114 results about "Computational RAM" patented technology

A method and apparatus for debugging of OS kernel and applications software that does not require use of a hardware probe; can debug both user-mode programs and a significant body of the OS kernel code; allows the OS to continue servicing exceptions while debugging; leverages OS built-in device drivers for communicating devices to communicate with the host debugger; and can debug a production version of the OS kernel. When debugging is required, the running OS kernel dynamically loads a software-based debug agent on demand whereby such debug agent dynamically modifies the running production OS kernel code and data to intercept debugging traps and provide run-control. To provide debugging of loadable module, the debug agent implement techniques to intercept the OS module loading system call; set breakpoints in the loaded module initialization function; calculate the start address of the debugged module in memory; and asynchronously put the system under debug. By structuring command loop to execute in non-exception mode, and devising a process to transfer execution from the debug agent exception handler to the debug agent command loop and back, the debug agent can communicate with the host debugger using interrupt-driven input/output devices as well as allowing the system to service interrupts while under debug.

A digital computer performs read-modify-write (RMW) processing on each bit of a row of memory in parallel, in one operation cycle, comprising: (a) addressing a memory, (b) reading each bit of a row of data from the memory in parallel, (c) performing the same computational operation on each bit of the data in parallel, using an arithmetic logic unit (ALU) in a dedicated processing element, and (d) writing the result of the operation back into the original memory location for each bit in the row.

A compute memory system can include a memory array and a controller that generates N-ary weighted (e.g., binary weighted) access pulses for a set of word lines during a single read operation. This multi-row read generates a charge on a bit line representing a word stored in a column of the memory array. The compute memory system further includes an embedded analog signal processor stage through which voltages from bit lines can be processed in the analog domain. Data is written into the memory array in a manner that stores words in columns instead of the traditional row configuration.

One embodiment of the present invention sets forth a technique providing an optimized way to allocate and access memory across a plurality of thread/data lanes. Specifically, the device driver receives an instruction targeted to a memory set up as an array of structures of arrays. The device driver computes an address within the memory using information about the number of thread/data lanes and parameters from the instruction itself. The result is a memory allocation and access approach where the device driver properly computes the target address in the memory. Advantageously, processing efficiency is improved where memory in a parallel processing subsystem is internally stored and accessed as an array of structures of arrays, proportional to the SIMT/SIMD group width (the number of threads or lanes per execution group).

The invention discloses a multi-resolution precondition method for analyzing aerial radiation and electromagnetic scattering problems in electromagnetic simulation. The method is a method for generating a multi-resolution basis function by using a geometrical mode on a laminar grid constructed in a grid aggregation mode and further generating multi-resolution preconditions, wherein the multi-resolution basis function is formed by linear combination of a classical vector triangle basis function (RWG), and can be conveniently applied to the conventional moment method electromagnetic simulation program to effectively improve the behavior of a matrix formed in the moment method electromagnetic simulation process so as to realize acceleration of the iterative solution process of a matrix equation and fulfill the purpose of accelerating the moment method electromagnetic simulation process. Meanwhile, the multi-resolution pre-processing technology can also be conveniently combined with a quick algorithm such as a quick multi-pole algorithm. The method has the advantages of short calculation time and capability of ensuring high precision of the program and low demand of a computing memory, and can effectively improve the computing efficiency of the conventional electromagnetic simulation.

A memory management method for use in a printing process in which a time required for rendering is calculated, for managing printing data in units of band, calculating a size of the printing data in a memory, managing plural band rasters, varying the number of band rasters to be managed in the memory, converting an image to data for management, determining in accordance with a memory state whether or not data conversion is performed, and calculating a data size in the memory after the data conversion.

A system that identifies processes with a memory leak in a computer system. During operation, the system periodically samples memory usage for processes running on the computer system. The system then ranks the processes by memory usage and selects a specified number of processes with highest memory usage based on the ranking. For each selected process, the system computes a first-order difference of memory usage by taking a difference between the memory usage at a current sampling time and the memory usage at an immediately preceding sampling time. The system then generates a memory-leak index based on the first-order difference and a preceding memory-leak index computed at the immediately preceding sampling time.

