[Bug fortran/45159] Unneccessary temporaries

["burnus at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org>]

------- Comment #1 from burnus at gcc dot gnu dot org  2010-08-01 16:52 -------
! --- FOUR ----  (also occurs for Octopus)

      subroutine t2(b,c)
        implicit none
        REAL :: b(3,3),c(3)
        c = matmul(b, c)
      end subroutine

That's a pattern I see quite often: LHS variable on the RHS as second (or
first) argument to MATMUL. However, no temporary is needed - but one probably
needs to handle this carefully and it depends how one generates the MATMUL
internally. While most implementation ways should work, one probably can
program it such that avoiding a temporary will lead to wrong results ...
Additionally, one should confirm that it works with-fexternal-blas

As a variant: The same but with one or two items TRANSPOSEd - or CONJugated.


! --- FIVE --- (Found in octopus, www.tddft.org [GPL])

  if(any(shape(a).ne.shape(b))) then

That generates two temporaries - one for each SHAPE.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45159