At 04:04 AM 3/9/2005, Jeff Squyres wrote:
We agreed yesterday to try augmenting laminfo in the current version. As
the architecture line would be the same in a 32-bit or 64-bit build, it
doesn't do much but confirm the possible ambiguity. Right now, trying
stuff like 'mpif77 -V' ,'mpicc -v' to give a clue about compilers and
options set has a chance, along with verifying by 'which mpicc' ,'which
mpirun' that they come from the same installation.
Thanks,
>Sorry -- had to go build a LAM 7.0.6 to test. :-)
>
>Yes, we expanded laminfo a lot in 7.1.x. I guess the only string that
>*may* help you in the 7.0.x line is the architecture line...?
>
>I would say your best bet would be to put it in laminfo somewhere. You
>might want to compare the 7.1.x laminfo source with your source; it
>shouldn't be hard to do. Heck, you could even easily add a line that just
>prints the sizeof(void*) and not try to draw on LAM's configuration
>information.
>
>
>On Mar 8, 2005, at 9:50 AM, Tim Prince wrote:
>
>>At 06:25 AM 3/8/2005, Jeff Squyres wrote:
>>Yes, having the information in laminfo should work. The newest version
>>of lam which I have installed at the moment is 7.0.3, which complains on
>>'laminfo -all.' I don't see anything in laminfo which gives the desired
>>information on 32- or 64- bit build mode; I'd try implementing
>>something, if that is reasonable to do.
>>
>>>Can you just use the laminfo command instead? (the lamboot -V info is
>>>only maintained for hysterical raisins)
>>>
>>>laminfo was intended to be the vehicle for all this kind of information
>>>-- it offers it in both prettyprint and easily parsable formats. For
>>>example, try:
>>>
>>>laminfo -all -parsable | grep pointer
>>>
>>>
>>>On Mar 8, 2005, at 9:21 AM, Tim Prince wrote:
>>>
>>>>Several of us agreed we would like to try adding an indication of how
>>>>lam was built to the lamboot message. Specifically, for an x86-64
>>>>system, we would like to have it confirm whether it is a 32- or 64-bit
>>>>lam. Could someone point me to where the message is generated, so I
>>>>could experiment?
>>>>
Tim Prince
|
