[PATCH] [lnt] Correct delta and percent-change of geometric mean.

[Daniel Sanders via llvm-commits]

Thanks. Committed in r245425.

At this point the values calculated from the geometric mean are calculated correctly but the geometric mean value itself is still misleading without the other patch ('[lnt] Values for failed tests should be treated as missing') since values from failed tests will be included in the mean. That patch also deals with this problem on individual values. What made me notice it was an apparent 200,000% regression in a recent comparison that turned out to be sphereflake failing in a couple seconds in the previous run and failing after 5000s in the current run.
 
> -----Original Message-----
> From: Chris Matthews [mailto:chris.matthews at apple.com]
> Sent: 18 August 2015 16:48
> To: Daniel Sanders
> Cc: llvm-commits at lists.llvm.org
> Subject: Re: [PATCH] [lnt] Correct delta and percent-change of geometric
> mean.
> 
> LGTM. Thanks!
> 
> > On Aug 18, 2015, at 7:13 AM, Daniel Sanders via llvm-commits <llvm-
> commits at lists.llvm.org> wrote:
> >
> > <0002-lnt-Correct-delta-and-percent-change-of-geometric-me.patch>