Richard Stacpoole (rstacpol@bigpond.net.au) on 2/6/03 wrote:
---------------------------
>Are chip design methodologies starting to reach a point where automated design
>tools etc (I am refering specifically to the comments on the Power4+/970) are becoming
>good enough to stand up to full custom design? I am thinking of quicker turn-around
>to new techniques, ability to react to market changes faster etc.
>Is there a parallel to the old battles over writing in assembly to using higher level languages?

Yes, the analogy works in many ways. Of course it is also
breaks down in certain aspects.


>
>A second quick point. Could Mr DeMone include the compiler used to generate the
>scores in the summary table. I mainly use gcc, use free OS's so would like to reach a better comparison.

As clearly stated in the text, those scores are estimates,
not measurements. These estimates are extrapolated from
the published performance of existing, similar processor
designs. As such it would reflect the use of the best
compiler for each MPU family (e.g. TRU64 compiler for
Alpha, Sun compiler for SPARC, HP/Intel compilers for IA64
etc).