At this point the onus seems to be on the asm.js-specific-optimizations-aren't-needed crowd to show that they can get performance that's even remotely close to optimized asm.js when using only non-asm.js-specific optimizations. If so, yeah, implement that! That'd be the best of all possible worlds; but it seems unlikely that loosely-constrained language X is as optimizable (with a roughly similar amount of resources) as tightly-constrained language Y. To my understanding, it's not a matter of current JavaScript engines needing to be "fixed" so much as just that they operate on a different class of problems than asm.js optimizers do, and thus the range of possible achievements is different, too.