clang: Build with PGO and ThinLTO

[ptr1337]

Benchmarks:

@ptr1337 , 9950X

PGO ThinLTO Clang, modprobed.db kernel, ThinLTO enabled
________________________________________________________
Executed in  175.51 secs    fish           external
   usr time   39.54 mins    0.00 micros   39.54 mins
   sys time    2.74 mins  133.00 micros    2.74 mins

GCC built Clang from znver4 repository
________________________________________________________
Executed in  195.74 secs    fish           external
   usr time   49.87 mins    0.00 micros   49.87 mins
   sys time    2.78 mins  142.00 micros    2.78 mins

Yuby, 5800X

CachyOS kernel + ThinLTO + Nvidia + modprobed

clang 18.1.8-2 8:21
clang 18.1.8-3 6:56

@1Naim Mesa ThinLTO Compilation

Optimized clang + llvm
Executed in   16.77 mins    fish           external
   usr time  165.51 mins    0.00 micros  165.51 mins
   sys time    4.39 mins  658.00 micros    4.39 mins

Unchanged clang + llvm
Executed in   20.26 mins    fish           external
   usr time  212.03 mins    0.42 millis  212.03 mins
   sys time    4.69 mins    1.43 millis    4.69 mins

[ptr1337]

@1Naim

sudo pacman -U https://archive.cachyos.org/clang-pgo-thinlto/clang-18.1.8-3-x86_64_v3.pkg.tar.zst

[1Naim]

save my cpu....

[ptr1337]

YUBY from Discord had following results:

CachyOS kernel + ThinLTO + Nvidia + modprobed

clang 18.1.8-2 8:21
clang 18.1.8-3 6:56

[ptr1337]

Made a kernel compilation, with modprobed db + ThinLTO enabled.
The packaging time and pahole is there also included, but its quite faster:

PGO ThinLTO Clang, modprobed.db kernel, ThinLTO enabled
________________________________________________________
Executed in  175.51 secs    fish           external
   usr time   39.54 mins    0.00 micros   39.54 mins
   sys time    2.74 mins  133.00 micros    2.74 mins

GCC built Clang from znver4 repository
________________________________________________________
Executed in  195.74 secs    fish           external
   usr time   49.87 mins    0.00 micros   49.87 mins
   sys time    2.78 mins  142.00 micros    2.78 mins

Means a 10% improvement for the whole kernel. This seems like a major win.

[1Naim]

Building mesa with clang + ThinLTO

Optimized clang + llvm
Executed in   16.77 mins    fish           external
   usr time  165.51 mins    0.00 micros  165.51 mins
   sys time    4.39 mins  658.00 micros    4.39 mins

Unchanged clang + llvm
Executed in   20.26 mins    fish           external
   usr time  212.03 mins    0.42 millis  212.03 mins
   sys time    4.69 mins    1.43 millis    4.69 mins