The assorted tests on AWFY didn't update for about 1.5 days. When it returned, misc-basic-array, misc-basic-fpops, misc-basic-intops, misc-bugs-636096-model2d and misc-bugs-847389-jpeg2000 seem to have regressed. The closest range I could find by looking at all machines, is http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=96b964d758c8&tochange=f1a74d04600a 32 bit, 64 bit and ARM are all affected.

There were also regressions on many of the asm.js micro-bench and asm.js apps benchmarks on the 24th for 'Firefox (no asmjs)'. I don't know if these are related but it falls in the same date range with a changset overlap with the regression range from comment 1. changeset for these regressions looks to be http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=d828c528e31e&tochange=f1a74d04600a possibly Bug 774006?? Regressed: ubench-fannkuch, ubench-fasta, ubench-memops, ubench-primes, ubench-skinning apps-box2d-workload1, apps-box2d-workload2, apps-box2d-workload3, apps-box2d-workload4, apps-bullet-workload1, apps-bullet-workload2, pps-bullet-workload3, apps-bullet-workload4, apps-zlib-workload1, apps-zlib-workload2, apps-zlib-workload3, apps-zlib-workload4

I guess this is likely related to the change I made recently to add a SetElementIC (Bug 774006), but if this is really the case then the question is why are these test ever using a SetElementIC. I made the SetElementIC to be generated before falling back to a CallSetElement, so if this is the case, then we should probably look at what is going wrong on these cases. I don't expect any of the asm.js benchmark to fall in the case of a SetElementIC as this would imply that we don't have enough type information on the object to use a known setter.

Just ran a bisect. And there are multiple sources of regressions, sigh :( misc-basic-array and misc-basic-fpops bug 875276: Don't create types for scripts until they are compiled by baseline. misc-basic-intops is definitely caused by something else

- misc-basic-intops is fixed. - most assorted will get fixed by bug 876607 - asmjs and jpeg2000 will get fixed by bug 876649 I think only the SS ones are left. Those are caused by bug 875276, but I'll relook when the depending bugs land to have a better picture.