Performance tips? about mpmath-2 HOT 7 CLOSED

open *it* up ... sorry, brain-o

I think you've misread the benchmark page. mpmath is faster than Python's 
decimal
module. It is *not* faster than Python's math module.

I don't understand the following:

    Based on your benchmark data, I expected an improvement in performance just
    for dropping it in (on equations that don't overflow with "math" and
    therefore don't need the extra precision).

Why would you drop it in for equations that *don't* overflow with "math"?

If I've understood what you're trying to do, you should use the math module / 
float
as far as possible and only switch to mpmath when standard floats overflow.

Could you please provide more details of what your code looks like?

I'm also curious about why it didn't overflow in Perl (were you using 
BigFloat?).

Thanks for a prompt reply.

Yes, I misread it.  Thanks.

My plugin for the GIMP creates images from equations that the user provides.  
Thus, I
don't know what functions are being used ahead of time.  I have a few solutions 
in
mind.  Thanks.

I couldn't tell you why exp() didn't overflow in Perl when it did with "math" in
Python.  I never looked into it because it always worked.  In case you want to 
look
into it further, I'm using ActivePerl 5.8.7 Build 813.  

Oh, and no, I did not use BigFloat...at least not explicitly.

Here's a possible approach to your problem:

import math
import mpmath

def evalmath(expr):
    try:
        r = eval(expr, math.__dict__)
    except OverflowError:
        r = eval(expr, mpmath.__dict__)
        try:
            r = float(r)
        except OverflowError:
            pass
    return r

>>> evalmath('sin(3)')
0.14112000805986721
>>>
>>> evalmath('exp(10000)')
mpf('8.8068182256629216e+4342')
>>>
>>> evalmath('exp(10000) / exp(10000)')
1.0

In other words, for highest speed float math is used by default. If overflow 
occurs,
mpmath is used. The result is converted back to a float if possible to make
subsequent operations fast; an mpf is only returned if the final result is too 
big.

Thanks.

I'm adding a description of this problem to the documentation.