Qualcomm Incorporated

Alain Artieri

Dexter Tamio Chun

Subbarao Palacharla

Jungwon Suh

Yanru Li

Deepti Vijayalakshmi Sriramagiri

Kedar Bhole

Moinul H Khan

Jean-Jacques Lecler

Vinod Chamarty

Raghu Sankuratri

Laurent Rene Moll

George Patsilaras

Shyamkumar Thoziyoor

Azzedine Touzni

Jean-Marie Tran

Andrew Edmund Turner

Tauseef Kazi

Girish Bhat

Lipeng Cao

Simon Peter William Booth

Samir Vijay Ginde

Rakesh Kumar Gupta

Christophe Avoinne

Vipul Gandhi

Hiral Nandu

Gopi Tummala

Younghoon Kim

Sparsh Singhai

Syed Minhaj Hassan

Jeong-Ho Woo

Pankaj Chaurasia

Carlo Spitale

Yen-Kuan Wu

Luigi Vinci

Pavan Kumar Thirunagari

Various embodiments include methods and devices for implementing a criterion aware cache replacement policy by a computing device. Embodiments may include updating a staling counter, writing a value of a local counter to a system cache in association with a location in the system cache for with data, in which the value of the local counter includes a value of the staling counter when (i.e., at the time) the associated data is written to the system cache, and using the value of the local counter of the associated data to determine whether the associated data is stale.

Multimedia compressed frame aware cache replacement policy

Data caching may include storing data associated with DRAM transaction requests in data storage structures organized in a manner corresponding to the DRAM bank, bank group and rank organization. Data may be selected for transfer to the DRAM by selecting among the data storage structures.

Dram-aware caching

Memory utilization in an SDRAM system may be improved by increasing memory bank group and memory bank interleaving. Memory bank group interleaving and memory bank interleaving may be increased by a memory controller generating a physical memory address in which the bank group address bits are positioned nearer the LSB of the physical memory address than the MSB. Alternatively, or in addition to positioning the bank group address bits in such a manner, memory bank group interleaving and memory bank interleaving may be increased by hashing the bank group address bits and bank address bits of the physical memory address with row address bits of the initial physical memory address, A rank address bit may also be involved in the hashing.

Memory bank group interleaving

Disclosed are techniques for generating a parity check matrix representing an error correcting code (ECC) for protecting a plurality of bits of a message. In an aspect, a method includes initializing a matrix M to store selected three-bit codes, selecting a first three-bit code from a set L of three-bit combinations of a number of bits n of the ECC that minimizes a sum of squared row weights of each row of the matrix M, comparing the first three-bit code with each of a plurality of error syndromes, and calculating, based on no comparison of the first three-bit code with each of the plurality of error syndromes, new error syndromes and storing the new error syndromes, wherein the new error syndromes are calculated by comparing the first three-bit code with each three-bit code in the matrix M, and storing the first three-bit code in the matrix M.

Optimized error-correcting code (ECC) for data protection

A method for dynamic memory scheduling with enhanced bank-group batching is described. The method includes determining a read-bank group-spread of each rank, as a number of bank-groups of each respective rank targeted by at least one read instruction. The method further includes determining a write-bank group-spread of each rank, as a number of bank-groups of each rank targeted by at least one write instruction. The method also includes stalling a current batch of read instructions in a rank when the read-bank group-spread of the rank is less than a predetermined value. The method further includes stalling a current batch of write instructions in a rank when the write-bank group-spread of the rank is less than the predetermined value.

Dynamic memory scheduling routine with enhanced bank-group batching

Errors can be introduced when data is transferred over a link between two entities such as between a host and a memory. Link error protection schemes can be implemented to detect and correct errors that occur on the link to enhance transmission reliability. However, these benefits are not without costs since such protection schemes increase both latency and power consumption. In one or more aspects, it is proposed to dynamically adjust the level of link error protection applied to match any change in the operating environment. For example, likelihood of link errors strongly correlates with the link speed. If the link speed is increased, a greater level of link error protection can be applied to counteract the increase in the link errors. If the link speed is decreased, the level of protection can be decreased so that latency and power consumption penalties can be minimized.

