A Learning Path For The Performance Optimization of GPUs

Context:

Although GPUs were originally designed to accelerate the rendering of 3D graphics, their parallel architecture makes them ideal hardware for massive computations. To fully realize the power of a GPU, a developer must tune applications to balance and monitor workloads on the CPU and GPU. Intel® VTune™ Profiler is a performance optimization software that can help developers with this task. This product has several profiling capabilities for the GPU user. As the documentation owner and content strategist for this product, I created this learning path to tie them all together.

The Work:

The GPU profiling workflow using VTune Profiler is a popular choice for application developers. However, there existed far too many pieces of content that contained relevant information. Older topics were in need of updates, while newer topics did not have enough traction to top the list of web searches.

I interviewed several GPU subject matter experts (SMEs) to learn about the workflow and the types of questions that developers frequently search for.

I knew I would create a new learning path. To make this path easily searchable, I used the same questions asked by developers to create steps and section titles in my content.

I did not want to duplicate content from the user guide and technical cookbook, so I linked to these topics in a ‘breadcrumb approach’.

I completed the learning path by embedding a popular webinar that demonstrated the steps in the article. Developers were quick to visit this page within a month!

This feedback from one of my reviewers sums up the impact of this work:

"As an end-user who ends up trying to find this information by <search engine>, I cannot tell you how much I appreciate having a single page I can bookmark that will then let me find the desired article, rather than having to dig through many levels of the table of contents. This will be a godsend for end-users!"

Result:

GPU Performance Optimization Learning Path