Dear all, I re-configure libmesh with --enable-parmesh (rather than --enable-everything?) and start from 50x50x50, things work fine. Thank you.
Regards, LL -----原始邮件----- 发件人: "Cody Permann" <[email protected]> 发送时间: 2015年2月9日 星期一 收件人: [email protected], libmesh-users <[email protected]> 抄送: 主题: Re: [Libmesh-users] Fw: Re: Re: A problem on memory blow up when using build_cube() for 3D mesh Don't get too focused on just the mesh memory usage. In my experience serial mesh works fine even for meshes of this size. You have other memory consumers to worry about like your system matrix and solution vectors too. Cody On Sat Feb 07 2015 at 7:31:07 PM <[email protected]> wrote: Thanks a lot. 1. I used the following configure option: ./configure --enable-everything --prefix=$LIBMESH_DIR --with-methods=opt Does --enable-everything include --enable-parmesh? Would you please recommend me an optimal option in fully using parallel computing? 2. If use uniformly_refine(2), the refinement is done after partition? I mean, every processor do refinement itself? Thank you. Regards, LL -----原始邮件----- 发件人: "Paul T. Bauman" <[email protected]> 发送时间: 2015年2月7日 星期六 收件人:[email protected] 抄送: "[email protected]" <[email protected]> 主题: Re: [Libmesh-users] A problem on memory blow up when using build_cube() for 3D mesh On Sat, Feb 7, 2015 at 2:32 AM, <[email protected]> wrote: Dear developers, I tried a 3D mesh of 200x200x200 by using build_cube(), on 128 cores. However, it runs out of memory. Make sure you use ParallelMesh for this size problem (the default from Mesh is SerialMesh unless you compiled libMesh with the --enable-parmesh option). SerialMesh will store the entire mesh on every processor while ParallelMesh will partition the mesh across processors. I read the details of build_cube(), it seems that a global 200x200x200 mesh is generated on every processor. If this is true, then the meshes are redundant. In addition to the ParallelMesh suggestion, you could try generating a smaller mesh, say 50x50x50, and call MeshRefinement(mesh).uniformly_refine(2) HTH, Paul -- Li Luo https://sites.google.com/site/rolyliluo/ Shenzhen Institutes of Advanced Technology Address: 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, P.R.China Tel: +86-15899753087 [email protected] -- Li Luo https://sites.google.com/site/rolyliluo/ Shenzhen Institutes of Advanced Technology Address: 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, P.R.China Tel: +86-15899753087 [email protected] ------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users -- Li Luo https://sites.google.com/site/rolyliluo/ Shenzhen Institutes of Advanced Technology Address: 1068 Xueyuan Avenue, Shenzhen University Town, Shenzhen, P.R.China Tel: +86-15899753087 [email protected] ------------------------------------------------------------------------------ Dive into the World of Parallel Programming. The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net/ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
