[Bug target/101325] New: [12 Regression] arm: Wrong code with MVE vcmpeqq intrinsic

[acoplan at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101325

            Bug ID: 101325
           Summary: [12 Regression] arm: Wrong code with MVE vcmpeqq
                    intrinsic
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: acoplan at gcc dot gnu.org
  Target Milestone: ---

For the following C testcase:

#include <arm_mve.h>
mve_pred16_t foo(int8x16_t v, int8x16_t w)
{
    return vcmpeqq (v, w);
}

GCC trunk with -O2 -march=armv8.1-m.main+mve -mfloat-abi=hard gives:

foo:
        vcmp.i8  eq, q0, q1
        vmrs     r0, P0 @ movhi
        and     r0, r0, #1
        bx      lr

but the and clears all but the lowest bit of the mask, which is incorrect. GCC
11 and 10 do not appear to be affected. Instead, they generate:

foo:
        vcmp.i8 eq, q0, q1
        vmrs     r0, P0 @ movhi
        uxth    r0, r0
        bx      lr