repmat in a mex file
Jozsef Molnar
I am trying to create my own repmat function in a C++ mex
file. It should be fairly simple but I still cannot make it
work properly. It works to replicate matrices horizontally
but does not work vertically. I have tried many different
things but no use. Anyone can help? Thanks in advance.
Below is the code and the output.
#include "mex.h"
#include "matrix.h"
// This is a MEX-file for MATLAB. JTS 19.9.2007
// REPMAT C++-function.
// Creates from matrix y[mF:mL][nL] and numbers repm,repn
// a larger result matrix z[(mL-mF+1)*repm][nL*repn]
void repmatcf6( mwSize mF, mwSize mL, mwSize nL, mwSize
repm, mwSize repn,const double *y, double *z )
{
mwSize i1,i2,j1,j2,count=0;
//for (count=0; count<(mL-mF+1)*repm*nL*repn; count++) {
for (i2=0; i2<repn ; i2++) {
for (i1=0; i1<repm ; i1++) {
for (j2=0; j2<nL ; j2++) {
for (j1=mF-1; j1<mL ; j1++) {
*(z+count)= *(y+j1+j2*(mL-mF+1));
//*(z+count)= *(y+j1*nL+j2);
count++;
}
}
}
}
}
/* the gateway function */
void mexFunction( int nlhs, mxArray *plhs[],
int nrhs, const mxArray *prhs[])
{
double *y,*z;
int mrowf,mrowl;
mwSize repm,repn;
mwSize mzrows,nzcols,ncols;
/* check for proper number of arguments */
/* NOTE: You do not need an else statement when using
mexErrMsgTxt
within an if statement, because it will never get to
the else
statement if mexErrMsgTxt is executed. (mexErrMsgTxt
breaks you out of
the MEX-file) */
if(nrhs!=5)
mexErrMsgTxt("5 inputs required.");
if(nlhs!=1)
mexErrMsgTxt("One output required.");
/* get the scalar input mf,ml,nl,repm,repn */
mrowf = (int)mxGetScalar(prhs[0]);
mrowl = (int)mxGetScalar(prhs[1]);
repm = mxGetScalar(prhs[2]);
repn = mxGetScalar(prhs[3]);
/* create a pointer to the input matrix y */
y = mxGetPr(prhs[4]);
/* get the dimensions of the matrix input y */
// mrows = mxGetM(prhs[4]);
ncols = mxGetN(prhs[4]);
mzrows = (mrowl-mrowf+1)* repm;
nzcols = ncols * repn;
/* set the output pointer to the output matrix z[(mL-
mF+1)*repm][nL*repn] */
plhs[0] = mxCreateDoubleMatrix(mzrows,nzcols, mxREAL);
/* create a C pointer to a copy of the output matrix */
z = mxGetPr(plhs[0]);
/* call the C subroutine */
repmatcf6(mrowf,mrowl,ncols, repm,repn, y,z);
}
And the output:
>> j=[1,2,3;4,5,6]
j =
1 2 3
4 5 6
>> z=repmatcf6(1, 2, 1,1,j)
z =
1 2 3
4 5 6
>> z=repmatcf6(1, 2, 1,2,j)
z =
1 2 3 1 2 3
4 5 6 4 5 6
>> z=repmatcf6(1, 2, 2,1,j)
z =
1 3 2
4 6 5
2 1 3
5 4 6
Bobby Cheng
Most usages of REPMAT can be replaced by a call to BSXFUN starting in
R2007a.
---Bob.
"Jozsef Molnar" <jom...@yahoo.com> wrote in message
news:fcrhus$a7n$1...@fred.mathworks.com...
Jozsef Molnar
This repmat would be only a part of a bigger mex file to
substitute a big for loop in matlab that I cannot avoid. I
will need to write a C code for kron and element-by-element
matrix multiplication as well.
My code works in Visual Basic but does not work in matlab.
Do you have any idea why?
Thanks!
J
"Bobby Cheng" <bch...@mathworks.com> wrote in message
<fcrmrh$9br$1...@fred.mathworks.com>...
Bobby Cheng
Can you post to this newsgroup the MATLAB code you want to turned into MEX?
Lastly, I am not sure why your code does not work. But looking at the
output, it seems that you are using row based storage in C instead of column
base storage that MATLAB uses.
---Bob.
"Jozsef Molnar" <jom...@yahoo.com> wrote in message
news:fcrrnt$948$1...@fred.mathworks.com...
Jozsef Molnar
Here is the matlab code:
The loop in the main part:
for i=1:hhs
k=hhspan(i,4);
negloglik=negloglik-msl_halt5(repmat(random((i-1)
*randdraws+1:i*randdraws,:),k,1),i,i);
end
and the function:
function[sumlik]=msl_halt5(random2,starthh,finhh)
%this function evaluates the simulated likelihood of
choices in data based
%on a parameter matrix for state variables theta and random
draws for each
%household/choice
global hhspan choicedata states randdraws hhs theta beta
charact
%compute probability of each choice
choicesmsl=size(choicedata);
span=hhspan(finhh,3)-hhspan(starthh,2)+1;
nohh=finhh-starthh+1;
t=hhspan(finhh,4); % #of weeks of all these HHs
stateutil=[zeros(span,1) states(hhspan(starthh,2):hhspan
(finhh,3),:)*theta'+[ones(span,1) charact(hhspan
(starthh,2):hhspan(finhh,3),:)]*beta'];
stateutil=kron(stateutil,ones(randdraws,1));
util=stateutil+[zeros(randdraws*nohh*t,1) random2];
temp2=exp(util);
temp=repmat(sum(temp2')',1,choicesmsl(1,2));
lnpi=log(temp2./temp)/randdraws;
lnpi1=lnpi.*kron(choicedata(hhspan(starthh,2):hhspan
(finhh,3),:),ones(randdraws,1)); %probability of actual
choices
sumlik=sum(sum(lnpi1(:,:)));
"Bobby Cheng" <bch...@mathworks.com> wrote in message
<fcuiop$9vv$1...@fred.mathworks.com>...