@Amanieu I remembered one other reason I'd really like optional support for AT&T syntax on x86: with such support available, it wouldn't be that hard to automatically translate
llvm_asm! to the new
asm_att! available or similar (or alternatively an
asm_intel! for Intel syntax) would make it relatively easy to port. (And, for that matter, easier to port C code as well.)
(I'm also increasingly debating if the default syntax on x86 should be AT&T for consistency with other targets and then we can have an
asm_intel!, or if the default syntax on x86 should be Intel and we can have an
Would it be excessively painful to plumb through an option for intelsyntax?
@Josh Triplett I've actually addressed that in the RFC: you can implement AT&T syntax on top
asm! by using
.att_syntax and adding
% before register template placeholders.
.intel_syntax noprefix still allows prefixes. Would you consider just unconditionally using the prefixes in substitution?
.intel_syntax noprefix does not actually allow prefixes. I tested this.
I've written code relying on that in assembly. Perhaps LLVM's assembler just doesn't support it?
Maybe you only used
.intel_syntax instead of