Thanks for the clarification. It looks good to me. Does anyone have any comments on the various competing proposed fixes? As for me, I like this one with int operations and no cache.
Thanks,
Brian
On Dec 23, 2013, at 6:02 PM, Dmitry Nadezhin wrote:
> We have
> first/0x1p57 - 1 < JAVA{first>>>57} && JAVA{first>>>57} <= first/0x1p57.
> […]
> Hence overflow is detected correctly in this case too.
