AMD Piledriver/Trinity A10-5800K Compiler Tuning
Similar to C-Ray, Smallpt saw improvements out of the instruction sets introduced in Barcelona and the first-generation Bulldozer processors, but bdver2 wasn't meaningful for the AMD A10-5800K Piledriver APU.
The Opus audio encoding and Nero2D benchmarks also didn't benefit from targeting "bdver2" with GCC 4.7.2 on the AMD A10-5800K compared to the "bdver1" micro-architecture target.
Having GCC 4.7.2 take advantage of the FMA3, BMI, TBM, and F16C instruction sets by using the -march=bdver2 flag for the AMD A10-5800K APU wasn't of any measurable benefit for the open-source Linux benchmarks tested that ranged from audio encoding to ray-tracing to image manipulation and molecular biology. The only potential improvement spotted with bdver2's additions were image blurring within GraphicsMagick.
When AMD rolls out their new Bulldozer 2 Vishera CPUs it will be tested again from the GCC 4.8 code-base that's presently under development to look for any further compiler optimizations. There's also LLVM/Clang compiler testing to do from AMD's Piledriver processor cores.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.