[Bug fortran/87597] [6/7/8/9 Regression] wrong result with matmul inlining

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87597

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> ---
If I look at the dump, i.e.
   -fdump-tree-original for -O0 -ffrontend-optimize
I get the following for
    B = matmul(A,gamma5)

Re-written in Fortran:

  do var1 = 0, 3
    do var2 = 0, 3
      do var3 = 0, 3
        b(var1,var3) = b(var1,var3) + a(var2,var3) * gamma5(var1+1,var2+1)
      end do
    end do
  end do

The "+1" for gamma5 looks wrong!

From the original dump:

  b[NON_LVALUE_EXPR <__var_1_do> * 4 + NON_LVALUE_EXPR <__var_3_do>] =
b[NON_LVALUE_EXPR <__var_1_do> * 4 + NON_LVALUE_EXPR <__var_3_do>] +
a[NON_LVALUE_EXPR <__var_2_do> * 4 + NON_LVALUE_EXPR <__var_3_do>] *
gamma5[(__var_1_do + 1) * 4 + (__var_2_do + 1)];