<strike>This is necessary for implementing floating point intrinsics. (cc #1)</strike> It is also necessary for 128 bit integer operations. Edit: this isn't actually necessary for implementing floating point intrinsics, as libm provides the necessary symbols.
