Operation method and device of sequence accumulator

Abstract

The invention provides an operation method and device of a sequence accumulator. The method comprises the following steps of: sequentially inputting value of sequences stored in a data cache into a first input end of a floating point adder; sequentially adding the value of the sequence in the first input end to the value of the second input end of the floating point adder, and outputting a resultto the output end of the floating point adder; sequentially inputting the value output by the output end into a second input end; in response to the fact that all the values of the sequence are inputinto the floating point adder, inputting the value output by the output end into the input end of the data cache, and the value in the data cache into the first input end and the second input end in sequence; and outputting a result by the output end of the floating point adder after a threshold period. The assembly line waiting period caused by the delay of the adder can be effectively reduced, and the invention has the advantages of short operation period, high resource utilization rate, small assembly line interval, small cache occupation and the like, and can be flexibly applied to different scenes.

Description

technical field [0001] This field relates to the computer field, and more specifically relates to a sequence accumulator operation method and device. Background technique [0002] Accumulation and summation of data is a common mathematical operation and one of the most important basic units of various complex signal processing algorithms. Modern advanced digital devices, circuits, equipment, etc., all include the design of accumulators. The accumulator is implemented by feeding back the output of the adder to the input, and the accumulation of a series of sequences is completed through multiple iterations. For floating-point operations (or multi-bit fixed-point data), since the calculation of the floating-point adder is more complicated, it needs to go through five steps: alignment, mantissa operation, result normalization, rounding processing, and overflow judgment. Generally, there will be 7 to 14 steps. Stage pipeline delay, that is to say, two floating-point numbers A a...

AI Technical Summary

Problems solved by technology

The delay of the adder will cause the blockage of the pipeline when designing the accumulator. The traditional accumulator adopts a waiting method when dealing with the delay of the adder, so it takes a long time to complete the accumulation operation.

[0003]The existing accumulator operation cycle is long: it takes n*k clock cycles to complete the accumulation of a floating-point sequence with a length of n, which is critical for many scenarios that are demanding on operation time Unable to apply; low resource utilization: the floating-point adder is only used once in k cycles, and the utilization rate is low; the pipeline interval is large: after a sequence is input, due to the long operation cycle, if there are subsequent sequences to be calculated, it must wait for the upper Completion of all calculations at one time can only be input, and the waiting interval is relatively large; the cache occupies a large amount: the input of the data cache is continuously input, and when reading, one data is read at intervals of k, and the size of the entire cache should be basically the same as that of the data The sequence is of the same size

On the one hand, when n is large, the cache space occupied is also large; on the other hand, when designing the accumulator, it is necessary to know the maximum number of sequences to be input in advance and limit the maximum number of sequences to be input. When the sequence to be input is greater than When the value is limited, the design of the accumulator needs to be readjusted, which is inflexible in application

Images