Putting a stake in the ground
Performance matters more than ever in the browser space, and even as the Web grows up and gets bigger we need to continue to push harder on getting smaller and faster. New tools and frameworks continue to evolve and provide more fodder for the performance team to chew on, and we’re going to do our part even as we continue to push on innovation in the browser. As a result, the Firefox team is going to put a really big stake in the ground on performance, with the bulk of the focus on cleaning up our own house.
We’re starting with three straightforward goals:
- Cut app overhead on pageload by 50% against Firefox 2. Measuring “pure” Gecko performance against total Firefox performance is a hard problem, but one well worth solving. Pageload is the sum of a lot of parts, the more we can fix those parts, the better.
- Improve Txul (new window opening) performance by 20% against Firefox 2.
- Improve Ts (overall app startup time) performance by 10% against Firefox 2.
These goals feel right, and should be within our reach. There’s a lot of hard work involved between now and these goals, but I believe we have the people and, thanks to the perf team (especially Robert Sayre and his awesome work with dtrace), the tools to start moving faster here.
Wow. It’s great that you’re setting these goals aggressively and publicly. I can’t wait to see some of the results.
September 28th, 2007 at 7:46 pmWill these performance improvements only apply to new computers, or will older systems be tested as well?
September 28th, 2007 at 10:50 pmAlan: Performance improvements are usually not computer specific (with a few exceptions like hardware accelerated graphics, or use of specialized instructions, neither of which seem to apply much here).
September 28th, 2007 at 11:20 pmAre there similar goals for Gecko? It seems strange to have a goal for Tp overhead (due to Firefox chrome) without also having a goal for overall Tp or chromeless Tp.
September 29th, 2007 at 7:34 pmHi,
I think Performance improvements could be computer specific. Indeed, most of the time, you have to trade off between memory and processor. And on low-ends computers, memory is not so cheap
October 2nd, 2007 at 6:37 pmGood luck!
October 3rd, 2007 at 10:13 am