Benchmarking AMD's New AOCC Compiler For Ryzen

[sdack]

They should open this project, if it gives no extra performance (with a single exception). ...

They should close the project and find something else to do. *lol*

The AOCC compiler doesn't look like it was worth the effort.

[mlau]

the only interesting result seem to be GCC being relatively bad optimizing for ryzen whereas Clang/AMD compiler does lot better in certain type of jobs

Depends on the mtune setting, really. Gcc's znver1 model is largely copied from bdver4 (Carrizo/Excavator), and that sucks on Zen. -march=znver1 -mtune=haswell gives much better results. Noone has submitted a more accurate zen scheduling/cost model yet.

[chuckula]

the only interesting result seem to be GCC being relatively bad optimizing for ryzen whereas Clang/AMD compiler does lot better in certain type of jobs

That's not really a valid conclusion to draw from these results without having an appropriate baseline of a bunch of other processors. For example, LLVM tends to do better at some workloads for pretty much all processors, while GCC tends to do better at other workloads for pretty much all processors. The RyZen hardware aspect of it isn't that important.

[chuckula]

It's not much of a surprise, RyZen performs about as well as it's going to in Linux anyway and the lack of real number crunching hardware (only 1/2 speed support for AVX and no AVX-512 at all) can't just be magically optimized away in a compiler.

The good news is that you have no need to feel compelled to get AMD's proprietary compiler and jump through the burning hoops to get it enabled just to run RyZen.

[jrch2k8]

the only interesting result seem to be GCC being relatively bad optimizing for ryzen whereas Clang/AMD compiler does lot better in certain type of jobs

[tildearrow]

They should open this project, if it gives no extra performance (with a single exception).

Also I think I'll be honest now.
What happened to the x264 test? Does it not work? (if so I might have the fix)
I really doubt anyone would like to convert videos for watching on an ancient DV camera.

[phoronix]

Phoronix: Benchmarking AMD's New AOCC Compiler For Ryzen

This week AMD released AOCC 1.0, the AMD Optimizing C/C++ Compiler, with initial support for Ryzen/Zen CPUs. In this article are our first benchmarks of the AOCC compiler compared to the GCC 6/7/8 and LLVM Clang 4/5 compilers.

http://www.phoronix.com/vr.php?view=24662