Re: [PATCH 1/3] RISC-V: prefer SLT{,U} aliases for SLTI{,U}

["Maciej W. Rozycki" <macro@orcam.me.uk>]

On Thu, 26 Jan 2023, Jan Beulich wrote:

> This, I'm afraid, is again a biased view. ISA and assembly syntax aren't
> necessarily connected. See the (odd in my personal view) AT&T syntax which
> has been the favorite one in the Unix world for x86, yet is entirely
> disconnected from the ISA specifications coming from the relevant x86
> vendors (all using so called Intel syntax, which really dates back to
> MASM / TASM).

 I don't think this example is really good here.  AFAIK the AT&T syntax 
predates what Microsoft invented with MASM and said company is quite known 
for not respecting established standards, whether formal or de facto.

 NB I find the MASM dialect odd and excessively elaborate with its "WORD 
PTR", etc. constructs even though I knew it first (back in ~1991).  Most 
assembly language dialects I came across are actually closer to x86 AT&T 
than to MASM.

> [...] Of course an architecture is free to define a "preferred"
> assembly language, but I guess that's still distinct from the ISA spec,
> and I'm unaware of a specific assembly syntax spec for RISC-V (which
> Andrew's reply looks to support).

 However I think syntax shouldn't be changed in an ad-hoc manner, because 
we want consistency and existing software to continue working.  We've had 
RISC-V support 6 years now in upstream binutils, so one can't really say 
it's still in its infancy anymore.

 FWIW,

  Maciej