Skip to content

passing matrix as argument to a C function

5 messages · Christos Hatzis, Brian Ripley, Dirk Eddelbuettel

#
Hi,
 
Although this is not directly an R-related question, it is relevant as I am
trying to port some R code to C to speed things up in a computation.

I am working through my first attempts to generate and link compiled C code
in R.  I could make the 'convolve' function to work and similar functions
that take vectors as arguments.  In my application I need to pass a couple
of matrices to the C function and could not figure out how to make it work
in C.
 
As an example, I tried to code first a simple matrix-vector product
function.  Here is my code:

void matrix_mult( double **A, double *b, double *ab, int *nrow, int *ncol )
{
    int i, j, nr = *nrow, nc = *ncol;
    for( i = 0; i < nr; i++ )
    {
        ab[i] = 0.0;
        for( j = 0; j < nc; j++ )
            ab[i] += A[i][j] * b[j];
    }
} 
 
As I understand, passing the matrix A as (double **) is not standard C and
does not compile (to try things out I am using Microsoft Visual C++ compiler
in Windows XP).  I tried to find the C code for crossprod for some hints,
but could not locate it in the R source.  But probably this does not use the
.C interface.

Is there a way this can be done with .C?  I suspect it will use in some way
the "vector of pointers to row vectors" concept, but I am not familiar
enough in C yet to figure it out.  Any hints on how in can be done with
.Call if easier?

Thank you.
-Christos  

Christos Hatzis, Ph.D.
Nuvera Biosciences, Inc.
400 West Cummings Park
Suite 5350
Woburn, MA 01801
Tel: 781-938-3830
www.nuverabio.com <http://www.nuverabio.com/>
#
The data in an R matrix is not stored as double **A, but as double *A: it 
is just a vector in R, with attributes.

There are lots of examples of manipulating R matrixes via .C: one is 
src/library/stats/src/kmeans.c.

The code for crossprod is not at all hard to find: it is in 
src/main/array.c.  However, is just a call to an LAPACK function.
The code there for matprod would be more informative.
On Sun, 3 Dec 2006, Christos Hatzis wrote:

            

  
    
#
Thank you.  The code in kmeans.c is indeed informative.  One needs to
remember that an R matrix is a vector with attributes.

-Christos 

-----Original Message-----
From: Prof Brian Ripley [mailto:ripley at stats.ox.ac.uk] 
Sent: Sunday, December 03, 2006 3:25 AM
To: Christos Hatzis
Cc: r-devel at r-project.org
Subject: Re: [Rd] passing matrix as argument to a C function

The data in an R matrix is not stored as double **A, but as double *A: it is
just a vector in R, with attributes.

There are lots of examples of manipulating R matrixes via .C: one is
src/library/stats/src/kmeans.c.

The code for crossprod is not at all hard to find: it is in
src/main/array.c.  However, is just a call to an LAPACK function.
The code there for matprod would be more informative.
On Sun, 3 Dec 2006, Christos Hatzis wrote:

            

  
    
#
On 3 December 2006 at 11:47, Christos Hatzis wrote:
| Thank you.  The code in kmeans.c is indeed informative.  One needs to
| remember that an R matrix is a vector with attributes.

In case there is any chance that you could turn your part of the source from
C into C++, try RcppTemplate from CRAN --- it makes interfacing R and C++ a
whole lot easier.  I cannot recommend it highly enough. 

But if you don't yet know C++ it provides yet another learning curve so you
may not want to go there quite yet...

Dirk
#
Thank you, Dirk.  I'll keep this in mind.

-Christos 

-----Original Message-----
From: Dirk Eddelbuettel [mailto:edd at debian.org] 
Sent: Sunday, December 03, 2006 12:47 PM
To: christos at nuverabio.com
Cc: 'Prof Brian Ripley'; r-devel at r-project.org
Subject: Re: [Rd] passing matrix as argument to a C function
On 3 December 2006 at 11:47, Christos Hatzis wrote:
| Thank you.  The code in kmeans.c is indeed informative.  One needs to 
| remember that an R matrix is a vector with attributes.

In case there is any chance that you could turn your part of the source from
C into C++, try RcppTemplate from CRAN --- it makes interfacing R and C++ a
whole lot easier.  I cannot recommend it highly enough. 

But if you don't yet know C++ it provides yet another learning curve so you
may not want to go there quite yet...

Dirk

--
Hell, there are no rules here - we're trying to accomplish something. 
                                                  -- Thomas A. Edison