Has anyone tested the difference between Mozilla's Firefox build and a local fully-optimized for your machine build?
Or something similar with another piece of open source software.
I wonder how much the "Gentoo-model" of compiling everything yourself affects everyday performance.
One big gap is that release builds of Firefox tend to have had profile-guided optimization applied, among other things, so you need to make sure you take steps like that correctly or your custom build will end up slower.
It's probably possible to produce a custom build for your machine that runs faster, whether by using custom instruction sets, or feeding different data to the optimizer when it's doing PGO. However, the risk you run there is that this could introduce bugs - Firefox and Chrome both regularly run into bugs caused by the optimizer or linker, and in some cases the fix is to stop the optimizer/linker from fiddling with that code.
Everything on my Gentoo system runs faster than in the binary ebuild equivalent.
That being said, things like choosing the wrong preemption mode can introduce latency problems, etc. Most users experience higher throughput and higher latency when comparing Windows to Gentoo. I believe the higher user interface latency is the general case when going from Windows to Linux.
On initial blush I am so far impressed. Sync pulled over all my add-ons and Firefox info and it is definitely quicker at rendering large tables (my one notable speed complaint with Firefox vs. Chrome). With everything working via Sync I'm going to give it a fair shot.
From the FAQ: Can I see your mozconfig/build environment/configuration files?
The answer is: No, this is my baby.
While I'm quite happy with FF as my primary browser now that historical memory issues are mostly resolved, I'd be interested in building a FF release from source and having build-time options. The Pale Moon source code being distributed via "ge.tt" seems designed to obfuscate in a world of git. I'll pass.
A few people have asked about getting a detailed run-down of my build
clearly not always with good intentions.
How could you copy a build process with bad intentions? The code doesn't belong to him/her and him/her aren't attempting to profit from Pale Moon. I'm genuinely puzzled.
As a web developer with some extensive Firefox familiarity, please, please, please do not use Pale Moon. Please do not suggest to others that they use Pale Moon.
It has bugs that other versions of Firefox do not. Because he does not openly share his build configuration/etc it's not possible for you to fix those bugs.
I've had to spend a lot of time trying to troubleshoot issues Pale Moon users had with my HTML5 applications, and in every case the solution was for them to use a properly built and maintained browser. :(
Is there any sort of proof of this "optimization" actually contributing to performance? The state of compiler optimization being what it is, I find it hard to imagine flag tweaking from the default -O3 or whatever can make an actual significant difference. The secretiveness of his build seems designed to obfuscate third-party replication of his results.
Seems more a mouthpiece for the this fellow's self-aggrandizement than anything else.
reply