A system for binary code instrumentation to reduce effective memory latency comprises a processor and memory coupled to the processor. The memory comprises program instructions executable by the processor to implement . a code analyzer configured to analyze an instruction stream of compiled code executable at an execution engine to identify, for a given memory reference instruction in the stream that references data at a memory address calculated during an execution of the instruction stream, an earliest point in time during the execution at which sufficient data is available at the execution engine to calculate the memory address. The code analyzer generates an indication of whether the given memory reference instruction is suitable for a prefetch operation based on a difference in time between the earliest point in time and a time at which the given memory reference instruction is executed during the execution.

The invention discloses a synchrony relation based parallel dynamic data race detection system which comprises a filter module, a recording module, a vector clock computation module and a parallel detection module. The filter module is responsible for monitoring memory access operation of each thread of a to-be-detected program in the running process and filtering redundant memory accesses and accesses with data races impossibly occurring, and rest accesses after filtration are written in record files corresponding to the threads by the recording module. The vector clock computation module reads out the memory access records of the program from the record files, and computes vector clocks of the memory accesses. The parallel detection module partitions a detection task by the vector clocks into small tasks to be distributed to multiple worker threads, and is responsible for summarizing detection results. The system only records the memory accesses with the races possibly occurring and merges and compresses access intervals, so that the records are greatly decreased; meanwhile, a data race detection algorithm is high in degree of parallelism, and multi-core hardware can be fully utilized to realize detection acceleration.

Embodiments of the invention disclose a memory usage feedback method and device. The method includes: detecting memory usage value of application software in a mobile terminal and memory residual value in the mobile terminal; calculating a ratio of the memory usage value to a memory peak as memory usage of the application software; judging whether the memory usage is larger than a threshold or not; if yes, transmitting a memory release command to the application software; if not, transmitting a memory applying command to the application software. The memory peak is a sum of the memory usage value and the memory residual value. The memory usage feedback method and device according to the scheme provided by the embodiments has the advantages that stable operation of the application software in the mobile terminal can be guaranteed, utilization rate of memory by the application software is increased, and user experience is improved.

An apparatus for managing memory in a real-time embedded system and a method of allocating, deallocating and managing memory in a real-time embedded system. The apparatus includes a defragmentation unit performing a defragmentation task according to a predetermined priority to collect together memory fragments, and a memory manager allocating or deallocating a predetermined area of memory upon request of a task, and calculating a memory fragmentation rate of the memory to determine a priority of the defragmentation task. The method of managing memory in a real-time embedded system includes determining whether the conditions under which the memory is used vary, and if the condition vary, calculating a memory fragmentation rate of the memory to determine a priority of the defragmentation task according to the memory fragmentation rate.

A method for operating a semiconductor device includes receiving a memory request for a memory; calculating a memory bandwidth such that the memory bandwidth is at least large enough to support allocation of the memory in accordance with the memory request; creating a memory path for accessing the memory using a memory hierarchical structure wherein a memory region that corresponds to the memory path is a memory region that is allocated to support the memory bandwidth; and performing memory interleaving with respect to the memory region that corresponds to the memory path.

The invention relates to a Swindow-Transform image denoising method and a Swindow-Transform image denoising system based on channel attention. According to the method, a noise image is input into a trained and optimized denoising network model, and a shallow feature extraction network in the denoising network model firstly extracts shallow feature information such as noise and channels of the noise image; inputting the extracted shallow feature information into a deep feature extraction network in a denoising network model to obtain deep feature information, and inputting the shallow feature information and the deep feature information into a reconstruction network of the denoising network model for feature fusion to obtain a pure image. The problems that in the prior art, an image denoising method based on a deep convolutional neural network is prone to losing input noise image details, and high computation memory and time consumption are caused are solved.

