andrew@frip.WV.TEK.COM (Andrew Klossner) (04/10/91)
[]
"I feel the 88k has a slight edge by not having the internals
of the pipeline visible to the compiler."
A decent 88k compiler will still keep its eye on pipeline internals or
the performance of its generated code will be poor. The obvious
example is a load;store with no intervening instructions, which wastes
at best two cycles and at worst a dozen or more.
The 88k scoreboard hold system just allowed 88k developers to get
started with stupid compilers while awaiting good compilers. And boy,
have I seen some stupid compilers ...
-=- Andrew Klossner (uunet!tektronix!frip.WV.TEK!andrew) [UUCP]
(andrew%frip.wv.tek.com@relay.cs.net) [ARPA]