53 results about "HashClash" patented technology

The invention discloses a method for selecting a hash function, and a method and a device for storing and searching a routing table. The method for selecting the hash function comprises the followingsteps: using the current alternative hash function to hash data to be hashed; decoding a hashed result; and accumulating an encoding result until a carry is not generated in the process of the accumulation, and selecting the alternative hash function as a hash function for normal use. The method for storing the routing table comprises the following steps: dividing the routing table into a next-stage node pointer part and a matching item part for storage; and selecting the hash function by using the method provided by the embodiment of the invention to select the hash function. The method for searching the routing table comprises the following steps: taking an IP address to be checked to carry out searching directly according to the directly-stored length of the next-stage node pointer partfor storing the routing table; and reading a matching item code according to a searching result. The method can avoid hash conflict, and effectively reduce the storage resources occupied by the routing table.

The invention discloses an index tree building method and device. The index tree building method comprises the steps that a keyword and corresponding index information of the keyword are built, wherein the keyword comprises at least one Chinese character; the keyword is divided into a first etymon sequence, searching is conducted on an index tree built in advance according to the first etymon sequence, whether a path which starts from a root node and enables an etymon sequence composed of passed nodes to be matched with the first etymon sequence exists in the index tree is judged; if the path does not exist, a last matched node of the first etymon sequence in the index tree is obtained, a corresponding sub index tree is built for an etymon sequence which is not matched successfully under the last matched node, and the keyword and the corresponding index information of the keyword are stored into the last node of the sub index tree. The invention further provides a Chinese vocabulary searching method and device; a Chinese vocabulary is divided into CXME to build a vertical index tree; the Chinese vocabulary searching method and device have the advantages of being easy to build, fast to search, and small in Hash collision.

An efficient virus detection, malware detection, and packet filtering system in a mobile device by providing optimized hash functions from a server to a mobile device that reduce hash collisions during the virus detection, malware detection, and packet filtering in a system-on-chip configuration.

The invention discloses a hash table construction method and system for a nonvolatile memory. The hash table construction method is characterized in that a hash table is constructed and is logically constructed into an inverted complete binary tree, all leaf nodes of the binary tree are addressable units, and all non-leaf nodes are unaddressable units and serve as standby units for the leaf nodes to deal with hash conflicts; all the non-leaf nodes on a path from the leaf nodes to root nodes are used for storing conflict elements of the hash conflicts at the leaf nodes; furthermore, a plurality of layers at the bottom of the binary tree are deleted, and only rest layers at the top are remained; and each element in the hash table corresponds to two different hash positions which are obtained through computation by use of two different hash functions. The invention also provides the hash table construction system for the nonvolatile memory. According to the technical scheme, the constructed hash table cannot cause any additional writing, and high space utilization ratio and low request delay are achieved.

The invention provides an SCM (Storage Class Memory)-based Key-Value log-type local-storage method. The storage method comprises the following steps: an SCM dividing step, wherein partition dividing is carried out on an SCM according to functions, that is, the SCM is divided into five different logic partitions of a superblock area, a checkpoint area, a page management area, a static-hash-table area and a data storage area; a multilayer memory management step, wherein three layers of memory management modules are built, the memory is managed at different layers, and dynamic allocation and reclaimation of pages are realized; and a high-concurrency hash-table building step, wherein a static hash table is used to index key-value objects, and a dynamically allocated array is used to store hash-conflicting elements. According to the method realized in the invention, storage space is reasonably scheduled and allocated according to hardware characteristics of a heterogeneous memory medium, autilization rate of the storage medium is increased, service life of hardware is prolonged, and concurrency and data throughput of a hybrid key-value storage system are improved.

The invention discloses a method and device for writing a MAC address, and a computer-readable storage medium; the method includes: judging whether the MAC address sent to the chip is successfully written into the MAC address table of the chip; if not, judging Whether the MAC address is a preset important MAC address; if so, calculate the position of the hash bucket of the MAC address of the chip corresponding to the MAC address; put the first dynamic MAC address in the hash bucket Replace with the said MAC address. The invention can ensure the robustness of the network and the ease of maintenance of the equipment. This prevents important network forwarding traffic from being abnormal due to MAC address hash collisions.

The present invention provides a sparse neural network accelerator and its implementation method. The accelerator mainly includes a PE array, an output memory, and a scheduler module. The PE array is divided into multiple PE groups, and each PE group and the corresponding output memory form an associated group. The number of PEs in the association group is equal to the number of output memories; each PE unit in the association group can access any output memory in the association group; PE units in any association group are calculated according to the input activation value and weight value Multiple output results are obtained and written into corresponding multiple output memories according to preset rules; the scheduler module schedules the sequence of output activation values ​​to reduce the probability of hash collisions. The invention divides the original PE array into multiple PE groups, and forms an associated group architecture with the corresponding output memory, which greatly reduces the area of ​​the output memory and reduces power consumption; the scheduler module reduces the probability of hash conflicts and greatly improves the Computational performance of the entire system.