[Bug target/109072] New: [12/13 Regression] SLP costs for vec duplicate too high since g:4963079769c99c4073adfd799885410ad484cbbe

["tnfchris at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109072

            Bug ID: 109072
           Summary: [12/13 Regression] SLP costs for vec duplicate too
                    high since g:4963079769c99c4073adfd799885410ad484cbbe
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: tnfchris at gcc dot gnu.org
                CC: rsandifo at gcc dot gnu.org
  Target Milestone: ---
            Target: aarch64*

The following example

---

#include <arm_neon.h>

float32x4_t f (float32x4_t v, float res)
{
  float data[4];
  data[0] = res;
  data[1] = res;
  data[2] = res;
  data[3] = res;
  return vld1q_f32 (&data[0]);
}

---

compiled with -Ofast fails to SLP starting with GCC 12.

This used to generate:

f:
        dup     v0.4s, v1.s[0]
        ret

and now generates:

f:
        fmov    w5, s1
        fmov    w1, s1
        fmov    w4, s1
        fmov    w0, s1
        mov     x2, 0
        mov     x3, 0
        bfi     x2, x5, 0, 32
        bfi     x3, x1, 0, 32
        bfi     x2, x4, 32, 32
        bfi     x3, x0, 32, 32
        fmov    d0, x2
        ins     v0.d[1], x3
        ret

The SLP costs went from:

  Vector cost: 2
  Scalar cost: 4

to:

  Vector cost: 12
  Scalar cost: 4

it looks like it's no longer costing it as a duplicate but instead 4 vec
inserts.

bisected to:

commit g:4963079769c99c4073adfd799885410ad484cbbe
Author: Richard Sandiford <richard.sandiford@arm.com>
Date:   Tue Feb 15 18:09:33 2022 +0000

    vect+aarch64: Fix ldp_stp_* regressions

    ldp_stp_1.c, ldp_stp_4.c and ldp_stp_5.c have been failing since
    vectorisation was enabled at -O2.  In all three cases SLP is
    generating vector code when scalar code would be better.

....