Is this thread safe?

[Walter Bright]

dsimcha wrote:
> Never mind, I see at least one reason why my last post would be incorrect.  The
> length variable could be set out-of-order before all of the factorials up to
> length are calculated.  Is there any simple way to write this function so it's
> thread-safe without using synchronized just to access cached results?

No. There are three choices:

1) synchronize it

2) make thread-local versions

3) insert fences with the inline assembler