Firefox Is Now Built With Clang+LTO Everywhere, Sizable Performance Wins For Linux

[phoronix]

Phoronix: Firefox Is Now Built With Clang+LTO Everywhere, Sizable Performance Wins For Linux

Firefox nightly builds are now built with the LLVM Clang compiler on all major platforms and the Linux build in particular is also now utilizing PGO optimizations too. Faster Firefox is coming thanks to this compiler work...

http://www.phoronix.com/scan.php?page=news_item&px=Firefox-Clang-LTO-All-Platforms

[szymon_g]

Is Firefox for Windows run on wine on Linux still faster than native Linux version?

[ernstp]

You can check with about:buildconfig

[Michael_S]

I still don't blame them for focusing on Windows first, when the great majority of their users are on Windows. But it's nice to see this happen.

Is Firefox for Windows run on wine on Linux still faster than native Linux version?

Oh yikes, that was true at one point?

[_Alex_]

It'd be better if we had performance metrics comparisons to GCC 8.2 / GCC 9(pre) builds instead of GCC 6.

[pyler]

Clang enables cross-language inlining with Rust which is another huge performance step.

[RussianNeuroMancer]

Even Chromium is more near to hardware video decoding than Firefox.

[Brane215]

My specimen was compiled with gcc-8.2.0 and non-LTO ( Gentoo)

This is good moment to try LTO on newest version.

BTW, is some good shortcut for checkiung the validity of LTO compiles ?

It took me quite some time to loca<te LTO breakage in ffmpeg that mnifestedd as SEGV during play of outube content iwth M4A for audio.

After ffmpeg recompile with non-LTO flags, such content plays normally...

[xpris]

OpenMandriva build Firefox with Clang for while. I tested some time ago on OpenMandriva LX3, Firefox from the system repo distribution built with Clang and the prebuild binary version from mozilla.com with GCC. As it turned out, the version from Clang worked much better and noticeably faster. But these were only feelings of use. I did not make benchmarks.

[discordian]

BTW, is some good shortcut for checkiung the validity of LTO compiles ?

It took me quite some time to loca<te LTO breakage in ffmpeg that mnifestedd as SEGV during play of outube content iwth M4A for audio.

After ffmpeg recompile with non-LTO flags, such content plays normally...

Short answer: No

There are different reasons for breakage, broken code that just happens to work aslong you compile it separately and bugs in the compilers/linkers LTO implemenation (different optimization flags might be another reason, but I would count not warning about those effects a compiler bug).
broken code might give you warnings (enable -Wall), but compiler bugs are just that.

You best bet would be using an elaborate test-suite, wherever possible