[Python-Dev] ineffective optimization: method tables

[Neil Schemenauer]

Well, I wasted a fair amount of my time for no apparent gain.
The idea was to have function pointer tables indexed by type that
could be used for common operations.  First, how to do we index
things by type?  Here's my solution:

    #define PyType_UNKNOWN 0
    #define PyType_NONE 1
    #define PyType_INT 2

    #define PyType_Ord(t) ((t)->tp_ord)
    #define PyObject_TypeOrd(o) PyType_Ord((o)->ob_type)

Here is an example of methods for PyObject_IsTrue:

    int
    int_istrue(PyObject *v)
    {
        return ((PyIntObject *)v)->ob_ival != 0;
    }

    int
    none_istrue(PyObject *v)
    {
        return 0;
    }

    inquiry istrue_table[] =  {
           PyObject_IsTrue,        /* PyType_UNKNOWN */
           none_istrue,            /* PyType_NONE */
           int_istrue,             /* PyType_INT */
    };

    #define PyObject_IS_TRUE(v) istrue_table[PyObject_TypeOrd(v)](v)

There is a patch at:

    http://arctrix.com/nas/python/method_table1.diff

I did have 2-D tables for binary operations but since they were
quite sparse I took them out in favor of arrays and case
statements.  Unfortunately, all my benchmarks show this patch to
be ineffective in terms of speeding up the interpreter.  Anyone
know why?

  Neil