A Significant Linux 4.7 Kernel Performance Regression Has Now Been Resolved

[theriddick]

Did they fix the radeon performance regression also? Or was that just a broken powerplay thing?

[W.Irrkopf]

It seems to be mostly specific to the SPEC Java tests. Unfortunately, last check, the SPEC tests were still rather strict about things that I don't support them in PTS aside from SPECViewPerf.

And part of the discussion thread emphazises that these tests are horribly written: https://lkml.org/lkml/2016/6/22/133

[Michael]

Any chance that your benchmarks Michael could have helped identify the performance regression? If not, why not?

It seems to be mostly specific to the SPEC Java tests. Unfortunately, last check, the SPEC tests were still rather strict about things that I don't support them in PTS aside from SPECViewPerf.

[pierce]

Any chance that your benchmarks Michael could have helped identify the performance regression? If not, why not?

[phoronix]

Phoronix: A Significant Linux 4.7 Kernel Performance Regression Has Now Been Resolved

There was a 30~40% drop in some of the SPEC Java benchmarks when using the Linux 4.7 development code, but fortunately this regression has now been discovered and addressed...

http://www.phoronix.com/scan.php?page=news_item&px=Linux-4.7-SPEC-Java-Regression