[Bug rtl-optimization/98791] New: ICE in paradoxical_subreg_p (in ira) with SVE, LTO

[Bug rtl-optimization/98791] New: ICE in paradoxical_subreg_p (in ira) with SVE, LTO

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

            Bug ID: 98791
           Summary: ICE in paradoxical_subreg_p (in ira) with SVE, LTO
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: acoplan at gcc dot gnu.org
  Target Milestone: ---

The following fails:

$ cat test.c
extern char a[];
extern long b[];
void c(unsigned d, unsigned long long g[][13]) {
  for (int f = 0; f < 24; f += d - 45810) {
    a[f] = g[0][2];
    b[f] = 0;
  }
}
static unsigned long long e[1][13];
int main() { c(45811, e); }
$ aarch64-elf-gcc -flto -O1 -ftree-vectorize -march=armv8.2-a+sve
--param=aarch64-autovec-preference=3 test.c
during RTL pass: ira
test.c: In function 'main':
test.c:10:27: internal compiler error: in paradoxical_subreg_p, at rtl.h:3170
   10 | int main() { c(45811, e); }
      |                           ^
0xab2093 paradoxical_subreg_p(machine_mode, machine_mode)
        /home/alecop01/toolchain/src/gcc/gcc/rtl.h:3170
0xab2093 narrower_subreg_mode(machine_mode, machine_mode)
        /home/alecop01/toolchain/src/gcc/gcc/rtl.h:3200
0xab2093 update_costs_from_allocno
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:1410
0xab26c5 update_costs_from_copies
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:1480
0xab3790 assign_hard_reg
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:1971
0xabc506 pop_allocnos_from_stack
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:2707
0xabc506 color_allocnos
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:3251
0xabc506 color_pass
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:3363
0xaa5ab6 ira_traverse_loop_tree(bool, ira_loop_tree_node*, void
(*)(ira_loop_tree_node*), void (*)(ira_loop_tree_node*))
        /home/alecop01/toolchain/src/gcc/gcc/ira-build.c:1776
0xab8649 do_coloring
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:3514
0xab8649 color
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:4885
0xab8649 ira_color()
        /home/alecop01/toolchain/src/gcc/gcc/ira-color.c:5014
0xa9f999 ira
        /home/alecop01/toolchain/src/gcc/gcc/ira.c:5654
0xa9f999 execute
        /home/alecop01/toolchain/src/gcc/gcc/ira.c:5964
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <https://gcc.gnu.org/bugs/> for instructions.
lto-wrapper: fatal error: aarch64-elf-gcc returned 1 exit status
compilation terminated.
/home/alecop01/toolchain/build-aarch64-elf/install/lib/gcc/aarch64-elf/11.0.0/../../../../aarch64-elf/bin/ld:
error: lto-wrapper failed
collect2: error: ld returned 1 exit status

The backtrace appears to be the same as PR97092, although that PR doesn't
appear to have regressed (at least for the original testcase).

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE, LTO

[ktkachov at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

ktkachov at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ktkachov at gcc dot gnu.org
             Status|UNCONFIRMED                 |NEW
           Keywords|                            |ice-on-valid-code
            Summary|ICE in paradoxical_subreg_p |[11 Regression] ICE in
                   |(in ira) with SVE, LTO      |paradoxical_subreg_p (in
                   |                            |ira) with SVE, LTO
   Target Milestone|---                         |11.0
   Last reconfirmed|                            |2021-01-22
      Known to work|                            |10.2.1
     Ever confirmed|0                           |1
           Priority|P3                          |P1

--- Comment #1 from ktkachov at gcc dot gnu.org ---
Confirmed on trunk.

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE, LTO

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

--- Comment #2 from Alex Coplan <acoplan at gcc dot gnu.org> ---
Here is a non-LTO testcase that fails with -O -ftree-vectorize
-march=armv8.2-a+sve --param=aarch64-autovec-preference=3:

extern char a[], b[];
short c, d;
long *e;
void f() {
  for (int g; g < c; g += 1) {
    a[g] = d;
    b[g] = e[g];
  }
}

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[acoplan at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

--- Comment #3 from Alex Coplan <acoplan at gcc dot gnu.org> ---
And here is a testcase (using SVE intrinsics) that ICEs without the param:

#include <arm_sve.h>
extern char a[11];
extern long b[];
void f() {
  for (int d; d < 10; d++) {
    a[d] = svaddv(svptrue_b8(), svdup_u8(0));
    b[d] = 0;
  }
}

i.e. with just -O -ftree-vectorize -march=armv8.2-a+sve.

