LLVM/Clang Performance On Intel Sandy Bridge
Phoronix: LLVM/Clang Performance On Intel Sandy Bridge
Lately there's been a lot of compiler benchmarks on Phoronix, particularly when looking at the performance on AMD's Bulldozer architecture and their FX-8150 system. However, here's some more compiler benchmarks, but this time under Intel Sandy Bridge...
It looks like smallpt is multithreaded on GCC and singlethreaded on Clang.
Yep. GCC supports OpenMP, Clang doesn't.
Originally Posted by Ex-Cyber
Where is EKOPath?
In Phoronix once much talked about thing was EKOPath compiler being open-sourced. Now where is it? Why don't Michel adding EKOPath to this compiler benchmarks? Why is that once what is so talked about, now is cold dust?
The nightly build on their website hasn't been updated since July... and that build has a broken installer.
Originally Posted by Tiger_Coder
I'm not seeing it written on the benchmark page (it should be): when these tests are compiled, is it just the app being tested that's compiled with the named compiler, or all dependencies as well? (I don't think Clang yet builds a working Linux kernel, but if it does, then it'd be noteworthy to know if the whole OS was compiled with each compiler or not -- specific versions and all.)
It's just the app, given that the Phoronix tests don't even try different optimization levels other than that which was set by default I'd say there's zero chance you'll see dependencies compiled using the corresponding compiler.
Originally Posted by elanthis
I built it from github.com/pathscale/path64-suite a while back but the code it generated was awfully slow. That said I don't know how representative this github version is of the proprietary compiler in terms of optimizations, maybe there's still alot more work to do.
Originally Posted by Michael