[Bug target/94072] [10 Regression] ICE: SIGSEGV due to infinite recursion in expand_expr/expand_expr_real_1 with -msve-vector-bits=512 since r10-4197

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94072

--- Comment #3 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:1aa22b1916a493ac46453d96e0c78ca47bcaeda3

commit r10-7301-g1aa22b1916a493ac46453d96e0c78ca47bcaeda3
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Tue Mar 17 14:43:08 2020 +0000

    c-family: Tighten vector handling in type_for_mode [PR94072]

    In this PR we had a 512-bit VECTOR_TYPE whose mode is XImode
    (an integer mode used for four 128-bit vectors).  When trying
    to expand a zero constant for it, we hit code in expand_expr_real_1
    that tries to use the associated integer type instead.  The code used
    type_for_mode (XImode, 1) to get this integer type.

    However, the c-family implementation of type_for_mode checks for
    any registered built-in type that matches the mode and has the
    right signedness.  This meant that it could return a built-in
    vector type when given an integer mode (particularly if, as here,
    the vector type isn't supported by the current subtarget and so
    TYPE_MODE != TYPE_MODE_RAW).  The expand code would then cycle
    endlessly trying to use this "new" type instead of the original
    vector type.

    2020-03-20  Richard Sandiford  <richard.sandiford@arm.com>

    gcc/c-family/
            PR middle-end/94072
            * c-common.c (c_common_type_for_mode): Before using a registered
            built-in type, check that the vectorness of the type matches
            the vectorness of the mode.

    gcc/testsuite/
            PR middle-end/94072
            * gcc.target/aarch64/pr94072.c: New test.