Linus Torvalds' Latest Commentary Against -O3'ing The Linux Kernel

[phoronix]

Phoronix: Linus Torvalds' Latest Commentary Against -O3'ing The Linux Kernel

This shouldn't be too surprising considering some of Linus Torvalds past commentary about compiler optimizations and bad experiences long ago with GCC, but Linus Torvalds is not interested in seeing a tunable Kconfig option for using the -O3 compiler optimization level for building the Linux kernel without substantial justification...

https://www.phoronix.com/scan.php?page=news_item&px=Linus-Against-O3-Kernel

[Jannik2099]

Yes, let's drop any debate of modernizing anything in linux, ever, because 20 years ago things weren't worth it.

It's time for Torvalds to either snap back to reality, and accept that things change over time, or step down and put someone in his place who is actually in touch with todays technical situation.

[birdie]

Yes, let's drop any debate of modernizing anything in linux, ever, because 20 years ago things weren't worth it.

It's time for Torvalds to either snap back to reality, and accept that things change over time, or step down and put someone in his place who is actually in touch with todays technical situation.

Linus has asked for solid data, you reply he's a Luddite. That kinda sums up the whole debate about -O3.

Oh, and the option to compile it using -O3 is all yours. Go for it. Leave us, apprehensive mortals, in safety and moderation. I don't care about point zero performance improvements but I surely want my kernel to be as reliable as humanly possible.

[robimarko]

Yes, let's drop any debate of modernizing anything in linux, ever, because 20 years ago things weren't worth it.

It's time for Torvalds to either snap back to reality, and accept that things change over time, or step down and put someone in his place who is actually in touch with todays technical situation.

I dont get what is wrong in Linus-es reply?
He wants data that to back up the commit and has explained why he is against it in general.
Anyway, nothing is preventing anybody from compiling with -O3

[Jannik2099]

not using O3 doesn't improve reliability, it hurts it. You're just hoping that the compiler will not blow up your shitty code at O2, but there's NEVER a guarantee for that.

Linux has a serious code rot problem, and sidestepping modern compiler techniques / blaming them is not helpful.

[birdie]

not using O3 doesn't improve reliability, it hurts it. You're just hoping that the compiler will not blow up your shitty code at O2, but there's NEVER a guarantee for that.

Linux has a serious code rot problem, and sidestepping modern compiler techniques / blaming them is not helpful.

[CochainComplex]

Is Linus requesting a phoronix article?
"Those explanations had better be more than "let people try". They should have in-depth actual performance numbers for a real load, not some made-up "bigger is better" logic."

[jeisom]

I think they should add it, as experimental, and direct problems to the compiler’s bug trackers if it is not reproducible with -O2 or supply a patch that fixes said bug. Otherwise “don’t tell us”

[erniv2]

not using O3 doesn't improve reliability, it hurts it. You're just hoping that the compiler will not blow up your shitty code at O2, but there's NEVER a guarantee for that.

Linux has a serious code rot problem, and sidestepping modern compiler techniques / blaming them is not helpful.

Nope GCC had/has serious compiler issues with -ffast-math wich is part of -O3 and simply breaks standards, wich is why everyone and his mother builds stuff with -O2

[Jannik2099]

I dont get what is wrong in Linus-es reply?
He wants data that to back up the commit and has explained why he is against it in general.
Anyway, nothing is preventing anybody from compiling with -O3

It wasn't phrased very constructively, to say the least.

O3 is not just about performance. Static analysis tooling for the kernel is still in its infancy, so O3 will most likely discover a fair share of bugs that have been lingering in the code base for god knows how long.