[Bug target/32803] New: -Os: shorter load immediates for x86_64

["dean at arctic dot org" <gcc-bugzilla at gcc dot gnu dot org>]

gcc currently uses "mov $imm32,reg" and "mov $imm64,reg" for loading non-zero
immediates... these cost 5 and 11 bytes resp.  for space optimizations there
are shorter choices.

for immediates -128..127 the following generates 64-bit sign-extended results
in 3 bytes:

  push $imm8
  pop %rax

the following generates immediates 0..255 with only 4 bytes (so given the above
sequence it's useful for 128..255):

  xor %eax,%eax
  mov $X,%al

the following generates powers of 2 in only 7 bytes (worthwhile only for powers
of 2 above 31):

  xor %eax,%eax
  bts $N,%eax

-dean


-- 
           Summary: -Os: shorter load immediates for x86_64
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: dean at arctic dot org
 GCC build triplet: x86_64-unknown-linux-gnu
  GCC host triplet: x86_64-unknown-linux-gnu
GCC target triplet: x86_64-unknown-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32803