Hi Edgar, Yeah I finally figured out that by MPI_2DOUBLE_COMPLEX they didn't mean it was two doubles side my side (real and imag) to make a complex. Once I sorted things out with MPI_DOUBLE_COMPLEX it seems to work great. Thanks!
Corey On 9/4/07, Edgar Gabriel <gabr...@cs.uh.edu> wrote: > you are probably using the wrong data type. MPI_2DOUBLE_COMPLEX is like > a structure of two 'double complex' data types for MPI_MINLOC/MAXLOC > operations in MPI_Reduce/Allreduce. > > You should try MPI_DOUBLE_COMPLEX. Resp, if this still doesn't work, > compare the size of your complex data type from complex.h (e.g. using > the sizeof operator) and the size of the MPI_DOUBLE_COMPLEX data type > (e.g. using MPI_Type_size). If they don't match you probably need to use > MPI_COMPLEX. > > > Corey Putkunz wrote: > > Hi everyone, > > > > An interesting problem I've discovered with MPI_Read_file (part of the > > MPI I/O interface). When trying to read a 256^3 complex array (by > > complex I mean "complex" defined in complex.h) into a single > > dynamically allocated array (whos memory is successfully allocated). > > To read I call: > > > > MPI_File_read ( file_in, data, S, MPI_2DOUBLE_COMPLEX, &status ); > > > > (where S=256x256x256) though the call fails and I can't figure out > > why. If I reduce the amount of data MPI_File_read attempts to read the > > call works. I can find this issue anywhere else on the net. Please > > help anyone! > > > > (System is running on an Intel machine, Fedora 7) > > > > Thanks. Corey > > _______________________________________________ > > users mailing list > > us...@open-mpi.org > > http://www.open-mpi.org/mailman/listinfo.cgi/users > > _______________________________________________ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users >
