Currently, 32-bit ARM is the only platform which does not have InlineIntrinsics enabled by default. This is the case because it is not implemented in the template interpreter. Other platforms do this by simply calling into the SharedRuntime implementations of the math intrinsics.

I propose to do the same for 32-bit ARM to be able to enable InlineIntrinsics by default.

Since the java.lang.StrictMath implementation currently calls into fdlibm, which has basically the same implementation for the operations as the shared runtime has, the benefit might not be as big as expected, but it should save a couple of method calls, which might be especially beneficial within the interpreter on small and not so powerful ARM devices.