> But his/her definition of "partial execution" only considered state changes to the programmer visible CPU architecture state (i.e., the user level register se

> But his/her definition of "partial execution" only considered state changes to the programmer visible CPU architecture state (i.e., the user level register set and the flags register). Their definition ignored the cache, because at the time the cache was considered simply a transparent optimization system that did not effect the values of contents of the CPU architectural state. You're putting words in other people's mouths here.