DYNAMICALLY DISABLING SPECULATIVE PREFETCH WHEN HIGH PRIORITY DEMAND FETCH OPPORTUNITY USE IS HIGH

A method of limiting, in a digital processor, low-priority utilization of a resource in favor of high-priority utilization of the resource comprises determining a value predictive of high-priority utilization of the resource. Low-priority utilization of the resource is inhibited if the determined predictive value is greater than a threshold. On the other hand, if the predictive value is less than or equal to the threshold, then low-priority utilization of the resource is allowed. In a preferred embodiment, the predictive value is derived by counting the number of actual high-priority utilizations of the resource out of the last N opportunities in which the resource could have been utilized for a high-priority need. Preferably, recent utilizations are given more weight than others. In a preferred embodiment, the resource comprises one of main memory, instruction cache memory, or data cache memory. High-priority utilization comprises high-priority access to memory and low-priority utilization comprises memory prefetching. Low priority utilizations are throttled when the threshold is exceeded. Counting and comparing are done with a shift register, counting logic, and a comparator; or with a counter and a comparator; or with two shift registers in which case no comparator is needed.