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

--- Comment #4 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Andre Simoes Dias Vieira
<avieira@gcc.gnu.org>:

https://gcc.gnu.org/g:4c31a3a6d31b6214ea774d403bf8ab7ebe1ea862

commit r11-7321-g4c31a3a6d31b6214ea774d403bf8ab7ebe1ea862
Author: Andre Vieira <andre.simoesdiasvieira@arm.com>
Date:   Mon Feb 22 13:41:46 2021 +0000

    ira: Make sure allocno copies are ordered [PR98791]

    gcc/ChangeLog:
    2021-02-22  Andre Vieira  <andre.simoesdiasvieira@arm.com>

            PR rtl-optimization/98791
            * ira-conflicts.c (process_regs_for_copy): Don't create allocno
copies
            for unordered modes.

    gcc/testsuite/ChangeLog:
    2021-02-22  Andre Vieira  <andre.simoesdiasvieira@arm.com>

            PR rtl-optimization/98791
            * gcc.target/aarch64/sve/pr98791.c: New test.

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
                 CC|                            |law at gcc dot gnu.org

--- Comment #5 from Jeffrey A. Law <law at gcc dot gnu.org> ---
Fixed by Andre's patch on the trunk.

[Bug rtl-optimization/98791] [11 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

avieira at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to work|10.2.1                      |
      Known to fail|                            |10.2.1
             Status|RESOLVED                    |REOPENED
         Resolution|FIXED                       |---

--- Comment #6 from avieira at gcc dot gnu.org ---
Hi Jeffrey,

I was leaving thos open to remind me to backport the fix to gcc-10. I see the
ticket falsely claims it works for gcc-10. Reopening for backport.

[Bug rtl-optimization/98791] [10 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[law at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

Jeffrey A. Law <law at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[11 Regression] ICE in      |[10 Regression] ICE in
                   |paradoxical_subreg_p (in    |paradoxical_subreg_p (in
                   |ira) with SVE               |ira) with SVE

--- Comment #7 from Jeffrey A. Law <law at gcc dot gnu.org> ---
In that case, make sure to update the Summary/Title so that it appears on the
right regression lists :-)

[Bug rtl-optimization/98791] [10 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

--- Comment #8 from avieira at gcc dot gnu.org ---
Aye my bad there, Thanks for the change.

[Bug rtl-optimization/98791] [10 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

--- Comment #9 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Andre Simoes Dias Vieira
<avieira@gcc.gnu.org>:

https://gcc.gnu.org/g:9b0bbe19afb2a6800e95f61cc57206e069a400d6

commit r10-9443-g9b0bbe19afb2a6800e95f61cc57206e069a400d6
Author: Andre Vieira <andre.simoesdiasvieira@arm.com>
Date:   Mon Feb 22 13:41:46 2021 +0000

    ira: Make sure allocno copies are ordered [PR98791]

    gcc/ChangeLog:
    2021-02-22  Andre Vieira  <andre.simoesdiasvieira@arm.com>

            PR rtl-optimization/98791
            * ira-conflicts.c (process_regs_for_copy): Don't create allocno
copies
            for unordered modes.

    gcc/testsuite/ChangeLog:
    2021-02-22  Andre Vieira  <andre.simoesdiasvieira@arm.com>

            PR rtl-optimization/98791
            * gcc.target/aarch64/sve/pr98791.c: New test.

    (cherry picked from commit 4c31a3a6d31b6214ea774d403bf8ab7ebe1ea862)

[Bug rtl-optimization/98791] [10 Regression] ICE in paradoxical_subreg_p (in ira) with SVE

[avieira at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98791

avieira at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|REOPENED                    |RESOLVED
         Resolution|---                         |FIXED

--- Comment #10 from avieira at gcc dot gnu.org ---
Closing now as backport is done.