MIPS: bitops: Only use ins for bit 16 or higher
authorPaul Burton <paul.burton@mips.com>
Tue, 1 Oct 2019 21:53:26 +0000 (21:53 +0000)
committerPaul Burton <paul.burton@mips.com>
Mon, 7 Oct 2019 16:42:39 +0000 (09:42 -0700)
commit3d2920cf4fd41a27730083ef395a0c49d4750474
treeb39ad6da2c40a4ee7c29892f319cf7938d5bd6d6
parentfe7cd97e68fac186492847f8eda0eff8bcfb0cbc
MIPS: bitops: Only use ins for bit 16 or higher

set_bit() can set bits 0-15 using an ori instruction, rather than
loading the value -1 into a register & then using an ins instruction.

That is, rather than the following:

  li   t0, -1
  ll   t1, 0(t2)
  ins  t1, t0, 4, 1
  sc   t1, 0(t2)

We can have the simpler:

  ll   t1, 0(t2)
  ori  t1, t1, 0x10
  sc   t1, 0(t2)

The or path already allows immediates to be used, so simply restricting
the ins path to bits that don't fit in immediates is sufficient to take
advantage of this.

Signed-off-by: Paul Burton <paul.burton@mips.com>
Cc: linux-mips@vger.kernel.org
Cc: Huacai Chen <chenhc@lemote.com>
Cc: Jiaxun Yang <jiaxun.yang@flygoat.com>
Cc: linux-kernel@vger.kernel.org
arch/mips/include/asm/bitops.h