Compiling The Linux Kernel With LLVM's Clang Matured In 2022

[phoronix]

Phoronix: Compiling The Linux Kernel With LLVM's Clang Matured In 2022

Over the past few years it's become possible to compile the mainline Linux kernel with LLVM/Clang compared to the long-standing dependence on using the GCC compiler. While it's been possible for 3+ years to use the mainline Linux kernel and mainline Clang for building a working x86_64 and AArch64 kernel, the process and support continues to mature...

https://www.phoronix.com/news/Clang-Built-Linux-2022-Recap

[dimko]

Compiled kernel for all of my life using GCC(gentoo fetishist), and last kernel or two with LLVM, did not notice difference, as user, what so ever.

[Quackdoc]

I just want faster clean compile times,

[dimko]

I just want faster clean compile times,

I dont care for fast compiles. Or clean ones.
I care for speed and stability of resulting kernel. I have 16 coresx2(hyperthreading). I throw hardware at the problem.
So far did not notice any difference between GCC and CLANG

[partcyborg]

Android kernels have only supported clang for the past few major versions. Glad to see support in the mainline improve.

[Quackdoc]

I dont care for fast compiles. Or clean ones.
I care for speed and stability of resulting kernel. I have 16 coresx2(hyperthreading). I throw hardware at the problem.
So far did not notice any difference between GCC and CLANG

I mostly compile on my 2600 that does dual duty as my main work PC, that and I do clean builds fairly often, mostly because im an idiot and keep forgetting though. but regardless, clean compile increases are what I mainly care about right now.

[ms178]

Good job, Nathan! As a user from the first hour of being able to build the Kernel with Clang, it has been quite a journey. And as I throw in Polly into the mix, this makes it even more challanging. Thanks to you, Nick and all the LLVM/Clang devs for making this a reality.

[Grinch]

Have there been any recent'ish benchmark comparisons between a GCC compiled kernel and a Clang/LLVM compiled one ?

[CHECK_12345]

⬤ Specs

CPU: Ryzen 7 1700, 2.5GHz at 0.768V and 1.55GHz at 0.612V, VCORE SOC set to 0.684V
GPU: XFX RX 580 4GB being able to use 1150MHz core clock at 0.865V and 1750 MHz memory clock at 0.865V
Storage: 2TB 7200RPM 3.5" HDD
Boot: Triple-boot: Windows 10, Ubuntu, Arch Linux

⬤ Still currently using GCC

Still compiling my minimized Linux Kernels with GCC and with initrd disabled, all things built-in, had to include firmware files from /lib/firmware which increased size of kernel (AMD GPU firmware files to make my XFX RX 580 4GB GPU work and USB wifi firmware file to make my USB wifi work) and got total size of kernel to 7.7MB.

⬤ Currently using "-Ofast", "-march=native", "-pipe" by editing the makefile when compiling Linux Kernel

-Ofast increases size of kernel
-march=native decreases size of kernel

[CochainComplex]

Kernel built works also quite fine with AOCC 4 which is based on LLVM 14.0.6