Oh, interesting. Do you have a pointer to those tests? I'd like to take a look to see if they're doing something similar.
I'm fully aware that this can't be made to work for 100% of cases, but my hope is that it will be sufficiently stable for a large subset. Ojan On Thu, Aug 26, 2010 at 11:15 PM, Mads Ager <[email protected]> wrote: > Ojan, > > just for your information we have disabled almost all > order-of-magnitude tests from the mozilla tests because there are > constant flakiness issues. With a garbage collected/JIT compilation > system it seems to me that something like this is bound to become > flaky. > > Cheers, -- Mads > > On Thu, Aug 26, 2010 at 9:48 PM, Ojan Vafai <[email protected]> wrote: > > These are order-of-magnitude tests. They just try to assert that a > > micro-benchmark is linear, quadratic, etc. This makes it so they can be > run > > as part of the normal layout test suite and the results mostly do not > depend > > on the machine you're running them on. > > > > The inspiration for this was to give a way of adding tests when fixing > bugs > > around the order-of-magnitude of a given chunk of code. > > > > Make sense? > > > > Ojan > > > > On 9:01 am, Erik Kay <[email protected]> wrote: > >> > >> Ojan, > >> I thought the DOMperf testing framework that we developed does a lot of > >> the work to help compensate (or at least illustrate) variability due to > >> GC. I thought there was more of a push to start using this for new > >> benchmarks as well. Is there a reason why that doesn't work for your > >> case? > >> Erik > > > > -- > > v8-dev mailing list > > [email protected] > > http://groups.google.com/group/v8-dev > > > > -- > v8-dev mailing list > [email protected] > http://groups.google.com/group/v8-dev > -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
