Oh, and I think that doing something smart on the hydrogen level for this exact problem is not worth it. But it can happen as a nice side effect of bounds check optimizations.
-- Vitaly On Thu, Jan 20, 2011 at 6:33 PM, Vitaly Repeshko <[email protected]> wrote: > Please take another look. > > > Thanks, > Vitaly > > On Thu, Jan 20, 2011 at 6:03 PM, <[email protected]> wrote: >> >> http://codereview.chromium.org/6303013/diff/1/src/ia32/lithium-codegen-ia32.cc >> File src/ia32/lithium-codegen-ia32.cc (right): >> >> http://codereview.chromium.org/6303013/diff/1/src/ia32/lithium-codegen-ia32.cc#newcode2758 >> src/ia32/lithium-codegen-ia32.cc:2758: if (!Smi::IsValid(const_index)) { >> We shouldn't generate code at all for this instruction in this case. Can >> we just do an unconditional deoptimization and return? Or catch this at >> a higher level and never build this graph? >> >> http://codereview.chromium.org/6303013/ >> > -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
