high performance computing on graphics processing units: hgpu.org

Field-Programmable Gate Arrays (FPGAs) have been increasingly deployed in datacenters and there has been a lot of focus on tools that help the development of FPGA applications. Among the most important tools are performance monitors that provide visibility into the state of the hardware. As the application platforms scale from one FPGA to many FPGAs, such as the well-known examples demonstrated on the Microsoft Catapult platform, it is no longer feasible to rely on low-level FPGA tools such as embedded logic analyzers. This thesis presents Pharos, a lightweight, generic performance monitor for multi-FPGA systems. We show that Pharos is able to measure cycle-accurate unidirectional point-to-point latency between multiple network-connected FPGAs. Pharos is also capable of monitoring network traffic and conducting throughput measurements, as well as tracing events across the datacenter.