Newsgroups: comp.parallel.mpi From: David Espadas Valladares Subject: Re: MPI_RECV Organization: Dpto. Arquitectura de Computadores. Universidad Complutense Date: Tue, 07 Jul 1998 15:16:01 +0200 Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <35A21F91.8768F70@eucmos.sim.ucm.es> I think that the receiving process may be requesting a message that has not been sent yet. So, When you do a print or something like that, you give the message time to be sent from sending process. Try to place a barrier just before the MPI_Recv ("MPI_Barrier"), and also just before the sending on the sending process. This will syncronize both calls, and I think it should work. If it doesn't, try placing the receiving barrier still before, so being sure the receiving process arrives at the receiving call when the sending has completed. Anyway, as you do not specify what machine are you using, I do not know if it buffers communications by default (I think almost all modern machines do, but...). If it doesn't, I'm quite sure the problem is what I tell you; if it does, the explanation is maybe more sophisticated and, well, I don't have any easy answer for that by now ;-). Aaron Focacci wrote: > > I'm trying to do a series of receives inside a loop. Before the call to > MPI_RECV, there is a print statement. If the print statement is there, > everything works fine. If the print statement is removed, I get the > message: > > p0_351: (3.865000) xx_shmalloc: returning NULL; requested 262184 bytes > p0_351: (3.865000) p4_shmalloc returning NULL; request = 262184 bytes > > Does anyone know what's going on? -- _____________________________________________________________________________________ David Espadas Valladares Dpto. de Arquitectura de Computadores y Automatica Facultad de CC. Fisicas Phone: +34 91 394 43 64 Universidad Complutense Fax : +34 91 394 46 87 Ciudad Universitaria S/N E-mail: despadas@eucmos.sim.ucm.es 28040 Madrid - SPAIN d.espadas@computer.org _____________________________________________________________________________________