Cesare, Thanks for your quick, and apologies for my slow, reply.
On Sat, 2007-08-25 at 13:09 +0000, Cesare Tirabassi wrote: > By giving a quick look at the runtime libraries, I don't think the > problem is related to these: > > dapper: > atlas3-base (3.6.0-20) > lapack3 (3.0.20000531a-6ubuntu2) > refblas3 (1.2-8ubuntu1) > > edgy: > atlas3-base (3.6.0-20.2) > lapack3 (3.0.20000531a-6ubuntu2) > refblas3 (1.2-8ubuntu1) > > the only differences between the two atlas3-blase libraries being in the > packaging (and of course the base libraries against which they were > compiled). > > Unless there is a copy&paste problem, these results: > Yes there was a copy&paste mistake, but both results were good. > athlon/dapper max= 2.28048070053e-16 min= -4.583786292e-16 GOOD > pentium/edgy max= 2.28048070053e-16 min= -4.583786292e-16 GOOD > > seem to confirm it. Maybe, but the 64 bit libraries will presumably have been compiled with a different compiler (at least it will have been the 64 bit version of gcc) and possibly different flags. For example I have known gcc -O3 ... to cause a problem on some architectures and not others with the same source, and I'm guessing (wildly) that this is a possible cause. > I guess in both cases you were using the 32bit version? No, 64 bit versions on the 64 bit hardware. > It could actually be an hardware problem, I really hope not, it would be a *MAJOR* PITA. > would it be possible to test this on the same machine with the two different > versions (edgy and dapper)? > Perhaps some of the test cases above covers this? > To summarise the relevant results plus a couple more: Machine 1: Opteron 246 Warty: max= 3.62882467131e-16 min= -3.61473004307e-16 GOOD Machine 2: Athlon 64 3500+ Dapper: max= 0.0599265087162 min= -0.13143867524 BAD Edgy: max = 0.97897895303 min= -0.935360320766 BAD Feisty: max = 0.97897895303 min= -0.935360320766 BAD Edgy and Feisty were run from live cd, and yes, the results were identical - it wasn't a cut&paste issue this time. I'm not sure how far forward this takes things although the difference in the results for dapper and edgy on the same machine presumably suggests the involvement of software to some extent. If you know of any working repositories for breezy I'll try that as well. Cheers, Mark -- Dr. Mark Poolman Senior Research Fellow Cell Systems Modelling Group School of Life Science Oxford Brookes University -- Error in linalg.svd for large matrix https://bugs.launchpad.net/bugs/134277 You received this bug notification because you are a member of Ubuntu Bugs, which is the bug contact for Ubuntu. -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
