“Obviously there will be differences and caveats with how we support each Having met that challenge and moved the PC-focused product out of beta, the next step for the RAD team working on Telemetry is to take final versions of the toolkit to console: “As with any multiplatform programming task, each platform presents its own unique challenges,” confirms Hook. So the largest challenge has just been making all this data gather and visualise quickly, when there isn’t one ‘right’ way to use it.” “One customer will want tens of thousands of profile zones, another is dumping GBs of logging messages, while another will send millions of samples of user data. “The biggest challenge has been scalability,” he admits. A SENSE OF SCALEĭeveloping the tech hasn’t been without challenges, of course, but those hurdles are the very things that have defined the development of Telemetry into what it is today, according to Hook.
“Many customers find significant problems within their first hour of using Telemetry often problems they were unaware of or had an intuitive feel about but no real evidence to help track it down.”įurthermore, Telemetry uses a client/server architecture, so information can stream to a single repository for analysis by others. In fact, visualisation is pitched as one of Telemetry’s key strengths.Īccording to Hook, gaining a clear visualisation of data is paramount to understanding what a given game is really doing: “Having code unfold into visual representations of execution really helps programmers identify problem areas in terms of performance, thread interactions, and memory usage. Such a function enables developers to see, for example, dumped out custom game data, along with thread locking data, OS context switches, memory allocation and function times, all along the same timeline. Practically speaking this means you can find spikes in performance, not just areas that ‘on average’ took a long time but which may actually be fast frame-to-frame.” ABOUT TIMEĪnother string to RAD’s profiling bow is its ability to display all gathered data on the same timeline, allowing users to reconcile vastly different types of data. Telemetry, insists Hook, also promises to outperform other profiling systems by showing information ‘over time’: “This is a crucial advantage over seeing an aggregate summary of events for an entire duration. “And it also means we can take many more readings-per-second than a sampling profiler.”
“This gives the programmer very focused information about what’s taking a long time, where, and even why based on surrounding context,” explains Telemetry’s chief developer Brian Hook.
0 kommentar(er)
