摘要:
A full-rail digital-read CIM circuit enables a weighted read operation on a single row of a memory array. A weighted read operation captures a value of a weight stored in the single memory array row without having to rely on weighted row-access. Rather, using full-rail access and a weighted sampling capacitance network, the CIM circuit enables the weighted read operation even under process variation, noise and mismatch.
摘要:
Techniques and mechanisms for performing in-memory computations with circuitry having a pipeline architecture. In an embodiment, various stages of a pipeline each include a respective input interface and a respective output interface, distinct from said input interface, to couple to different respective circuitry. These stages each further include a respective array of memory cells and circuitry to perform operations based on data stored by said array. A result of one such in-memory computation may be communicated from one pipeline stage to a respective next pipeline stage for use in further in-memory computations. Control circuitry, interconnect circuitry, configuration circuitry or other logic of the pipeline precludes operation of the pipeline as a monolithic, general-purpose memory device. In other embodiments, stages of the pipeline each provide a different respective layer of a neural network.
摘要:
A method, operable on a processing device, for merging calendar entries may include receiving a plurality of calendar entries each associated with entry identification data. The method may also include comparing by the processing device at least a portion of the entry identification data associated with each of the calendar entries. The method may additionally include merging, by the processing device, the calendar entries based at least in part on comparing of at least the portion of the entry identification data associated with each of the at least two calendar entries. The method may further include comparing at least one time attribute associated with each of the calendar entries and comparing at least one textual attribute associated with each of the calendar entries and basing merging the calendar entries additionally on comparing the time attributes and the textual attributes.
摘要:
Embodiments disclosed herein provide sparse adder circuits comprising Ling type propagate and generate circuits and sparse carry circuits to efficiently add first and second operands to one another.
摘要:
A system to process multiplier X and multiplicand Y may include multiplication of a least-significant bit of X and a least-significant w bits of Y to generate a least-significant w bits of product Z. The system may further include determination of whether a least-significant bit of product Z is 1, addition of a least-significant w bits of modulus M to the least-significant w bits of product Z if the least-significant bit of product Z is 1, multiplication of the least-significant bit of X and bits 2w-1:w of Y to generate bits 2w-1:w of product Z, and addition of bits 2w-1:w of modulus M to bits 2w-1:w of product Z if the least-significant bit of product Z is 1. Multiplying the least-significant bit of X and bits 2w-1:w of Y may occur at least partially contemporaneously with multiplying the least-significant bit of X and the least-significant w bits of Y, determining if the least-significant bit of product Z is 1, and adding the least-significant w bits of modulus M to the least-significant w bits of product Z if the least-significant bit of product Z is 1.
摘要:
A current mirror multi-channel leakage monitor circuit and method measures die leakage and generates digital keeper control bits to control a process compensated dynamic circuit. The leakage monitor enables high resolution on-chip leakage measurements in multiple locations on a die, thereby saving test time and enabling both die to die and within die process compensation.
摘要:
A dynamic bus architecture is provided. This may include an encoding circuit coupled to a bus line and a decoder circuit coupled to the bus line. The encoder circuit may receive an input signal and generate an encoded signal on the bus line. The decoder circuit may receive the encoded signal from the bus line and generate the original unencoded signal. The encoder circuit may include a first flip-flop circuit to store a previous input signal from the bus line based on a clocking signal from the bus line. Additionally, the decoder circuit may include a second flip-flop circuit having a clock input to receive the encoded signal from the bus line as a clocking input.
摘要:
A Viterbi decoder includes an ACS unit that performs state metric updates for every symbol cycle. State metric updates involve adding the state metrics corresponding to a likely input symbol to the respective branch matrix, comparing the results of the additions to determine which is smaller, and selecting the smaller result for the next state metric. The ACS unit includes two parallel adders followed by a parallel comparator that generates a multiplexer-select signal. The outputs of the parallel adders are input into a multiplexer and the multiplexer-select signal is input into the multiplexer for a decision.
摘要:
A register file contains a local bit trace and a driving signal trace as well as a plurality of data cells coupled to the local bit trace. A device is coupled to the driving signal trace and the local bit trace to intelligently charge and float the local bit trace. The intelligent charging and floating is facilitated by determination of a selection of one of the data cells.
摘要:
A variable keeper strength based process-compensated dynamic circuit and method provides a robust digital way to overcome the intrinsic parameter variation present in manufactured die. Using a process-compensated dynamic circuit, the wide robustness and delay distribution becomes narrower which improves performance without sacrificing worst-case robustness. The strength of the keeper is programmed depending on the amount of die leakage. The keeper will have an optimal strength for the best and worst case leakage, allowing better performance with improved worst-case robustness.