JiscMail Logo
Email discussion lists for the UK Education and Research communities

Help for COMP-FORTRAN-90 Archives


COMP-FORTRAN-90 Archives

COMP-FORTRAN-90 Archives


COMP-FORTRAN-90@JISCMAIL.AC.UK


View:

Message:

[

First

|

Previous

|

Next

|

Last

]

By Topic:

[

First

|

Previous

|

Next

|

Last

]

By Author:

[

First

|

Previous

|

Next

|

Last

]

Font:

Proportional Font

LISTSERV Archives

LISTSERV Archives

COMP-FORTRAN-90 Home

COMP-FORTRAN-90 Home

COMP-FORTRAN-90  2005

COMP-FORTRAN-90 2005

Options

Subscribe or Unsubscribe

Subscribe or Unsubscribe

Log In

Log In

Get Password

Get Password

Subject:

Re: Flattening nested loops

From:

Stuart Midgley <[log in to unmask]>

Reply-To:

Fortran 90 List <[log in to unmask]>

Date:

Thu, 23 Jun 2005 10:44:06 +0800

Content-Type:

text/plain

Parts/Attachments:

Parts/Attachments

text/plain (126 lines)

This is a common piece of code that I use to copy a segment from 1  
arbitrary dimensional array to another array of the same  
dimensiononality (but may be different size)



   function offset(pos, lb, strd)
     !
     ! compute the offset into a flattened array
     !
     integer, dimension(:), intent(in) :: pos, lb, strd
     integer :: offset, i

     offset = 1

     do i = 1, ndims
        offset = offset + (pos(i)-lb(i))*strd(i)
     end do
   end function offset






   real(8), allocatable(:) :: array1, array2
   integer, allocatable(:) :: inc, lb1, ub1, lb2, ub2
   integer :: i, ii, imax, jj, ndims

   ndims=something

   allocate(array1(some large size), array2(another large size))
   allocate(inc(ndims), lb1(ndims), ub1(ndims))
   allocate(lb2(ndims), ub2(ndims))

   lb1=something
   ub1=something

   strd1(1)=1

   do i = 2, ndims
      strd1(i) = (ub1(i)-lb1(i)+1)*strd1(i-1)
   end do


   lb2=something
   ub2=something

   strd2(1)=1

   do i = 2, ndims
      strd2(i) = (ub2(i)-lb2(i)+1)*strd2(i-1)
   end do


   imax = offset(ub1, lb1, strd1)
   ii = imax
   inc = lb1

   do while (ii <= imax)
      i = offset(inc, lb1, strd1)
      jj = offset(inc, lb2, strd2)

      do ii = i, i+ub1(1)-lb1(1)
         array2(jj) = array1(ii)
         jj = jj + 1
      end do

      do k = 2, ndims
         inc(k) = inc(k) + 1
         if (inc(k) <= ub2(k)) exit
         inc(k) = lb1(k)
      end do
   end do


Notionally lb1/ub1 and lb2/ub2 are the upper and lower bounds of the  
two arrays.  So by adjusting the lb's and ub's in

   imax = offset(ub1, lb1, strd1)
   ii = imax
   inc = lb1

you can choose a different segment of the arrays.  The example I have  
provided copys ALL of array1 into array2 (though array2 may be of  
different size).

I use to have the function offset defined as a cpp macro so that it  
was inlined... but most compilers can inline a function just as well.

Things to note:  for speed, you need to try and keep your memory- 
stores contiguous - hence the array over ii/jj is contiguous.  If you  
stride over cache lines while doing a memory-store, you will actually  
induce a cache-read, cache-modify, cache-write.  Which is costly :)   
Much better just to do a cache-read, lost of cache-modifies, cache- 
write and fush a full/modified cache line.

I use modifications of this piece of code all through my MPI-AMR code  
so I can do 2 and 3D problems without having to recompile.  I use it  
in my fortran scripting language so I can have interpreted arrays of  
arbitrary size and dimensionality.  In all the performance testing I  
have done, this code snippet is as fast as hard wiring the number of  
dimensions.  Deep down the compiler has to be doing the same thing,  
as memory is a single dimensional address range.

Of course, doing this sort of thing screws with your mind...  
especially when you start getting into some hairy mathematics :)  It  
is much easier to let the compiler do it.

Stu.



--
Dr Stuart Midgley
Industry Uptake Program Leader
iVEC, 'The hub of advanced computing in Western Australia'
26 Dick Perry Avenue, Technology Park
Kensington WA 6151
Australia

Phone: +61 8 6436 8545
Fax: +61 8 6436 8555
Email: [log in to unmask]
WWW:  http://www.ivec.org

Top of Message | Previous Page | Permalink

JiscMail Tools


RSS Feeds and Sharing


Advanced Options


Archives

December 2023
February 2023
November 2022
September 2022
February 2022
January 2022
June 2021
November 2020
September 2020
June 2020
May 2020
April 2020
December 2019
October 2019
September 2019
March 2019
February 2019
January 2019
November 2018
October 2018
September 2018
August 2018
July 2018
May 2018
April 2018
March 2018
February 2018
January 2018
December 2017
November 2017
October 2017
September 2017
August 2017
July 2017
June 2017
May 2017
April 2017
March 2017
January 2017
December 2016
November 2016
October 2016
September 2016
August 2016
July 2016
June 2016
May 2016
April 2016
March 2016
February 2016
December 2015
November 2015
October 2015
September 2015
August 2015
June 2015
April 2015
March 2015
January 2015
December 2014
November 2014
October 2014
August 2014
July 2014
May 2014
April 2014
March 2014
February 2014
January 2014
December 2013
November 2013
July 2013
June 2013
May 2013
April 2013
February 2013
January 2013
December 2012
November 2012
October 2012
September 2012
August 2012
July 2012
June 2012
April 2012
March 2012
February 2012
January 2012
December 2011
November 2011
October 2011
September 2011
August 2011
July 2011
June 2011
May 2011
April 2011
March 2011
February 2011
January 2011
December 2010
November 2010
October 2010
August 2010
July 2010
June 2010
March 2010
February 2010
January 2010
December 2009
October 2009
August 2009
July 2009
June 2009
March 2009
February 2009
January 2009
December 2008
November 2008
October 2008
September 2008
August 2008
July 2008
June 2008
May 2008
April 2008
March 2008
February 2008
December 2007
November 2007
October 2007
September 2007
August 2007
July 2007
June 2007
May 2007
April 2007
March 2007
January 2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997


JiscMail is a Jisc service.

View our service policies at https://www.jiscmail.ac.uk/policyandsecurity/ and Jisc's privacy policy at https://www.jisc.ac.uk/website/privacy-notice

For help and support help@jisc.ac.uk

Secured by F-Secure Anti-Virus CataList Email List Search Powered by the LISTSERV Email List Manager