> To answer the second part of your response, the correct way to do this yourself would be to memcpy the double and then inspect the result directly. Any compil

> To answer the second part of your response, the correct way to do this yourself would be to memcpy the double and then inspect the result directly. Any compiler worth its salt will end up generating the assembly code you want (namely, no call to memcpy, just a couple of arithmetic shifts and masks). What does a compiler "worth its salt" do with the memcpy at "-O0"?