Reviewers: ulan, Message: Please take a look.
I removed the double_int_runtime bail out since this would only be used when we encounter potential subnormals in the int_exponent case. Instead we just bail
out to power_double_double. I also cleared up the runtime functions to better mirror the generated code. http://codereview.chromium.org/8831008/diff/1/src/ia32/code-stubs-ia32.cc File src/ia32/code-stubs-ia32.cc (left): http://codereview.chromium.org/8831008/diff/1/src/ia32/code-stubs-ia32.cc#oldcode3135 src/ia32/code-stubs-ia32.cc:3135: __ and_(exponent, Immediate(kClearSignBitMask)); This is a screw-up I did earlier. Obviously this code does not result in absolute, but triggered a bail out, which is why this mistake made no visible difference. http://codereview.chromium.org/8831008/diff/1/src/x64/code-stubs-x64.cc File src/x64/code-stubs-x64.cc (left): http://codereview.chromium.org/8831008/diff/1/src/x64/code-stubs-x64.cc#oldcode2186 src/x64/code-stubs-x64.cc:2186: __ andl(scratch, Immediate(kClearSignBitMask)); Ditto in ia32. Description: Tweaks on Math.pow (ia32 and x64). Please review this at http://codereview.chromium.org/8831008/ SVN Base: https://v8.googlecode.com/svn/branches/bleeding_edge Affected files: M src/assembler.cc M src/ia32/code-stubs-ia32.cc M src/runtime.cc M src/x64/code-stubs-x64.cc -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
