Features include:
a) subvolume generation and saving
b) Support for Fuzzy Volume generation
c) function value interpolation APIs
d) lots of improvements to UI and standard set of bug fixes..
To get a break of what i generally do, i am (re) learning a bit of windows programming and with it am writing a small set of C++ APIs for doing windows programming on my old Win98 machine. Its called weon and is downloadable from http://tovganesh.googlepages.com/weon.tar.gz
To create and open a window its as simple as:
include "weon.h"
CWeonApp app;
WEON_APP_MAP(app)
weon is under development and right now you cant do much with it. but i promise to make it simple to read and use :)
And with the help of APR and RK we have written a small set of APIs for doing MPI style programming purely using Python. Called pimpi its available from: http://tovganesh.googlepages.com/pimpi.tar.gz
A small test program that uses pimpi looks like this:
import pimpi
import sys
pimpi.init(sys.argv)
noOfProcessors = pimpi.size()
myRank = pimpi.rank()
print "Total number of processors: ", noOfProcessors
print "My rank: ", myRank
if (myRank == 0):
for i in range(1, noOfProcessors):
pimpi.send(i, "hi")
else:
print pimpi.recv()
data = pimpi.bcast(0, "bhello")
print data, myRank
data = pimpi.bcast(0, repr((1,2,3)))
print data, myRank
pimpi.finalize()
To run the above code you use:
python pimpiboot.py test.py
Note that pimpiboot.py reads a file called "node.list" that contains the nodes which you want to use in parallel. You need to setup your cluster environment for password less rsh/ssh access for pimpi to work correctly.
... look back here for more updates on these codes :)
No comments:
Post a Comment