Abstract: Techniques for selecting one of a plurality of heterogeneous memory units for placement of blocks of data (e.g., memory pages), based on both reliability and performance, are disclosed. A “cost” for each data block/memory unit combination is determined, based on the frequency of access of the data block, the latency of the memory unit, and, optionally, an architectural vulnerability factor (which represents the level of exposure of a particular memory data value to memory faults such as bit flips). A memory unit is selected for the data block for which the determined cost is the lowest, out of all memory units considered, and the data block is placed into that memory unit.

Abstract: Some embodiments of a processing device include one or more power supply monitors to provide one or more counts representative of one or more operating frequencies of one or more circuit blocks based on a voltage supplied to the circuit block(s). Some embodiments of the processing device also include a system management unit to determine an initial voltage supplied to the circuit block(s) based on a target count and to reduce the voltage supplied to the circuit block(s) from the initial voltage in response to the count(s) generated by the power supply monitor(s) exceeding the target count.

Abstract: A method and apparatus for managing power in a thermal couple aware system includes determining a candidate configuration mapping based upon one or more criteria, the candidate configuration mapping being a mapping of performance for a candidate configuration of processor sockets in the thermal couple aware system. The candidate configuration mapping is evaluated by comparing the candidate configuration mapping to a stored configuration. If the evaluated candidate configuration mapping provides a better metric than the stored configuration, the stored configuration is updated with the evaluated candidate configuration mapping, and programming instructions are executed in accordance with the candidate configuration mapping if no other configuration mappings are to be determined.

Abstract: Systems, apparatuses, and methods for moving the interconnect protocol configuration registers into the main memory space of a node. The region of memory used for storing the interconnect protocol configuration registers may also be made cacheable to reduce the latency of accesses to the interconnect protocol configuration registers. Interconnect protocol configuration registers which are used during a startup routine may be prefetched into the host's cache to make the startup routine more efficient. The interconnect protocol configuration registers for various interconnect protocols may include one or more of device capability tables, memory-side statistics (e.g., to support two-level memory data mapping decisions), advanced memory and interconnect features such as repair resources and routing tables, prefetching hints, error correcting code (ECC) bits, lists of device capabilities, set and store base address, capability, device ID, status, configuration, capabilities, and other settings.

Abstract: The described embodiments comprise a selection mechanism that selects a resource from a set of resources in a computing device for performing an operation. In some embodiments, the selection mechanism performs a lookup in a table selected from a set of tables to identify a resource from the set of resources. When the resource is not available for performing the operation and until another resource is selected for performing the operation, the selection mechanism identifies a next resource in the table and selects the next resource for performing the operation when the next resource is available for performing the operation.

Abstract: A method, a system and a computer-readable medium for writing to a cache memory are provided. The method comprises maintaining a write count associated with a set, the set containing a memory block associated with a physical block address. A mapping from a logical address to the physical address of the block is also maintained. The method shifts the mapping based on the value of the write count and writes data to the block based on the mapping.

Abstract: A method, system, and computer program product synchronize a group of workitems executing an instruction stream on a processor. The processor is yielded by a first workitem responsive to a synchronization instruction in the instruction stream. A first one of a plurality of program counters is updated to point to a next instruction following the synchronization instruction in the instruction stream to be executed by the first workitem. A second workitem is run on the processor after the yielding.

Abstract: The arrival time of an asynchronous signal from an asynchronous domain at a synchronizer circuit of a synchronous domain is modified by injecting synchronous domain timing into an additional last stage of the asynchronous logic function generating the asynchronous signal. That reduces the probability of metastability by increasing the probability that the asynchronous signal will arrive at the synchronizer at a time that can guarantee the setup time for the flip-flop(s) of the synchronizer.

Abstract: A system for saving the architectural state of a processor is described. The system performs a save state operation, which involves, for each sector in a set of sectors of the architectural state, determining whether the architectural state for the sector has already been saved to a memory, and saving the architectural state for the sector to the memory when the architectural state for the sector has not already been saved to the memory. Each sector in the set of sectors comprises a different and separate portion of the architectural state of the processor. The system determines whether the architectural state for a given sector has already been saved to the memory by checking a needs-rinsing flag for the given sector. The needs-rinsing flag for the given sector is asserted upon modifying the given sector and cleared following the save state operation.

