A Few Worrisome Regressions Appear In Ubuntu 15.04 vs. 15.10 Performance

With Ubuntu 15.10 set to be released later this month, I've started preparing for a variety of Linux performance comparisons involving the Wily Werewolf. This morning I ran some Ubuntu 15.04 vs. 15.10 benchmarks on one of my frequent test beds and it's revealed a few significant changes in some of the benchmarks.

From the Intel Xeon E5-2687W v3 Haswell system with AMD FirePro V7900 (Cayman) graphics, I ran some Ubuntu 15.04 x86_64 vs. Ubuntu 15.10 x86_64 (the latest daily ISO) benchmarks on it. For the most part, the performance was close to the same as would largely be expected given just a six-month time span and the hardware having been already well supported by Linux for a while. Both installs were done cleanly and testing the out-of-the-box configuration.

However, in some of the benchmarks there is some rather significant performance changes:
Some of the scientific OpenMP workloads are much slower now with Ubuntu 15.10... Perhaps due to the GCC 5 upgrade? However, on other systems/distributions I don't recall ever running into such a regression for these tests on the latest GNU Compiler Collection.
Compile times are slower.
Some other tests are also clearly slower.

I'm still investigating these issues and running Ubuntu 15.04 vs. 15.10 performance benchmarks on other systems, but just sharing these preliminary results to make this weekend performance investigation worthwhile and seeing if any Phoronix readers have encountered similar slowdowns if moving to Ubuntu Wily. All of the initial performance results in full for this system on Ubuntu 15.04 vs. 15.10 can be found via this OpenBenchmarking.org result file.