Re: svds test failures

[Marco atzeri]

On 9/2/2011 3:20 AM, Tatsuro MATSUOKA wrote:
> Sorry I sent the previous mail before it is completed.
>
> --- On Fri, 2011/9/2, Tatsuro MATSUOKA  wrote:
>
> Hello
>
> The test results perhaps depend on the kind of blas.
> Today I built OpenBLAS-a546694 and replace from altas-3.9.19 with 
> OpenBLAS-a546694.
>
> For OpenBLAS
> %! randn ('state',42);      % Initialize to make normest function reproducible
> %! rand ('state', 42)
>
> give
> warning: returning fewer singular values than requested
> warning: try increasing the value of sigma
>    ***** testif HAVE_UMFPACK
>   [u2,s2,v2,flag] = svds(A,k,0,opts);
>   s2 = diag(s2);
>   assert(flag,!1);
>   assert(s2, s(k:-1:1), 1e-10);
> !!!!! test failed
> assert (s2,s (k:-1:1),1e-10) expected
>     38.060
>     38.034
>     38.034
>     38.015
>     38.015
>     38.004
>     38.004
> but got
>     38.034
>     38.034
>     38.015
>     38.015
>     38.004
>     38.004
> Dimensions don't match
>
> The seed 42 gave no error for atlas-3.9.19.

similar on cygwin for 42

lapack-3.2.2 fails
atlas-3.9.47 passes

>   ****The results are
>      ATLAS OpenBLAS
> 35 OK       OK
> 36 OK       OK
> 37 fail       OK
> 38 OK       OK
> 39 OK       OK
> 40 OK       OK
> 41 OK       OK
> 42 OK       fail
> 43 fail       OK
> 44 OK       OK
> 45 OK       OK
> 46 OK       OK
> 47 OK       fail
> 48 OK       fail
> 49 fail       OK
> 50 OK       OK
> **************

the test is really depending on
system library

   ATLAS    Lapack
   3.9.47   3.2.2

 35 OK       OK
 36 OK       fail
 37 OK       OK
 38 fail     fail
 39 OK       fail
 40 OK       OK
 41 OK       OK
 42 OK       fail
 43 OK       OK
 44 OK       fail
 45 OK       OK
 46 OK       OK
 47 OK       fail
 48 fail     fail
 49 OK       OK
 50 OK       OK
> **************

we should look for a less critical test case

> Regards
>
> Tatsuro

Marco