(j3.2006) Question about zero-sized array.

Daniel C Chen cdchen
Thu Jun 22 10:22:42 EDT 2017


Thanks Walt and Tom!

I think my confusion is I still couldn't find a place in the standard that
specifies the behavior you guys described.

Daniel

XL Fortran Development, Fortran Standard Representative
IBM Toronto Software Lab
Phone: 905-413-3056
Tie: 969-3056
Email: cdchen at ca.ibm.com
http://www.ibm.com/software/awdtools/fortran/xlfortran



From:	Walt Brainerd <walt.brainerd at gmail.com>
To:	fortran standards email list for J3 <j3 at mailman.j3-fortran.org>
Date:	06/21/2017 07:24 PM
Subject:	Re: (j3.2006) Question about zero-sized array.
Sent by:	j3-bounces at mailman.j3-fortran.org



I think Tom correctly answered your Q3.
The result of the MATMUL is a rank-1 size 7 array of 0s,
since each element is the sum of an empty array.

You can generate all the code you want as long as you get the right answer,
but as Tom says, if somehow you figure out that the MATMUL result is all 0,
it certainly could be optimized away. So the whole assignment can be
optimized away (again, if all that can be determined by the compiler).


On Wed, Jun 21, 2017 at 1:18 PM, Clune, Thomas L. (GSFC-6101) <
thomas.l.clune at nasa.gov> wrote:

        On Jun 21, 2017, at 2:51 PM, Daniel C Chen <cdchen at ca.ibm.com>
        wrote:



        Sure. I agree with Q1 and Q2.

        As for Q3,
        1. is the statement standard conforming?
        y(:) = MATMUL( a(:,:), x(:) ) + 2*y(:)



  Yes.



        2. if it is standard conforming, would it be equivalent to
        y(:) = 2*y(:)



  Yes.


        or it should be no code gen for the whole assignment?


  The array y _does_ have elements. ? The expression changes their values.
  The compiler would need to ?know? that all of the elements are initially
  0 to skip the assignment, and this has nothing to do with zero-sized
  things.

  - Tom





        Thanks,

        Daniel

        XL Fortran Development, Fortran Standard Representative
        IBM Toronto Software Lab
        Phone: 905-413-3056
        Tie: 969-3056
        Email: cdchen at ca.ibm.com
        http://www.ibm.com/software/awdtools/fortran/xlfortran

        <graycol.gif>Walt Brainerd ---06/21/2017 01:43:49 PM---Let me try
        to help (maybe it won't) with a different point of view. I always
        thought it was not the




        From: Walt Brainerd <walt.brainerd at gmail.com>
        To: fortran standards email list for J3 <j3 at mailman.j3-fortran.org>
        Date: 06/21/2017 01:43 PM
        Subject: Re: (j3.2006) Question about zero-sized array.
        Sent by: j3-bounces at mailman.j3-fortran.org





        Let me try to help (maybe it won't) with a different point of view.
        I always thought it was not the best to say that x is always
        defined,
        but since the standard does . . .

        For Q1, I would say the value of x is the empty array.

        For Q2, I think you are asking the wrong question.
        An *element* of the array can have the value 0.0 or 1.0, but
        x has no elements, and the question is meaningless.
        Also Q3.

        Crystal clear?

        On Wed, Jun 21, 2017 at 9:37 AM, Daniel C Chen <cdchen at ca.ibm.com>
        wrote:
              Thanks John and Malcolm for your response. I didn't make the
              questions clear enough so I will try again.

              As for Q1,
              non-allocatable case:
              real :: x(0) ! x is always defined by the standard. What is
              its value at this point?
              allocatable case:
              real, allocatable :: x(:)
              allocate(x0)) ! at this point, x is defined by allocation.
              Again, what is the value after allocation?

              As for Q2, I actualy meant to ask 'x' rather than 'a'.
              x = 0.0 ! Does x have value 0.0?
              x = 1.0 ! x Does x have value 1.0?

              As for Q3, if x in Q2 is defined with the value 0.0 or 1.0
              and used in the following statement respectively,
              a(:,:) = 1.0
              y(:) = MATMUL( a(:,:), x(:) ) + 2*y(:) ! Does the value of x
              affect the result?

              Thanks,

              Daniel

              XL Fortran Development, Fortran Standard Representative
              IBM Toronto Software Lab
              Phone: 905-413-3056
              Tie: 969-3056
              Email: cdchen at ca.ibm.com
              http://www.ibm.com/software/awdtools/fortran/xlfortran

              <graycol.gif>"Malcolm Cohen" ---06/20/2017 07:52:35
              PM---Daniel Chen asks: > The standard specifies, "Zero-sized
              arrays and zero-length strings are


              From: "Malcolm Cohen" <malcolm at nag-j.co.jp>
              To: "'fortran standards email list for J3'" <
              j3 at mailman.j3-fortran.org>
              Date: 06/20/2017 07:52 PM
              Subject: Re: (j3.2006) Question about zero-sized array.
              Sent by: j3-bounces at mailman.j3-fortran.org



              Daniel Chen asks:
              > The standard specifies, "Zero-sized arrays and zero-length
              strings are
              > always defined".
              > It also says "Allocation of a zero-sized array or
              zero-length character
              > variable causes the array or variable to become defined"
              >
              > Question1: Does it have initial value at all?

              John Reid opines:
              >Well, yes. It always has the same value, including
              initially.

