Re: [PATCH][AArch64] Improve dup pattern

[Wilco Dijkstra <Wilco.Dijkstra@arm.com>]

James Greenhalgh wrote:
>
> Have you tested this in cases where an integer dup is definitely the right
> thing to do?

Yes, this still generates:

  #include <arm_neon.h>
  void f(unsigned a, unsigned b, uint32x4_t *c)
  {
    c[0] = vdupq_n_u32(a);
    c[1] = vdupq_n_u32(b);
  }

	dup	v1.4s, w0
	dup	v0.4s, w1
	str	q1, [x2]
	str	q0, [x2, 16]
	ret

The reason is that the GP to FP register move cost is typically >= 5, while
the additional cost of '?' is just 1.

> And similar cases? If these still look good, then the patch is OK - though
> I'm still very nervous about the register allocator cost model!

Well it's complex and hard to get working well... However slightly preferring one
variant works alright (unlike using '*' which results in incorrect costs).

Wilco