fortran - store multi dimensional arrays to be operated on along all the dimensions -

-

foreword

the fortran program i'm writing should deal 1d, 2d , 3d problems depending on ndims, can 1, 2 or 3 , read input file.

in these cases quantity/ies of interest can stored in arrays (one named phi)

  1. of rank dims (allocatable(:) or allocatable(:,:) or allocatable(:,:,:)),
  2. or in arrays of rank 3 (allocatable(:,:,:) third dimension set equal 1 in 2d or both second , third dimensions equal 1 in 1d);

both cases explained in this answer. first approach seems more elegant me, in following assume second one, simpler.

these quantities have operated on several subroutines (e.g. mysub) along ndims dimensions (along "pencils" should give graphic idea), should call like

select case (ndims)  ! 3d case case (3)   j = ...     k = ...       call mysub(phi(:,j,k))     end   end   = ...     k = ...       call mysub(phi(i,:,k))     end   end   = ...     j = ...       call mysub(phi(i,j,:))     end   end  ! 2d case case (2)   j = ...     k = ...       call mysub(phi(:,j,1))     end   end   = ...     k = ...       call mysub(phi(i,:,1))     end   end  ! 1d case case (1)   j = ...     k = ...       call mysub(phi(:,1,1))     end   end end select

actual question

can suggest me (or me to devise!) different way of store phi (maybe involving derived data types?) can collapse preceding code follows?

do id = 1, ndims   call mysub2(phi,id) end

(here mysub2 acts on mysub's place.)

so question how should store phi, can substitute first code second one?

maybe return foreword , decide follow point 1., in case easier write generic interface. be, think, way "hide" select case do. of 2 (select case/generic interface) more efficient?

are these 2 ways face problem?

perhaps have misunderstood, think answer specific question not make changes storage or declaration of phi @ all.

in original code, 3 dimensional data (differentiating rank of data rank of array used store data) processed in slices along first dimension, second, third. 2 dimensional data processed along first, second, , 1 dimensional data processed along first only.

so id going 1 number of dimensions in data, consider following implementation of mysub2:

subroutine mysub2(phi, id)   type(pink_elephant), intent(in) :: phi(:,:,:)   integer, intent(in) :: id    integer :: i, j, k    select case (id)   case (1)     j = ...       k = ...         call mysub(phi(:,j,k))       end     end    case (2)     = ...       k = ...         call mysub(phi(i,:,k))       end     end    case (3)     = ...       j = ...         call mysub(phi(i,j,:))       end     end    end select end subroutine mysub2

~~

generic interfaces can resolved @ "compile time" - specific procedure (not type bound) or binding (type bound) invoked particular call statement or function reference can determined looking @ declarations in code.

if have situation "runtime" information going affect choice of procedure, there has other executable mechanism, other or additional resolution of generic, comes play - if statement, select case, dynamic dispatch, etc, etc, etc.

asking whether generic resolution more efficient executable decision therefore not particularly meaningful - different things.


Comments

Post a Comment

Popular posts from this blog

sql - invalid in the select list because it is not contained in either an aggregate function -

-
i can't figure out rewrite sql statement avoid error below msg 8120, level 16, state 1, line 26 column 'staffmember.staffmemberid' invalid in select list because not contained in either aggregate function or group clause. sql statment select (select [dbo].[udf_readstaffmemberbasic](sm.staffmemberid) xml path ('staffmembersummary'), root ('items'), type ), count(sm.staffmemberid) totalresults, ceiling(cast(count(sm.staffmemberid) decimal)/ @pagesize) totalpages staffmember sm inner join staffmembertoadditionalrole ar on sm.staffmemberid = ar.staffmemberid (sm.firstname '%' + @name + '%' or sm.surname '%' + @name + '%') , (@floorid null or sm.floorid = @floorid) , (@directorateid null or sm.directorateid = @directorateid) , (@additionalroleid null or ar.additionalroleid = @additionalroleid) order sm.staffmemberid offset @page * @pagesize rows fe...
Read more

Angularjs unit testing - ng-disabled not working when adding text to textarea -

-
i trying test angularjs component using karma, mocha, power-assert. have textarea , button in component, button disabled based on length of text in textarea. when run component in browser works perfectly. can't figure out how test functionality. here's bit of code. inquiryform.js function inquiryform($scope) { // plan add logic here } angular.module('myapp').component('inquiryform', { controller: inquiryform, controlleras: 'inquiryform', templateurl: 'inquiryform.html' }); inquiryform.html <div class="contact"> <div>thanks contacting us.</div> <form> <textarea ng-model="inquiryform.inquirytext" name=""></textarea> <input type="submit" value="send" ng-disabled="!inquiryform.inquirytext.length"> </form> </div> inquiryformtest.js describe('inquiry form ', () => { let $r...
Read more

How to start daemon on android by adb -

-
can me please, head swollen... i trying run daemon on android emu/device command: adb -s <device_name> shell su -c /dir/daemon <port_number> but nothink happens, , no errors! if do: adb -s <device> shell and form shell cmdline: su -c /dir/daemon <port_number> than work fine. try use shell-script , run: adb -s <device_name> shell sh su -c /dir/script.sh <port_number> and try generate script qt code port number , use: adb -s <device_name> shell sh su -c /dir/script.sh but not helped... problem if enter adb shell previous run daemon - work. problem on different device/emu/iso_image different command format not work. example: adb -s <device_name> shell su -c /dir/daemon <port_number> // work @ emu, not @ iso and vice: adb -s <device_name> shell su -c "/dir/daemon <port_number>" // work @ iso etc. everybody can answer what's matter? sorry english adb shell su -c ...
Read more