Re: [pipe] d60337eff1: phoronix-test-suite.noise-level.0.activity_level 144.0% improvement

On Thu, Nov 7, 2019 at 1:03 AM lkp report check [off-list ref] wrote:

FYI, we noticed a 144.0% improvement of phoronix-test-suite.noise-level.0.activity_level due to commit:

commit: d60337eff18a3c587832ab8053a567f1da9710d2 ("[RFC PATCH 04/11] pipe: Use head and tail pointers for the ring, not cursor and length [ver #3]")

That sounds nice, but is odd. That commit really shouldn't change
anything noticeable. David, any idea?

               Linus

Linus Torvalds [off-list ref] wrote:

quoted

FYI, we noticed a 144.0% improvement of phoronix-test-suite.noise-level.0.activity_level due to commit:

commit: d60337eff18a3c587832ab8053a567f1da9710d2 ("[RFC PATCH 04/11] pipe: Use head and tail pointers for the ring, not cursor and length [ver #3]")

That sounds nice, but is odd. That commit really shouldn't change
anything noticeable. David, any idea?

Yeah, it does sound odd - the only thing that springs particularly to mind is
that maybe it's an effect of one of the preceding patches that affects the
waitqueue stuff.

TBH, I'm not sure what the Phoronix test suite is actually testing from the
report.  "noise-level-1.1.0" seems more a reporting script than an actual
test.

David