LGTM with comments (and what Vitaly said)
http://codereview.chromium.org/7230006/diff/6001/src/json-parser.h File src/json-parser.h (right): http://codereview.chromium.org/7230006/diff/6001/src/json-parser.h#newcode381 src/json-parser.h:381: if (c0_ == 'e' || c0_ == 'E') { Why do you think this is faster? The AsciiAlphaToLower should be inlined, so this just becomes: if ((c0_ | 0x20) == 'e') { http://codereview.chromium.org/7230006/diff/6001/src/json.js File src/json.js (right): http://codereview.chromium.org/7230006/diff/6001/src/json.js#newcode281 src/json.js:281: builder.push(value ? %QuoteJSONString(value) : '""'); use (value !== "") as condition. http://codereview.chromium.org/7230006/diff/6001/src/runtime.cc File src/runtime.cc (right): http://codereview.chromium.org/7230006/diff/6001/src/runtime.cc#newcode3173 src/runtime.cc:3173: int to = offsets.at(j++); Is this really faster? Any optimizing compiler worth its salt should be able to strength reduce this if it wants to (and it shouldn't want to, since the offsets.at(i * 2 + 1) code should be implementable as a single opcode. http://codereview.chromium.org/7230006/ -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
