Multi-Core ARM and Speed vs power thoughts

[Tony Hardie-Bick]

thx1138 wrote:
>  From my point of observation the MultiCore Coretex A9 ARM platform does
> open up some new dimensions. However, there are several problems bothering
> the semiconductor business at present. Software tools for Linux are not
> evolving fast enough to keep up with the demand on code optimization. This
> is evolving but is it fast enough in the Open Source community?

 From a contributor's point-of-view, there is a competition between fast 
progress sponsored by companies like TI and Intel, who will do their best to 
convince people to use platform-specific optimised code, in TI's case, (the DSP 
in) OMAP, in Intel's case, Moblin. Actually what these are are two highly 
specific and serious strategies by companies guarding their space by pushing 
into new ones.

For example, with Moblin, there is only one "contributor" with the bucks and 
interest in porting to ARM, and that's Nokia. They just did a deal with Intel, 
and will play nice. Intel will release the code for the OpenGL part of Moblin, 
only when they need to under terms of the GPL, which is, on the same day they 
distribute binaries. One can see all this as strategic, and understand why.

 From a contributor's point-of-view, why indulge this? Silicon is fast and 
getting faster. Your PC is unbelievably fast. Write code that's portable. Ti, 
Intel, Nokia and others will not seriously undermine their credibility, but, 
IMHO, certainly TI and Intel could do a little better, and are, very mildly, 
being disingenuous.

Code need not get stuck in one domain, and that's the key to a big open source 
tidal wave in any domain. It will be interesting to see, how such inspirations 
play out in the new world of mobile, which is the most interesting space, 
future-wise, for a developer.

Tony (HB)

(Funny talking about this stuff on music-bar...)