Abstract: Error checking and correcting (ECC) may be performed in an on-chip memory where an error is corrected by a controller and not the on-chip memory. The controller may be flagged to show that an error has occurred and where it has occurred in the memory. The controller may access ECC bits associated with the error and may fix incorrect data. The error checking may be done in parallel with read operations of the memory so as to lower latency.

Abstract: Systems and methods to manage memory on a spin transfer torque magnetoresistive random-access memory (STT-MRAM) are provided. A particular method may include determining a performance characteristic using relationship information that relates a bit error rate to at least one of a programming pulse width, a temperature, a history-based predictive performance parameter , a coding scheme, and a voltage level also associated with a memory. The performance characteristic is stored and used to manage a write operation associated with the memory.

Abstract: Systems and methods to manage memory on a spin transfer torque magnetoresistive random-access memory (STT-MRAM) are provided. A particular method may include determining a performance characteristic using relationship information that relates a bit error rate to at least one of a programming pulse width, a temperature, a history-based predictive performance parameter , a coding scheme, and a voltage level also associated with a memory. The performance characteristic is stored and used to manage a write operation associated with the memory.

Abstract: Systems and methods to manage memory on a spin transfer torque magnetoresistive random-access memory (STT-MRAM) are provided. A particular method of managing memory includes determining a temperature associated with the memory and determining a level of write queue utilization associated with the memory. A write operation may be performed based on the level of write queue utilization and the temperature.

Abstract: Systems and methods to manage memory on a spin transfer torque magnetoresistive random-access memory (STT-MRAM) are provided. A particular method of managing memory includes determining a temperature associated with the memory and determining a level of write queue utilization associated with the memory. A write operation may be performed based on the level of write queue utilization and the temperature.

Abstract: Errors on a dynamic random access memory (“DRAM”) having an error correcting decoder (“ECC”) can be detected by the ECC when reading a row of the DRAM. The ECC includes error correcting code logic. If errors are detected that cannot be corrected by the ECC logic, test control logic determines weak cell information for the row, evaluates the errors using the weak cell information, and may correct the errors. The weak cell information may include weak cell locations and failure values.

Abstract: A method, system and memory controller are provided for implementing memory devices with sub-bank architecture in a computer system. An array is divided into sub-blocks having odd bit lines and even bit lines. The sub-blocks are alternated with rows of sense amplifiers; wherein a particular row of sense amplifiers connects only to odd bit lines and a next row of sense amplifiers connects only to even bit lines. More than one word line for a sub-block is allowed to be active at the same time, where a first active word line will select memory cells connected to even bit lines and a second active word line will select memory cells connected to odd bit lines.

Abstract: According to one embodiment, a memory system includes a plurality of memory devices and a memory controller operatively coupled to the memory devices. The memory controller is configured to partition write data into a plurality of data blocks, where each data block is associated with one of the memory devices. The memory controller is further configured to generate an instance of a local error-correcting code (ECC) corresponding to each data block, and merge each data block with the corresponding instance of the local ECC to form an encoded data block for each memory device. Additionally, the memory controller is configured to write each encoded data block to the memory devices such that each memory device stores one of the data blocks with the corresponding instance of the local ECC. A global ECC and a local ECC of the global ECC can also be included in the memory system.

Abstract: This disclosure includes a method for preventing errors in a DRAM (dynamic random access memory) due to weak cells that includes determining the location of a weak cell in a DRAM row, receiving data to write to the DRAM, and encoding the data into a bit vector to be written to memory. For each weak cell location, the corresponding bit from the bit vector is equal to the reliable logic state of the weak cell and the bit vector is longer than the data.

Abstract: A method, system and memory controller are provided for implementing memory devices with sub-bank architecture in a computer system. An array is divided into sub-blocks having odd bit lines and even bit lines. The sub-blocks are alternated with rows of sense amplifiers; wherein a particular row of sense amplifiers connects only to odd bit lines and a next row of sense amplifiers connects only to even bit lines. More than one word line for a sub-block is allowed to be active at the same time, where a first active word line will select memory cells connected to even bit lines and a second active word line will select memory cells connected to odd bit lines.

