Abstract:
The present disclosure is directed to systems and methods of implementing a neural network using in-memory mathematical operations performed by pipelined SRAM architecture (PISA) circuitry disposed in on-chip processor memory circuitry. A high-level compiler may be provided to compile data representative of a multi-layer neural network model and one or more neural network data inputs from a first high-level programming language to an intermediate domain-specific language (DSL). A low-level compiler may be provided to compile the representative data from the intermediate DSL to multiple instruction sets in accordance with an instruction set architecture (ISA), such that each of the multiple instruction sets corresponds to a single respective layer of the multi-layer neural network model. Each of the multiple instruction sets may be assigned to a respective SRAM array of the PISA circuitry for in-memory execution. Thus, the systems and methods described herein beneficially leverage the on-chip processor memory circuitry to perform a relatively large number of in-memory vector/tensor calculations in furtherance of neural network processing without burdening the processor circuitry.
Abstract:
Described is an apparatus for a voltage controlled nano-magnetic random number generator. The apparatus comprises: a free ferromagnetic layer; a fixed ferromagnetic layer positioned in a non-collinear direction relative to the free ferromagnet layer; and a first terminal coupled to the free ferromagnetic layer, the first terminal to provide a bias voltage to the free ferromagnetic layer. Described is also an integrated circuit comprising: a random number generator including a magnetic tunnel junction (MTJ) device with non-collinearly positioned free and fixed ferromagnetic layers; and a circuit to provide an adjustable bias voltage to the free ferromagnetic layer, the circuit to control variance of current generated by the random number generator.
Abstract:
Field effect transistor structures are described that are formed using germanium nanowires. In one example, the structure has a germanium nanowire formed on a substrate along a predetermined confinement orientation, a first doped region of the nanowire at a first end of the nanowire to define a source, a second doped region of the nanowire at a second end of the nanowire to define a drain, and a gate dielectric formed over the nanowire between the source and the drain.
Abstract:
Described is an apparatus 1T-1 Magnetic Tunnel Junction (MTJ) Spin Hall Magnetic Random Access Memory (MRAM) bit-cell and array, and method of forming such. The apparatus comprises: a select line; an interconnect with Spin Hall Effect (SHE) material, the interconnect coupled to a write bit line; a transistor coupled to the select line and the interconnect, the transistor controllable by a word line; and a MTJ device having a free magnetic layer coupled to the interconnect.