Status: New
Owner: ----
New issue 746 by [email protected]: Inline fast-pass MUL binary
operation on ARM
http://code.google.com/p/v8/issues/detail?id=746
MUL is called many times in the v8 crypto benchmark. If we inline a
fast-pass MUL, the function call overhead could be avoided. We propose a
change in GenericBinaryOperation(codegen-arm.cc), that emits the fast-pass
MUL instructions for SMI. The other changes include updating the original
MUL instructions in Generate, and checking under which conditions we could
do inlining. Our current change gives 8.8% improvement on crypto.
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
