I'm fine with extending the typefeedback oracle to guess that any receiver of .call is the standard JSFunction-map in case of absent type info. That's not worse than putting in a soft-deopt. As long as it's in the right place and might not cause weird side effects, it's all good ;)
On Thu, Jun 12, 2014 at 1:14 PM, Vyacheslav Egorov <[email protected]> wrote: > > 1. For recognizing .call in general to call tryinlinebuiltinmethodcall > when the > receiver of .call is known > > Having support for .call in general would be cool (even if the receiver is > not known, but just known to be a function), because right now there is an > unexpected discrepancy in the performance between call and apply > > There is even an issue https://code.google.com/p/v8/issues/detail?id=3063 > > I had a patch somewhere to support .call but never submitted it. > > Vyacheslav Egorov > > > On Thu, Jun 12, 2014 at 12:14 PM, <[email protected]> wrote: > >> Ok, sounds good :) >> >> I assume it should be 2 patches, right?: >> >> 1. For recognizing .call in general to call tryinlinebuiltinmethodcall >> when the >> receiver of .call is known >> >> 2. Adding kArraySlice case to tryinlinebuiltinmethodcall >> >> >> https://codereview.chromium.org/327173002/ >> >> -- >> -- >> v8-dev mailing list >> [email protected] >> http://groups.google.com/group/v8-dev >> ---You received this message because you are subscribed to the Google >> Groups "v8-dev" 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/d/optout. >> > > -- -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev --- You received this message because you are subscribed to the Google Groups "v8-dev" 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/d/optout.
