Hi from the look of your code, I dont know whether you have initialised the
MPI code before trying to derive the new datatype. I also hope that all the
executing processes call the datatype derivation code. Just take a look at
your just in case ...
Cheers
Leslie Nkansa Osei-Bonsu
Graduate Student
Computational Science and Engineering
Department of Informatics
Technische Universität München
>From: Karlos Bohan <kartle_at_[hidden]>
>Reply-To: Karlos Bohan <kartle_at_[hidden]>, General LAM/MPI mailing
>list <lam_at_[hidden]>
>To: lam_at_[hidden]
>Subject: LAM: User Derived Type Problem
>Date: Fri, 18 Feb 2005 11:59:04 +0000
>
>Hello,
>
>I'm a final year student trying currently working on my thesis in
>which I'm attempting to implement a global optimization algorithm in
>parallel.
>
>I'm new to LAM and parallel programming in general.
>
>I'm attempting to define a struct in C which will be passed to the
>different processors on the cluster I'm implementing it on.
>
>The stuct has three elements,
>
>1. A multidimensional Array of type double, which holds the current
>population of solution vectors
>2. A vector array of type double which holds the best current vector
>3. A double which holds the result of the target function on the best
>current vector
>
>When I started out I was using the source code for a general
>Master/Slave skeleton at:
>
>http://www.lam-mpi.org/tutorials/one-step/ezstart.php
>
>I originally attempted to fit the code for my algorthim into this
>skeleton, however this proved extremely difficult to compile and debug
>and so I've now reverted to building up my own version from scratch
>with the idea of having as few compile errors as I go, and building up
>the program eliminating them as I go.
>
>I've already run into problems defining the MPI_Datatype which I'll
>need to pass, this is before I can even start to actually use it!!
>
>The code for defining it is as follows and is based on section 5.6
>"Using MPI Derived Types for User-Defined Types" of the NCSA tutorial
>reccomended on the LAM/MPI site
>
>The Code I have is as follows for the user defined struct Kstruct:
>
>//Kstruct definition:
>struct kstruct{
> double population[MAXPOP][MAXDIM];
> double bestPopVector[MAXDIM];
> double bestCostValue;
>};
>
>typedef struct kstruct Kstruct;
>
>//This is the code then running in Main()
>
>
>Kstruct masterKstruct; /* a representative variable of this type */
> int lena[3]; /* the three arrays used to describe an MPI derived
>type */
> MPI_Aint loca[3]; /* their size reflects the number of
>components in SparseElt */
> MPI_Datatype typa[3];
>
> MPI_Aint baseaddress;
> MPI_Datatype MPI_Kstruct; /* a variable to hold the MPI type
>indicator for SparseElt */
> /* set up the MPI description of SparseElt */
> MPI_Address(&masterKstruct, &baseaddress);
> lena[0] = 3; MPI_Address(&masterKstruct.population,&loca[0]);
> loca[0] -= baseaddress; typa[0] = MPI_DOUBLE;
> lena[1] = 2; MPI_Address(&masterKstruct.bestPopVector,&loca[1]);
> loca[1] -= baseaddress; typa[1] = MPI_DOUBLE;
> lena[2] = 1; MPI_Address(&masterKstruct.bestCostValue ,&loca[2]);
> loca[2] -= baseaddress; typa[2] = MPI_DOUBLE;
>
> MPI_Type_struct(3, lena, loca, typa, &MPI_Kstruct);
> MPI_Type_commit(&MPI_Kstruct);
>
> MPI_Status status;
>
> MPI_Init(&argc, &argv);
> MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
> MPI_Comm_rank(MPI_COMM_WORLD,&myid);
>
>....
>
>When I try to compile this program I get the following error:
>
>mpitest2.c:234: parse error before `status'
>
>
>Could anyone tell me where I'm going wrong with this?
>
>Also, assuming I can get this working, would it be fair to say that
>comparison operations on the struct once its defined as an MPI type
>will be relatively straight forward as all I'll be doing then is
>comparison functions?
>
>Many thanks
>k
>_______________________________________________
>This list is archived at http://www.lam-mpi.org/MailArchives/lam/
_________________________________________________________________
Don't just search. Find. Check out the new MSN Search!
http://search.msn.com/
|
