[llvm-bugs] [Bug 40968] New: AArch64 fails to select the correct addp

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=40968

            Bug ID: 40968
           Summary: AArch64 fails to select the correct addp
           Product: libraries
           Version: trunk
          Hardware: PC
                OS: Linux
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: Backend: AArch64
          Assignee: unassignedbugs at nondot.org
          Reporter: david.green at arm.com
                CC: arnaud.degrandmaison at arm.com, jpaquette at apple.com,
                    llvm-bugs at lists.llvm.org, peter.smith at linaro.org,
                    t.p.northover at gmail.com, Ties.Stuij at arm.com

This code:

typedef __attribute__((neon_vector_type(2))) float float32x2_t;
int printf(const char *, ...);
static __inline__ __attribute__((__always_inline__, __nodebug__)) float32x2_t
vcreate_f32(unsigned long __p0) {
  float32x2_t __ret;
  __ret = (float32x2_t)(__p0);
  return __ret;
}
int main() {
  float32x2_t c;
  float32x2_t d;
  float32x2_t e;
  c = vcreate_f32(0x3f69925b3e663ec8UL);
  d = vcreate_f32(0x3f4375073f409923UL);
  e = __builtin_neon_vpadd_v(d, c, 9);
  float f =  __builtin_neon_vget_lane_f32(e, 0);
  printf("%g\n", f);
}

Should print ~1.51584. It seems to currently print "8.77658e+37".


This started to fail at -O0 (with global isel) after rL355344, but I think
that's just because it no longer falls back to isel. It appears to be selecting
ADDP (i32x2), not FADDP (f32x2).

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20190305/0d2fe69a/attachment-0001.html>