"Read table" and field symbols

Question

Hello all,
I have a (hopefully simple) question. Let's suppose I want to do the following:
loop at itab1 assigning <fs1>.
  read table itab2 assigning <fs2> with table key <fs1>-field1.
  if sy-subrc eq 0.
    move <fs2>-field1 to ls_out-field1.
  endif.
  read table itab3 assigning <fs3> with table key <fs1>-field2.
  if sy-subrc eq 0.
    move <fs3>-field1 to ls_out-field2.
  endif.
endloop.
It's also possible to do the following instead:
loop at itab1 assigning <fs1>.
  unassign: <fs2>,
            <fs3>.
  read table itab2 assigning <fs2> with table key <fs1>-field1.
  read table itab3 assigning <fs3> with table key <fs1>-field2.
  if <fs2> is assigned.
    move <fs2>-field1 to ls_out-field1.
  endif.
  if <fs3> is assigned.
    move <fs3>-field1 to ls_out-field2.
  endif.
endloop.
My question is: is it "better" to check sy-subrc after each read, and then do a move statement within that if statement, or is it better to check whether the field symbol is assigned each time I want to fill in a field and read everything at the same time?
I'm only really concerned about the case where you don't want to drop out of the loop after the read (so if the read fails you just don't fill any data in). I'm also discounting potential human error (otherwise the sy-subrc method wins hands-down).

Answer 1

Just to close this one (albeit somewhat belatedly). I asked Harry Dietz (the ABAP Performance blogging fella) what he would suggest, and he says sy-subrc all the way. Checking that an integer is zero is considerably faster than checking for the assignment, or otherwise, of a field symbol. Furthermore, whilst the field symbol method has the potential to be faster than the sy-subrc method given some compiler optimizations, unfortunately these optimizations don't exist in the ABAP compiler - they're there in the Java compiler though! Shame really. Apparently the ABAP compiler is a pretty simple beast.
Anyway - thanks Harry for clearing this one up!