[PATCH] atomicity for mips patch

Paul Koning pkoning@equallogic.com
Wed Jul 3 11:31:00 GMT 2002


Excerpt of message (sent 3 July 2002) by Eric Christopher:
> On Wed, 2002-07-03 at 11:12, Paul Koning wrote:
> > Excerpt of message (sent 3 July 2002) by Eric Christopher:
> > > 
> > > > How about a way to enable sc/ll for other targets by a specific
> > > > switch?  That will help for embedded systems where the target is known
> > > > to be sane.
> > > 
> > > For the *-elf or a particular os?
> > 
> > I'm not sure how to answer that.  One example is an OS which supports
> > MIPS-1 platforms so it cannot in general count on ll/sc, but we're
> > running that same on a platform that does have ll/sc so it makes sense
> > to enable it through an explicit override (not based on the os name or
> > the like in the target string).
> 
> Don't suppose you have the source and would consider trapping and
> emulating the instructions? This is why I turned off atomicity by
> default. IF Phil and Benjamin want, we could add a "no really, i want
> atomic instructions please" configure flag, but...

But it's hard to push for spending that effort on emulation when I
know that I will never use it on my systems.

Jason's note precisely captures the scenario I had in mind.

	paul



More information about the Gcc-patches mailing list