13 GPU profiling and tools

In this chapter, we will cover the tools and the different workflows that you can use to accelerate your application development. We show you how profiling tools for the GPU can be helpful. Additionally, we discuss how to deal with the challenges of using profiling tools when working on a remote HPC cluster.

The profiling tools continue to change and improve, so we’ll focus on the methodology rather than the details of any one tool. The main takeaway of this chapter will be understanding how to create a productive workflow when using complex GPU profiling tools.

13.1  Profiling tools overview

The available profiling tools allow for quicker optimization, improving hardware utilization, and a better understanding of the application performance and hotspots. We discuss how profiling tools expose bottlenecks and assist you in attaining better usage of hardware.

13.2  How to select a good workflow

13.3  Example problem: shallow water simulation

13.4  A sample of a profiling workflow

13.4.1    Run the shallow water code

13.4.2    Profile the CPU code:

13.4.3    Add OpenACC compute directives

13.4.4    Add data movement directives

13.4.5    The Nvidia Nsight suite of tools can be a powerful development aid

13.4.6    Code XL for the AMD GPU ecosystem

13.5  Don’t get lost in the swamp: focus on the important metrics

13.5.1    Occupancy

13.5.2    Issue efficiency

13.5.3    Achieved Bandwidth

13.6  Containers and Virtual Machines provide alternate workflows

13.6.1    Docker containers as a workaround

13.6.2    Virtual machines using Virtual Box

13.7  Cloud options: a flexible and portable capability

13.8  Further Explorations

13.8.1    Additional Reading

13.8.2    Exercises