Abstract: An embedded DRAM (eDRAM) having multi-use refresh cycles is described. In one embodiment, there is a multi-level cache memory system that comprises a pending write queue configured to receive pending prefetch operations from at least one of the levels of cache. A prefetch queue is configured to receive prefetch operations for at least one of the levels of cache. A refresh controller is configured to determine addresses within each level of cache that are due for a refresh. The refresh controller is configured to assert a refresh write-in signal to write data supplied from the pending write queue specified for an address due for a refresh rather than refresh existing data. The refresh controller asserts the refresh write-in signal in response to a determination that there is pending data to supply to the address specified to have the refresh.

Abstract: A design structure for an embedded DRAM (eDRAM) having multi-use refresh cycles is described. In one embodiment, there is a multi-level cache memory system that comprises a pending write queue configured to receive pending prefetch operations from at least one of the levels of cache. A prefetch queue is configured to receive prefetch operations for at least one of the levels of cache. A refresh controller is configured to determine addresses within each level of cache that are due for a refresh. The refresh controller is configured to assert a refresh write-in signal to write data supplied from the pending write queue specified for an address due for a refresh rather than refresh existing data. The refresh controller asserts the refresh write-in signal in response to a determination that there is pending data to supply to the address specified to have the refresh.

Abstract: A method for implementing dynamic refresh protocols for DRAM based cache includes partitioning a DRAM cache into a refreshable portion and a non-refreshable portion, and assigning incoming individual cache lines to one of the refreshable portion and the non-refreshable portion of the cache based on a usage history of the cache lines. Cache lines corresponding to data having a usage history below a defined frequency are assigned to the refreshable portion of the cache, and cache lines corresponding to data having a usage history at or above the defined frequency are assigned to the non-refreshable portion of the cache.

Abstract: A hardware description language (HDL) design structure embodied on a machine-readable data storage medium includes elements that when processed in a computer aided design system generates a machine executable representation of a device for implementing dynamic refresh protocols for DRAM based cache. The HDL design structure further includes a DRAM cache partitioned into a refreshable portion and a non-refreshable portion; and a cache controller configured to assign incoming individual cache lines to one of the refreshable portion and the non-refreshable portion of the cache based on a usage history of the cache lines; wherein cache lines corresponding to data having a usage history below a defined frequency are assigned by the controller to the refreshable portion of the cache, and cache lines corresponding to data having a usage history at or above the defined frequency are assigned to the non-refreshable portion of the cache.

Abstract: A method of integrating a hybrid architecture in a set associative cache having a first type of memory structure for one or more ways in each congruence class, and a second type of memory structure for the remaining ways of the congruence class, includes determining whether a memory access request results in a cache hit or a cache miss; in the event of a cache miss, determining whether LRU way of the first type memory structure is also the LRU way of the entire congruence class, and if not, then copying the contents of the LRU way of the first type memory structure into the LRU way of the entire congruence class, and filling the LRU way of the first type memory structure with a new cache line in the event of a cache miss; and updating LRU bits, depending upon the results of the memory access request.

Abstract: A memory cache comprising, a data sector having a sector ID, wherein the data sector stores a data entry, a primary directory having a primary directory entry, wherein a position of the primary directory entry is defined by a congruence class value and a way value, and a secondary directory corresponding to the data sector having a secondary directory entry corresponding to the data sector, wherein the secondary directory entry include, a primary ID field corresponding to the way value, and a sector ID field operative to identify the sector ID.

Abstract: A method for reducing the number of times in-flight loads must be searched by store instructions in a multi-threaded processor. A load issue for a thread t_old is frozen for a number of cycles. A t13 new load instruction is rejected. A notification is sent to the rest of the processor that the t_new load instruction has been rejected. A load reorder queue (LRQ) of a t_old is snooped for any load which comes from a cache line L accessed by the load instruction and then forces such loads to be re-executed. Ownership of line L is changed to thread t_new.

Abstract: A method of power management of a system of connected components includes initializing a token allocation map across the connected components, wherein each component is assigned a power budget as determined by a number of allocated tokens in the token allocation map, monitoring utilization sensor inputs and command state vector inputs, determining, at first periodic time intervals, a current performance level, a current power consumption level and an assigned power budget for the system based on the utilization sensor inputs and the command state vector inputs, and determining, at second periodic time intervals, a token re-allocation map based on the current performance level, the current power consumption level and the assigned power budget for the system, according to a re-assigned power budget of at least one of the connected components, while enforcing a power consumption limit based on a total number of allocated tokens in the system.

Abstract: A hardware based method for determining when to migrate cache lines to the cache bank closest to the requesting processor to avoid remote access penalty for future requests. In a preferred embodiment, decay counters are enhanced and used in determining the cost of retaining a line as opposed to replacing it while not losing the data. In one embodiment, a minimization of off-chip communication is sought; this may be particularly useful in a CMP environment.

Abstract: A two level branch history table (TLBHT) is substantially improved by providing a mechanism to prefetch entries from the very large second level branch history table (L2 BHT) into the active (very fast) first level branch history table (L1 BHT) before the processor uses them in the branch prediction process and at the same time prefetch cache misses into the instruction cache. The mechanism prefetches entries from the very large L2 BHT into the very fast L1 BHT before the processor uses them in the branch prediction process. A TLBHT is successful because it can prefetch branch entries into the L1 BHT sufficiently ahead of the time the entry is needed. This feature of the TLBHT is also used to prefetch instructions into the cache ahead of their use. In fact, the timeliness of the prefetches produced by the TLBHT can be used to remove most of the cycle time penalty incurred by cache misses.

