🚀 go-pugleaf

In article <danpop.809177909@rscernix> Dan Pop, Dan.Pop@mail.cern.ch
writes:
>>programs don't fit in cache?  If this is true then the external
>>bus is generally the bottleneck. 
>
>Not necessarily.  You can write programs where the external bus is the
>bottleneck, but this is not automatically true in the general case.
>I/O bound and cpu bound programs are at least as common as memory bound
>programs.
>

Basically what you're implying is that real-life programs could often fit
in
the internal cache?  Come on!  What about the fact that the OS is
constantly
switching processes and servicing interrupts?  What about the fact that
most
programs consist of *some* sections of high temporal locality, but
regularly
branch outside?

Yes, not *every* piece of code is automatically limited by the external
bus
bottleneck, but that is generally the case.  It only gets worse at higher
clock speeds.

For instance, the Alpha processor uses a 128-bit data bus on the 21064 and
21164 implementations.  It would not get near its current performance
levels
without that.  That gives a theoretical bandwidth limit of 1.5 GB per
second
at 300MHz.(using the maximum 100MHz data bus)  At that high clockspeed,
the
processor would be severely slowed down on cache misses if the data bus
were
64 bits.

Steve Bakke
Computer Engineering
University of Cincinnati