[0/5] x86: (mainly) limit suffix emission for stack accessing and branch insns

Message ID c15c2759-ed4b-1e02-0b71-51467737a6b1@suse.com
Headers show
Series
  • x86: (mainly) limit suffix emission for stack accessing and branch insns
Related show

Message

Jan Beulich July 14, 2020, 10:11 a.m.
Forever since the introduction of x86-64 support I was puzzled by the
emission of e.g. "retq" or "pushq" when the assembler accepts suffix-
less (and hence easier readable) mnemonics just fine. Mirror assembler
behavior to the disassembler, in that suffixes only get emitted for
otherwise ambiguous cases, but without considering odd things like 16-
bit pushes or branches as rendering the default 64-bit forms ambiguous.
(The 16-bit forms are still distinguishable, as for them the suffixes
remain.) Note that this also extends to 32- and 16-bit modes.

The last patch is just cleanup that I didn't want to re-base ahead of
this series.

1: x86: avoid attaching suffixes to unambiguous insns
2: x86-64: adjust stack insn test case
3: x86: make PUSH/POP disassembly uniform
4: x86-64: Intel64 adjustments for conditional jumps
5: x86: move putop() case labels to restore alphabetic sorting

Jan