Game Tools Tune-Up: Optimize Your Pipeline Through Usability
May 6, 2009 Page 1 of 4
[How do you perfect your custom-created game tools? Sega and LucasArts veteran and Robotic Arm Software founder Goodman discusses tips and tactics for methodology when you roll your own tools.]
In game development, it is often some of the most technical (programmers) making tools for the some of the least technical (everyone else). It's no wonder that many of those "least technical" struggle with the tools that are handed to them.
The tools themselves are designed by the people who understand the underlying systems, but not necessarily those who will be using them.
Usability determines how easy software is to use, how quickly the desired results can be achieved by its users and how error-free those results are. Techniques to test and improve usability have been successfully used in other types of software development, and are slowly making their way into ours.
There are probably tools in your pipeline that could benefit from usability techniques, but as a developer, you'll want to know which tools would benefit the most before investing in them.
In this article, I'll discuss how to find the bottlenecks holding up your production, measure the usability of your tools using proven techniques and streamline your entire development process.
Mapping the Pipeline
The first step to solve any problem is understanding what the problem is and why it's occurring. We need to know where in the development pipeline our process is being stalled. To do this, we must map the problem-space.
There are many ways to view the development pipeline. Some may view it as a checklist of assets and responsibilities, but this doesn't represent the dependencies between the developers.
Those dependencies create an environment where developers are wasting time waiting for someone else to finish a task, which requires managerial effort.
To understand these dependencies, it is more convenient to look at the pipeline as a graph showing workflow between various resources and what each produces. Consider the example workflow for character creation in the below figure.
In this illustration, the arrows point to the dependent task -- the task that depends on the one the arrow is coming from. So, modeling depends on character design, texturing and rigging depend on modeling, and so on.
This represents a very small, incomplete piece of a typical game development pipeline. Your actual pipeline will be much more complex, including more areas of development (level design, UI, gameplay programming, and so on), resources performing multiple processes, and multiple dependencies for a single task. The stages that hold up the development pipeline should be looked at for optimization through tool improvement, as well as other means.
Assuming that each phase of character development needs to be complete before continuing to the next, the first character could be fully implemented in 24 days. We get this by adding up the number of days in the longest pathway (the primary development pipeline), from character design to completion.
Texturing and effects are left out of the calculation, since their pathways diverge from the main, and the time for completing those pathways are less than or equivalent to the primary one.
Character Design + Modeling + Rigging + Animation + Behavior = 24 Days
That doesn't mean that it takes 24 calendar days to complete every character. This is an assembly line; four other characters are being worked on while the current one is being finished -- one at each stage in the pipeline. So how many calendar days does it take to complete a new character after the first? Exactly the time it takes for the longest stage -- 10 days.
Ten characters completed in just under six calendar months may sound great. Of course, if just the animation time were cut in half, we could double the character output by creating a new character every five days.
In any assembly line, efficiency is achieved by breaking down tasks into component parts so that every stage of manufacturing takes an equal amount of time. That's also the most efficient model for developing game assets.
Unfortunately, each stage in the pipeline requires unique skills, and it is often difficult to predict the exact timing for each stage. So, work backs up behind one developer, while another sits idly by waiting for the first one to finish.
There are many tactics for getting around this issue, all of which are employed by game companies in one extent or another. Managerial overhead and individual efficiency is often traded for development pipeline bandwidth.
Still, if efficiency can be increased in the areas of the pipeline that cause slowdown, pipeline bandwidth will also increase, with little overhead. One method for achieving this efficiency is through tool improvements.
Page 1 of 4