From help-request at octave dot org Sun Feb  6 10:31:33 2005
Subject: Re: randn newbie question
From: Paul Kienzle <pkienzle at users dot sourceforge dot net>
To: Hamish Allan <hamish at jamkit dot com>
Cc: help at octave dot org
Date: Sun, 6 Feb 2005 11:36:29 -0500


--Apple-Mail-1-551570306
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
	charset=US-ASCII;
	format=flowed

On Feb 5, 2005, at 11:55 AM, Hamish Allan wrote:

> Why does:
>
> plot(randn(2,100)', '+')
>
> which looks like this in Matlab:
>
> http://igor.gold.ac.uk/~map01ra/matlab_randn.jpeg
>
> look like this in Octave:
>
> http://igor.gold.ac.uk/~map01ra/octave_randn.jpeg
>
> ???

The reason is that I messed up the test for X86 architecture
in randmtzig.c, and you are compiling little endian rather than
big endian architecture.

I've fixed this in CVS.  Before making a new release, I want
to put some distributional tests around the random number
generator code so nothing so egregious happens again.

Please try the attached patch.

- Paul


--Apple-Mail-1-551570306
Content-Transfer-Encoding: 7bit
Content-Type: application/octet-stream;
	x-unix-mode=0644;
	name="randmt.patch"
Content-Disposition: attachment;
	filename=randmt.patch

Index: randmtzig.c
===================================================================
RCS file: /cvsroot/octave/octave-forge/FIXES/randmtzig.c,v
retrieving revision 1.9
retrieving revision 1.10
diff -c -p -r1.9 -r1.10
*** randmtzig.c	8 Nov 2004 11:41:25 -0000	1.9
--- randmtzig.c	6 Feb 2005 16:19:28 -0000	1.10
*************** USE_INLINE uint64_t randi53(void)
*** 335,341 ****
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32()&0x1FFFFF;
! #if defined(HAVE_X86)
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
--- 335,341 ----
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32()&0x1FFFFF;
! #if HAVE_X86
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
*************** USE_INLINE uint64_t randi54(void)
*** 350,356 ****
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32()&0x3FFFFF;
! #if defined(HAVE_X86)
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
--- 350,356 ----
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32()&0x3FFFFF;
! #if HAVE_X86
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
*************** USE_INLINE uint64_t randi64(void)
*** 365,371 ****
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32();
! #if defined(HAVE_X86)
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
--- 365,371 ----
  {
      const uint32_t lo = randi32();
      const uint32_t hi = randi32();
! #if HAVE_X86
      uint64_t u;
      uint32_t *p = (uint32_t *)&u;
      p[0] = lo;
*************** USE_INLINE double randn (void)
*** 552,558 ****
         * have something to do with this.
         */
  #if defined(ALLBITS)
! # if defined(HAVE_X86)
        /* 53-bit mantissa, 1-bit sign, x86 32-bit architecture */
        double x;
        int si,idx;
--- 552,558 ----
         * have something to do with this.
         */
  #if defined(ALLBITS)
! # if HAVE_X86
        /* 53-bit mantissa, 1-bit sign, x86 32-bit architecture */
        double x;
        int si,idx;

--Apple-Mail-1-551570306--



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------