Performance/MemShrink/Meetings/2011-11-08

We had a free ranging discussion about what MemShrink should look at and what our goals are. njn has a more polished summary of the discussion on his blog.

• Nick is wondering about the big picture. What sort of bugs do we want to fix? What do we do about the continually increasing number of MemShrink bugs? Should we keep fighting fires, or focus on digging into larger scale projects?
• What is our goal? How can we improve things for users? Smaller slimming efforts (like 50mb) probably make more difference on mobile than on desktop. Could we focus efforts on addon and web developers?
• Memory usage degrading over a long period of time?
• Should we work on coming up with benchmarks? A way to measure progress? Some kind of graph to measure things? Set a target, even if it is imperfect? How much memory do we use after pushing a day's worth of pages?
• Is memory usage a primary or a secondary characteristic? Things like pause times getting long can be more bothersome than the browser using an extra 100mb of memory.
• One metric: In 32-bit Windows, how many copies of Gmail can we open?
• If we have some kind of accurate way to simulate a day's worth of browsing, what do we want to measure? What is bad? Heap going beyond a certain size rather than average memory? Do we want to measure all pauses? The last X pauses?
• On mobile, a goal is to reduce peak memory usage to avoid being killed, whereas on desktop a goal is to ensure performance doesn't degrade over time.