Subject: Re: [OMPI users] Linking MPI applications with pgi IPA
From: Brock Palen (brockp_at_[hidden])
Date: 2009-06-22 15:47:22
I will have to try this latter, in theory if the library is compiled
with the options it will have the needed information.
ST (the makers of PGI) got back and I quote:
"You can try -Mipa=fast,inline,safe"
"I think for now, all libraries are safe. It is more a matter of how
aggressively does a compiler's inter-procedural analysis behave in
perhaps re-ordering function calls.
Today, we don't do that. But we are leaving our options open in the
future, if we find we could benefit performance by call reordering
safely via compiler-gathered ipa information."
I have not found time to test this though.
Brock Palen
www.umich.edu/~brockp
Center for Advanced Computing
brockp_at_[hidden]
(734)936-1985
On Jun 22, 2009, at 7:38 AM, Jeff Squyres wrote:
> Can you compile Open MPI with the same compiler options?
>
>
> On Jun 19, 2009, at 2:57 PM, Brock Palen wrote:
>
>> When linking application that are being compiled and linked with
>> the -
>> Mipa=fast,inline option, the IPA stops with errors like this case
>> with amber:
>>
>> The following function(s) are called, but no IPA information is
>> available:
>> mpi_allgatherv_, mpi_gatherv_, mpi_bcast_, mpi_wait_, mpi_get_count_,
>> mpi_recv_, mpi_isend_, mpi_gather_, mpi_allreduce_, mpi_abort_,
>> mpi_finalize_, mpi_send_
>> Linking without IPA
>>
>> Is there a way to tell the compiler its ok to ignore the MPI library
>> and do IPA for everything else?
>>
>>
>> Brock Palen
>> www.umich.edu/~brockp
>> Center for Advanced Computing
>> brockp_at_[hidden]
>> (734)936-1985
>>
>>
>>
>> _______________________________________________
>> users mailing list
>> users_at_[hidden]
>> http://www.open-mpi.org/mailman/listinfo.cgi/users
>>
>
>
> --
> Jeff Squyres
> Cisco Systems
>
> _______________________________________________
> users mailing list
> users_at_[hidden]
> http://www.open-mpi.org/mailman/listinfo.cgi/users
>
>