The boosts when a single core can run over clocked are less likely to be easily measured, but you can measure thermal throttling very simply with the thermal logging of pmset. With 8 cores on many MacBook Pro and dozens of cores on the iMac Pro - you're boiling a ton of complexity down to one number. Since the code and each core of a CPU can and will change hundreds of times a second based on ephemeral load factors, power optimizations that consider what's visible on the screen, what network data arrives, the idea that a modern CPU even has one "common" clock rate at any one point in time seems to vastly over-simplify reality. CPU interrupts on macOS are shaped in intervals of 150 ms and much of this detail is public from WWDC 2013 and later on power management, App Nap ( Session 209 in particular is both good and approachable) and battery life optimizations on macOS. The API is otherwise unlikely to be of interest to Mozilla developers.I have to think this is a bit of an X Y question in that "What are you going to do once you get this number?" and want to answer that directly, but let's dive a bit into what you're trying to measure. Unfortunately, the Gecko profiler takes 1000 samples per second on desktop and is CPU intensive and so is likely to skew the RAPL estimates significantly, so the API integration was removed. At one point the Gecko Profiler used this API on Windows to implement experimental package power estimates. Version 3.0 (available on Mac and Windows, but not on Linux) also exposes an API from which the same measurements can be extracted programmatically. (An energia dashboard can be seen here please note that the data has not been updated since early 2014.) This feature has been used in energia, Roberto Vitillo's tool for systematically measuring differential power usage between different browsers. Intel Power Gadget can also log these results to a file. Specificially, the temperature is a proxy measurement that is affected by processor power consumption, rather than one that affects it, which makes it even less useful than most proxy measurements. This is interesting, but again not useful for power profiling purposes.
0 Comments
Leave a Reply. |