Modular-logarithmic coprocessor for massive arithmetic calculations

Бесплатный доступ

The paper presents a conceptual design of an IP module of mathematical coprocessor. It consists of a set of processing cores of the same kind which perform single-cycle scalar, or vector operations with real numbers. The processed data is represented in the modular logarithmic format that provides two levels of translating the original numbers, namely: the modular level instead of the conventional positional system and the logarithmic level instead of the floating point format. As a result of the research and development, new scientific and technical solutions are proposed that implement the proposed methods of computing and coding data. Owing to this feature a coprocessor has a higher performance, a higher accuracy and a higher level of reliability, as compared to the known analogs. Convert codes in modular-logarithmic number system and vice versa does not introduce significant time delays in a large stream of input data by offering hardware solutions pipelined process of interpolation of the logarithm function and conversion of residual classes system codes. A prototype coprocessor is an FPGA-based IP module. Companies developing general-purpose processors are the target market for this design.

Еще

Residue number system, logarithmic number system, reconfigurable architecture, highly reliable computing

Короткий адрес: https://sciup.org/147160617

IDR: 147160617   |   DOI: 10.14529/cmse170202

Статья научная