Jeff Squyres wrote:
>On Jun 24, 2005, at 5:12 PM, Dave Diller wrote:
>
>
>
>>I am trying to use LAM on a 64-bit Linux cluster running RedHat
>>2.4.21. I find that the version of LAM that comes bundled with
>>the operating system does not function properly, specifically the
>>applications always hang in the method MPI_Init.
>>
>>
>
>Wow -- bummer. :-(
>
>
>
>>I tried downloading LAM and building it myself, and it works OK,
>>so I am confident that it is not an application error. The
>>code is very simple, basically a Hello World.
>>
>>
>
>Just to be absolutely positively sure, can you send the code? Is it
>from the examples in the LAM tarball?
>
>
>
>>I'm guessing that this is some kind of a library problem, but
>>it seems that other people must have seen it too.
>>
>>The output from the command "lamboot -V" is as follows:
>>
>>LAM 6.5.9/MPI 2 C++ - Indiana University
>>
>> Arch: x86_64-redhat-linux-gnu
>> RPI: usysv
>>
>>The output from the command "uname -a" is as follows:
>>
>>Linux a1 2.4.21-27.ELsmp #1 SMP Wed Dec 1 21:53:57 EST 2004 x86_64
>>x86_64
>>x86_64 GNU/Linux
>>
>>I need to solve this problem, rather then working around it
>>by building my own version of LAM, because our customers
>>will be using the pre-installed version of LAM.
>>
>>
>
>I'm afraid that we can't really help much here:
>
>1. That's a *really* old version of LAM. It really isn't supported by
>us anymore.
>
>
Nor does Red Hat use it any longer in their current releases.
Applications which I deal with have dropped 6.5.9 in favor of 7.0.3,
with a source tarball supplied to customers. --enable-shared has become
a necessity, and Red Hat does not support that. I know of no situations
which would force Red Hat to support 6.5.9, regardless of its inclusion
in paid supported installations. Even the source version of 6.5.9 has
failed us on certain RHEL4 versions, which come with 7.0.6.
>2. We have neither control nor knowledge of how Red Hat bundles LAM
>with its shipping versions. You'll probably need to bring this up with
>Red Hat support (feel free to CC me on the conversation; I might be
>able to help with the technical content of the conversation).
>
>
>
That would be interesting, if there were a conclusion which could be
quoted publically.
|
