The optimized binary operations currently use C style overflow checking.
LLVM supports several provides several intrinsics for arithmetics with
overflow which have access to registers and flags which are not exposed in
C. Using these intrinsics may be faster than our current method.
Preliminary research showed that they are not supported by clang. One
possibility would be to create the necessary function by hand in the global
module initialisation.
Original issue reported on code.google.com by joerg...@gmail.com on 14 Jan 2010 at 10:42
Original issue reported on code.google.com by
joerg...@gmail.com
on 14 Jan 2010 at 10:42