Marcus von Appen wrote:

> On, Mon Aug 21, 2006, Alex Holkner wrote:
> 
> [...]
>  
>   
> 
> >In the more general case, the small amount of Python wrapper that exists 
> >on every Pygame-ctypes function will indeed make it perform slower than 
> >its Pygame counterpart.  I would argue that if this small performance 
> >penalty is significant for a particular project, then that project is 
> >not really suited to be implemented in Python.
> >    
> >
> 
> If a small penalty of e.g. 0.5ms in a function exists and this function
> is called 50 times in your project code you'll have a total penalty of
> 2.5 ms already. Now let's assume, this function is (or must be) called
> periodically...
> If the author now mourns about that, are you going to tell him, that
> his projects is not suited to be implemented in Python?
>   
> 
Of course your argument can be applied to any arbitrary-sized problem, 
but I thought I would clarify the figures.  The overhead of 
Pygame-ctypes that is directly a result of using ctypes (see below) is 
0.22 microseconds---several orders of magnitude less than your example.  
For comparison, the cost of a single addition is 0.07 microseconds.

Tests conducted on AMD64 3500, GCC 3.4.3, Linux 2.6.14, Python 2.5 beta 
3 (64-bit), Pygame-ctypes r938, Pygame r938, SDL 1.2.11.

Pygame:
python2.5 -OO -m timeit -s 'import pygame; pygame.init()' 
'pygame.time.get_ticks()'
1000000 loops, best of 3: 0.47 usec per loop

SDL-ctypes:
python2.5 -OO -m timeit -s 'import SDL; 
SDL.SDL_Init(SDL.SDL_INIT_TIMER)' 'SDL.SDL_GetTicks()'
1000000 loops, best of 3: 0.694 usec per loop

Pygame-ctypes: (performs additional checking)
python2.5 -OO -m timeit -s 'import pygame; pygame.init()' 
'pygame.time.get_ticks()'
100000 loops, best of 3: 2.3 usec per loop

Simple addition only:
python2.5 -OO -m timeit '1+1'
10000000 loops, best of 3: 0.0714 usec per loop

> Saying things like that basically means, that you do not understand the
> pitfalls of software dependencies and how to deal with them.
>   
> 
Alex.