Abstract: A method for accelerating machine learning on a computing device is described. The method includes accessing a neural network. The method also includes splitting the neural network into N sub-neural networks. The method further includes hosting the N sub-neural networks in M inference accelerators. The method also includes scheduling the N sub-neural networks in the M inference accelerators. The method further includes executing the N sub-neural networks in the M inference accelerators.

Abstract: Exemplary embodiments provide for compressing, storing, retrieving and decompressing paged code from mass storage devices. By evaluating the size of compressed virtual pages relative to the storage page (read unit) of the mass storage device into which the compressed virtual pages are to be stored, decisions can be made which facilitate later read out and decompression of those compressed virtual pages. According to exemplary embodiments, a virtual page can be stored uncompressed, compressed but undivided or compressed and subdivided into a plurality of parts based on an evaluation.

Abstract: Exemplary embodiments provide for compressing, storing, retrieving and decompressing paged code from mass storage devices. By evaluating the size of compressed virtual pages relative to the storage page (read unit) of the mass storage device into which the compressed virtual pages are to be stored, decisions can be made which facilitate later read out and decompression of those compressed virtual pages. According to exemplary embodiments, a virtual page can be stored uncompressed, compressed but undivided or compressed and subdivided into a plurality of parts based on an evaluation.

Abstract: A memory management system for managing memory of a processing system having a primary memory and at least one secondary memory is disclosed. The memory is managed by optimizing the number of writes required by swapping one or more relevant pages of an application from the primary memory to the at least one secondary memory. The system comprises of a dynamic memory manager for allocating memory to the application from a memory pool and having a first table containing virtual addresses and chunk sizes of memory allocated by the application. The system further comprises of a swap manager having a second table containing the physical addresses of the primary memory pages and information whether the pages are allocated or not. The system further comprises of a memory management unit having a third table containing a mapping information of the physical addresses and the virtual addresses of the one or more physical pages used by the application and information whether the page is dirty or not.