[CWB] CWB4 and Ziggurat

Stefan Evert stefanML at collocations.de
Sat Oct 10 09:47:08 CEST 2015


> On 9 Oct 2015, at 20:45, Yannick Versley <yversley at gmail.com> wrote:
> 
> I'd suggest to do it in Python if the Python interpreter weren't quite as bad as it is.  Slower than R, seriously?
> Slower than R would be weird. See here for a semi-scientific comparison
> http://www.johndcook.com/blog/2014/06/20/benchmarking-c-python-r-etc/ 

Sure, but it's unlikely that we're ever going to solve the "stochastic neoclassical growth model, the workhorse of modern macroeconomics" with CQP.

And I find it a little suspicious that CPython is much faster on their Windows machine (even if you take relative times compared to C++). Usually, people on mailing lists (mostly R, I have to admit) ask why their scripting code is slower on Windows than Linux.

Here's what I get with a simple count loop of the form

	for x in xrange(N+1):
		sum = sum + x


python/simple_count.py               7.47 Mops/s  (20.0M ops in 2.68 s)
py3/simple_count.py                  7.62 Mops/s  (20.0M ops in 2.62 s)
R/simple_count.R                    15.59 Mops/s  (20.0M ops in 1.28 s)
perl/simple_count.perl              18.17 Mops/s  (40.0M ops in 2.20 s)
lua/simple_count.lua                97.12 Mops/s  (100.0M ops in 1.03 s)

Now tell me that less than 8 million loops per second isn't just sad.  I was going to say that even bash scripting has got to be faster than this, but …

bash/simple_count.sh                 0.07 Mops/s  (0.2M ops in 2.92 s)


Best, and thanks for the additional pointers to interesting software packages,
Stefan





More information about the CWB mailing list