From: KIRK, B. (JSC-E. (NASA) <ben...@na...> - 2004-10-27 21:42:26
|
The PETSc 2.1.3 install... Is that from the Debian package or = installed 3rd-party? libMesh expects the configuration in ${PETSC_DIR}/bmake/${PETSC_ARCH}/packages to work. This is what PETSc itself uses if you run 'make BOPT=3Dg testexamples_uni' in $PETSC_DIR. = If that configuration does not work for some reason then there can be a problem. Honestly, I'm not sure what a good way to handle this is. As a hack we = can look for 'MPE_' symbols in the petsc libs and explicitly add -lmpe in = that case, but it is a pretty ugly approach and will likely break something = else. Probably better if you simply add -lmpe to ${PETSC_DIR}/bmake/${PETSC_ARCH}/packages As for the build problems in src/apps, that is just laziness on my = part. Most of those files were temporary codes designed to solve a specific problem and should have never been checked in to the main branch. Nevertheless, some of them could be fixed to compile again, and I'll = look into that. Finally, for the problems in your applications: =20 ----------------------------------------------- I've never actually used the MeshFunction class, so perhaps I can defer = this one to another developer? If there are no takers I'll look at it = further... The UCDIO object only implements reading a mesh at the moment. This is because I don't have any documentation on the format, so I'm not sure exactly *what* to write. I use UCD primarily for importing meshes from Gridgen, a commercial grid generation package. If you could point me = to some documentation on the UCD format it should be fairly simple to = support writing it. -Ben -----Original Message----- From: lib...@li... [mailto:lib...@li...] On Behalf Of Michael Schindler Sent: Wednesday, October 27, 2004 3:41 PM To: lib...@li... Subject: Re: [Libmesh-users] libmesh-0.4.3-rc2 now available. Hello Benjamin, and list, On 26.10.04, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > This release has been tested on a number of platforms. Please report = > any usability issues so that they may be addressed in libMesh 0.4.3. I tried to use libmesh on a debian sarge machine with two compilers = (mainly Intel icc version 7) with some success. This is a great project! Nevertheless, it was quite hard to get everything to work. I think, = there are some details that should be improved for the convenience of the = users. You will find the outcome of my tries in the following attachments: - for more details on my configure flags see attachment "make_echo" - for warnings during the compilation of the library see = "compil_library" - on a debian woody machine with petsc version 2.1.3 I had troubles = with the mpich libraries snooped from petsc. You will find details in "compil_library_2.1.3" - I tried to compile the applications in src/apps --> see "compil_applications" - Then, the examples --> see "execut_examples" - And finally, there are some bugs in the library I have already reported some months ago. They have not yet been fixed. see "my_programs" There is much to be done, I think -- but you asked for it ;-) Best greetings from Augsburg, Michael. --=20 "A mathematician is a device for turning coffee into theorems" Paul Erd=F6s. |
From: Michael S. <m-s...@us...> - 2004-10-28 13:38:10
|
On 27.10.04, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > The PETSc 2.1.3 install... Is that from the Debian package or installed > 3rd-party? That is debian package petsc2.1.3 with the version 2.1.3-1.woody.1 > libMesh expects the configuration in > ${PETSC_DIR}/bmake/${PETSC_ARCH}/packages to work. This is what PETSc > itself uses if you run 'make BOPT=g testexamples_uni' in $PETSC_DIR. If > that configuration does not work for some reason then there can be a > problem. Good hind. I will try that. > Honestly, I'm not sure what a good way to handle this is. As a hack we can > look for 'MPE_' symbols in the petsc libs and explicitly add -lmpe in that > case, but it is a pretty ugly approach and will likely break something else. > Probably better if you simply add -lmpe to > ${PETSC_DIR}/bmake/${PETSC_ARCH}/packages I think this is a problem in petsc. It would be sufficient to state (e.g at the libmesh homepage that this version needs a patch in ${PETSC_DIR}/bmake/${PETSC_ARCH}/packages (or people should path their Make.common.in which has the same effect. > As for the build problems in src/apps, that is just laziness on my part. > Most of those files were temporary codes designed to solve a specific > problem and should have never been checked in to the main branch. > Nevertheless, some of them could be fixed to compile again, and I'll look > into that. > > Finally, for the problems in your applications: > ----------------------------------------------- > I've never actually used the MeshFunction class, so perhaps I can defer this > one to another developer? If there are no takers I'll look at it further... Alright > The UCDIO object only implements reading a mesh at the moment. This is > because I don't have any documentation on the format, so I'm not sure > exactly *what* to write. I use UCD primarily for importing meshes from > Gridgen, a commercial grid generation package. If you could point me to > some documentation on the UCD format it should be fairly simple to support > writing it. I just wanted to try it out to get an alternative to gmv (and have no documentation, sorry) Is there, actually, a better way to get pictures from the simulation results than writing the equation system to a gmv-file and viewing with that? The (free) gmv I am using cannot make any reasonable shapshot or printout. Could you point me to some better program? Michael. -- "A mathematician is a device for turning coffee into theorems" Paul Erdös. |
From: John P. <pet...@cf...> - 2004-10-28 14:28:36
|
Michael Schindler writes: > > results than writing the equation system to a gmv-file and viewing > with that? The (free) gmv I am using cannot make any reasonable > shapshot or printout. Could you point me to some better program? Hi, You might want to get a newer version. GMV has supported making "snapshots" for some time, I believe it is in the first pull-down menu. The file type is "rgb" which you can easily convert to jpg or png using the GIMP. -J |