[Bug tree-optimization/61743] [5 Regression] Complete unroll is not happened for loops with short upper bound

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61743

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #15 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The testcases fail on x86_64 with -m32:
grep 'loop with . iterations completely unrolled' pr61743-1.c.128t.cunroll 
pr61743-1.c:25:5: note: loop with 8 iterations completely unrolled
pr61743-1.c:16:5: note: loop with 8 iterations completely unrolled
pr61743-1.c:16:5: note: loop with 3 iterations completely unrolled
pr61743-1.c:29:5: note: loop with 2 iterations completely unrolled
pr61743-1.c:24:3: note: loop with 4 iterations completely unrolled
grep 'loop with . iterations completely unrolled' pr61743-2.c.128t.cunroll 
pr61743-2.c:25:5: note: loop with 8 iterations completely unrolled
pr61743-2.c:16:5: note: loop with 8 iterations completely unrolled
pr61743-2.c:16:5: note: loop with 3 iterations completely unrolled
pr61743-2.c:29:5: note: loop with 2 iterations completely unrolled
pr61743-2.c:24:3: note: loop with 4 iterations completely unrolled
(latest trunk).  With -m64 they work fine.