On 09.10.2020 13:04, Torsten Paul wrote:
On 09.10.20 12:55, David wrote:
Is it possible to compile and run OpenSCAD on an RPi and/or
a Jetson Nano/Xavier?
Yes, if Qt is available with Desktop OpenGL support.
That's by default the case for most arm64 system, e.g. latest
(still beta?) version of Raspberry Pi OS 64-bit.
Otherwise it's still possible but requires a custom Qt build
which is shown to work on an old 32bit Chromebook via Software
Rendering.
See also ARM64 AppImage on the download page
https://www.openscad.org/downloads.html#snapshots
If you have an OpenSCAD RPI executable for Raspberry PI4 you could try
this example for n=2 to n=16
module manyballs(n)
{
$fn=25;
delta = 45;
for(i=[0:1:n-1]) {
x = idelta;
for(j=[0:1:n-1]) {
y = jdelta;
for(k=[0:1:n-1]) {
z = k*delta;
translate([x,y,z])sphere(25);
}
}
}
}
manyballs(n=7);
Just for the fun of it I built https://github.com/arnholm/xcsg to 32bit
raspbian for RPI4 and ran the equivalent "manyballs" samples with xcsg
for n=2 to n=16.
It worked for n up to 9 and failed for higher values, probably because
of 32bit memory addressing limitations (not sure).
timings
xcsg, manyballs(n=9), RPI4 32bit raspbian 1m 47.75s
xcsg, manyballs(n=9), Ubuntu 18.04 64bit 0m 42.659s
So the Ubuntu/64bit workstation runs only about 2.5 times faster than my
RPI4/Raspbian/32bit for this kind of problem. Actually an interesting
result.
Comparing OpenSCAD 2020.10.05.nightly, and xcsg for n=7 case
openscad manyballs(n=7), Ubuntu 18.04 64bit 12m 59.87s
xcsg, manyballs(n=7), Ubuntu 18.04 64bit 0m 12.50s
xcsg, manyballs(n=7), RPI4 32bit raspbian 0m 48.72s
Carsten Arnholm
On 09.10.2020 13:04, Torsten Paul wrote:
> On 09.10.20 12:55, David wrote:
>> Is it possible to compile and run OpenSCAD on an RPi and/or
>> a Jetson Nano/Xavier?
>
> Yes, if Qt is available with Desktop OpenGL support.
>
> That's by default the case for most arm64 system, e.g. latest
> (still beta?) version of Raspberry Pi OS 64-bit.
>
> Otherwise it's still possible but requires a custom Qt build
> which is shown to work on an old 32bit Chromebook via Software
> Rendering.
>
> See also ARM64 AppImage on the download page
> https://www.openscad.org/downloads.html#snapshots
If you have an OpenSCAD RPI executable for Raspberry PI4 you could try
this example for n=2 to n=16
module manyballs(n)
{
$fn=25;
delta = 45;
for(i=[0:1:n-1]) {
x = i*delta;
for(j=[0:1:n-1]) {
y = j*delta;
for(k=[0:1:n-1]) {
z = k*delta;
translate([x,y,z])sphere(25);
}
}
}
}
manyballs(n=7);
Just for the fun of it I built https://github.com/arnholm/xcsg to 32bit
raspbian for RPI4 and ran the equivalent "manyballs" samples with xcsg
for n=2 to n=16.
It worked for n up to 9 and failed for higher values, probably because
of 32bit memory addressing limitations (not sure).
timings
xcsg, manyballs(n=9), RPI4 32bit raspbian 1m 47.75s
xcsg, manyballs(n=9), Ubuntu 18.04 64bit 0m 42.659s
So the Ubuntu/64bit workstation runs only about 2.5 times faster than my
RPI4/Raspbian/32bit for this kind of problem. Actually an interesting
result.
Comparing OpenSCAD 2020.10.05.nightly, and xcsg for n=7 case
openscad manyballs(n=7), Ubuntu 18.04 64bit 12m 59.87s
xcsg, manyballs(n=7), Ubuntu 18.04 64bit 0m 12.50s
xcsg, manyballs(n=7), RPI4 32bit raspbian 0m 48.72s
Carsten Arnholm
