MATLAB Answers

Greg
0

Accessing Non-Scalar Arrays with ceval

Asked by Greg
on 11 Jun 2012
Hi,
I'm trying to use the Matlab "mat" and "mx" C-libraries to use .mat-files inside of a Matlab function I'm compiling with the Matlab Coder. I can do this successfully for a scalar variable, but I'm having trouble getting non-scalar arrays in. Because coder.ceval can only return a scalar, my best guess of how to do this at the moment is to use the C "memcpy" function with a coder.wref input.
Here's what I've got right now, referencing a .mat-file "testmat.mat" that has one variable in it, a 1x4 double:
y = ones(1,4);
SS_Table_File = coder.opaque('MATFile *');
rr = ['r',char(0)];
SS_Table_file = coder.ceval('matOpen','testmat.mat',rr);
Xp = coder.opaque('mxArray *');
XX = ['X',char(0)];
Xp = coder.ceval('matGetVariable',SS_Table_file,XX);
coder.ceval('matClose',SS_Table_file);
yp = coder.opaque('double *');
yp = coder.ceval('mxGetPr',Xp);
coder.ceval('memcpy',coder.wref(y),yp,int32(4*8));
This currently compiles OK but then produces a segfault (or some other fatal error that crashes Matlab) when I run it.

  0 Comments

Sign in to comment.

1 Answer

Answer by Greg
on 13 Jun 2012
 Accepted Answer

Apparently the above code would actually work OK if size_t were a 32-bit integer (which I had assumed it was). On my 64-bit linux machine that defines size_t as an unsigned long int (a uint64), I have to do something like this since the Matlab coder doesn't support 64-bit integers:
sz = coder.opaque('size_t');
sz = coder.ceval('(size_t)',4*8);
coder.ceval('memcpy',coder.wref(y),yp,sz);

  0 Comments

Sign in to comment.