              No.? Being allocatable or a pointer it is initially
              undefined.? ?initially? means at the beginning of execution;
              at that point the array does not have any shape.

              Daniel continues::
              > allocate(a(7,0),x(0),y(7))
              > a(:,:) = 0.0 !Q2: Is this legal? What is the value of a?

              John replies:
              >Yes, you are broadcasting a scalar to every element of the
              array. a has
              >the value it always has when the size is zero.

              ?it always has? could be a little misleading here; being an
              array, the shape is part of the value, so the value is the
              7x0 empty array and not the 0x23 empty array.? I?m not sure
              we have all this explicitly written down anywhere though,
              except for derived types, where we say the array bounds of a
              component are part of the value of the structure.? Outside of
              a structure it is the shape that?s important rather than the
              bounds.

              Cheers,
              --
              ..............Malcolm Cohen, NAG Oxford/Tokyo.

              From: j3-bounces at mailman.j3-fortran.org [
              mailto:j3-bounces at mailman.j3-fortran.org] On Behalf Of John
              Reid
              Sent: Wednesday, June 21, 2017 12:53 AM
              To: fortran standards email list for J3 <
              j3 at mailman.j3-fortran.org>
              Subject: Re: (j3.2006) Question about zero-sized array.



              Daniel C Chen wrote:
              > Hello,
              >
              > The standard specifies, "Zero-sized arrays and zero-length
              strings are
              > always defined".
              > It also says "Allocation of a zero-sized array or
              zero-length character
              > variable causes the array or variable to become defined"
              >
              > Question1: Does it have initial value at all?

              Well, yes. It always has the same value, including initially.

              > Question2: Can it appear as the LHS of an assignment?

              Yes.

              > Question3: Can it be used in an expression and have a
              defined result?

              Yes.
              >
              > Please consider the following example:
              >
              > program test
              > real, dimension(:,:), allocatable:: a
              > real, dimension(:), allocatable :: x,y
              >
              > allocate(a(7,0),x(0),y(7))
              > a(:,:) = 0.0 !Q2: Is this legal? What is the value of a?

              Yes, you are broadcasting a scalar to every element of the
              array. a has
              the value it always has when the size is zero.

              > x(:) = 0.0

              Yes, you are again broadcasting a scalar to every element of
              the array.

              > y(:) = 1.0
              > y(:) = MATMUL( a(:,:), x(:) ) !Q3: Is this legal? what is
              the value of y?

              Yes. Every element of y has the value 0.0.

              > end program test
              >
              > Non-allocatable version:

              Nothing is reallocated in the allocatable version, so this is
              the same.

              Cheers,

              John.

              > program test2
              > real :: a(7,0)
              > real :: x(0),y(7)
              >
              > a = 0.0 !Q2: is this legal? What is the value of a?
              > x = 0.0
              > y = 1.0
              > y = MATMUL( a, x ) !Q3: is this legal? what is the value of
              y?
              > end program test2
              >
              > Thanks,
              >
              > Daniel
              >
              > XL Fortran Development, Fortran Standard Representative
              > IBM Toronto Software Lab
              > Phone: 905-413-3056
              > Tie: 969-3056
              > Email: cdchen at ca.ibm.com
              > http://www.ibm.com/software/awdtools/fortran/xlfortran
              >
              >
              > _______________________________________________
              > J3 mailing list
              > J3 at mailman.j3-fortran.org
              > http://mailman.j3-fortran.org/mailman/listinfo/j3
              >
              _______________________________________________
              J3 mailing list
              J3 at mailman.j3-fortran.org
              http://mailman.j3-fortran.org/mailman/listinfo/j3


              Disclaimer


              The Numerical Algorithms Group Ltd is a company registered in
              England and Wales with company number 1249803. The registered
              office is: Wilkinson House, Jordan Hill Road, Oxford OX2 8DR,
              United Kingdom.

              This e-mail has been scanned for all viruses and malware, and
              may have been automatically archived by Mimecast Ltd, an
              innovator in Software as a Service (SaaS) for business.
              _______________________________________________
              J3 mailing list
              J3 at mailman.j3-fortran.org
              http://mailman.j3-fortran.org/mailman/listinfo/j3




              _______________________________________________
              J3 mailing list
              J3 at mailman.j3-fortran.org
              http://mailman.j3-fortran.org/mailman/listinfo/j3





        --
        Walt Brainerd_______________________________________________
        J3 mailing list
        J3 at mailman.j3-fortran.org
        http://mailman.j3-fortran.org/mailman/listinfo/j3



        _______________________________________________
        J3 mailing list
        J3 at mailman.j3-fortran.org
        http://mailman.j3-fortran.org/mailman/listinfo/j3


  _______________________________________________
  J3 mailing list
  J3 at mailman.j3-fortran.org
  http://mailman.j3-fortran.org/mailman/listinfo/j3




--
Walt Brainerd_______________________________________________
J3 mailing list
J3 at mailman.j3-fortran.org
http://mailman.j3-fortran.org/mailman/listinfo/j3


-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.j3-fortran.org/pipermail/j3/attachments/20170622/281abbaf/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: graycol.gif
Type: image/gif
Size: 105 bytes
Desc: not available
Url : http://mailman.j3-fortran.org/pipermail/j3/attachments/20170622/281abbaf/attachment-0001.gif 



More information about the J3 mailing list