On Saturday 26 May 2007 12:05, dwight elvey wrote:
I think most software people would agree but that hard
fact is that
main memory is way behind processor speeds. This means that the
more compact the instruction stream is, the faster the processor can
run. We are quickly reaching the limits of clever caching, even for
CISC processors like the X86 machines.
I'm not saying that RISC machines are dead, just the opposite. Every
X86 machine today has a RISC engine inside.
I expect that the next generation processor will be even more
CISC like and less RISC like to the external software.
Memory bandwidth will dominate choices.
Just my thoughts
Dwight
This is one reason why POWER/PowerPC has "code compression", which is
lets the runtime be compressed using a huffman-based compression scheme,
which provides a speed-up similar to (or better than) CISC
architectures.
Pat
--
Purdue University ITAP/RCAC ---
http://www.rcac.purdue.edu/
The Computer Refuge ---
http://computer-refuge.org