[Python-Dev] Re: Python 2.1 slower than 2.0

[Mark Hammond]

"M.-A. Lemburg" wrote:
> what results do you get on Windows ?

Win2k, dual 800, relatively quiet!

Python 2.0

F:\src\Python-2.0\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.847605
This machine benchmarks at 11798 pystones/second

F:\src\Python-2.0\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.845104
This machine benchmarks at 11832.9 pystones/second

F:\src\Python-2.0\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.846069
This machine benchmarks at 11819.4 pystones/second

F:\src\Python-2.0\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.849447
This machine benchmarks at 11772.4 pystones/second

Python from CVS today:

F:\src\python-cvs\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.885801
This machine benchmarks at 11289.2 pystones/second

F:\src\python-cvs\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.889048
This machine benchmarks at 11248 pystones/second

F:\src\python-cvs\PCbuild>python ..\lib\test\pystone.py
Pystone(1.1) time for 10000 passes = 0.892422
This machine benchmarks at 11205.5 pystones/second


Although I deleted Tim's earlier mail, from memory this is pretty similar in
terms of performance lost.  I'm afraid I have no idea what your benchmarks
are or how to build them <wink>, but did check that the optimizer is set for
"mazimize for speed" (/O2).  Other compiler options gave significantly
smaller results (no optimizations around 8500, and "optimize for space"
(/O1) at around 10000).  Other fiddling with the optimizer couldn't get
better results than the existing settings.

Mark.