facing a quite slower importing of complex shapes, I wonder if multithreading the conversion from shape to shell could speed up that process or, as alternative approach, the rewriting of those functions as C code.

Hi Turro,
there should be no changes in master branch which have slowed down the imports. Do you register that importing got slower or are you importing bigger shapes bigger than before ?

> Multiprocessing/Multithreading: May be better, i have to check how good this works with FC.
> C-code: Sure faster, but i have no idea about implementing it into FC.

Additional:
I am working on a big update within the devel branch. A2plus creates "xxx.a2p" files during importing of files. The files contain all resulting work of an import process (shape, colours, toponames) and speed up a second time import by factor 3-4.

[Turro75]

Hi Turro,
there should be no changes in master branch which have slowed down the imports. Do you register that importing got slower or are you importing bigger shapes bigger than before ?

> Multiprocessing/Multithreading: May be better, i have to check how good this works with FC.
> C-code: Sure faster, but i have no idea about implementing it into FC.

Additional:
I am working on a big update within the devel branch. A2plus creates "xxx.a2p" files during importing of files. The files contain all resulting work of an import process (shape, colours, toponames) and speed up a second time import by factor 3-4.

Hi Kbwbe,

Yes I'm talking about complex objects with many curved surfaces which creates tons of faces.

by analyzing a2p_muxassembly.py it looks like the slowness is mainly caused by Part.Makeshell(faces), other iterations look such as creating the faces list so I don't expect a big improvement in parallelizing them. Am I wrong?

Regarding C code.... rewriting the workbench in C IMHO requires too much work compared to the benefit, it would be interesting writing complex calculation task in a .dll and .so and importing it in python, it could be helpful in replacing some numpy or placement complex iterations within the solver (if any).

threading across cores is quite easy in python, from https://docs.python.org/3/library/multiprocessing.html

Code: Select all

from multiprocessing import Pool

def f(x):
    return x*x

if __name__ == '__main__':
    with Pool(5) as p:
        print(p.map(f, [1, 2, 3]))
        
        

[kbwbe]

Hi Turro,

Yes I'm talking about complex objects with many curved surfaces which creates tons of faces.

by analyzing a2p_muxassembly.py it looks like the slowness is mainly caused by Part.Makeshell(faces), other iterations look such as creating the faces list so I don't expect a big improvement in parallelizing them.

I am not quit sure, but Part.Makeshell(faces) cannot be parallelized, as OCC is only using one core/process. Please tell me if i am wrong.

Regarding C code.... rewriting the workbench in C IMHO requires too much work compared to the benefit,

Yes...

it would be interesting writing complex calculation task in a .dll and .so and importing it in python, it could be helpful in replacing some numpy or placement complex iterations within the solver (if any).

1) I do not have any experience in interfacing python with C/C++, that's my problem
2) The ".dll" and ".so" files would have to be compiled for each OS, probably by the user. Using pure python is much more easy for many users.

threading across cores is quite easy in python

I know, i have been working with that on other projects.

Are you using the toponaming option of A2plus? Using this needs some computation time.

[kbwbe]

Hi people,

A2plus V0.4.33 has a new feature (thanks to a user request). Now you can click a part and move it around under rule of all constraints.

Have a look at this video:

phpBB [video]

Have fun.

Merry christmas to all,
Klaus

P.S: Find the assembly attached (for your own experiments)

[Christoph]

Hi people,

A2plus V0.4.33 has a new feature (thanks to a user request). Now you can click a part and move it around under rule of all constraints.

I'm impressed, thank you very much creating this feature!

[dan-miel]

A2plus V0.4.33 has a new feature (thanks to a user request). Now you can click a part and move it around under rule of all constraints.

kbwbe:
I haven't tried this yet but it is something that I wanted to ask for but, again a but, the animations were so slow I never dreamed that this would be possible. Fantastic!

Dan

[roerich_64]

Hi Klaus,

Also from me: thank you for the new future:-)

Have a nice Christmas with your family and a happy new year:

IMG-20191222-WA0001.jpg (174.55 KiB) Viewed 2189 times

Bests
Walter

[uwestoehr]

A2plus V0.4.33 has a new feature

Documented: A2plus_Workbench#Dragging

[kbwbe]

A2plus function "importPart" has been broken today for a short time on master branch. Please update to new version V0.4.34a
Sorry!

[Syres]

Using v0.4.35, when deleting constraints I was occasionally getting the following error in the Report View but not consistently:

Code: Select all

Traceback (most recent call last):
  File "C:\Users\JPS\AppData\Roaming\FreeCAD\Mod\A2plus\a2p_viewProviderProxies.py", line 126, in getIcon
    FreeCAD.ActiveDocument.getObject(self.constraintObj_name).Type
<class 'AttributeError'>: 'NoneType' object has no attribute 'Type'

I modified my a2p_viewProviderProxies.py file at line 124 from:

Code: Select all

    def getIcon(self):
        return a2plib.A2P_CONSTRAINTS_ICON_MAP[
            FreeCAD.ActiveDocument.getObject(self.constraintObj_name).Type
            ]

to:

Code: Select all

    def getIcon(self):
        if FreeCAD.ActiveDocument.getObject(self.constraintObj_name).Type != None:
            return a2plib.A2P_CONSTRAINTS_ICON_MAP[
                FreeCAD.ActiveDocument.getObject(self.constraintObj_name).Type
                ]

and the error hasn't reappeared since.

OS: Windows 7 SP 1 (6.1)
Word size of OS: 64-bit
Word size of FreeCAD: 64-bit
Version: 0.19.19060 (Git)
Build type: Release
Branch: master
Hash: 72065b0d3960ec9b4aa90cce1a6e810e8bab68c4
Python version: 3.6.8
Qt version: 5.12.1
Coin version: 4.0.0a
OCC version: 7.3.0
Locale: English/United Kingdom (en_GB)