Explore Continuous CA with Ready

[Hektor]

Great, now it works, but most of the patterns give an error:

Code: Select all

An error occurred when running the simulation:

OpenCL_RD::Update : clEnqueueNDRangeKernel failed: Invalid work group size

Maybe I have the wrong version of OpenCL?

[Andrew]

An error occurred when running the simulation:
OpenCL_RD::Update : clEnqueueNDRangeKernel failed: Invalid work group size

Can you give an example of one pattern (not from the CPU-only folder) that fails and one that succeeds?

Do things improve if you use Action > Select OpenCL Device and select the CPU device?

Also show us the output from Action > Show OpenCL Diagnostics. Here's the info from my iMac:

Code: Select all

Found 1 platform(s):
Platform 1:
2304 : FULL_PROFILE
2305 : OpenCL 1.0 (Jan  2 2011 18:00:11)
2306 : Apple
2307 : Apple
2308 : 

Found 2 device(s) on this platform.
Device 1:
4139 : ATI Radeon HD 6970M
4140 : AMD
4141 : 1.0
4142 : FULL_PROFILE
4143 : OpenCL 1.0 
4144 : cl_APPLE_gl_sharing
Device 2:
4139 : Intel(R) Core(TM) i5-2400 CPU @ 3.10GHz
4140 : Intel
4141 : 1.0
4142 : FULL_PROFILE
4143 : OpenCL 1.0 
4144 : cl_khr_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_byte_addressable_store cl_APPLE_gl_sharing cl_APPLE_SetMemObjectDestructor cl_APPLE_ContextLoggingFunctions 

[Andrew]

Attached is a .vti file that implements a 3D cellular automaton using rule 4,5/5. This is the notation used in the applet created by Carter Bays and his colleagues at http://www.cse.sc.edu/~bays/d4d4d4/ (warning: the applet is very slow to load).

The initial pattern has a glider in one corner and a random blob in the middle.

Tim, creating this file revealed a couple of things we should add to the next version of Ready:

- An option to specify the rendered cell shape (octahedron, cube, sphere?).
- An easier way for overlays to initialize a single cell; eg. setcell3D x="5" y="3" z="2" value="1"
(have a look at the ugly way I created the glider).

I also noticed a bug: Load conway_life.vti then save it as another vti file. Loading the new file fails due to ampersand errors in the kernel code.

[Pilou]

Open the file 3D-Ca.vti inside ParaView works fine

I obtain a very big cube 100*100*100 made with many little Cubes

[Wojowu]

The ability to change look of cells (cubes, spheres) instead of mixed soup of cells would make 3D discrete CA better. Until then good choice is 3DLife.py from Golly Scripts Database

[Tim Hutton]

We've posted a new version, 0.2.2, that should hopefully fix the work group size problems mentioned above, and the save/reload issue. Please give it a go and let us know if it works.
https://code.google.com/p/reaction-diffusion/

[Wojowu]

Everything now works great, but pulsate.vti significantly slows down my computer. I know it may be fault of my hardware, but I want to ask: is that happening only to me?

[Hektor]

Everything runs fine for me now.

but pulsate.vti significantly slows down my computer

pulsate is a bit slow for me too, but I don't think it's ready's fault...