Ilya Mkrtchan

The Residue Number System (RNS) is a non-positional number system that allows parallel computations without transfers between digits. However, some operations in RNS require knowledge of the positional characteristic of a number. Among these operations is the conversion from RNS to the positional number system. The methods of reverse conversion for general form moduli based on the Chinese remainder theorem and the mixed-radix conversion are considered, as well as the optimized methods for special form moduli. In this paper, a method is proposed that develops the authors’ ideas based on the modified mixed-radix conversion and reference points. The modified method based on the mixed-radix conversion in this case makes it possible to replace the operation of finding the residue of division by a large modulo with the sequential calculation of the residue. The method of reference points allows to reduce the size of the stored information compared to the use of ROM to store all the residues of RNS. The application of this approach makes it possible to find a balance between the speed of the calculation and the hardware used, by varying the number of moduli of one method and the other.

Modern computational tasks are often required to not only guarantee predefined accuracy, but get the result fast. Optimizing calculations using floating point numbers, as opposed to integers, is a non-trivial task. For this reason, there is a need to explore new ways to improve such operations. This paper presents analysis and comparison of various floating point formats - float, posit and bfloat. One of the promising areas in which the problem of using such values can be considered to be the most acute is neural networks. That is why we pay special attention to algorithms of linear algebra and artificial intelligence to assess efficiency of new data types in this area. The research results showed that software implementations of posit16 and posit32 have high accuracy, but they are not particularly fast; on the other hand, bfloat16 is not much different from float32 in accuracy, but significantly surpasses it in performance for large amounts of data and complex machine learning algorithms. Thus, posit16 can be used in systems with less stringent performance requirements, as well as in conditions of limited computer memory; and also in cases when bfloat16 cannot provide required accuracy. As for bfloat16, it can speed up systems based on the IEEE 754 standard, but it cannot solve all the problems of conventional floating point arithmetic. Thus, although posits and bfloats are not a full fledged replacement for float, they provide (under certain conditions) advantages that can be useful for implementation of machine learning algorithms.