This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Question about insn lengths


DJ Delorie <dj@redhat.com> writes:

> If you don't mind changing the assembler, you can add a pseudo-op
> that says "this is the end of a loop, pad to 6 words if needed"
> 
> loop_top:
> 	...
> 	.loopend loop_top
> 	...
> 
> Perhaps a generic "optional padding here" pseudo would be more useful.

Well, it doesn't seem to be difficult to fix the situation in the
assembler. In fact, the binutils I have (patched for r5900, naturally)
currently emits a warning if the loop is too short.

However, I've seen other places where CPU errata is handled by gcc,
which made me think there is a policy that gcc should always generate
correct assembly for the target processor. You wouldn't know if it is
so? Inserting the nops in the assembler might be easier, even.

-- 
Pontus Lidman, pontus@lysator.liu.se, Software Engineer
No matter how cynical you get, it's impossible to keep up.
Scene: www.dc-s.com | MUD: tyme.envy.com 6969 | irc: irc.quakenet.eu.org


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]