http://codereview.chromium.org/8968028/diff/2001/src/objects-inl.h File src/objects-inl.h (right):
http://codereview.chromium.org/8968028/diff/2001/src/objects-inl.h#newcode1222 src/objects-inl.h:1222: ASSERT(current->IsSmi() || current-IsTheHole()); On 2011/12/29 15:54:48, Vyacheslav Egorov wrote:
current->IsHole()
Done. http://codereview.chromium.org/8968028/diff/2001/src/objects.cc File src/objects.cc (right): http://codereview.chromium.org/8968028/diff/2001/src/objects.cc#newcode9253 src/objects.cc:9253: if (has_smi_only_elements) { There's no harm in going to smi elements for very large arrays when the flag isn't on. We already do this for very large array literals. The smi_only_arrays flag has become a bit of a misnomer, it now means "enable the smi/double features that are known to still cause performance regressions". On 2011/12/29 15:18:41, Jakob wrote:
I think you need an additional "&& FLAG_smi_only_arrays" here,
otherwise it'll
transition to SMI elements even if the flag is off.
http://codereview.chromium.org/8968028/diff/2001/test/mjsunit/regress/regress-1849.js File test/mjsunit/regress/regress-1849.js (right): http://codereview.chromium.org/8968028/diff/2001/test/mjsunit/regress/regress-1849.js#newcode1 test/mjsunit/regress/regress-1849.js:1: // Copyright 2009 the V8 project authors. All rights reserved. On 2011/12/29 15:18:41, Jakob wrote:
2011
Done. http://codereview.chromium.org/8968028/diff/2001/test/mjsunit/regress/regress-1849.js#newcode30 test/mjsunit/regress/regress-1849.js:30: // Flags: --allow-natives-syntax See the other comment. On 2011/12/29 15:18:41, Jakob wrote:
--smi-only-arrays ? And then you'll need to put the assertion behind
an if()
that checks if the flag is in effect.
http://codereview.chromium.org/8968028/ -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
