Description To investigate a statistical-runtime-feedback-driven optimisation that will improve performance of multicore interconnected processors with applications in medical imaging and biometric recognitionWhereas in the 1990s the speed of computer hardware doubled every 18 months while the software technologies remained largely stable, in the early 2000s the raw hardware speed reached its maximum and any further improvements in performance could only be achieved by increasing the amount of hardware being used by a program at any given time. Consequently software has stopped being hardware-agnostic. The age-old maxim: “if you are concerned about the speed of your program, wait for a faster machine” is no longer relevant. If one does wait, the machine will become larger instead of faster: more processor cores per processor, more multicore processors, more interconnect, more possible configurations and yet more ways of using all these. To convert “more” into “faster” the software has to be guided by hardware models. Worse still, these models need to be taken on board by either the (human) software developer or otherwise some development tools without involving the developer too much. The former increases the cost of software production and the time to market. The latter, we hope, can be effective in reducing both. This project is investigating statistical models of both software and hardware together, which are its key innovations. By doing so the investigators hope to demonstrate that a high degree of automation in matching the software with the present day complex, sophisticated hardware, thus improving its performance can be done more or less automatically.