Prev: What is the program of excel solver in MatlaB?
Next: Normality test Question
From: Chaopeng on 9 Aug 2010 23:31
Dear Friends,

We are troubled by the fact that %VAL construct doesn't seem to work with Mat-file 'stand-alone' programs. Our compiler has been set to work perfectly if not %VAL construct was used. However, we were getting used to write programs with %VAL in our previous mex files as it passes the addresses thus avoiding duplicating the data. While %VAL works happily in a mex file, in a mat-file program the compiler complains:

Error 1 Error: The type of the actual argument differs from the type of the dummy argument. [%VAL] C:\temp\test.f90 12

We are wondering if someone has worked through this before and made %VAL work with mat-file programs? It really doesn't make sense why %VAL doesn't work in program but works in a mexFunction. Here is the attached simple test program. Basically, it create a variable, and then tries to modify it inside a program.

We really appreciate your input!

test.F90:

#include "fintrf.h"
program paws
implicit none
mwPointer mxCreateDoubleMatrix, mxGetPr
mwSize m, n
integer*4 ComplexFlag
mwPointer pr, ptr2
m = 1
n = 1
ComplexFlag = 0
pr = mxCreateDoubleMatrix(m, n, ComplexFlag)
ptr2 = mxGetPr(pr)
CALL change_value(%val(ptr2))
end program paws


subroutine change_value(pr)
implicit none
real*8 pr
pr = 1000
end subroutine
From: Chaopeng on 10 Aug 2010 09:52
Anyone worked with Mat-file programs before?


"Chaopeng " <shakeham(a)gmail.com> wrote in message <i3qh5o$a7$1(a)fred.mathworks.com>...
> Dear Friends,
>
> We are troubled by the fact that %VAL construct doesn't seem to work with Mat-file 'stand-alone' programs. Our compiler has been set to work perfectly if not %VAL construct was used. However, we were getting used to write programs with %VAL in our previous mex files as it passes the addresses thus avoiding duplicating the data. While %VAL works happily in a mex file, in a mat-file program the compiler complains:
>
> Error 1 Error: The type of the actual argument differs from the type of the dummy argument. [%VAL] C:\temp\test.f90 12
>
> We are wondering if someone has worked through this before and made %VAL work with mat-file programs? It really doesn't make sense why %VAL doesn't work in program but works in a mexFunction. Here is the attached simple test program. Basically, it create a variable, and then tries to modify it inside a program.
>
> We really appreciate your input!
>
> test.F90:
>
> #include "fintrf.h"
> program paws
> implicit none
> mwPointer mxCreateDoubleMatrix, mxGetPr
> mwSize m, n
> integer*4 ComplexFlag
> mwPointer pr, ptr2
> m = 1
> n = 1
> ComplexFlag = 0
> pr = mxCreateDoubleMatrix(m, n, ComplexFlag)
> ptr2 = mxGetPr(pr)
> CALL change_value(%val(ptr2))
> end program paws
>
>
> subroutine change_value(pr)
> implicit none
> real*8 pr
> pr = 1000
> end subroutine
From: James Tursa on 10 Aug 2010 11:04
"Chaopeng " <shakeham(a)gmail.com> wrote in message <i3qh5o$a7$1(a)fred.mathworks.com>...
> Dear Friends,
>
> We are troubled by the fact that %VAL construct doesn't seem to work with Mat-file 'stand-alone' programs. Our compiler has been set to work perfectly if not %VAL construct was used. However, we were getting used to write programs with %VAL in our previous mex files as it passes the addresses thus avoiding duplicating the data. While %VAL works happily in a mex file, in a mat-file program the compiler complains:
>
> Error 1 Error: The type of the actual argument differs from the type of the dummy argument. [%VAL] C:\temp\test.f90 12
>
> We are wondering if someone has worked through this before and made %VAL work with mat-file programs? It really doesn't make sense why %VAL doesn't work in program but works in a mexFunction. Here is the attached simple test program. Basically, it create a variable, and then tries to modify it inside a program.
>
> We really appreciate your input!
>
> test.F90:
>
> #include "fintrf.h"
> program paws
> implicit none
> mwPointer mxCreateDoubleMatrix, mxGetPr
> mwSize m, n
> integer*4 ComplexFlag
> mwPointer pr, ptr2
> m = 1
> n = 1
> ComplexFlag = 0
> pr = mxCreateDoubleMatrix(m, n, ComplexFlag)
> ptr2 = mxGetPr(pr)
> CALL change_value(%val(ptr2))
> end program paws
>
>
> subroutine change_value(pr)
> implicit none
> real*8 pr
> pr = 1000
> end subroutine