Abstract: A system and method for efficient cache data access in a large row-based memory of a computing system. A computing system includes a processing unit and an integrated three-dimensional (3D) dynamic random access memory (DRAM). The processing unit uses the 3D DRAM as a cache. Each row of the multiple rows in the memory array banks of the 3D DRAM stores at least multiple cache tags and multiple corresponding cache lines indicated by the multiple cache tags. In response to receiving a memory request from the processing unit, the 3D DRAM performs a memory access according to the received memory request on a given cache line indicated by a cache tag within the received memory request. Rather than utilizing multiple DRAM transactions, a single, complex DRAM transaction may be used to reduce latency and power consumption.

Abstract: A communication device includes a data source that generates data for transmission over a bus, and a data encoder that receives and encodes outgoing data. An encoder system receives outgoing data from a data source and stores the outgoing data in a first queue. An encoder encodes outgoing data with a header type that is based upon a header type indication from a controller and stores the encoded data that may be a packet or a data word with at least one layered header in a second queue for transmission. The device is configured to receive at a payload extractor, a packet protocol change command from the controller and to remove the encoded data and to re-encode the data to create a re-encoded data packet and placing the re-encoded data packet in the second queue for transmission.

Abstract: A processor prunes state information based on information provided by software, thereby reducing the amount of state information to be stored prior to the processor entering a low-power state. The software, such as an operating system or application program executing at the processor, indicates one or more registers of the processor as storing data that is no longer useful. When preparing to enter the low-power state, the processor omits the indicated registers from the state information stored to memory.

Abstract: A method of enhancing performance of an application executing in a parallel processor and a system for executing the method are disclosed. A block size for input to the application is determined. Input is partitioned into blocks having the block size. Input within each block is sorted. The application is executed with the sorted input.

Abstract: A method, a device, and a non-transitory computer readable medium for performing memory management in a graphics processing unit are presented. Hints about the memory usage of an application are provided to a page manager. At least one runtime memory usage pattern of the application is sent to the page manager. Data is swapped into and out of a memory by analyzing the hints and the at least one runtime memory usage pattern.

Abstract: Embodiments of a peripheral component are described herein. Embodiments provide alternatives to the use of an external bridge integrated circuit (IC) architecture. For example, an embodiment multiplexes a peripheral bus such that multiple processors in one peripheral component can use one peripheral interface slot without requiring an external bridge IC. Embodiments are usable with known bus protocols.

Abstract: A method of way prediction for a data cache having a plurality of ways is provided. Responsive to an instruction to access a stack data block, the method accesses identifying information associated with a plurality of most recently accessed ways of a data cache to determine whether the stack data block resides in one of the plurality of most recently accessed ways of the data cache, wherein the identifying information is accessed from a subset of an array of identifying information corresponding to the plurality of most recently accessed ways; and when the stack data block resides in one of the plurality of most recently accessed ways of the data cache, the method accesses the stack data block from the data cache.

Abstract: In response to determining an operation is a dependent operation, a mapper of a processor determines the source registers of the operation from which the dependent operation depends. The mapper translates the dependent operation to a new operation that uses as its source operands at least one of the determined source registers and a source register of the dependent operation. The new operation is independent of other pending operations and therefore can be executed without waiting for execution of other operations, thus reducing execution latency.

Abstract: A compute server accomplishes physical address to virtual address translation to optimize physical storage capacity via thin provisioning techniques. The thin provisioning techniques can minimize disk seeks during command functions by utilizing a translation table and free list stored to both one or more physical storage devices as well as to a cache. The cached translation table and free list can be updated directly in response to disk write procedures. A read-only copy of the cached translation table and free list can be created and stored to physical storage device for use in building the cached translation table and free list upon a boot of the compute server. The copy may also be used to repair the cached translation table in the event of a power failure or other event affecting the cache.

Abstract: Described is a method and system to efficiently compress and stream texture-space rendered content that enables low latency wireless virtual reality applications. In particular, camera motion, object motion/deformation, and shading information are decoupled and each type of information is then compressed as needed and streamed separately, while taking into account its tolerance to delays.