There is a 19% regression on octane-splay on x86 and x64 MacOSX shell slaves. http://arewefastyet.com/#machine=28&view=single&suite=octane&subtest=Splay&start=1421422389&end=1421624846 Combing the x86 and x64 regression ranges I get: http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3d0dd2ae7e6c&tochange=cafc08288982

So putting NI on terrence, since the regression range shows only patches from him.

Thanks, Hannes! Splay does a ton more scanning of live objects than our other benchmarks, so I'm fairly certain that this is the extra budget check I added in bug 1111361. I'll try backing out that push and we can see what happens to splay.

Nope, looks like I'm wrong: the backout is on awfy and the performance did not change. Cycling to 1119537.

Thanks a lot. Confirmed fixed on awfy by the backout http://hg.mozilla.org/integration/mozilla-inbound/rev/aa05cd133a76