On Fri, 2011-12-16 at 10:09 +0100, Richard Herveille wrote:
> What would the benefit of this be?? You get one additional register,
> but it breaks the entire way the OR is setup (R0 being a '0'
> register). 
> I don't think having 1 extra register outweighs the disadvantages.

I'm not saying it is a bad thing at all. If I were designing the
architecture from scratch I would hard wire r0 to '0' since it is so
useful.

However it is an incompatible change, because future code will not be
sure to work on older machines. So for example Samsung would not be able
to use the new compiler and libraries on their current range of DTV
chips.

It is precisely because it is such a useful change that is dangerous.
Loads of code (including the compiler) will take advantage of it, and
that code will almost certainly break on old chips.

If we want OpenRISC to be credible, obsoleting all existing designs at a
stroke is not a good plan!


Jeremy

-- 
Tel:      +44 (1590) 610184
Cell:     +44 (7970) 676050
SkypeID: jeremybennett
Email:   [email protected]
Web:     www.embecosm.com

_______________________________________________
OpenRISC mailing list
[email protected]
http://lists.openrisc.net/listinfo/openrisc

Reply via email to