The invention discloses a Linux platform physical memory information monitoring system, which can monitor use conditions of physical memories under a Linux platform and optimize use of Linux platform physical memory resources. By adoption of the means of kernel patching and a memory monitoring processing module, static monitoring and dynamic monitoring of overall conditions and detailed use conditions of the physical memories in the Linux platform can be performed and monitoring of the following physical memory information in the Linux platform is realized, for example, acquisition of the number of distributive designated stage memory blocks, acquisition of memory details to compute external fragmentation rate of the memories and so on, thereby the invention has good compatibility and suitability. The system of the invention is particularly suitable for an embedded type Linux platform, wherein, various memory resources are limited but complex memory operations are needed to be provided.

The invention discloses a compression method of a deep neural network, which comprises the following steps of: network parameter trimming: trimming the network through pruning, deleting redundant connections, and reserving the connection with the maximum information amount; Training quantification and weight sharing: quantifying the weight, enabling the plurality of connections to share the same weight, and storing the effective weight and the index; And obtaining a compression network by using Huffman coding and bias distribution of effective weights. According to the method, the precision ofthe compressed network is improved through modes of modifying pruning, weight sharing and the like, the calculation memory space is greatly reduced, and the running speed is greatly increased; Therefore, the calculation amount and the memory of the large-scale network are effectively reduced, so that the large-scale network can effectively run on limited hardware equipment.

An input-pattern aware reference generation system for a memory cell array having a plurality of word lines crossing a plurality of bit lines includes an input counting circuit, a reference array, and a reference word line control circuit. The input counting circuit receives the input signal of the memory cell array, discovers input activated word lines according to the input signal and generates a number signal representing a number of the input activated word lines. The reference array includes a plurality of reference memory cells storing a predetermined set of weights. The reference word line control circuit is electrically connected between the input counting circuit and the reference array. Moreover, the reference word line control circuit controls the reference array to generate a plurality of reference signals being able to distinguish candidates of the computational result of the bit lines in the memory cell array.

Evaluating memory allocation in a multi-node computer including calculating, in dependence upon a normalized measure of page frame demand, a weighted coefficient of memory affinity, the weighted coefficient representing desirability of allocating memory from the node, and allocating memory may include allocating memory in dependence upon the weighted coefficient of memory affinity.

Implementations are provided herein for optimizing the usage of cluster resources in a cluster of nodes operating as a distributed file system. A node relative capacity table can be generated that inventories the total capacity of each node within the cluster of nodes. Each node can then be dynamically monitored for usage of node resources. A node available capacity table can be dynamically populated with the amount of available capacity each node has for compute, memory usage, and network bandwidth. When clients connect to the distributed file system, they can be directed to have their requests serviced by nodes with greater available capacity based on policy.

The invention discloses a page generating and display method based on a JSON (java script object notation) format. The page generating and display method comprises the following steps: receiving a page loading request; obtaining a page configuration file based on the JSON format; calculating quantity of memory resources to be occupied by the page; obtaining the quantity of preset reserved memory resources; calculating total amount of memory needs; calculating the quantity of current available memory resources; if the total amount of memory needs is smaller than the quantity of current available memory resources, loading the page in a full loading mode, otherwise, converting the page configuration file based on the JSON format into the page configuration file based on an XML (extensive markup language) format and loading the page in a partial loading mode. The page generating and display method is suitable for needs of different memory conditions, so that the system breakdown probability is reduced, the memory resource judging granularity is relatively small, the judging accuracy for the practical memory available resources is relatively high, the practical page loading speed is increased, and the defect that the JSON format cannot be partially loaded is overcome.

An embodiment of the invention provides a global memory sharing method, a global memory sharing device and a communication system to improve independence of sub-operation systems and reduce the pressure of a global memory management service. The method includes the steps: counting the page swap-out rate in the rotation period of tasks in predetermined times and the memory residence time ratio in the page swapping period by the sub-operation systems; calculating memory pressure indexes according to the page swap-out rate and the memory residence time ratio by the sub-operation systems; issuing application to the global memory management service by the sub-operation systems if the memory pressure indexes are larger than a memory pressure threshold. Under the architecture of a multi-operation system, the sub-operation systems can finish more adaptive work, global memory management service complexity is reduced, and system performances are improved. Besides, as the sub-operation systems can independently apply for and release memories, the utilization rate of a global memory can be increased, and the architecture of the multi-operation system has better expandability.

