http://codereview.chromium.org/6368097/diff/1/src/x64/lithium-codegen-x64.cc File src/x64/lithium-codegen-x64.cc (right):
http://codereview.chromium.org/6368097/diff/1/src/x64/lithium-codegen-x64.cc#newcode1973 src/x64/lithium-codegen-x64.cc:1973: Heap::kHeapNumberMapRootIndex); On 2011/02/04 12:57:55, Lasse Reichstein wrote:
indentation.
Done. http://codereview.chromium.org/6368097/diff/1/src/x64/lithium-codegen-x64.cc#newcode1986 src/x64/lithium-codegen-x64.cc:1986: CpuFeatures::Scope scope(SSE3); On 2011/02/04 12:57:55, Lasse Reichstein wrote:
Fisttp isn't any "more powerful" than cvttsd2si in X64 mode, since we
have a
64-bit version of that too. I.e., the valid ranges are the same. Use
the 64-bit
cvttsd2si in all cases, and avoid the FPU stack.
Done. http://codereview.chromium.org/6368097/diff/1/src/x64/lithium-codegen-x64.cc#newcode2000 src/x64/lithium-codegen-x64.cc:2000: __ fincstp(); Entire block of code deleted. On 2011/02/04 12:57:55, Lasse Reichstein wrote:
Use __ fstp(0); to drop the top of the FPU stack.
http://codereview.chromium.org/6368097/diff/1/src/x64/lithium-codegen-x64.cc#newcode2014 src/x64/lithium-codegen-x64.cc:2014: __ cvttsd2si(input_reg, xmm0); On 2011/02/04 12:57:55, Lasse Reichstein wrote:
Use the 64-bit version and compare to 0x8000000000000000 instead.
Done. http://codereview.chromium.org/6368097/ -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
