Ефективність порозрядної конвеєризаціїї обчислень у FPGA-компонентах систем критичного застосування

• Дата: 2018
• Автори: Nikul, V. V., Drozd, A. V., Drozd, J. V., Ozeransky, V. S.
• Формат: Стаття
• Мова: Ukrainian
• Опубліковано: PE "Politekhperiodika", Book and Journal Publishers 2018
• Теми: система критичного застосування; прихована несправність; цифровий компонент; матрична структура; порозрядний конвеєр; FPGA-проектування
• Онлайн доступ: https://www.tkea.com.ua/index.php/journal/article/view/TKEA2018.4.03
• Назва журналу: Technology and design in electronic equipment

The use of computer systems like safety-related systems to ensure the functional safety of high-risk objects has raised them in the development of resources to the level of diversification. At the same time, the digital components, traditionally designed on the basis of array structures, remained at the lower level of a replication of resources. This discrepancy has created a problem of the hidden faults that can be accumulated in digital components during a prolonged normal mode and reduce their fault tolerance and system functional safety in the most critical emergency mode. Bitwise pipelines related to the level of resource diversification allow solving the problem of the hidden faults, but they are compelled to compete with the array structures, for which a powerful supporting infrastructure has been created for many decades.The paper presents a comparative analysis of the efficiency of FPGA-basedbitwise pipelines and matrix structures. Studies have shown that bitwise pipelines exhibit high efficiency exceeding that of the matrix structures in terms of both performance and energy consumption, even when designing a matrix orientation on CAD. Since such orientation reduces the advantages of bitwise pipelines, a method is proposed to increase their efficiency, which improves their throughput and energy consumption, whilemaking a cеrtain concessionsto the matrix orientation of CAD. Thiswould beparticularly importantduring the transitional period, while the traditions of matrix domination are to be overcome and the of bitwise pipeline computing is to be formed.