Thread Subject: Float data to a MEX function

Hi!
I was aondering: is there a way to pass float data to a MEX function? does mxGetPr only work with doubles?
Many thanks in advance,
Cesare

"Cesare " <cmfornewsgroup@gmail.com> wrote in message <h43qap$9ot$1@fred.mathworks.com>...
> Hi!
> I was aondering: is there a way to pass float data to a MEX function? does mxGetPr only work with doubles?

Simply cast mxGetPr to float*, or cleaner way use mxGetData (you might need to cast still).

Bruno

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <h43rlu$6oc$1@fred.mathworks.com>...
> "Cesare " <cmfornewsgroup@gmail.com> wrote in message <h43qap$9ot$1@fred.mathworks.com>...
> > Hi!
> > I was aondering: is there a way to pass float data to a MEX function? does mxGetPr only work with doubles?
>
> Simply cast mxGetPr to float*, or cleaner way use mxGetData (you might need to cast still).
>
> Bruno

Hi!
thanks very much. So there is no way to do it without a casting I guess. Is that very time consuming if done inside a C application?
Best,
Cesare

"Cesare " <cmfornewsgroup@gmail.com> wrote in message <h441dn$ffg$1@fred.mathworks.com>...
> "Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <h43rlu$6oc$1@fred.mathworks.com>...
> > "Cesare " <cmfornewsgroup@gmail.com> wrote in message <h43qap$9ot$1@fred.mathworks.com>...
> > > Hi!
> > > I was aondering: is there a way to pass float data to a MEX function? does mxGetPr only work with doubles?
> >
> > Simply cast mxGetPr to float*, or cleaner way use mxGetData (you might need to cast still).
> >
> > Bruno
>
> Hi!
> thanks very much. So there is no way to do it without a casting I guess. Is that very time consuming if done inside a C application?

Cast a pointer costs 0 in CPU runtime, it was done during compilation.

Bruno

But what if I have the opposite situation: my MEX-written function is suppose to accept a double and the user passes a long (this is actually causing my MEX function to crash). Is there a workaround for that?
Thanks again,
Cesare


"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <h4422a$qg6$1@fred.mathworks.com>...
> "Cesare " <cmfornewsgroup@gmail.com> wrote in message <h441dn$ffg$1@fred.mathworks.com>...
> > "Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <h43rlu$6oc$1@fred.mathworks.com>...
> > > "Cesare " <cmfornewsgroup@gmail.com> wrote in message <h43qap$9ot$1@fred.mathworks.com>...
> > > > Hi!
> > > > I was aondering: is there a way to pass float data to a MEX function? does mxGetPr only work with doubles?
> > >
> > > Simply cast mxGetPr to float*, or cleaner way use mxGetData (you might need to cast still).
> > >
> > > Bruno
> >
> > Hi!
> > thanks very much. So there is no way to do it without a casting I guess. Is that very time consuming if done inside a C application?
>
> Cast a pointer costs 0 in CPU runtime, it was done during compilation.
>
> Bruno

"Cesare " <cmfornewsgroup@gmail.com> wrote in message <h4439o$ffu$1@fred.mathworks.com>...
> But what if I have the opposite situation: my MEX-written function is suppose to accept a double and the user passes a long (this is actually causing my MEX function to crash). Is there a workaround for that?

If your MEX accepts double only, then you should check the input using mxGetClassID. More or less strict verification is really at the discretion of programmer/user.

Bruno

Thanks again! I'm sorry, but I still got a problem. I'll try to explain it better.
In my code, the variable R is declared as a double pointer

double *R;

R is supposed to point to the numerical content of prhs[0]

R = mxGetPr(prhs[0]);

The problem comes when the user passes a matlab single matrix as first input, instead of a double one. The crash does not happen at this point but later in the code probably cause, since R was declared as a double pointer, I end up accessing values in the first input which do not exist. Is there a fast easy way to solve it?
Thanks a lot again,
Cesare

"Bruno Luong" <b.luong@fogale.findmycountry> wrote in message <h4443t$al7$1@fred.mathworks.com>...
> "Cesare " <cmfornewsgroup@gmail.com> wrote in message <h4439o$ffu$1@fred.mathworks.com>...
> > But what if I have the opposite situation: my MEX-written function is suppose to accept a double and the user passes a long (this is actually causing my MEX function to crash). Is there a workaround for that?
>
> If your MEX accepts double only, then you should check the input using mxGetClassID. More or less strict verification is really at the discretion of programmer/user.
>
> Bruno

"Cesare " <cmfornewsgroup@gmail.com> wrote in message <h44ava$lia$1@fred.mathworks.com>...
> Thanks again! I'm sorry, but I still got a problem. I'll try to explain it better.
> In my code, the variable R is declared as a double pointer
>
> double *R;
>
> R is supposed to point to the numerical content of prhs[0]
>
> R = mxGetPr(prhs[0]);
>
> The problem comes when the user passes a matlab single matrix as first input, instead of a double one. The crash does not happen at this point but later in the code probably cause, since R was declared as a double pointer, I end up accessing values in the first input which do not exist. Is there a fast easy way to solve it?
> Thanks a lot again,
> Cesare

Why a bunch of back and forth question. Your Mex work with double, then you need to cast the input in double, either preprocessing from Matlab, either on your Mex.

Otherwise you need to build a Mex file that can accepts single and double, period.

Cast in Matlab is very easy to perform:

function output = dosomething(input)

if strcmp(class(input),'single')
  input=double(input);
end
output = mexsomething(input);
end