Regression in go_benchmark

[magistere]

go_benchmark time increased on ~30% from January, 15 to January, 22.
See http://speed.julialang.org/timeline/#/?exe=2,1,3&base=none&ben=go_benchmark&env=1&revs=200&equid=off

[JeffBezanson]

The first regression around Jan 15-16 is "intentional" in that I added missing checks to mod and rem. The bigger regression around Jan 20 I'm not yet sure about.

[carlobaldassi]

Bisect blames 7b60d50

[JeffBezanson]

Just discovered the same thing. cc @timholy

[timholy]

Mostly a consequence of #5536. I had also called to_index twice on each index, doh! (Although that was much less important.)

[timholy]

Oh, and thanks for running git bisect. Between that and the profiler, it was trivial to figure out the problem(s).