[Bug rtl-optimization/101939] New: Register spilling when zero initialization of ymm array is defered past the declaration point.

[ts.tomeksopel at gmail dot com via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101939

            Bug ID: 101939
           Summary: Register spilling when zero initialization of ymm
                    array is defered past the declaration point.
           Product: gcc
           Version: 11.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: rtl-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ts.tomeksopel at gmail dot com
  Target Milestone: ---

using

    __m256i r[8] = {_mm256_setzero_si256()};

as opposed to

    __m256i r[8];
    for (int i = 0; i < 8; ++i) r[i] = _mm256_setzero_si256();

before a hot loop results in MUCH better codegen. See godbolt below.

https://godbolt.org/z/sbscKzso3