IPython

[shr]

IPython looks like a good interface for exploring Parallella possibilities. Ha anyone checked this out yet?

[nealmcb]

ipython seems to work fine when installed in my kicktarter parallella-16. E.g. after I did an install of the necessary packages, which included this (and perhaps a few more I'm not remembering):

$ sudo apt-get install ipython-notebook python-matplotlib

I see some interesting notebooks, e.g. from the ipython source itself here is a static web version of their plotting tutorial:

http://nbviewer.ipython.org/github/ipyt ... tlib.ipynb

You can run that locally by getting the source and firing it up:

$ git clone https://github.com/ipython/ipython.git
$ cd ipython/examples/Notebook directory and run
$ ipython notebook --pylab inline

and chose the matplotlib example.

There are lots more tutorials and examples of notebooks on the web:

https://github.com/ipython/ipython/wiki ... -Notebooks

Now if we can just get numpy and scipy using the Epiphany cores, we'll be in great shape!

[hochopeper]

IPython itself can be used for distributed computing but my understanding is it would be better at controlling a cluster of Parallellas and the individual Parallella boards using Numba or similar to accelerate the individual workloads. See the other thread in here about LLVM being necessary for Numba. Since Numba is the brainchild of one of the Numpy original developers I think it is a very interesting development! I've been playing around with IPython on a beaglebone (white) to get used to the Python on arm limitations recently, haven't hit any significant stumbling blocks for my projects, yet.



I use IPython Notebooks all the time for home and work projects so very much enjoy the workflow that it enables and having that with a Epiphany accelerated computation engine would really tick a lot of boxes IMO. I just need to learn a bit more FPGA stuff because I've got some more ADC and DAC interfaces to hook up.