[Bug other/82627] New: [8 regression] r253809 breaks test cases gcc.dg/graphite/interchange-3.c -7.c, -9.c, and uns-interchange-9.c

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82627

            Bug ID: 82627
           Summary: [8 regression] r253809 breaks test cases
                    gcc.dg/graphite/interchange-3.c -7.c, -9.c, and
                    uns-interchange-9.c
           Product: gcc
           Version: 8.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: other
          Assignee: unassigned at gcc dot gnu.org
          Reporter: seurer at gcc dot gnu.org
  Target Milestone: ---

gcc.dg/graphite/interchange-3.c
gcc.dg/graphite/interchange-7.c
gcc.dg/graphite/interchange-9.c
gcc.dg/graphite/uns-interchange-9.c

The above test cases started failing with r253809 on powerpc64 be.  They run
fine on le.

They all fail the same way:

PASS: gcc.dg/graphite/interchange-9.c execution test
FAIL: gcc.dg/graphite/interchange-9.c scan-tree-dump graphite "tiled"

I compared the r253808 and 9 dumps and in the r253808 one there's a "not tiled"
output line missing in the other.

. . .
data dependences (
{ S_3[i1, 110] -> S_4[i1] : 0 <= i1 <= 110; S_9[110] -> S_6[]; S_4[i1] ->
S_9[i1] : 0 <= i1 <= 110; S_9[i1] -> S_3[1 + i1, 0] : 0 <= i1 <= 109; S_3[i1,
110] -> S_5[1 + i1] : 0 <= i1 <= 109; S_9[i1] -> S_5[1 + i1] : 0 <= i1 <= 109;
S_9[i1] -> S_9[1 + i1] : 0 <= i1 <= 109; S_5[i1] -> S_9[i1] : 0 <= i1 <= 110;
S_3[i1, i2] -> S_3[i1, 1 + i2] : 0 <= i1 <= 110 and 0 <= i2 <= 109; S_3[i1,
110] -> S_3[1 + i1, 0] : 0 <= i1 <= 109; S_11[] -> S_5[0]; S_4[i1] -> S_3[1 +
i1, 0] : 0 <= i1 <= 109; S_5[i1] -> S_3[i1, 0] : 0 <= i1 <= 110 }
)
not tiled
[scheduler] original schedule:
domain: "{ S_5[i1] : 0 <= i1 <= 110; S_9[i1] : 0 <= i1 <= 110; S_11[]; S_3[i1,
i2] : 0 <= i1 <= 110 and 0 <= i2 <= 110; S_6[]; S_4[i1] : 0 <= i1 <= 110 }"
. . .