AOCC 3.1 Compiler Performance Against Clang 12, GCC 11 On AMD EPYC

[phoronix]

Phoronix: AOCC 3.1 Compiler Performance Against Clang 12, GCC 11 On AMD EPYC

Following the recent benchmarks seeing how AMD's new AOCC 3.1 compiler has brought some performance improvements over the prior AOCC 3.0 release that introduced initial Zen 3 optimizations, here are some benchmarks looking at how that latest AMD Optimizing C/C++ Compiler performance compares to the upstream LLVM Clang 12 compiler for which it is based as well as against GCC 11 as the latest GNU compiler release that remains common to Linux systems.

https://www.phoronix.com/vr.php?view=30395

[oleid]

Clang 12 maintained about 1% faster performance on the geo mean over GCC 11

No surprise, 12 > 11.

But jokes aside, is this only a case of GCC lacking optimizations for Zen or has clang better optimizations/CPU cost/performance heuristics nowadays?

[onlyLinuxLuvUBack]

As a reference bar, could you just throw a new drive on an do clear os to see out of box what a milan does with whatever gcc version they are using there.

[DanglingPointer]

hmmm... perhaps I should try out defaulting CC to AOCC.

Can linux be built with AOCC ? Has anyone tried it?

[CochainComplex]

hmmm... perhaps I should try out defaulting CC to AOCC.

Can linux be built with AOCC ? Has anyone tried it?

You can build the Kernel with it.

if you are on debian based distro:

Install AOCC should bin installed to /opt/AMD/a.../

source /opt/AMD/.../setenv.sh
make LLVM=1 LLVM_IAS=1 menuconfig
make LLVM=1 LLVM_IAS=1 -j N deb-pkg

with setenv.sh it is quite a drop-in replacement for clang. As long something can be built with LLVM it can be built with AOCC3.1
(I have had issues with AOCC3.0 but not with 3.1 so far - znver2)

Michael Thankyou for this Benchmark. It supports my observations.