llvm-dev Digest, Vol 196, Issue 39


I continue to believe M680x0 is a mistake.

I disagree. I think the M680x0 name fits LLVM's naming scheme nicely.

Your own patch series uses "[M68K]".

I think that should be "[M680x0]".

The triple name is m68k-linux-gnu etc.

Yes, but the triple name for "SystemZ" is also s390x. For "SPARC", the triplet
name is "sparc", yet the backend is called "Sparc" which is unusual either since
"SPARC" is an acronym, not a name.

Everyone calls it either m68k or just 68k.. As for GCC there are only four
instances of m680x0: 3 in install.texi and 1 in m68k.md, but the port is

Well, in gcc the backend names are for one all lower-case [1] and in the case
of PowerPC, the backend name is even completely different ("rs6000").

So please let's not use this name for LLVM; it's longer, more confusing and

It it? I think it's actually nice as it easily makes clear we're talking about the
backend itself and not the architecture.

We are still using the proper GNU arch name in the triplet, i.e. m68k-linux-gnu.

just generally clumsy in comparison to the de-facto standard that everyone knows
(even this thread's subject has "(M68k)" in the subject, acknowledging that that's
a name people recognise).

People also don't agree whether it's called "x64", "x86_64" or "amd64", do they? :wink:

m680x0 is definitely not a valid GNU target triple though; config.sub will let you
have m68010 etc instead of m68k, but not a literal x in there.

Yes, but the GNU target triple name is not necessarily matching the backend name
and the backend name is normally invisible to the user anyway.

I think "M68K" looks weird (as compared to a lower case "m68k").