Intel® Fortran Compiler
Build applications that can scale for the future with optimized code designed for Intel® Xeon® and compatible processors.
29403 ディスカッション

How to create Fortran array referencing SafeArrayAccessData

MWind2
新規コントリビューター III
6,806件の閲覧回数

VBArray INOUT length known
integer(C_INTPTR_T) :: dptr
...
iRes=SafeArrayAccessData(VBArray,dptr)
dptr address is starting point of one dimensional c array of real{8}

I want to put it in a Fortran array so I can see it in debugger for one reason, but I want the Fortran array to reference the underlying data and not be a copy. Intention would be changes to data in Fortran array would be preserved in dptr array.

0 件の賞賛
1 解決策
Steve_Lionel
名誉コントリビューター III
6,744件の閲覧回数

Two errors. First, you omitted the size argument I specified to C_F_POINTER. Second, you treated the array as if it were zero-origin, but the array you get from C_F_POINTER is one-origin. (This could be finessed with pointer assignment redeclaring the bounds, but that seems unnecessary.)

I attach my modified df.f90. I didn't change anything else.

元の投稿で解決策を見る

23 返答(返信)
Steve_Lionel
名誉コントリビューター III
5,900件の閲覧回数

Something like this:

use, intrinsic :: iso_c_binding
...
real(c_double), pointer :: array_data(:)
...
call c_f_pointer(transfer(dptr,c_null_ptr),array_data,[array_size])

where array_size is the number of elements in the array (can be a variable or a constant). After this, array_data is your one-dimensional array, no copy done.

MWind2
新規コントリビューター III
5,886件の閲覧回数

I seem to be getting an Exception thrown at 0x00007FF88E912D83 (df.dll) in EXCEL.EXE: 0xC0000005: Access violation writing location 0x0000000000000000 because of the "c_null_ptr" in the transfer.  Data is "adptr"(changed from dptr as dptr used as another pointer).

Steve_Lionel
名誉コントリビューター III
5,882件の閲覧回数

C_NULL_PTR as the second argument to TRANSFER is just the "mold" - it provides the type to interpret the source (first argument) as.  It is not accessed as data. If you got an access violation, then adptr was zero. I suggest adding some logging to your code (write to a file perhaps.)

MWind2
新規コントリビューター III
5,871件の閲覧回数

I attached a vs2022 folder with all inside. Excel is 64 bit and projects. 

Steve_Lionel
名誉コントリビューター III
6,745件の閲覧回数

Two errors. First, you omitted the size argument I specified to C_F_POINTER. Second, you treated the array as if it were zero-origin, but the array you get from C_F_POINTER is one-origin. (This could be finessed with pointer assignment redeclaring the bounds, but that seems unnecessary.)

I attach my modified df.f90. I didn't change anything else.

MWind2
新規コントリビューター III
5,856件の閲覧回数

Thank you! I spent all day trying for magic incantations. Intention here was to show that a Excel vba array of doubles used in calculations could be easily sent to a fortran dll and the results sent back in a position(s) in the array for Re: Reading Data from an Excel Spreadsheet? - Page 2 - Intel Communities. One change was made in formula 

array_data(4)=PI*array_data(1)*array_data(1)*array_data(2)*array_data(3)/4.0 for weight of cylinder agree with others I had Option 0 on in vba, but array in fortran was [1,4]. Without mucking it up, how to set array_data to be [0,3]?

Steve_Lionel
名誉コントリビューター III
5,842件の閲覧回数

Add a new array pointer declaration as follows:

 

real(8), pointer :: array_data(:), array_data1(:)

 

then replace the C_F_POINTER call with these two lines:

 

    call c_f_pointer(cptr,array_data1,[4])
    array_data(0:) => array_data1

 

I had thought I could do:

 

array_data(0:) => array_data

 

but that didn't work, for reasons still unclear to me (it caused array_data to become disassociated - this MIGHT be a bug, and I will investigate further.)

Edit: I can't reproduce the problem in a simple test. Shrug....

MWind2
新規コントリビューター III
5,828件の閲覧回数

Thank you! Sometimes I'm "cset" in my ways if permitted, and I would have still been looking for Rosetta stone.

While trying to find things that applied in books, I was thinking that most write to the overview, not that there is anything wrong with that, but a compendium of use questions and answers would be more useful at times. It seems forums nearly do the job, but there are problems with that like too much archiving or deleting of still relevant material, besides signal to noise.

