Bottleneck in conversion of integer to NumPy Scalar

What's wrong 

For scalar operations Numpy first try to extract the underlying C value from a Python Integers. It causes bottleneck because it first converts the Python scalar into its matching NumPy scalar (e.g. PyLong -> int32) and then it extracts the C value from the NumPy scalar.

Avoiding conversion

Hence avoiding this conversation improve speed significantly.  I have avoided conversion for known integer type but extracting its value directly.

For byte, short, int, long

#if PY_VERSION_HEX >= 0x03000000
    if(PyLong_CheckExact(a)){
        *arg1 = PyLong_AsLong(a);
        return 0;        
    }
#else
    if (PyInt_CheckExact(a)){
        *arg1 = PyInt_AS_LONG(a);
        return 0;
    }
#endif

Performance

• 1.09 Array * Array
• 1.09 Array + Array
• 1.18 np.int * np.int
• 1.14 np.int + np.int
• 1.08 PyInt * Array
• 1.07 np.int * Array
• 1.10 PyInt + Array
• 1.09 np.int + Array
• 1.08 PyInt * PyInt
• 1.09 PyInt + PyInt
• 2.53 PyInt * np.int
• 2.69 PyInt + np.int
• 1.06 np.int * np.int 
• 1.13 np.int + np.int
• 7.29 PyInt < np.int
• 2.38 np.int ** 2.     

Where,

Array = np.array([2234,32342])
PyType = 3
NyType = Array[0]

More

• PR for this enhancement is #3567
• Profiled Speedup Datasheet is http://goo.gl/U9R3DS
• Also Raul has made improvement for float dtype at #2941