I'm actually only trying to see if I'm hitting the machine precision in my computations, so a longdouble version would be enough. I tried to compile a longdouble version of PETSC 3.1, but it wouldn't build a special blas/lapack for it when passing --download-c-blaslapack. How do you get a blaslapack for longdouble ? Sylvain ________________________________________ From: Roy Stogner [[email protected]] Sent: Tuesday, June 26, 2012 3:11 AM To: Derek Gaston Cc: Sylvain Vallaghe; [email protected] Subject: Re: [Libmesh-users] Quadruple precision
This might be more than just a lark for me too. It looks like long double solver support has moved from "requires a tricky blas/lapack build" in PETSc 3.1 to "requires a tricky blas/lapack build and a couple edits in PETSc config python scripts" in PETSc 3.2 to "requires a tricky blas/lapack build, several edits in PETSc config python scripts, and more edits then I've been able to successfully identify in PETSc source code" in PETSc 3.3. So quad precision is now the *only* extended precision option available when compiling against current PETSc. --- Roy On Mon, 25 Jun 2012, Derek Gaston wrote: > We're definitely interested in trying quadruple precision as well... we can > definitely help test some of this if you send in any > patches... > Derek > > On Mon, Jun 25, 2012 at 6:25 PM, Roy Stogner <[email protected]> wrote: > > On Mon, 25 Jun 2012, Sylvain Vallaghe wrote: > > > I'm trying to compile libmesh with quadruple precision. So far I > > compiled PETSC with float128, but the compilation of libmesh crashes > > right away. Is there something special to do in order to compile > > libmesh with quad precision ? > > Sadly, the first thing to do is "add quad precision support to > libMesh". I know I put some of the infrastructure in there on a lark, > and it shouldn't be a big step up from the single/double/"triple" > precision independence which is relatively well tested, but IIRC there > was still work to be done. > > If you want to help us with that work, I'd be happy to continue it. > To start I'd want to know what version of PETSc are you using, what > configuration commands you used for PETSc and libMesh, and what > compilation errors you hit for libMesh, so I can see if I can > replicate the problem. > --- > Roy > > ------------------------------------------------------------------------------ > Live Security Virtual Conference > Exclusive live event will cover all the ways today's security and > threat landscape has changed and how IT managers can respond. Discussions > will include endpoint security, mobile security and the latest in malware > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ > _______________________________________________ > Libmesh-users mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/libmesh-users > > > > ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
