On Sun, Apr 7, 2013 at 11:02 PM, Jakob Kummerow <[email protected]> wrote: > This is expected. V8 has a special optimization for "Math.floor of a > division". > > The only way of "fixing" this would be to remove that optimization, in which > case the "/2" case would become about as slow as the "*.5" case. That's not > what you want, is it? ;-) > > Background: > All of this is due to JavaScript, where every number is a double. As you > know, integers are faster to handle on the CPU, so JS engines try to detect > cases where they can internally represent/manipulate numbers as integers > without violating JS semantics. Essentially, in the "floor of div" case V8 > decides that it can treat the numbers as integers, and further that division > by constant 2 can be done by bit shift, so in optimized code it actually > emits a bit shift instruction. In the "*.5" case, on the other hand, a > regular floating-point multiplication is performed, and the resulting > floating-point number is then rounded (well, floored) to integer, which is a > bit slower.
Is there anything intrinsically unsafe in the example the OP posted? If Math.floor is untainted (i.e. not monkey-patched by the user) and the multiplier is a literal, it should be safe in most cases to convert it to integer math, right? The benefits may not outweigh the increased complexity, of course, but it doesn't look like it's inherently impossible. (I'm not volunteering a patch, by the way. :-)) -- -- v8-users mailing list [email protected] http://groups.google.com/group/v8-users --- You received this message because you are subscribed to the Google Groups "v8-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
