[Bug rtl-optimization/98037] New: ICE in dse.c:find_shift_sequence for large non-integer modes

[Bug rtl-optimization/98037] New: ICE in dse.c:find_shift_sequence for large non-integer modes

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98037

            Bug ID: 98037
           Summary: ICE in dse.c:find_shift_sequence for large non-integer
                    modes
           Product: gcc
           Version: 10.2.1
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rsandifo at gcc dot gnu.org
  Target Milestone: ---
            Target: aarch64*-*-*

Compiling the following test on AArch64 with -march=armv8.2-a+sve
-msve-vector-bits=512:

-----------------------------------------------------------------
typedef __SVInt8_t vec __attribute__((arm_sve_vector_bits(512)));
struct pair { vec v[2]; };
void use(pair &);
vec f(pair p) { vec v = p.v[1]; use(p); return v; }
-----------------------------------------------------------------

triggers:

foo.c:4:51: internal compiler error: in smallest_mode_for_size, at
stor-layout.c:356
    4 | vec f(pair p) { vec v = p.v[1]; use(p); return v; }
      |                                                   ^
... smallest_mode_for_size(poly_int<2u, unsigned long>, mode_class)
        .../gcc/stor-layout.c:356
... smallest_int_mode_for_size(poly_int<2u, unsigned long>)
        .../gcc/machmode.h:879
... find_shift_sequence
        .../gcc/dse.c:1731

because there is no integer mode big enough to represent the
access size.

[Bug rtl-optimization/98037] ICE in dse.c:find_shift_sequence for large non-integer modes

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98037

rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2020-11-27
             Status|UNCONFIRMED                 |ASSIGNED
   Target Milestone|---                         |10.3
           Assignee|unassigned at gcc dot gnu.org      |rsandifo at gcc dot gnu.org

[Bug rtl-optimization/98037] ICE in dse.c:find_shift_sequence for large non-integer modes

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98037

--- Comment #1 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Richard Sandiford <rsandifo@gcc.gnu.org>:

https://gcc.gnu.org/g:f835e9f6562dda9c8a1384be2c9d4e45c112ed8e

commit r11-5580-gf835e9f6562dda9c8a1384be2c9d4e45c112ed8e
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Mon Nov 30 17:15:47 2020 +0000

    dse: Cope with bigger-than-integer modes [PR98037]

    dse.c:find_shift_sequence tries to represent a store and load
    back as a shift right followed by a truncation.  It therefore
    needs to find an integer mode in which to do the shift right.
    The loop it uses has the form:

      FOR_EACH_MODE_FROM (new_mode_iter,
                          smallest_int_mode_for_size (GET_MODE_BITSIZE
(read_mode)))

    which implicitly assumes that read_mode has an equivalent integer mode.
    As shown in the testcase, not all modes have such an integer mode.

    This patch just makes the code start from the smallest integer mode and
    skip modes that are too small.  The loop already breaks at the first
    mode wider than word_mode.

    gcc/
            PR rtl-optimization/98037
            * dse.c (find_shift_sequence): Iterate over all integers and
            skip modes that are too small.

    gcc/testsuite/
            PR rtl-optimization/98037
            * gcc.target/aarch64/sve/acle/general/pr98037.c: New test.

[Bug rtl-optimization/98037] ICE in dse.c:find_shift_sequence for large non-integer modes

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98037

--- Comment #2 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-10 branch has been updated by Richard Sandiford
<rsandifo@gcc.gnu.org>:

https://gcc.gnu.org/g:fb9ee3e7419178e6530c09b99272022d062b50f3

commit r10-9108-gfb9ee3e7419178e6530c09b99272022d062b50f3
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Wed Dec 2 16:20:34 2020 +0000

    dse: Cope with bigger-than-integer modes [PR98037]

    dse.c:find_shift_sequence tries to represent a store and load
    back as a shift right followed by a truncation.  It therefore
    needs to find an integer mode in which to do the shift right.
    The loop it uses has the form:

      FOR_EACH_MODE_FROM (new_mode_iter,
                          smallest_int_mode_for_size (GET_MODE_BITSIZE
(read_mode)))

    which implicitly assumes that read_mode has an equivalent integer mode.
    As shown in the testcase, not all modes have such an integer mode.

    This patch just makes the code start from the smallest integer mode and
    skip modes that are too small.  The loop already breaks at the first
    mode wider than word_mode.

    gcc/
            PR rtl-optimization/98037
            * dse.c (find_shift_sequence): Iterate over all integers and
            skip modes that are too small.

    gcc/testsuite/
            PR rtl-optimization/98037
            * gcc.target/aarch64/sve/acle/general/pr98037.c: New test.

    (cherry picked from commit f835e9f6562dda9c8a1384be2c9d4e45c112ed8e)

[Bug rtl-optimization/98037] ICE in dse.c:find_shift_sequence for large non-integer modes

[rsandifo at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98037

rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> changed:

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

--- Comment #3 from rsandifo at gcc dot gnu.org <rsandifo at gcc dot gnu.org> ---
Fixed on trunk and GCC 10 branch.