Re: Request for test

[Sergei Steshenko]

--- On Fri, 1/21/11, Isak Delberth Davids <address@hidden> wrote:

From: Isak Delberth Davids <address@hidden>
Subject: Re: Request for test
To: "CdeMills" <address@hidden>
Cc: address@hidden
Date: Friday, January 21, 2011, 11:55 PM

On 13 January 2011 22:10, CdeMills <address@hidden> wrote:



Hello,



I would like to have the following test performed on machines similar to

mine. I openened a bug report at https://savannah.gnu.org/bugs/?32120



Test: do twice

[1+i 2+i ; 3+i 4+i ] * [5+i ; 6+i]



Expected:

ans =



  15 + 14i

  37 + 18i



Wrong result

ans =



17 + 11i

39 + 11i



It seems that this occurs under Ubuntu 10.04 LTS on i686 machines, with

library libatlas3gf-sse2, version 3.6.0-24ubuntu1: i386, installed. Could

peoples having access to I686 machines running either Ubuntu 10.04, either

debian stable, test if

/usr/bin/octave -f

  generates the wrong result, AND

LD_PRELOAD= /usr/lib/libatlas.so.3gf   /usr/bin/octave -f

gives the right result ? This would indicate bugs in libatlas-sse2 "ZEMM"

implementation.



Thanks in advance



Pascal Dupuis





--

View this message in context: 
http://octave.1599824.n4.nabble.com/Request-for-test-tp3216568p3216568.html

Sent from the Octave - General mailing list archive at Nabble.com.

_______________________________________________

Help-octave mailing list

address@hidden

https://mailman.cae.wisc.edu/listinfo/help-octave


I hope my reply is not too late, it took me some time to react ... sorry

I run Ubuntu 10.04 on a old laptop, when I do "cat /etc/issue" in a terminal, I 
get "Ubuntu 10.04.1 LTS \n \l" and from "uname -a" I get "Linux delbert 
2.6.32-27-generic #49-Ubuntu SMP Wed Dec 1 23:52:12 UTC 2010 i686 GNU/Linux"


And here is my test result on octave-3.2.3
octave:1> [1+i 2+i ; 3+i 4+i ] * [5+i ; 6+i]
ans =

   15 + 14i
   37 + 18i

octave:2> [1+i 2+i ; 3+i 4+i ] * [5+i ; 6+i]
ans =

   15 + 14i

   37 + 18i

I hope this contributes --- if I did not missed a point.

Cheers,
IDD



-----Inline Attachment Follows-----

_______________________________________________
Help-octave mailing list
address@hidden
https://mailman.cae.wisc.edu/listinfo/help-octave

Self-built octave-3.0.5 with self-built "everything", including
SSE2-enabled ATLAS (32 bit machine):

"
octave:1> [1+i 2+i ; 3+i 4+i ] * [5+i ; 6+i]
ans =

   15 + 14i
   37 + 18i

octave:2> [1+i 2+i ; 3+i 4+i ] * [5+i ; 6+i]
ans =

   15 + 14i
   37 + 18i

octave:3>              
".

Regards,
  Sergei.