Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #92275 > unrolled thread

Python Random vs. Cython C Rand for Dice Rolls

Back to article view | Back to comp.lang.python

Contents

  Python Random vs. Cython C Rand for Dice Rolls "C.D. Reimer" <chris@cdreimer.com> - 2015-06-07 10:17 -0700
    Re: Python Random vs. Cython C Rand for Dice Rolls Steven D'Aprano <steve@pearwood.info> - 2015-06-08 04:33 +1000
      Re: Python Random vs. Cython C Rand for Dice Rolls "C.D. Reimer" <chris@cdreimer.com> - 2015-06-07 11:56 -0700
      Re: Python Random vs. Cython C Rand for Dice Rolls Laura Creighton <lac@openend.se> - 2015-06-08 19:33 +0200
      Re: Python Random vs. Cython C Rand for Dice Rolls Sturla Molden <sturla.molden@gmail.com> - 2015-06-08 23:44 +0200

#92275 — Python Random vs. Cython C Rand for Dice Rolls

Greetings,

I've revisited my misbegotten childhood by translating the programs from 
"BASIC Computer Games" by David H. Ahl into Python. This is mostly an 
exercise in unraveling spaghetti code with all those GOTO statements 
going all over the place. The dice program caught my attention in 
particular for a stripped down version in Python and Cython to test the 
random number generator functionality.

http://www.atariarchives.org/basicgames/showpage.php?page=57

Here are my scripts to roll a pair of dice 50,000,000 times (this number 
creates a noticeable delay on my 3GHz quad processor). The Python script 
uses random, the Cython script uses C rand. Besides the obvious speed 
difference, the results are quite different.

This is the Python script that takes ~197 seconds to complete.

import random, time

startTime = time.time()

f = [0] * 12

for i in range(50000000):

     a = random.randint(1,6)

     b = random.randint(1,6)

     f[(a + b) - 1] += 1

print "\nTOTAL SPOTS","\tNUMBER OF TIMES\n"

for i in range(1,12):

     print ' ' + str(i + 1), '\t\t ', f[i]

print '\n', time.time() - startTime


This the Cython script that is imported into a test scripts similar to 
the Python script that takes ~1.6 seconds to complete.

cdef extern from "stdlib.h":

     int c_libc_rand "rand"()

def roll(int x):

     cdef:

         int a, b, i

         int f[12]

     for i in range(x):

         a = c_libc_rand() % 6 + 1

         b = c_libc_rand() % 6 + 1

         f[(a + b) - 1] += 1

     return f

Here's the console output.

PS Z:\projects\programming\python\basic_games\fastdice> python slowdice.py

TOTAL SPOTS     NUMBER OF TIMES

  2                1388086
  3                2776286
  4                4165556
  5                5555869
  6                6940547
  7                8335864
  8                6945446
  9                5556470
  10               4169549
  11               2777972
  12               1388355

197.006999969
PS Z:\projects\programming\python\basic_games\fastdice> python 
test_fastdice.py

TOTAL SPOTS     NUMBER OF TIMES

  2                1389911
  3                -2144697697
  4                4168249
  5                35008856
  6                6944907
  7                512318212
  8                6945597
  9                342017362
  10               4167485
  11               2775806
  12               1388465

1.63799977303

The Python random shows a uniform bell curve with low numbers at the 
ends and the peak in the middle, which is similar to the text in the 
book for the BASIC program. The Cython C rand is over all the place 
(especially with a negative number), which reminds me how bad the random 
number generator was on my 1MHz C64 in the day.

Is there something in the Cython code that I need to change and/or find 
a better C random number generator?

Thanks,

Chris R.

[toc] | [next] | [standalone]

#92285

On Mon, 8 Jun 2015 03:17 am, C.D. Reimer wrote:

> Greetings,
> 
> I've revisited my misbegotten childhood by translating the programs from
> "BASIC Computer Games" by David H. Ahl into Python. This is mostly an
> exercise in unraveling spaghetti code with all those GOTO statements
> going all over the place. The dice program caught my attention in
> particular for a stripped down version in Python and Cython to test the
> random number generator functionality.
> 
> http://www.atariarchives.org/basicgames/showpage.php?page=57
> 
> Here are my scripts to roll a pair of dice 50,000,000 times (this number
> creates a noticeable delay on my 3GHz quad processor). The Python script
> uses random, the Cython script uses C rand. Besides the obvious speed
> difference, the results are quite different.

C rand is not even close to random. The technical term for it is "shite".


-- 
Steve

[toc] | [prev] | [next] | [standalone]

#92291

On 6/7/2015 11:33 AM, Steven D'Aprano wrote:
> C rand is not even close to random. The technical term for it is "shite".

Looking through the BASIC book, I remembered all the tricks needed to 
get a half-way decent number generator on a 1MHz processor back in the 
day. Either the numbers start repeating after a while or get stuck on a 
particular number. That's one of the challenges of converting these old 
games from BASIC to Python. A lot of the stuff done in BASIC was meant 
to get around the hardware limitations of the early computers.

Chris R.

[toc] | [prev] | [next] | [standalone]

#92341

Better C random number generator.
http://www.pcg-random.org/download.html

Laura

[toc] | [prev] | [next] | [standalone]

#92350

On 08/06/15 19:33, Laura Creighton wrote:

> Better C random number generator.
> http://www.pcg-random.org/download.html

Or for something less minimalistic, just grab randomkit.c and 
randomkit.h from NumPy, which implements the same Mersenne Twister as 
Python. That is what I usually do to get fast random numbers from Cython

https://github.com/numpy/numpy/tree/master/numpy/random/mtrand


Sturla

[toc] | [prev] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web