[SciPy-user] Benchmark data

[Arnd Baecker]

Hi,

enclosed is a short example of a benchmarking test for basic
functions. This makes use of the whole machinery
already provided by scipy.test, in particular
of `measure` which is also used in the tests of fft.

On my PIII 1.2 GHz laptop I get:
python test_bench.py
Importing test to scipy
Importing base to scipy
Importing basic to scipy
Python 2.3.5 (#2, Sep  4 2005, 22:01:42)
[GCC 3.3.5 (Debian 1:3.3.5-13)]
Optimization flags: -DNDEBUG -g -O3 -Wall -Wstrict-prototypes
CPU info: getNCPUs has_mmx has_sse is_32bit is_Intel is_Pentium
is_PentiumII is_PentiumIII is_i686 is_singleCPU
Numeric: 23.8
numarray: 1.1.1
scipy: 0.8.2.1623
Running pystone...
Pystone(1.1) time for 50000 passes = 2.61
This machine benchmarks at 19157.1 pystones/second

==== bench arange ====
      size        Numeric     numarray        scipy
         10          0.12         1.77         0.18  (secs for 10000
calls)
        100          0.20         1.75         0.25  (secs for 10000
calls)
       1000          0.44         0.89         0.41  (secs for 5000 calls)
      10000          0.77         0.20         0.66  (secs for 1000 calls)
     100000          0.79         0.05         0.67  (secs for 100 calls)
.
==== bench function: exp  ====
      size        Numeric     numarray        scipy
         10          0.10         0.29         0.17  (secs for 10000
calls)
        100          0.28         0.45         0.32  (secs for 10000
calls)
       1000          0.99         1.03         0.98  (secs for 5000 calls)
      10000          1.93         1.82         1.83  (secs for 1000 calls)
     100000          2.13         1.81         1.72  (secs for 100 calls)
.
==== bench function: sin  ====
      size        Numeric     numarray        scipy
         10          0.10         0.28         0.16  (secs for 10000
calls)
        100          0.20         0.38         0.24  (secs for 10000
calls)
       1000          0.62         0.67         0.60  (secs for 5000 calls)
      10000          1.19         1.08         1.07  (secs for 1000 calls)
     100000          1.30         1.15         1.09  (secs for 100 calls)
.
----------------------------------------------------------------------
Ran 6 tests in 36.768s

So, apart from arange, it looks very good for scipy!

Maybe some developer can have a look over the code
(which is a combination of bench.py, the tests in fftpack
and some additions),
and if it is of sufficient interest it could maybe serve
as a basis for adding more routines from all over scipy
to get detailed performance coverage?

Best,

Arnd
-------------- next part --------------
A non-text attachment was scrubbed...
Name: test_bench.py
Type: text/x-python
Size: 6912 bytes
Desc: 
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20051210/4b8d7cc4/attachment.py>