Abstract: A method for accessing data in memory comprising, receiving address bits associated with a data item including a first tag, an index, and a sector ID from a requestor, associating the index with a congruence class in a primary directory, determining whether the first tag matches a second tag in a plurality of tags in the congruence class, wherein the each tag of the plurality of tags uniquely identifies a cache line associated with a primary ID in the congruence class, defining the primary ID of the second tag of the primary directory that matches the first tag, determining whether the primary ID and the sector ID are present in a secondary directory entry having a one to one correspondence with a sector in a data array, and sending the data item from the sector to the requestor.

Abstract: A system to prevent livelock. An outcome of an event is predicted to form an event outcome prediction. The event outcome prediction is compared with a correct value for a datum to be accessed. An instruction is appended with a real event outcome when the outcome of the event is mispredicted to form an appended instruction. A prediction override bit is set on the appended instruction. Then, the appended instruction is executed with the real event outcome.

Abstract: A hardware based method for determining when to migrate cache lines to the cache bank closest to the requesting processor to avoid remote access penalty for future requests. In a preferred embodiment, decay counters are enhanced and used in determining the cost of retaining a line as opposed to replacing it while not losing the data. In one embodiment, a minimization of off-chip communication is sought; this may be particularly useful in a CMP environment.

Abstract: A method for reducing the number of load instructions in the load reorder queue (LRQ) that are searched when a load instruction is executed by a processor, including dispatching the load instructions; inserting the load instructions in the LRQ in program order; clearing a load received data field; executing the load instructions; checking load reorder queue (LRQ) entries; re-executing the load instruction of the matching LRQ entry; continuing execution; getting the load data; setting the load received data field; comparing a load sequence number (LSQN) of each load instruction to a snoop_safe register contents; ANDing all the load received data bits if the LSQN is greater in magnitude to the snoop_safe; setting the snoop_safe register to the LSQN of the load instruction; searching the LRQ entry; and setting a load_peril_snoop register to the LRQ index value where the first load instruction younger to the snoop_safe was found.

Abstract: A device for controlling power parameters in a microprocessor includes a resource activation control unit for controlling the maximum power of the microprocessor and two or more resources. The resource activation control unit controls the activation of the resources such that the consumed and dissipated power of the microprocessor does not exceed a power bound which is configurable to a predetermined value below the maximum power.

Abstract: A novel trace cache design and organization to efficiently store and retrieve multi-path traces. A goal is to design a trace cache, which is capable of storing multi-path traces without significant duplication in the traces. Furthermore, the effective access latency of these traces is reduced.

Abstract: A method for reducing the number of load instructions in the load reorder queue (LRQ) that are searched when a load instruction is executed by a processor, including dispatching the load instructions; inserting the load instructions in the LRQ in program order; clearing a load received data field; executing the load instructions; checking load reorder queue (LRQ) entries; re-executing the load instruction of the matching LRQ entry; continuing execution; getting the load data; setting the load received data field; comparing a load sequence number (LSQN) of each load instruction to a snoop_safe register contents; ANDing all the load received data bits if the LSQN is greater in magnitude to the snoop_safe; setting the snoop_safe register to the LSQN of the load instruction; searching the LRQ entry; and setting a load_peril_snoop register to the LRQ index value where the first load instruction younger to the snoop_safe was found.

Abstract: A novel trace cache design and organization to efficiently store and retrieve multi-path traces. A goal is to design a trace cache, which is capable of storing multi-path traces without significant duplication in the traces. Furthermore, the effective access latency of these traces is reduced.

Abstract: A system and method of providing a cache system having a store-in policy and affording the advantages of store-in cache operation, while simultaneously providing protection against soft-errors in locally modified data, which would normally preclude the use of a store-in cache when reliability is paramount. The improved store-in cache mechanism includes a store-in L1 cache, at least one higher-level storage hierarchy; an ancillary store-only cache (ASOC) that holds most recently stored-to lines of the store-in L1 cache, and a cache controller that controls storing of data to the ancillary store-only cache (ASOC) and recovering of data from the ancillary store-only cache (ASOC) such that the data from the ancillary store-only cache (ASOC) is used only if parity errors are encountered in the store-in L1 cache.

Abstract: A method for reducing the number of load instructions in the load reorder queue (LRQ) that are searched when a load instruction is executed by a processor, including dispatching the load instructions; inserting the load instructions in the LRQ in program order; clearing a load received data field; executing the load instructions; checking load reorder queue (LRQ) entries; re-executing the load instruction of the matching LRQ entry; continuing execution; getting the load data; setting the load received data field; comparing a load sequence number (LSQN) of each load instruction to a snoop_safe register contents; ANDing all the load received data bits if the LSQN is greater in magnitude to the snoop_safe; setting the snoop_safe register to the LSQN of the load instruction; searching the LRQ entry; and setting a load_peril_snoop register to the LRQ index value where the first load instruction younger to the snoop_safe was found.