Re: strange error with matrix largern then 2GB

[David Bateman]

Daniel Heiserer wrote:

> I created a large matrix and then tried to multiply this matrix. This
> results in a dimension mismatch. There must be a bug.
> Did anybody have the same experience?
>  
> My machine is x86_64 linux box with octave 2.9.3
>  
> best regards, daniel
>  
>  
> N=647088*3;
>  tic;x=rand(N,211);toc
> whos
>  
>    rwd x          1941264x211               2147482616  matrix
>    rwd x1             200x3                       4800  matrix
>    rwd x2               3x1                         24  matrix
>    rwd x3             200x1                       1600  matrix
>  
> Total is 409607509 elements using 2147489056 bytes
>  
> octave-2.9.3:11> y=rand(211,1);
> octave-2.9.3:12> tic;x1=x*y;toc
> error: Array::Array (const Array&, const dim_vector&): dimension mismatch
> error: Array::Array (const Array&, const dim_vector&): dimension mismatch
> error: Array::Array (const Array&, const dim_vector&): dimension mismatch
> panic: Segmentation fault -- stopping myself...
> attempting to save variables to `octave-core'...
> save to `octave-core' complete
> Segmentation fault


The 64bit support is very new. How was this binary compiled? Is it
linked to an external lapack/atlas binary? If yes is the int/long type
of this lapack/atlas binary match the same in octave? 2.9.3 is also very
old, perhaps this is already fixed in 2.9.6. Can you test a newer version?

D.

-- 
David Bateman                                address@hidden
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary