[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Chageset]: Add the spaugment function
From: |
John W. Eaton |
Subject: |
Re: [Chageset]: Add the spaugment function |
Date: |
Thu, 03 Apr 2008 10:39:19 -0400 |
On 3-Apr-2008, Jaroslav Hajek wrote:
| > | maximum absolute error 1.655e-07 exceeds tolerance 1e-10
| >
| > This looks like a case where we just need to change the tolerance.
| >
| > jwe
| >
|
| another such case is tol = 10*eps in testqr from test/test_linalg.m.
| The attached changeset makes it 100*eps. Strangely enough, the test
| passed when I was linking against the single-threaded version of ACML,
| but failed when with the multi-threaded version. Apparently the ST is
| computing the result, but the MT is just guessing it :)
|
| Anyway, such tests should probably be scaled by the norm of the
| factorized matrix,
| at least those used agaist random matrices of various sizes.
I applied the patch.
Thanks,
jwe