Link-Time Optimization To Speed Up The Linux Kernel

[RealNC]

Isn't GCC-specific behavior bad?

No. It's designed that way. The GCC-specific code isn't an accident or oversight.

Shouldn't it be ANSI/ISO-compatible?

Nope. Linux is specifically targetting GCC as it's compiler.

[uid313]

No. It's designed that way. The GCC-specific code isn't an accident or oversight.

Maybe those GCC-specific extensions should be sent to ISO for standarization.

[curaga]

Yeah, and be standardized in C25 in 2025...

[uid313]

Yeah, and be standardized in C25 in 2025...

Better late than never?

[BO$$]

I heard a few years ago that Microsoft c++ compiler kicks g++ compiler ass. Is this statement still correct? Or was it incorrect even back then?

[elanthis]

i'd like to see someone try compile linux with;
* gcc
* microsoft compiler
* llvm / clang / dragonegg
* pcc, tcc
* intel c compiler
* open64

hahahahahahahhahahaha good fucking luck, buddy.

(For more context, Microsoft does not support or enhance their C compiler anymore, which hasn't been able to compile many modern C codebases in ages. It's stuck at the 23-year-old C89 standard and is receiving zero further development support. Microsoft tells anyone who runs into problems to compile as C++ or to go find another compiler vendor. I'd be mildly surprised if Microsoft's compiler can even successfully parse more than 25% of the headers in Linux or even a single source file.)

Maybe those GCC-specific extensions should be sent to ISO for standarization.

Many of the extensions are just not really general-C kinds of things. http://gcc.gnu.org/onlinedocs/gcc/C-Extensions.html - if you look through that list, and especially at the ones the Linux kernel depends on, they're not really applicable in the kind of general sense that C prefers. Some of them place extra restrictions on the compiler's code generator which would make them unusable on "small system" C compilers, others are very tightly bound to the linker and require an ELF-like system, some are hardware or platform specific, some are specific to GCC's internals, and yet others are just of dubious nature that may have already been superceded by C11 or C++ equivalents. (Note that many of those extensions actually are just C99 or C11 features, with the "extension" part being that GCC accepts them even when compiling in older standards modes.)

[uid313]

Many of the extensions are just not really general-C kinds of things. http://gcc.gnu.org/onlinedocs/gcc/C-Extensions.html - if you look through that list, and especially at the ones the Linux kernel depends on, they're not really applicable in the kind of general sense that C prefers. Some of them place extra restrictions on the compiler's code generator which would make them unusable on "small system" C compilers, others are very tightly bound to the linker and require an ELF-like system, some are hardware or platform specific, some are specific to GCC's internals, and yet others are just of dubious nature that may have already been superceded by C11 or C++ equivalents. (Note that many of those extensions actually are just C99 or C11 features, with the "extension" part being that GCC accepts them even when compiling in older standards modes.)

Does the Linux kernel take advantage of C99 and C11?

[RealNC]

Does the Linux kernel take advantage of C99 and C11?

It doesn't care if it's in C89, C99 or C11, as long as it's in GCC.

[uid313]

It doesn't care if it's in C89, C99 or C11, as long as it's in GCC.

Alright.
But does it use any of the new C99 or C11 features?

[RealNC]

Alright.
But does it use any of the new C99 or C11 features?

I've seen 'inline' in there, at least.