[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [pygame] PyOpenGL


On Aug 8, 2006, at 8:10 PM, Simon Wittber wrote:

On 8/9/06, Bob Ippolito <bob@xxxxxxxxxx> wrote: 
Eh? What does scipy or numarray have to do with it? Totally different
packages. Especially scipy.


They all provide facilities for array manipulation.

import Numeric
import numarray
import scipy
n = 1500
a_numeric = Numeric.arange(1., n)
a_numarray = numarray.arange(1., n)
a_scipy = scipy.arange(1., n)

The arrays you get (from scipy trunk anyway) are provided by numpy, but were provided by other packages in the past. SciPy doesn't actually implement any of that stuff, it's simply provided in the namespace as a convenience for users.

On a different node, I've attached CSV file with results from this
benchmark program, with a third test for scipy.

http://arbutus.mcmaster.ca/dmc/numpy/numcomp.py

Interestingly numarray consistently beats numeric and scipy with
arrays of length greater than 1500. With arrays smaller than 1500,
SciPy and Numeric are often on par with each other.

Beating Numeric's performance on large arrays was the whole point of numarray. I'd imagine that beating both Numeric and numarray in performance is an eventual goal of numpy, but I wouldn't expect a whole lot given its age. However, it seems that the community is gravitating towards numpy and away from both numarray and Numeric, so it may not take all that long.

-bob