710042Concurrent Proof of ConceptClosedGRD Proof of ConceptInnovate UKIn a market research study into software development for multi-core and heterogeneous systems that include GPU processors, Concurrent Systems identified: + A strong need to make use of the increased compute capabilities these systems provide to meet current and future application performance points + These systems are too difficult to program for most developers + There is an opportunity for Concurrent Systems to provide tools to the developer for manycore and heterogeneous application programming Applications undergo constant improvements in features and performance to maintain their competitive edge and to meet growing customer expectations. There is therefore a demand to keep pace with the latest hardware technology releases and to obtain the best performance from them. Today, and for the foreseeable future, this means using many-core and heterogeneous systems. Many-core processors have created a technology gap between the hardware and the software that runs on them. It is common to have developers who are experts in their field rather than specialists in many-core program design. Improved productivity tools and programming models will ease development concerns and increase the speed of market adoption. There are a number of products on the market today that attempt to address parallelization either through code analysis and applying programming model pragmas, or through automated code generation. Concurrent Systems’ solution is an innovative approach to the problem. The underlining points to our philosophy on the optimal utilization of compute resources are: 1. Performance comes from parallelism 2. Efficiency comes from data locality To successfully address this market Concurrent Systems needs to apply these principles in development tools that will enable better products to be created for current and future manycore and heterogeneous architectures. A tool that assists the developer in locating bottlenecks and designing applications to account for architectural features, such as caching, SSE, SIMD units, creates a great opportunity for achieving high performance and scalability. There is a large number of existing applications that were developed taking into account hardware constraints of technology available at the time. These constraints are no longer relevant today. A lot of this code persists in applications without modification and this has a detrimental performance impact. These applications would benefit from refactoring. Tools that allow both novice and experienced developers to highlight the limiting factors of their application will result in software that can be developed rapidly and make efficient use of current and future architectural features.