Abstract: A method and apparatus for managing memory in an electronic system is described. The method includes determining a failure in an element of the memory array that is repairable by a redundant element. The method may further include using a latch to identify the redundant element. The method may also include that upon an event, using a value in the latch in an eFuse which subsequently selects the redundant element.

Abstract: A method and apparatus for managing memory in an electronic system is described. The method includes determining a failure in an element of the memory array that is repairable by a redundant element. The method may further include using a latch to identify the redundant element. The method may also include that upon an event, using a value in the latch in an eFuse which subsequently selects the redundant element.

Abstract: A processor and a method implemented by the processor to obtain computation results are described. The processor includes a unified reuse table embedded in a processor pipeline, the unified reuse table including a plurality of entries, each entry of the plurality of entries corresponding with a computation instruction or a set of computation instructions. The processor also includes a functional unit to perform a computation based on a corresponding instruction.

Abstract: A processor and a method implemented by the processor to obtain computation results are described. The processor includes a unified reuse table embedded in a processor pipeline, the unified reuse table including a plurality of entries, each entry of the plurality of entries corresponding with a computation instruction or a set of computation instructions. The processor also includes a functional unit to perform a computation based on a corresponding instruction.

Abstract: A method, system and memory controller for implementing memory hierarchy placement decisions in a memory system including direct routing of arriving data into a main memory system and selective injection of the data or computed results into a processor cache in a computer system. A memory controller, or a processing element in a memory system, selectively drives placement of data into other levels of the memory hierarchy. The decision to inject into the hierarchy can be triggered by the arrival of data from an input output (IO) device, from computation, or from a directive of an in-memory processing element.

Abstract: A method and system are provided for implementing enhanced flash storage control using reinforcement learning to provide enhanced performance metrics. A flash controller, such as a Reinforcement Learning (RL) flash controller, is coupled to a flash storage. The flash controller defines a feature set of flash parameters determined by a predefined one of a plurality of optimization metrics. The optimization metric is adapted dynamically based upon system workload and system state. The flash controller employing the feature set including at least one feature responsive to erase operations; computes a current system state responsive to the employed feature set; selects actions at each time step by sensing the computed current system state for performing an action to maximize a long term reward, and moves to another state in the system while obtaining a short-term reward for the performed action.

Abstract: This disclosure includes a method for correcting errors on a DRAM having an ECC which includes writing data to a DRAM row, reading data from the DRAM row, detecting errors in the data that cannot be corrected by the DRAM's ECC, determining erasure information for the row, evaluating the errors using the erasure information, and correcting the errors in the data.

Abstract: This disclosure includes a method for preventing errors in a DRAM (dynamic random access memory) due to weak cells that includes determining the location of a weak cell in a DRAM row, receiving data to write to the DRAM, and encoding the data into a bit vector to be written to memory. For each weak cell location, the corresponding bit from the bit vector is equal to the reliable logic state of the weak cell and the bit vector is longer than the data.

Abstract: A method for determining an optimized refresh rate involves testing a refresh rate on rows of cells, determining an error rate of the rows, evaluating the error rate of the rows; and repeating these steps for a decreased refresh rate until the error rate is greater than a constraint, at which point a slow refresh rate is set.

Abstract: A method for selection of a DRAM refresh timing in a DRAM memory system is disclosed. The method may include running a workload for a first number of refresh intervals using a first DRAM refresh timing and making a first workload throughput measurement for the first number of refresh intervals. The method may also include running the workload for a second number of refresh intervals using a second DRAM refresh timing and making a second workload throughput measurement for the second number of refresh intervals. The method may further include deciding if the first throughput measurement is greater than the second throughput measurement, and then selecting the first DRAM refresh timing as a selected DRAM refresh timing, or deciding if the second throughput measurement is greater than the first throughput measurement, then selecting the second DRAM refresh timing as the selected DRAM refresh timing.