Re: Re: build charm++ and namd

From: Peter Freddolino (petefred_at_ks.uiuc.edu)
Date: Thu Jul 31 2008 - 08:39:07 CDT

Hi Vlad,
-O3 almost certainly helps for some parts of namd since it's so floating
point heavy; I was just pointing out that it probably does little for
charm++, and heavier levels of optimization do occasionally have
downsides. This is a minor point.

You're probably right that there's something breaking in the interpreter
handling the build script, but I have no idea what (I've never run into
similar problems)... I just wanted to point out the PPL people in case
you didn't get any help from namd-l.

Sorry I can't be of more help.

Peter

Vlad Cojocaru wrote:
> Hi Peter,
>
> Well -O3 works fine for everything on the first machine I compiled. I
> compiled everything with -O3 and I didn't run into problems.... the
> namd executable is as fast as the one provided on the namd site of the
> corresponding platform ... so I didn't see any issue with it ...
>
> I don't know if I am right but I could guess that the problem I see
> has something to do with the shell... My guess is that one of the
> invocations of the cat command is not correctly processed by the shell
> ... And since the shell might be different on the new machine
> comparing to the one I tested before .... I'll look into that ....
>
> Of course I know that it could be that the charm people might
> understand better what's happening but before I subscribe to a new
> list I tried to see if one of the people routinely compiling NAMD has
> seen such a thing before ... If I cannot find the solution I'll see
> with the charm people ...
>
> Thanks
>
> vlad
>
> Peter Freddolino wrote:
>> Hi Vlad,
>> for nontrivial charm++ issues, you're more likely to have success
>> with the charm++ people (ppl_at_cs.uiuc.edu), most of whom don't read
>> namd-l.
>> BTW, is there any particular reason you're compiling with -O3?
>> charm++ isn't the sort of code that's likely to benefit from this,
>> and since charm++ is usually compiled at -O2, it's conceivable for
>> you to run into something that the higher optimization breaks (this
>> isn't your problem here, but I figured I'd note it).
>> Best,
>> Peter
>>
>> Vlad Cojocaru wrote:
>>> Just to follow up on the previous email, it seems that during build
>>> a strange thing happen.
>>> The 2 lines below appear in my build.log file (just before the
>>> errors I posted in my previous mail):
>>>
>>> cat pup_f.f90.sh >pup_f.f90
>>> chmod a+x pup_f.f90
>>>
>>> This obviously makes no sense at it cats the sh script
>>> (src/util/pup_f.f90.sh) into the fortran file and that's why I got
>>> the reported errors (the fortran compiler tries to compile an .sh
>>> script). However, the build script does not contain these 2 lines so
>>> I dont know where they come from ... Even more strange is that
>>> yesterday I build charm++ 6.0 on another opteron machine without any
>>> problem with exactly the same build I use here...
>>>
>>> Vlad
>>>
>>> Vlad Cojocaru wrote:
>>>> Dear NAMD users,
>>>>
>>>> I have a problem building charm++ (version 6.0) and namd on a
>>>> cluster with AMD Barcelona chips using intel 10.1.015 compiler.
>>>> My build command is:
>>>> ./build charm++ net-linux-amd64 icc10amd tcp --no-build-shared
>>>> -DCMK_OPTIMIZE=1 -O3 2>&1 | tee build.log
>>>>
>>>> Although the build appeared successful (megatest passed), I got the
>>>> error attached during the build.
>>>>
>>>> Is this something I should worry about? Has anybody seen this before ?
>>>>
>>>> Thanks
>>>>
>>>> Best wishes
>>>> vlad
>>>>
>>>
>>
>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:42 CST