Dynamic link error protection in memory systems

An integrated circuit (IC) is disclosed herein for power management using duty cycles. In an example aspect, the integrated circuit includes multiple power domains, each of which includes a respective power state controller. The power state controller acts as a bridge between global supply lines of the integrated circuit and local supply lines of the respective power domain. Global supply lines can include a first global power rail, a second global power rail, and a global clock tree. Local supply lines can include a local power rail and a local clock tree. In operation, a power state controller adjusts a power state of the respective power domain in accordance with a duty cycle. A timeslot corresponding to the duty cycle can be separated into multiple time periods with durations of the time periods being based on the duty cycle.

Power management using duty cycles

Aspects include computing devices, systems, and methods for implementing a cache maintenance or status operation for a component cache of a system cache. A computing device may generate a component cache configuration table, assign at least one component cache indicator of a component cache to a master of the component cache, and map at least one control register to the component cache indicator by a centralized control entity. The computing device may store the component cache indicator such that the component cache indicator is accessible by the master of the component cache for discovering a virtualized view of the system cache and issuing a cache maintenance or status command for the component cache bypassing the centralized control entity. The computing device may receive the cache maintenance or status command by a control register associated with a cache maintenance or status command and the component cache bypassing the centralized control entity.

Method and apparatus for virtualized control of a shared system cache

Aspects include computing devices, systems, and methods for dynamically partitioning a system cache by sets and ways into component caches. A system cache memory controller may manage the component caches and manage access to the component caches. The system cache memory controller may receive system cache access requests and reserve locations in the system cache corresponding to the component caches correlated with component cache identifiers of the requests. Reserving locations in the system cache may activate the locations in the system cache for use by a requesting client, and may also prevent other client from using the reserved locations in the system cache. Releasing the locations in the system cache may deactivate the locations in the system cache and allow other clients to use them. A client reserving locations in the system cache may change the amount of locations it has reserved within its component cache.

Method and apparatus for a shared cache with dynamic partitioning

Systems, methods, and computer programs are disclosed for method for reducing memory subsystem power. In an exemplary method, a system resource manager provides memory performance requirements for a plurality of memory clients to a double data rate (DDR) subsystem. The DDR subsystem and the system resource manager reside on a system on chip (SoC) electrically coupled to a dynamic random access memory (DRAM). A cache hit rate is determined of each of the plurality of memory clients associated with a system cache residing on the DDR subsystem. The DDR subsystem adjusts access to the DRAM based on the memory performance requirements received from the system resource manager and the cache hit rates of the plurality of memory clients.

Power-reducing memory subsystem having a system cache and local resource management

Systems, methods, and computer programs are disclosed for method for reducing memory subsystem power. In an exemplary method, a system resource manager provides memory performance requirements for a plurality of memory clients to a double data rate (DDR) subsystem. The DDR subsystem and the system resource manager reside on a system on chip (SoC) electrically coupled to a dynamic random access memory (DRAM). A cache hit rate is determined of each of the plurality of memory clients associated with a system cache residing on the DDR subsystem. The DDR subsystem controls a DDR clock frequency based on the memory performance requirements received from the system resource manager and the cache hit rates of the plurality of memory clients.

A cache controller adaptively partitions a shared cache. The adaptive partitioning cache controller includes tag comparison and staling logic and selection logic that are responsive to client access requests and various parameters. A component cache is assigned a target occupancy which is compared to a current occupancy. A conditional identification of stale cache lines is used to manage data stored in the shared cache. When a conflict or cache miss is identified, selection logic identifies candidates for replacement preferably among cache lines identified as stale. Each cache line is assigned to a bucket with a fixed number of buckets per component cache. Allocated cache lines are assigned to a bucket as a function of the target occupancy. After a select number of buckets are filled, subsequent allocations result in the oldest cache lines being marked stale. Cache lines are deemed stale when their respective component cache active indicator is de-asserted.

System and method for a shared cache with adaptive partitioning

An integrated circuit is provided with a low-power island including embedded memory power domains that may selectively couple to either an active-mode power supply voltage supplied on a first power rail or to a sleep-mode power supply voltage supplied on a second power rail.

Growing community of inventors

San Diego, CA, United States of America

Alain Artieri