[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ns] Validation on intel linux



I met the same problem in my Mandrake 7.0,Turbo Linux 6.0,and in redhat6.1,
there are only one problem on red test.Now with Redhat 6.2,it all works well,
pass all the tests.I don't know why,but I suggest you to use RH6.2 with intel.

Wu Wei


�� 00-5-18 10:53:00 
>Hi,
>
>I normally do my ns work on solaris boxes, but we recently got some 600MHz
>PCs in and I wanted to see how much faster than our (relatively old)
>solaris box they are.
>
>I compiled ns-2.1b6 on the system without any difficulty. I then ran the
>validation tests and found, to my surprise, that many of them failed. I
>had a look on the website and found Sally Floyds mail that briefly
>describes the reasons why results differ on intel linux platforms from
>those obtained on platforms that use IEEE floating point arithmetic.
>
>I was going to suggest putting the output of 'make test' somewhere on the
>website, so that people would know if their ns distribution is doing the
>same things as it does on other peoples linux boxes. This got me to
>thinking about the tests, and I couldn't think of many reasons why the
>tests would differ between different modern versions of linux.
>Consequently, it may be useful to put such output on the website to make
>people more confident that their version of ns is working in the same way
>as other peoples, but I think that nearly everybody who can run the binary
>on modern linux systems should get the same results.
>
>Here, we think in terms of using linux intel (lintel?) boxes for any
>heavy-duty work in the future, and I'm sure many organisations think like
>this. Hence, I think it's important to have a set of tests that work on
>such boxes. Perhaps it would be a good idea to distribute the test output
>as run on a lintel box with the distribution with the caveat that some of
>the tests have not been checked against the reference output. Over time
>people, I guess, would check results of tests run on the lintel boxes
>against the reference tests and (hopefully) ensure that the differences
>are only due to floating point operations. Such tests would also be useful
>for ensuring consistent behaviour (even if it's not certain to be
>accurate) - if the behaviour changes, then it could point to a bug -
>across different ns distributions (ns-2.1b6, ns-2.1b7 etc).
>
>Regards,
>Sean.
>
>-----
>Sean Murphy,			Email: [email protected]
>Teltec Ireland,			Phone: +353-1-7045080
>DCU, Dublin 9,			Fax:   +353-1-7045092
>Ireland.