comparing interpreter speed on two platforms

[Chris Liechti]

i have run a simple benchmark on two systems:

a standard PC:
1. PentiumPro 200Mhz (400BogoMips) Debian Linux 2.2 (Kernel 2.2-17) 64M RAM

and an embedded system:
2. PowerPC    ~60Mhz (68BogoMips)  HardHat Linux 2.0 (Kernel 2.4) 16MB RAM

the speed difference based on BogoMips (which are displayed when Linux is 
booted) is about 6.2.

when i compare the results of the benchmark programm which is entirely 
written in python i get differences of factor 20.

both target run Python 2.1.1, compiled from the same tarball with 
"./configure;make;make install", both compiled with GCC.

why is the python programm that much slower than the BogoMips comparison 
would imply?

(ideas: maybe "configure" finds some speedy modules on one system but not 
on the other, maybe the GCC compiler is not that efficient on PPC?
its not the memory. the embedded system has more than 3MB free during the 
tests)

the benchmark tested object instatiation, object identity, value assignment 
to a list, function calls and some math, every operation in a separate test 
with separate results (execution times, measured with "time.clock()"), 
performing 10000 loops.

any ideas would be apreciated, as my real application fails to run on the 
embedded system because its too slow, which is sad. (i have 20 Byte packets 
of data on the serial line which arrive 4 times a second, that is given and 
cannot be changed)

Chris

-- 
Chris <cliechti at gmx.net>