|
From: tom f. <tf...@al...> - 2010-01-31 06:57:49
|
Matt Funk <mat...@gm...> writes: > i have seen some stuff on the mailing list about this, however, it > hasn't gotten me anywhere. So basically i am wondering: > - is this warning is to be expected and thus to be ignored? As a general rule, no valgrind error is something you should ignore. That said, everyone's time is limited and thus you will need to make discretionary choices when you are inundated with errors. > - do i need to write a suppression file for this (is there one > available already) If you don't want to see it, then you need to write a suppression for it. This is the wrong forum to ask for available suppression files, for the most part. In this case, you want to go bother your MPI library vendor. -tom |
|
From: Dave G. <go...@mc...> - 2010-02-01 17:21:29
|
On Jan 31, 2010, at 1:00 AM, tom fogal wrote: > This is the wrong forum to ask for available suppression files, for > the most part. In this case, you want to go bother your MPI library > vendor. Matt, If you are using valgrind with MPICH2 then you should make sure that you configure MPICH2 with "--enable-g=dbg,meminit". This will cause the library to initialize communication buffers before they are passed to the write syscall and will use valgrind client requests to otherwise make valgrind happier. All of this comes at a slight performance penalty, so people typically build two versions of their MPI library: one for debugging and one for performance. Peeking ahead at your other mail thread, it looks like you are actually using MVAPICH2, so you may want to ask about this on their mailing list (http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss ). MVAPICH2 is derived from MPICH2, so the same configure flag may work. However I don't know that first-hand. As Julian pointed out on the other thread, compiling for TCP/IP might make it easier to debug as well. I think that you can do that as outlined here: http://mvapich.cse.ohio-state.edu/support/mvapich_user_guide-1.2rc1.html#x1-150004.4.6 -Dave |