Final Year Research Project

1. A novel multi-core benchmark application has been developed using the latest C++ features, this application built upon the previous mpbenchmark application. It outperformed the previous mpbenchmark implementations in C and Ada by 11%, 4% and 1% on desktop, Raspberry Pi 5 and 4 respectively in terms of run time and provided better speedup across threads. Moreover the novel benchmark application also utilised SIMD intrinsics which provided up to an astonishing 70% reduction in run time, this unprecedented improvement provided means of analysing the target CPU's SIMD operations ability. A publication describing the contribution of this project has been submitted to SBCCI(Symposium on Integrated Circuits and System Design) conference part of IEEE(Institute of Electrical and Electronics Engineers) organisation.

2. The development of the first multi-threaded MobileNet CNN(convolutional neural network) model. This enhanced CNN model has been used as a case study to assess the behaviour of multi-threaded applications running on multi-core processors under neutron radiation. A publication describing this CNN model has been submitted to IEEE RADECS("RADiation Effects on Components and Systems") conference.

3. The optimised redesign of the DeBaTE-FI platform resulted in up to 60% reduction in run time. Moreover a high performance C++ library was developed for telnet communication with microcontrollers. The enhanced DeBaTE-FI platform produced a 43.5% reduction in run time when testing 36 STM32F767ZI microcontrollers when tested in PhD research office in Loughborough University.

Dissertation

This final year dissertation was written using LaTex with TeXstudio IDE and LuaLaTex compiler. Please find the full dissertation in the file named "Individual_report.pdf".