On Wed, 6 May 2015 01:07:29 -0400, Robert A. Rosenberg wrote: >Aside from the "CPU-determined" number of bytes halt there is the >issue of page faults with the target (since you just read the source >I will assume it is still all in memory). I think this will be a >bigger factor than the need to test-cc/branch/reissue-the-move.
True. So would cache misses. In fact, I could imagine a CPU that would recognize a cache miss, initiate the fetch from memory and halt the instruction. In that case, the "CPU-determined" number of bytes would not necessarily be a fixed number. -- Tom Marchant
