Confusion with extremely basic monte carlo pi program

[Gryffin]

Hello all,

I am having the most peculiar behavior with an extremely simple VI that I do not understand.

I am writing a program to approximate Pi by randomly generating points within X: [0,1) Y:[0,1), calculating the ratio of points inside an inscribed circle and outside of it in a square. I understand the math and the theory perfectly fine.

What I don't understand.....:

This is my program. If i set "# of points" to 1 - ~40000, the program works fine and a relatively decent approximation of pi comes out (e.g. 3.1522). However, if I exceed ~40000 points, my program suddenly gives a clearly incorrect result, sometimes being ~0.5, sometimes being negative (which shouldn't be possible). I would like to know, why is this occurring and can it be rectified?

[paul_cardinale · Accepted solution]

The function "Boolean To (0,1)" is outputting an I16.  When the sum exceeds 32367, it wraps around.

Convert the output of "Boolean To (0,1)" to I32 inside the loop.

[Gryffin]

Wow what a simple reason.

Thank you very much!

[altenbach]

Taking this a bit further, here are some tips for the future:

• Instead of squaring, summing and rooting, you can use complex math primitives instead.
• It is inefficient to build a potentially huge array at the autoindexing output tunnel just to reduce it to a scalar a nanosecond later by summing all elements. It takes N times less memory to just accumulate the running sum in an initialized scalar shift register!
• Since you can initialize the shift register with a suitable datatype that can hold all possible numbers, you don't need to worry about the output of "boolean to (0,1)".

Here is a quick example:

(note that "re/im to complex" followed by "absolute value" would be more efficient because you don't care about the theta output. See also our NI week presentation form 2016.)