The paper talks about a lot of optimizations for describing
machine description efficiently in terms of storage and access time.
As
an extension one of the optimizations can be applied to store machine
description efficiently for processors with multiple symmetric clusters.
If there are two clusters implementing ADD instructions, we need to
store
resource constraint for only one and then introduce another OR level
in
the AND-OR tree which expresses the availability of many ADD resources
available. Also this should be the first entry after sorting as if
there
are no ADD resources available, then you dont have to check anyother
OR tree.
Note that the last one is true for a single cluster processor also.
If we
add another OR level (a bit) to indicate the resource itself, then
the
check eliminates unnecessary checks.
Subramanian Rajagopalan