Abrégé

The present application provides a neural network weight storage method and reading method, and a related device. The storage method comprises: solidifying the weight of a first layer of a neural network in the static memory of a static random access memory when a system is initialized; obtaining the cumulative length of weights of respective layers of the neural network in a cyclic buffer of the static random access memory; solidifying in the static memory the weight of a target layer among the layers according to the cumulative length of weights; performing the next operation of obtaining the cumulative length of weights, and repeating multiple times the operation of obtaining the cumulative length of weights until there is no target layer among the layers; and setting the size of the cyclic buffer according to the corresponding cumulative length of weights when there is no target layer among the layers. Embodiments of the present application can solve the problems of overflow and underflow of the cyclic buffer, such that throughput of a chip is optimized, thereby reducing the power consumption and cost of the chip.

Classes IPC  ?

• G06F 12/08 - Adressage ou affectation; Réadressage dans des systèmes de mémoires hiérarchiques, p.ex. des systèmes de mémoire virtuelle
• G06N 3/06 - Réalisation physique, c. à d. mise en œuvre matérielle de réseaux neuronaux, de neurones ou de parties de neurone

Abrégé

Processors, systems and methods are provided for thread level parallel processing. A processor may comprise a plurality of processing elements (PEs) that each may comprise a configuration buffer, a sequencer coupled to the configuration buffer of each of the plurality of PEs and configured to distribute one or more PE configurations to the plurality of PEs, and a gasket memory coupled to the plurality of PEs and being configured to store at least one PE execution result to be used by at least one of the plurality of PEs during a next PE configuration.

Classes IPC  ?

• G06F 15/00 - TRAITEMENT ÉLECTRIQUE DE DONNÉES NUMÉRIQUES Équipement de traitement de données en général
• G06F 15/76 - Architectures de calculateurs universels à programmes enregistrés