Re: qhull6 and octave

[Brad Barber]

Hi Tatsuro,

Did you make Petr's change (compiling the test program with -Dqh_QHpointer)?  
If not that's may be the source of your problem.  If it isn't, you'll need to 
provide more information since I don't have octave running here and can't use 
debian builds (its a Windows system).  

A good approach is to compare a trace of your test program, with the 
corresponding trace from Qhull.  Option 'T1' gives a high level view, option 
'T4' gives most of the details.  Option 'T5' includes memory management.  
There's various commands to turn on tracing according to a trigger event.

                                        --Brad
P.S: email to octave-help bounces.  What should the address be?

                                                
At 03:01 AM 12/1/2011, Tatsuro MATSUOKA wrote:
>Hello
>
>On MinGW，qhull2011 gives errors.
>(The qhull library was created using GCC-4.6.1)
>
>For example,   
>src\DLD-FUNCTIONS\convhulln.cc ......................... PASS    0/3    FAIL 3
>
>Â  ***** testif HAVE_QHULLÂ cube = [0 0 0;1 0 0;1 1 0;0 1 0;0 0 1;1 0 1;1 1 
>1;0 1 1];Â [h, v] = convhulln (cube);Â
> assert (size (h), [6 4]);Â Â h = sortrows (sort (h, 2), [1:4]);Â assert (h, 
> [1 2 3 4; 1 2 5 6; 1 4 5 8; 2 3 6 7; 3 4 7 8; 5 6 7 8]);Â assert (v, 1, 
> 10*eps);!!!!! test failedassert (size (h),[6, 4]) expected   6   4but got 
>   0   3values do not match  ***** testif HAVE_QHULL cube = [0 0 0;1 0 0;1 
> 1 0;0 1 0;0 0 1;1 0 1;1 1 1;0 1 1];Â [h, v] = convhulln (cube, "QJ");Â assert 
> (size (h), [12 3]);Â Â assert (sortrows (sort (h, 2), [1:3]), [1 2 4; 1 2 5; 
> 1 4 5; 2 3 4; 2 3 6; 2 5 6; 3 4 8; 3 6 7; 3 7 8; 4 5 8; 5 6 8; 6 7 8]);Â 
> assert (v, 1.0, 1e6*eps);!!!!! test failedassert (size (h),[12, 3]) expected 
>   12    3but got   0   3values do not match  ***** testif HAVE_QHULL 
> tetrahedron = [1 1 1;-1 -1 1;-1 1 -1;1 -1 -1];Â [h, v] = convhulln 
> (tetrahedron);Â h = sortrows (sort (h, 2), [1 2 3]);Â assert (h, [1 2 3;1 2 
> 4; 1 3 4; 2 3
> 4]);Â assert (v, 8/3, 10*eps);!!!!! test failedassert (h,[1, 2, 3; 1, 2, 4; 
> 1, 3, 4; 2, 3, 4]) expected   1   2   3   1   2   4   1   3   4  
> Â 2 Â  3 Â  4but got[](0x3)Dimensions don't match
>This error did not occur on qull2010.
>
>Regards
>
>Tatsuro--- On Thu, 2011/12/1, Brad Barber wrote:
>
>> Hi Petr,
>> 
>> Glad it worked.   Without seeing the whole build system that was used, 
>> here's my guess from what you showed me.
>> 
>> Some news first.  I've released qhull 2011.2 with several changes, mostly 
>> minor.   Octave may be interested in the fix to qh_new_qhull (call 
>> qh_prepare_output if !outfile).  qh_prepare_output handles options that are 
>> processed after building the convex hull (e.g., compute area and volume).  
>> There's a new, simple Makefile for static builds (in src/libqhull).   
>> 
>> Please let me know when the Debian builds are available.  The qhull 
>> download page is rather out-of-date.  Can someone produce a rpm build?  
>> I've also lost my archive/mirror site for old Qhull downloads.  Let me know 
>> if someone would like to host it.
>> Â  Â  http://www.qhull.org/download/
>> 
>> qhull5 was apparently built with qh_QHpointer=0  [user.h].  The global 
>> data structure for qhull is defined as a static array qh_qh [libqhull.h].  
>> qhull6 was built with qh_QHpointer=1.  The global data structure for qhull 
>> is allocated on the heap with a pointer from the global variable qh_qh.   
>> This allows for multiple global data structures that can be swapped by 
>> updating qh_qh.   Two other pointers/static structures are similarly 
>> defined (qh_qhmem for the memory allocator and qh_qhstat for statistics).  
>> 
>> The extra flexibility comes at some cost.  The last time I measured it, the 
>> cost was 4% overhead in space and 4% in time.  The C++ interface for qhull 
>> uses qh_QHpointer=1.    The same effect could be achieved by copying data 
>> from  statically defined structures to/from the heap.
>> 
>> The code references these representations with the 'qh', 'qhmem', and 
>> 'qhstat' macros.  That's why the code has an odd appearance.  These macros 
>> allow the data structures to be allocated elsewhere (e.g., thread-local 
>> storage or an extra function parameter).
>> 
>> Many thanks for keeping Qhull part of Octave.
>> 
>> Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  Â  --Brad
>> 
>> At 06:09 AM 11/30/2011, Petr Gajdos wrote:
>> >Hi Brad,
>> >
>> >very well, defining qh_QHpointer to 1, i. e.
>> >
>> >#include <stdio.h>
>> >#if defined QHULL5
>> >#include <qhull/qhull_a.h>
>> >#warning qhull5
>> >#else
>> >#define qh_QHpointer 1
>> >#include <libqhull/qhull_a.h>
>> >#warning qhull6
>> >#endif
>> >...
>> >
>> >gives same results for qhull6 as for qhull5.
>> >
>> >When you find time, could you please give me 
>> >an explanation (maybe cc octave-help@ too)?
>> >
>> >Have a nice day & thanks a lot,
>> >Petr
>> >
>> >
>> 
>> 
>
>
>
>-----
>No virus found in this message.
>Checked by AVG - www.avg.com
>Version: 10.0.1411 / Virus Database: 2092/4049 - Release Date: 11/30/11