The invention provides a path finding method based on an A star optimization algorithm. The method comprises the steps of: giving a grid chart of a map, determining the position coordinates of a corresponding obstacle point, and determining the coordinates of a starting point and a target point, constructing a linear function between the starting point and the target point, solving a key point where a straight line intersects with the grid, then solving related adjacent nodes, and judging whether nodes coinciding with the obstacle point exist in the nodes or not, if the node does not coincidewith the obstacle node, enabling the searched optimal path to be a straight line between the starting point and the target point of the automobile, if an obstacle point is encountered, internally calling an A star algorithm, and putting non-obstacle nodes around the point into an open list, judging whether the open list is empty or not, taking a point with the smallest f value out from the open list to serve as the next step of path finding, judging whether the point is a target point or not, and if so, determining that path finding succeeds and ending the algorithm, or otherwise, continuing path finding, setting the point as the current point, and continuing the path finding process. According to the method, the algorithm is optimized by preprocessing the obstacle, so that the node searchtime is shortened, and the calculation memory is reduced.

The invention discloses a method for testing the power consumption of a memory. The method comprises the following steps of connecting a resistor to the memory in series by using a memory increasing card; measuring current passing through a memory bank by the resistor; then multiplying the current by VDD (Voltage Drain Drain) of the memory to accurately calculate the power consumption of the memory; meanwhile, calculating values of the power consumption in different states by combining a memory pressure procedure under the system. Compared with the traditional method for testing the power consumption of the memory, the method for testing the power consumption of the memory, has the advantages that the accuracy is higher; the power consumption values of the memory in different states can be tested; the power consumption conditions of the memories from all manufacturers can be transversely compared, so the convenience is brought to the selection of customers; the simplicity and feasibility for operation are realized; higher practicality and better popularization and use prospect are obtained.

The invention discloses a memory exception processing method. The processing method comprises the steps of reading a memory error number of a target memory bank in a memory Error register; when the memory error number is greater than a preset value, executing a heat removal operation on the target memory bank; calculating a memory delay parameter, and writing the memory delay parameter into a memory controller; wherein the memory delay parameter is waiting time after the memory controller controls the target memory bank to receive a read-write command; and executing a hot addition operation onthe target memory bank, so that the memory controller continues to execute a read-write operation on the target memory bank by using the memory delay parameter. According to the invention, the errorrate of memory reading and writing can be reduced. The invention furthermore discloses a memory exception processing system, electronic equipment and a storage medium, which have the above beneficialeffects.

The invention discloses a memory alignment plugging method and system based on machine vision, equipment and a storage medium. The method comprises the steps of: driving a mobile camera to move to aninitial position through a mechanical arm, and obtaining the initial state of a memory bank; moving the mechanical arm to a grabbing position to grab the memory bank according to the deviation value of the initial state of the memory bank and a grabbing reference state; moving the mechanical arm to a photographing position of a fixed camera, calculating the post-grabbing state of the memory bank,and moving the mechanical arm to a position to be assembled according to the deviation value of the post-grabbing state of the memory bank and an assembling reference state; performing image acquisition on a memory slot, calculating the current state of the memory slot, and calculating a compensation state according to the current state of the memory slot and an insertion reference state of the memory slot; and calculating a current plugging mounting state according to the position to be assembled and the compensation state, and moving the mechanical arm to the current plugging mounting position to perform memory bank plugging mounting operation. According to the method, system, equipment and storage medium of the invention, the memory plugging mounting operation efficiency can be improved, accurate alignment of the memory bank and the memory slot is ensured, and collision accidents in the plugging mounting operation process are prevented.