https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66158
Bug ID: 66158
Summary: Use DO loop constraints for optimizing bounds checking
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: enhancement
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: tkoenig at gcc dot gnu.org
Target Milestone: ---
For the code
real, dimension(10) :: a
...
do i=1,10
a(i) = a(i) + 3.
end do
we can prove that there is no possible bounds checking violation,
and the compiler should not insert code to do so. It is more tricky with
do i=1,10
a(i) = a(i) + 3.
call foo(i)
end do
Sure, it is illegal to change i, but the whole point of bounds checking
is to catch buggy programs, so we should not eliminate the checks there.
Also, the code we generate ourselves, such as for inline matmul, we
can easily prove that no bounds check occurs.
If somebody does strange things like equivalencing a DO loop variable,
we should also be prepared for the worst.
