(Re-posting to dev-platform -- I had replied to Nick privately by mistake. On Fri, Feb 1, 2013 at 9:46 PM, Nicholas Nethercote <n.netherc...@gmail.com>wrote:
> On Sat, Feb 2, 2013 at 9:05 AM, Ehsan Akhgari <ehsan.akhg...@gmail.com> > wrote: > >> > >> Has anyone tried running PGO and non-PGO builds of the same changeset > >> to see how they feel during ordinary browsing? > > > > Not on the same changeset, but I don't expect us to find vast > differences to > > what we already know. Do you have any reason to think otherwise? > > I'm not sure what you mean by "what we already know". > I was mentioning the performance measurements performed by dmandelin and Vladan as mentioned on this thread. > My point, which arises from general skepticism about browser > benchmarks, is this: do the microbenchmark regressions correspond to > anything a user would notice during normal use? E.g. does a non-PGO > build feel more sluggish, or something? If so, I'd find that far more > convincing evidence than a benchmark result. > I think that some of the telemetry regressions can correlate to sluggishness that can be observed by the user. David measured a bunch of simple interactions in gmail for example, but he didn't notice any performance difference. However, small slow-downs can add up, and I don't think that it's reasonable to rely only on how the browser feels in regular usage when we're talking about low-level compiler optimizations. Cheers, -- Ehsan <http://ehsanakhgari.org/> _______________________________________________ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
