Baselining Functionality and Performance - 2022.1 English

Vitis Unified Software Platform Documentation: Application Acceleration Development (UG1393)

Document ID
UG1393
Release Date
2022-05-25
Version
2022.1 English

Methodology for Accelerating Data Center Applications with the Vitis Software Platform provides an overview of designing an application beginning with profiling the application to identify functions to accelerate, leading into recommended ways of developing C/C++ accelerators. As discussed in the this guide, it is very important to understand the architecture and performance of your application before you start any optimization effort. This is achieved by establishing a baseline for the application in terms of functions and performance.

Figure 1. Baselining Functions and Performance Flow

Identify Bottlenecks

The first step is to identify the bottlenecks of the your application running on your target platform. The most effective way is to run the application with profiling tools like the user profiling features described in Custom Profiling of the Host Application, or valgrind, callgrind, and GNU gprof. The profiling data generated by these tools show the call graph with the number of calls to all functions and their execution time.

Run Software and Hardware Emulation

Run software and hardware emulation on the accelerated application as described in Running Emulation, to verify functional correctness, and to generate profiling data on the host code and the kernels. Use Vitis analyzer to review the kernel compilation reports, profile summary, timeline trace, and device hardware transactions to understand the baseline performance estimate for timing interval, latency, and resource utilization, such as DSP and block RAM.

Build and Run the Application

The last step in baselining is building and running the application on an FPGA acceleration card, like one of the Alveo™ Data Center accelerator cards, as described in Running the Application Hardware Build. Analyze the reports from the Vitis compiler, and the profiling data from application execution to see the actual performance and resource utilization on hardware.

Tip: Save all the reports during the baseline process, so that you can refer back to them and compare results during optimization.