I don't see anything obviously wrong with the code. What compiler are you using? Have you looked in the compiler doc to see if there is a special compiler setting that must be set to use %VAL?

James Tursa
From: Chaopeng on 10 Aug 2010 11:44
we are using ifort
as I am searching in futile on the internet about the use %VAL, I found other people were able to use it in fortran programs that calls C functions which don't seem to need any special compiler options
http://docs.hp.com/en/B3908-90002/ch07s06.html

Here are some compiler options included in the mexopts.sh file. I don't see anything special that is related to this feature

FC='ifort'
FFLAGS='-fexceptions'
FFLAGS="$FFLAGS -O0 -static-intel -fpp -g -fPIC -fno-omit-frame-pointer -DMATLAB_MEX_FILE -I/$MATLAB/extern/include"
FLIBS="$RPATH $MLIBS -lm"
FOPTIMFLAGS='-O0'
FDEBUGFLAGS='-g'
#
LD='ifort'
LDEXTENSION='.mexa64'
LDFLAGS="-O0 -pthread -shared -Wl,--version-script,$TMW_ROOT/extern/lib/$Arch/$MAPFILE -Wl,--no-undefined"
LDOPTIMFLAGS='-O0'
LDDEBUGFLAGS='-g'

"James Tursa" <aclassyguy_with_a_k_not_a_c(a)hotmail.com> wrote in message <i3rpp5$3ck$1(a)fred.mathworks.com>...
> "Chaopeng " <shakeham(a)gmail.com> wrote in message <i3qh5o$a7$1(a)fred.mathworks.com>...
> > Dear Friends,
> >
> > We are troubled by the fact that %VAL construct doesn't seem to work with Mat-file 'stand-alone' programs. Our compiler has been set to work perfectly if not %VAL construct was used. However, we were getting used to write programs with %VAL in our previous mex files as it passes the addresses thus avoiding duplicating the data. While %VAL works happily in a mex file, in a mat-file program the compiler complains:
> >
> > Error 1 Error: The type of the actual argument differs from the type of the dummy argument. [%VAL] C:\temp\test.f90 12
> >
> > We are wondering if someone has worked through this before and made %VAL work with mat-file programs? It really doesn't make sense why %VAL doesn't work in program but works in a mexFunction. Here is the attached simple test program. Basically, it create a variable, and then tries to modify it inside a program.
> >
> > We really appreciate your input!
> >
> > test.F90:
> >
> > #include "fintrf.h"
> > program paws
> > implicit none
> > mwPointer mxCreateDoubleMatrix, mxGetPr
> > mwSize m, n
> > integer*4 ComplexFlag
> > mwPointer pr, ptr2
> > m = 1
> > n = 1
> > ComplexFlag = 0
> > pr = mxCreateDoubleMatrix(m, n, ComplexFlag)
> > ptr2 = mxGetPr(pr)
> > CALL change_value(%val(ptr2))
> > end program paws
> >
> >
> > subroutine change_value(pr)
> > implicit none
> > real*8 pr
> > pr = 1000
> > end subroutine
>
> I don't see anything obviously wrong with the code. What compiler are you using? Have you looked in the compiler doc to see if there is a special compiler setting that must be set to use %VAL?
>
> James Tursa
From: Chaopeng on 10 Aug 2010 12:00
it appears when using %VAL in general, the type is the same between actual and dummy argument. However, in this case actual argument is mwPointer (an integer) while dummy is real*8. However, we don't really care since we are merely concerned with the memory address. the mex compiler must somehow ignored the error message. We haven't been able to locate what that option is
 |  Next  |  Last
Pages: 1 2
Prev: What is the program of excel solver in MatlaB?
Next: Normality test Question