Performance question about math operations

[Jonathan Hogg]

On 2/7/2002 7:00, in article mailman.1025589663.1583.python-list at python.org,
"Andrew P. Lentvorski" <bsder at mail.allcaps.org> wrote:

> I have a VLSI layout editor written in Python.  At its core, it has to
> redraw a lot of polygons.  This requires a lot of coordinate conversion
> mathematics.  Essentially the following loop:

Could you post the actual code, or is it too long/proprietary? How are all
the magic numbers arrived at? Does it really look as simple as you
suggested?

> What is eating all that time?  And can I do anything about it?

Try disassembling the loop:

>>> def foo( n ):
...     i = 0
...     while i < 100000:
...         i = i + 1
... 
>>> import dis
>>> dis.dis( foo )
          0 SET_LINENO               1

          3 SET_LINENO               2
          6 LOAD_CONST               1 (0)
          9 STORE_FAST               1 (i)
          [...]

You could look out for LOAD_GLOBAL instructions which are expensive. Another
thread here recently discussed why and a couple of methods of speeding them
up.

If it turns out that your loop is as fast as it's going to get, your best
bet is to convert it to C instead. You can do this by writing an extension
module. There's some documentation on the Python website that should get you
started on writing extension modules.

Jonathan