MWind2
新規コントリビューター III
5,808件の閲覧回数

Is it possible to do this for the two dimensional array? So far I have had no success?

 

The bounds part works.

Steve_Lionel
名誉コントリビューター III
5,800件の閲覧回数

Look again at the VB.NET-Safearrays sample, from which you borrowed some code. It illustrates how to access arrays of any dimension.

MWind2
新規コントリビューター III
5,791件の閲覧回数

The looping into each one array goes fine, but combining or assigning those to a Fortran 2d array is where I am not getting.

Steve_Lionel
名誉コントリビューター III
5,771件の閲覧回数

Please zip and attach what you have now, and I'll take a look. 

MWind2
新規コントリビューター III
5,762件の閲覧回数

I've been using x1.xlsm, dfad and dcpp1, no progress with fortran.

Steve_Lionel
名誉コントリビューター III
5,754件の閲覧回数

You have a lot of code commented out, but I see this:

cptr=transfer(adptr,c_null_ptr)
  call c_f_pointer(cptr,array_data,[12])
  array_data(0:) => array_data1

That should be array_data1 in the second line. 

Where is the code where you're "assigning to a 2D Fortran array"?

MWind2
新規コントリビューター III
5,748件の閲覧回数

I was getting nowhere but locking exceptions, so this was what I was doing . Trying to do in fortran what was done in dcpp1 and so far getting nowhere.

Steve_Lionel
名誉コントリビューター III
5,741件の閲覧回数

I have not tried to edit your sources, but 1) You still have that error on the call to c_f_pointer, and 2) unlike the C++ code, you unlock the SafeArray before accessing it.

MWind2
新規コントリビューター III
5,730件の閲覧回数

When I use 

 

 

readloop:  do
  ires=SafeArrayGetElemsize(VBArray)
  ires = SafeArrayGetElement (VBArray, indexes(1), loc(adptr))
  !
  cptr=transfer(adptr,c_null_ptr)
  call c_f_pointer(cptr,array_data,[ilen])  
  call c_f_pointer(cptr,array_data1,[ilen])
  array_data(0:) => array_data1
  !
  !  array_data1(3)=PI*array_data1(0)*array_data1(0)*array_data1(1)*array_data1(2)
  do i = nbounds, 1, -1
     indexes(i) = indexes(i) + 1
     if (indexes(i) <= bounds(i)%ub) exit
       indexes(i) = bounds(i)%lb
     if (i == 1) exit readloop
  end do
  end do readloop
 iRes=SafeArrayUnlock(VBArray)

 

 

I get an element size of 8, which does not seem to be an array size. Maybe I'm reading it wrong but GetElement may be just a double.

MWind2
新規コントリビューター III
5,726件の閲覧回数

Got a fortran array with such but maybe order wrong:

 iRes=SafeArrayLock(VBArray)
  indexes = bounds%lb  ! Initialize to all lower bounds
  ilen=bounds(2)%ub-bounds(2)%lb
  ires=SafeArrayGetElemsize(VBArray)
  !ires = SafeArrayPtrOfIndex(VBArray, indx, cptr0)
  ires=SafeArrayAccessData(VBArray,adptr)
  !ires = SafeArrayGetElement (VBArray, indexes(1), loc(adptr))
  !
  cptr=transfer(adptr,c_null_ptr)
  call c_f_pointer(cptr,array_data,[12])  
  !call c_f_pointer(cptr,array_data1,[12])
  !array_data(0:) => array_data1
  !
  !  array_data1(3)=PI*array_data1(0)*array_data1(0)*array_data1(1)*array_data1(2)
 
  iRes=SafeArrayUnlock(VBArray)
MWind2
新規コントリビューター III
5,721件の閲覧回数

Not pretty but curious.

MWind2
新規コントリビューター III
5,678件の閲覧回数

I'm working on a general dimensional case now, but I found what I needed to do provided I know the ub,lb of each dimension with the following working for the 3X4:

                                 

 

 

call c_f_pointer(cptr,array_data,[ilen])  
  call c_f_pointer(cptr,array_data1,[ilen])
  array_data(0:) => array_data1
  a2d(0:2,0:3)=>array_data
  do i=0,2
      a2d(i,3)=PI*a2d(i,0)*a2d(i,0)*a2d(i,1)*a2d(i,2)/4
  end do

 

 

返信