Cluster for 3D
Posted: Sat Mar 12, 2022 8:34 pmHi!
Just to share that I've realized an application useful to visualize 3D functions in parametric form
x=x(k,t); y=y(k,t); z=z(k,t)
that is working using a double loop, one numeric (for i=0;i<width;i++) and the other one through the parallel calculation of cores both in a single board asset of 16 cores or using a cluster of Parallella boards using boards*16 cores (in theory up to 8 boards, 128 cores).
Using a single board (16 cores) or a cluster (for example my 64 cores cluster) it is possible to run multiple cycles in parallel: multiple of 16 cores if using a single board or multiple of 64 cores in a cluster of boards in order to process the evolution of trajectories of f(x,y,z) one by each core.
The Host program has a run.ini file in order to run options:
axis yes/no; frame yes/no; assonometry view yes/no; +-zoom value; rotation xy, zy, xz; boost x y or z; step length; number of parallel batteries of cores; number of boards used.
Each core run its own program to model and calculate its f( x(t,k) y(t,k) z(t,k) ) in a form of a template to realize the evolution of the parametric functions.
Actually I'm running and using a cluster of up to 4 and 5 boards with 64 and 80 cores in a cluster of epiphany chips via eLink (North/South gates).
I'm going to apply Runge Kutta applications for 3D derivation of f(x,y,z) through 3D trajectories.
I can only image what we could do if it would be possible to using an access to a Framebuffer or an HDMI output.... still I've not been able to use any HDMI version of the Bitstream.
I hope someone could be able to add an HDMI and a Frambuffer to visualize dinamic 3D functions.
I'm available to share cluster setting and my application to work with multi cores boards.
Just to share that I've realized an application useful to visualize 3D functions in parametric form
x=x(k,t); y=y(k,t); z=z(k,t)
that is working using a double loop, one numeric (for i=0;i<width;i++) and the other one through the parallel calculation of cores both in a single board asset of 16 cores or using a cluster of Parallella boards using boards*16 cores (in theory up to 8 boards, 128 cores).
Using a single board (16 cores) or a cluster (for example my 64 cores cluster) it is possible to run multiple cycles in parallel: multiple of 16 cores if using a single board or multiple of 64 cores in a cluster of boards in order to process the evolution of trajectories of f(x,y,z) one by each core.
The Host program has a run.ini file in order to run options:
axis yes/no; frame yes/no; assonometry view yes/no; +-zoom value; rotation xy, zy, xz; boost x y or z; step length; number of parallel batteries of cores; number of boards used.
Each core run its own program to model and calculate its f( x(t,k) y(t,k) z(t,k) ) in a form of a template to realize the evolution of the parametric functions.
Actually I'm running and using a cluster of up to 4 and 5 boards with 64 and 80 cores in a cluster of epiphany chips via eLink (North/South gates).
I'm going to apply Runge Kutta applications for 3D derivation of f(x,y,z) through 3D trajectories.
I can only image what we could do if it would be possible to using an access to a Framebuffer or an HDMI output.... still I've not been able to use any HDMI version of the Bitstream.
I hope someone could be able to add an HDMI and a Frambuffer to visualize dinamic 3D functions.
I'm available to share cluster setting and my application to work with multi cores boards.