SkeinPyPy - Skeinforge, but then fast!
February 14, 2012 09:41AM
Just dropping by from the Ultimaker groups/forums, to let you know I've made a pre-packaged version of Skeinforge. Called SkeinPyPy: [github.com]

This version is packaged to use "PyPy" which is a faster Python engine. This seriously cuts into the slicing time, i've measured up to 4x faster slicing.
Configuration is saved in $HOME/.skeinforge_pypy, so it does not mess with your normal skeinforge profiles.
The Skeinforge default configuration has been modified to be working for an Ultimaker with Marlin firmware.
The Width/Thickness settings have been changed into "Width" settings. This makes more sense for step extruders like the Ultimaker.
The "slowdown first layer" setting has been modifed to be "slowdown first X layers", so you gradually increase the speed over the first X layers. Giving a better bottom layer if you print without raft.

I have more plans, but this is it for now. Have fun with it. (The git source version is a bit ahead of the Alpha4, but the Alpha4 has been tested)
Re: SkeinPyPy - Skeinforge, but then fast!
February 15, 2012 05:23AM
Nice, will give it a test. Was about time I update my skeinforge anyway (still at 45 here).


Edit : Done some testing.

Works fine, but the absence of the display plugins is a problem, I like to visually inspect layers output in skeinlayer before printing.
I like the few changes to stock skeinforge parameters (Width instead of width over height ratio, 1st layer travel feed rate multiplier...)


Speed test for a moderately complicated STL (thingiverse # 1151 @ 0.2mm layers) :

SF45            first run 5m28s   2nd run 5m03s   3rd run 5m56s
SF49            first run 4m58s   2nd run 4m25s   3rd run 4m19s
SPPalpha4       first run 1m42s   2nd run 1m36s   3rd run 1m35s
Slic3r 0.6      first run 0m58s   2nd run 0m56s   3rd run 0m56s

Edit2 : adding Slic3r times with as similar settings as possible for reference.

So, nice speed improvement over stock SF ! Note that SF4x was run through "skeinforge_utilities/skeinforge_craft.py" which does a little less than "skeinforge.py", so it's not a perfect comparison.



The output gcode of SF49 and SPP are nearly identical, provided you set the different parameter to have the same behaviour. The difference lies in decimals, for instance first difference is :

< G1 X19.992 Y43.368 F1152.0 E2.5783 (SF49)
---
> G1 X19.992 Y43.367 F1152.0 E2.5783 (SPP)


Luckily that "error" doesn't seem to accumulate on the axis except for the extruder. Last difference :

< G1 E3152.478 (SF49)
---
> G1 E3154.0784 (SPP)

1.6mm difference out of 3100+ mm extruded is probably not a problem.
That might also be caused by me using SF49 instead of 48 in SPP.




Question, would that packaging work on a non-intel platform ? I got that really slow ARM ultralight laptop running linux which is handy when moving the reprap around, but it's a pain to skeinforge anything on that because of the time it takes.

Edited 2 time(s). Last edit at 02/15/2012 08:18AM by DeuxVis.


Most of my technical comments should be correct, but is THIS one ?
Anyway, as a rule of thumb, always double check what people write.
Sorry, only registered users may post in this forum.

Click here to login