Chris G. Demetriou wrote:

> nisimura@itc.aist-nara.ac.jp (Toru Nishimura) writes:
> > I'm pretty impressed by what Algorithmics Ltd. provides with their product
> > "SDE-MIPS" (enhanced GNU toolchain). 
> > 
> >       http://www.algor.co.uk/algor/info/sde-benefits.html
> > 
> > Is it possible or worth pursuing to incorporate the free portion of
> > their effort into NetBSD?
> 
> which portion would that be?
> 
> The parts which are the mods to the GNU tools?  Algorithmics should be
> prodded to give the code back to the FSF, if they have not already
> done so.  However, if they don't, I'm of the opinion that we should
> be working to reduce and finally _eliminate_ divergence from the GNU
> tools from our toolchain, which means that shouldn't really pursue
> using their stuff.
> 
> I seem to recall that they use their own custom assembler -- not gas
> -- but that's just to say that that isn't free.  8-)

GAS it is:

 * Assembler: GNU gas. Optional Algorithmics' assembler is useful for fixing up
   pipeline hazards in assembler code.

I was more interested in the previous paragraph:

 * 64-bit extensions: many MIPS CPUs are 64-bit machines which are 32-bit compatible.
   64-bit pointers are rarely useful outside big Unix sytems, but using 64-bit data can
   be worthwhile. SDE-MIPS generates 64-bit instructions for ``long long'' data, while
   keeping a valuable degree of compatibility with the 32-bit calling convention.

I thought we'd looked at this before in the past on the pmax before we
split off the mips stuff into a separate directory, but can't remember
why we didn't do anything...

Simon.