Abstract: Described herein are apparatus, system, and method for providing clock signal on demand. The method comprises determining an indication of clock signal usage in multiple hardware logic units; generating an enable signal according to the indication; and gating or un-gating the clock signal for clock islands of at least a hardware logic unit, of the multiple hardware logic units, in response to a logic level of the enable signal, wherein the clock islands are part of a global clock distribution network and are operable to be enabled or disabled independently.

Abstract: Described herein are method, apparatus, and system for reducing leakage power consumption. The method comprises determining an input vector for input to a logic unit, the input vector for generating a least leakage power dissipation in the logic unit; and applying the input vector to the logic unit when a clock signal associated with the logic unit is gated. The method results in reduced leakage power consumption for the logic unit when the logic unit is not active with performing its normal operation, i.e. when the logic unit is idle.

Abstract: Methods and apparatus relating to multi-level cache prefetch are described. In some embodiments, a data parking logic updates a prefetch request with one or more bits based on the status of a request queue. The one or more bits may in turn cause the corresponding prefetched data to be stored in one of at least two caches. Other embodiments are also described and claimed.

Abstract: A method is described that includes reading a cache tag and the cache tag's corresponding ECC from storage circuitry of a cache. The method also includes generating an ECC for a search tag. The method also includes calculating a hamming distance between a) the cache tag and its corresponding ECC and b) the search tag and its corresponding ECC. The method also includes determining if the cache tag matches the search tag by comparing the hamming distance against a threshold.

Abstract: Methods and apparatus relating to multi-level cache prefetch are described. In some embodiments, a data parking logic updates a prefetch request with one or more bits based on the status of a request queue. The one or more bits may in turn cause the corresponding prefetched data to be stored in one of at least two caches. Other embodiments are also described and claimed.

Abstract: Briefly, a method, main processing unit and a computer system to report a failure in a bit of a memory line by updating first and second counters for a first time and a second time correction of bit failures in a line, respectively. The updating of first and second counters after a third time correction of bit failure in the line is disabled.

Abstract: Briefly, a method, main processing unit and a computer system to report a failure in a bit of a memory line by updating first and second counters for a first time and a second time correction of bit failures in a line, respectively. The updating of first and second counters after a third time correction of bit failure in the line is disabled.