Abstract:
A voltage-controllable power-mode-aware (PMA) clock tree in an integrated circuit (IC) and a synthesis method and an operation method thereof are provided. The PMA clock tree includes at least two sub clock trees, at least two PMA buffers and a power mode control circuit. The at least two PMA buffers respectively delay a system clock and provide the delayed system clock to the sub clock trees as delayed clocks. The power mode control circuit respectively provides at least two first power information to at least two function modules to respectively determine the power modes of the function modules. The power mode control circuit respectively provides at least two second power information to the at least two PMA buffers to respectively determine the delay time of the PMA buffers.
Abstract:
A clock tree in a circuit and an operation method thereof are provided. The clock tree includes at least two sub clock trees, at least two voltage-controllable power-mode-aware (PMA) buffers and a power-mode control circuit. The PMA buffers delay a system clock to serve as the delayed clock, and provide respectively the delayed clock to the sub clock trees. The power-mode control circuit provides at least two first power information to at least two function modules respectively, wherein a power mode of each of the function modules is determined according to the first power information respectively. The power-mode control circuit provides at least two second power information to the PMA buffers respectively, wherein a delay time of each of the PMA buffers is determined according to the second power information respectively.
Abstract:
An apparatus and a method for neural network computation are provided. The apparatus for neural network computation includes a first neuron circuit and a second neuron circuit. The first neuron circuit is configured to execute a neural network computation of at least one computing layer with a fixed feature pattern in a neural network algorithm. The second neuron circuit is configured to execute the neural network computation of at least one computing layer with an unfixed feature pattern in the neural network algorithm. The performance of the first neuron circuit is greater than that of the second neuron circuit.
Abstract:
An apparatus and a method for neural network computation are provided. The apparatus for neural network computation includes a first neuron circuit and a second neuron circuit. The first neuron circuit is configured to execute a neural network computation of at least one computing layer with a fixed feature pattern in a neural network algorithm. The second neuron circuit is configured to execute the neural network computation of at least one computing layer with an unfixed feature pattern in the neural network algorithm. The performance of the first neuron circuit is greater than that of the second neuron circuit.
Abstract:
A data feature augmentation system and method for a low-precision neural network are provided. The data feature augmentation system includes a first time difference unit. The first time difference unit includes a first sample-and-hold circuit and a subtractor. The first sample-and-hold circuit is used for receiving an input signal and obtaining a first signal according to the input signal. The first signal is related to a first leakage rate of the first sample-and-hold circuit and the first signal is the signal generated by delaying the input signal by one time unit. The subtractor is used for performing subtraction on the input signal and the first signal to obtain a time difference signal. The input signal and the time difference signal are inputted to the low-precision neural network.
Abstract:
A neural circuit is provided. The neural circuit includes a neural array. The neural array includes a plurality of semiconductor components. Each of the semiconductor components stores a weighting value to generate a corresponding output current or a corresponding equivalent resistance. The neural array receives a plurality of input signals to control the semiconductor components in the neural array and respectively generates the output currents or changes the equivalent resistances. Since the semiconductor components are coupled to each other, output of the neural array may generate a summation current or a summation equivalent resistance related to the input signals and a weighting condition, so that a computing result exhibits high performance.
Abstract:
A clock tree in a circuit and an operation method thereof are provided. The clock tree includes at least two sub clock trees, at least two voltage-controllable power-mode-aware (PMA) buffers and a power-mode control circuit. The PMA buffers delay a system clock to serve as the delayed clock, and provide respectively the delayed clock to the sub clock trees. The power-mode control circuit provides at least two first power information to at least two function modules respectively, wherein a power mode of each of the function modules is determined according to the first power information respectively. The power-mode control circuit provides at least two second power information to the PMA buffers respectively, wherein a delay time of each of the PMA buffers is determined according to the second power information respectively.