It's probably not effective anymore and it constrains the amount of constant folding we can do.
Created attachment 233171 [details] the patch
Comment on attachment 233171 [details] the patch Turns out this is a 1% SunSpider regression. I guess I'll keep the optimization.