Re: __atomic_test_and_set() and mips o32 - help wanted

On Wed, 19 Nov 2025, Jason Thorpe wrote:

> > Conversely the compiler is happy to inline LL/SC sequences as they map 
> > directly to RTL and at worst you need to maintain emulation.  It seems 
> > like a good compromise.
> 
> I’m not disagreeing that inline ll/sc is more efficient on systems that 
> support it.  But if the compiler is targeted at a MIPS-I target, this is 
> generally considered an acceptable compromise.  Sure, if you build with 
> -mmips3 or whatever the flag is, inline it.  But if -mmips1, a libcall 
> doesn’t seem unreasonable.

 Nothing wrong with it in principle, especially if you do want to optimise 
for MIPS I.  Traps are cheap on MIPS however, and even more so with MIPS I 
and its simple four-stage execution pipeline, so it might be the case of 
perfect being the enemy of good enough.  There is a balance to be found 
here.

  Maciej