[Bug middle-end/115101] New: [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[Bug middle-end/115101] New: [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[liuhongt at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115101

            Bug ID: 115101
           Summary: [wrong code] with -O1 -floop-nest-optimize for
                    gcc.dg/graphite/interchange-8.c
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: liuhongt at gcc dot gnu.org
  Target Milestone: ---

When I'm working on turning cunrolli, I found if cunrollis is disabled, the
testcase will fail.
It can be reproduced with trunk for -O1 -floop-nest-optimize.

[Bug tree-optimization/115101] [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115101

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rguenth at gcc dot gnu.org
   Last reconfirmed|                            |2024-05-15
             Status|UNCONFIRMED                 |NEW
          Component|middle-end                  |tree-optimization
     Ever confirmed|0                           |1

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed.

[scheduler] original ast:
for (int c0 = 0; c0 <= 3; c0 += 1)
  for (int c1 = 0; c1 <= 3; c1 += 1) {
    for (int c2 = 0; c2 <= 3; c2 += 1) {
      for (int c3 = 0; c3 <= 1; c3 += 1)
        S_3(c0, c1, c2, c3);
      S_21(c0, c1, c2);
    }
    for (int c2 = 0; c2 <= 1; c2 += 1)
      for (int c5 = 0; c5 <= 3; c5 += 1)
        S_6(c0, c1, c5, c2);
    for (int c2 = 0; c2 <= 3; c2 += 1)
      S_23(c0, c1, c2);
  }

shows something goes wrong translating the sibling loop or the original AST
shown is in fact a transformed one already.  The transformed shown shows
an extra interchange:

[scheduler] AST generated by isl:
for (int c0 = 0; c0 <= 3; c0 += 1)
  for (int c1 = 0; c1 <= 3; c1 += 1) {
    for (int c2 = 0; c2 <= 3; c2 += 1) {
      for (int c3 = 0; c3 <= 1; c3 += 1)
        S_3(c0, c1, c2, c3);
      S_21(c0, c1, c2);
    }
    for (int c4 = 0; c4 <= 3; c4 += 1)
      for (int c5 = 0; c5 <= 1; c5 += 1)
        S_6(c0, c1, c4, c5);
    for (int c2 = 0; c2 <= 3; c2 += 1)
      S_23(c0, c1, c2);
  }

but the distribution of the sibling for (i...) loop looks confused already.

[Bug tree-optimization/115101] [11/12/13/14/15 regression] [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[sjames at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115101

Sam James <sjames at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[wrong code] with -O1       |[11/12/13/14/15 regression]
                   |-floop-nest-optimize for    |[wrong code] with -O1
                   |gcc.dg/graphite/interchange |-floop-nest-optimize for
                   |-8.c                        |gcc.dg/graphite/interchange
                   |                            |-8.c
      Known to work|                            |5.4.0
      Known to fail|                            |6.1.0
                 CC|                            |sjames at gcc dot gnu.org

--- Comment #2 from Sam James <sjames at gcc dot gnu.org> ---
on godbolt, it looks like gcc-5.4 is fine, and 6.1 isn't

[Bug tree-optimization/115101] [11/12/13/14/15 regression] [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115101

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |11.5

[Bug tree-optimization/115101] [12/13/14/15 regression] [wrong code] with -O1 -floop-nest-optimize for gcc.dg/graphite/interchange-8.c

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115101

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|11.5                        |12.5

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
GCC 11 branch is being closed.