Recent speed optimizations have entirely broken "const". <rdar://problem/5657450>
Created attachment 18708 [details] patch
Created attachment 18718 [details] same patch, rebased
Created attachment 18719 [details] same patch, rebased
Comment on attachment 18719 [details] same patch, rebased Can you test this in sunspider? r=me
(In reply to comment #4) > Can you test this in sunspider? I did test it in SunSpider. It showed 1.003x slower (0.3%) with const support, although there was enough noise and that's that I'm not sure it's right. We might be able to shave it even closer with some more effort, but I think that is good enough that this should be landed as-is.
Committed revision 29818.