fu@uicsrd.csrd.uiuc.edu (07/21/88)
Starting multiple operations and killing off the unwanted ones, when the
conditions are available, seems to be a typical trick in hardware
design. For instance, start computing one or more results in parallel
and kill off the unwanted ones when you have the conditions available.
For the cache, you would start using the data as soon as possible and
start constructing the memory access. When the hit/miss has been
resolved, kill off the appropiate operation. Clearly you need to do this
before something irreversible happens e.g. writing to a register, sending
the request etc. The VAX 88XX cache and TB kind-off do this.
******************************************************************************
John W. C. Fu
University of Illinois at Urbana-Champaign
UUCP: {ihnp4,seismo,pur-ee,convex}!uiucdcs!uicsrd!fu
ARPANET: fu%sneezy.csg.uiuc.edu or fu%uicsrd@a.cs.uiuc.edu
CSNET: fu%uicsrd@uiuc.csnet
BITNET: fu@uicsrd.csrd.uiuc.edu
*******************************************************************************