Abstract:
A computer processor that performs operations in a logarithmic number system (LNS) domain includes a log converter (20) which generates log signals, a data pipeline (22), a plurality of processing elements (231a-f) coupled to respective stages (24a-d) of the data pipeline, an inverse-log converter (28), and a programmable accumulator (232) that performs various summing operations to produce an output signal. An instruction, selected from a set of instructions, is decoded by a control unit (234) to configure the computer processor to perform operations on one or more data streams. Mathematical operations that can be performed by the processor include matrix multiplication, matrix-inversion, fast Fourier transforms (FFT), auto-correlation, cross-correlation, discrete cosine transforms (DCT), polynomial equations, and difference equations in general, such as those used to approximate infinite impulse response (IIR) and finite impulse response (FIR) filters. The computer processor can be used as a co-processor (340) in a general purpose computer system.
Abstract:
A method for storing instructions in a computer memory (106) is provided. The method is utilized in a multiprocessor computer which includes a plurality of processing elements (108, 110, 112). Instructions are stored in the computer memory (106) so that the processing elements (108, 110, 112) produce outputs according to an execution schedule.
Abstract:
A computer processor that performs operations in a logarithmic number system (LNS) domain includes an input log converter (20), a feedback log converter (303), a first data pipeline (304), a second data pipeline (306), a plurality of processing elements (26a-f) coupled to respective stages of the data pipelines, an inverse-log converter (28), and a programmable accumulator (232) which produces output signals. An instruction, selected from a set of instructions, is decoded by a control unit (235) to configure the computer processor to perform operations on one or more data streams. Mathematical operations that can be performed by the processor include matrix multiplication, matrix-inversion, fast Fourier transforms (FFT), auto-correlation, cross-correlation, discrete cosine transforms (DCT), polynomial equations, and difference equations in general, such as those used to approximate infinite impulse response (IIR) and finite impulse response (FIR) filters. The computer processor can be used as a coprocessor (340) in a general purpose computer system.
Abstract:
A circuit and method for computing an exponential signal x is provided. The circuit includes a logarithm converter (4) which converts an input signal to binary word that represents the logarithm of an input signal x. A first shift register shifts (8) the binary word in a bit-wise fashion to produce a first intermediate value; while a second shift register (28) shifts the binary word in a bit-wise fashion to produce a second intermediate value. The shift registers may be implemented using multiplexers. The shifting operations are equivalent to multiplying the intermediate values by a factor which is a power of two. The first intermediate value is either added to or subtracted from the second intermediate value to produce a combined value. An inverse-logarithm converter (34) converts the combined value to the exponential signal.
Abstract:
A method and system for a correlation operation in a digital signal processor (DSP) are provided. The correlation operation is performed in a logarithmic number system (LNS) domain. The DSP includes a plurality of processing elements (26a-d) that executes the correlation operation in a highly parallel fashion. The method can be implemented as a software program that directs a LNS based DSP to execute the correlation operation.
Abstract:
A method and system for a convolution operation in a digital signal processor (DSP) are provided. The convolution operation is performed in a logarithmic number system (LNS) domain. The DSP includes a plurality of processing elements (26a-d) that executes the convolution in a highly parallel fashion. The method can be implemented as a software program that directs a LNS based DSP to execute the convolution operation.