Ideas on Improving Development Tools to Benefit Performance on Multi-core CPUs

With maturing compiler technologies, compile-time analysis can be a very powerful tool for optimizing and monitoring code on any architecture. In combination with modern run-time analysis tools and existing performance programming interfaces extensively monitoring hardware counters, we will survey modern techniques of monitoring performance hazards: event analyzers and identifiers, performance tracking threads, dynamic hotspot analysis, performance profiles generation, optimizer outputs. The running process' performance data and sequence of performance events allow to trigger your application and OS for hazards, and cause interrupts. In such way, sequence of performance events can be your debugging breakpoint or a callback. This paper bridges the capabilities given by advanced performance monitoring with common software development infrastructure (debuggers, gcc, loader, process scheduler).

With hyper-threading and multi-core CPUs, logical CPUs end up sharing the resource at different levels. In such case applications will need more interfaces and tunables in the scheduler then just a priority. Task affinity will help a little in this case where application can ask to be run on specific subset of CPUs. However, there is scope for more generic interface between kernel sheduler and user apps, to come to a common understanding about the actual hardware capabilities and the application requirements. In the era of parallel and platform computing, higher granularity performance data and processor efficiency per functionality concept can be applied for making intelligent decisions on process scheduling. In the end, we will suggest particular improvements in performance and run-time tools as a reflection of proposed approaches and transition to platform-level optimization goals.