Just fyi - test of modified python core.

[- c o v e n t r y -]

This is just an FYI regarding my latest experiment to speed up python.

After doing some research and finding that gcc is capable of having 
pointers to labels in C/C++ code, I decided to try something: I modified 
the python interpreter's main loop to utilize an array of label pointers 
   whose indices were equal to the bytecode they were for.  Thus, in 
pseudo code, I change this:

switch (bytecode)
	case BYTECODE1:
		do something;
		break;

-----------------------------
into:

[label array for bytecodes]

goto labelarray[bytecode];	
	BYTECODE1:
		do something;
		goto end_loop;

-----------------------------

The results of this change were not as expected... as you can see from 
the pystone results below.

                            avg results
python2.2:                   7142.
python2.2-experimental:      7142.			

------------------------------------------
Conclusion: this gcc compiler trick is useless in the main loop, since 
gcc is apparently either A) generating code equivilent to a 
label-array-goto for the main switch or B) producing code that runs at 
equivilent speed.  I'm leaning towards A.

Ah well, at least it only took a few hours to find out ;)

-c