grale.renderers
This module describes the ‘renderers’ you can use. By default, calculating
a lens plane mapping or a mass density on a grid, will only use a single
processor core. Functions in the plotutil module
take a renderer argument, with which you can specify an external process
to be used for this purpose. Depending on the type selected, this can use
multiple threads, distributed processes using MPI, or the GPU
using OpenCL, each of which can lead to a significant speedup.
A specified renderer can either be one of the objects that are present in this module, or it can simply be a string that describes the kind of renderer to use. For a lens plane renderer these strings (case insensitive) and objects can be:
“Threads” or an instance of
ThreadsLensPlaneRenderer“MPI” or an instance of
MPILensPlaneRenderer“OpenCL” or an instance of
OpenCLLensPlaneRenderer“NC:portnumber” or an instance of
NetcatLensPlaneRenderer
For a mass density renderer these strings (case insensitive) and objects can be:
“Threads” or an instance of
ThreadsMassDensityRenderer“MPI” or an instance of
MPIMassDensityRenderer“NC:portnumber” or an instance of
NetcatMassDensityRenderer
Defaults
- grale.renderers.setDefaultMassRenderer(x)
Sets the default mass renderer, see the module documentation for allowed values.
- grale.renderers.getDefaultMassRenderer()
Returns the default mass renderer that was set using
setDefaultMassRenderer().
- grale.renderers.setDefaultLensPlaneRenderer(x)
Sets the default lens plane renderer, see the module documentation for allowed values.
- grale.renderers.getDefaultLensPlaneRenderer()
Returns the default lens plane renderer that was set using
setDefaultLensPlaneRenderer().
Lens plane renderers
- class grale.renderers.ThreadsLensPlaneRenderer(numProcesses=None, feedbackObject=None)
Renderer that uses multiple threads to speed up the calculation of a lens plane.
- class grale.renderers.MPILensPlaneRenderer(numProcesses=None, feedbackObject=None)
Renderer that uses multiple processes (using MPI) to speed up the calculation of a lens plane.
- class grale.renderers.OpenCLLensPlaneRenderer(feedbackObject=None)
Renderer that uses OpenCL to use your GPU to speed up the calculation of a lens plane.
- class grale.renderers.NetcatLensPlaneRenderer(port, feedbackObject=None)
Forward rendering commands to another process over a TCP connection using the netcat (nc) tool.
On the other machine, you can use socat to redirect stdout/stdin of one of the render programs over the connection. For example
export GRALE_NUMTHREADS=36 socat TCP4-LISTEN:9999,reuseaddr,fork EXEC:grale_lensplane_threads
would wait for incoming TCP connections on port 9999, and start the thread based lens plane renderer when an incoming connection was detected.
To run one of the MPI based renderers, socat cannot be used (pipes are used for the communication instead of stdin/stdout). Instead, there’s a helper program grale_socket_to_mpi.py, for example:
grale_socket_to_mpi.py 9998 mpirun grale_massdens_mpi
- __init__(port, feedbackObject=None)
Connect to localhost (127.0.0.1) on the specified port number. The idea is to use SSH tunnels to connect to another machine.
Mass density renderers
- class grale.renderers.ThreadsMassDensityRenderer(numProcesses=None, feedbackObject=None)
Renderer that uses multiple threads to speed up the calculation of a mass density map.
- class grale.renderers.MPIMassDensityRenderer(numProcesses=None, feedbackObject=None)
Renderer that uses multiple processes (using MPI) to speed up the calculation of a mass density map.
- class grale.renderers.NetcatMassDensityRenderer(port, feedbackObject=None)
Forward rendering commands to another process over a TCP connection using the netcat (nc) tool. See
NetcatLensPlaneRendererfor examples.- __init__(port, feedbackObject=None)
Connect to localhost (127.0.0.1) on the specified port number. The idea is to use SSH tunnels to connect to another machine.