#9798: Accelerate Polyhedron constructor and fix cddlib output ordering
----------------------------+-----------------------------------------------
Reporter: vbraun | Owner: mhampton
Type: defect | Status: needs_work
Priority: major | Milestone: sage-4.6
Component: geometry | Keywords:
Author: Volker Braun | Upstream: N/A
Reviewer: | Merged:
Work_issues: |
----------------------------+-----------------------------------------------
Changes (by drkirkby):
* cc: dimpase (added)
Comment:
FWIW I had a quick look at the code of that so-called ''portable random
number generator'', and did a few quick checks. Several things are odd
about it. None are serious issues, but make me skeptical of the code -
especially use of the word "portable".
I would certainly '''not''' trust its statistical properties, given the
pretty basic flaws I found in the coding, though I understand that does
not matter for this application.
Anyway, this is what I found.
* {{{portable_srand}}} is declared as void, but then tries to return 0.
That's a pretty basic programming error. Even GCC flags that as a warning.
* The local variable {{{type}}} is declared in {{{portable_srand}}} then
set to the value {{{rng_state.rand_type}}}. However, {{{type}}} is never
used after being set. So it was pointless declaring {{{type}}} and setting
it. It's anyone's guess if someone actually intended the algorithm to use
{{{{{{rng_state.rand_type}}} or not.
GCC can't detect the last problem, but !SunStudio detects this programming
error. That's one of the advantages of using another !SunStudio - it roots
out dubious code better than gcc.
I added a {{{main}}}, and compared the sequence of numbers generated on
several platforms. All these produced the same sequence for the 4 seeds I
tried.
* AIX 5.3 on PowerPC 32-bit (Since my RS/6000 7025 F50 is only 32-bit, I
could not try 64-bit)
* Linux on x86 32-bit
* Linux on x86 64-bit
* !OpenSolaris on x86 32-bit
* !OpenSolaris on x86 64-bit
* OS X on x86 32-bit
* OS X on x86 64-bit
* Solaris 10 on SPARC 32-bit
* Solaris 10 on SPARC 64-bit
It will not compile at all on my HP C3600 running HP-UX 11.11B, since the
RNG uses {{{stdint.h}}}, which is only part of the C99 standard and the
older HP-UX 11.11B is not C99 aware. Quite why we need C99 code for a RNG
I will never know. I was using portable random number generators well
before 1999.
So I'm not sure one could attach the name "portable" to this, as it
certainly presents issues on HP-UX. It may do on Cygwin and FreeBSD too.
In contrast, the GNU Scientific Library builds and passes all tests on HP-
UX, FreeBSD and Cygwin, so is more portable.
It should also be noted this RNG produces the same sequence if the seed is
0 or 1. For at least some other values I tried, the numbers changed as the
seed was changed. But for 0 or 1, the output is the same.
I cc'ed Dima on this, as I thought he might find this interesting!
Dave
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/9798#comment:47>
Sage <http://www.sagemath.org>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to
[email protected].
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en.
