Re: [encoding] Last Call Comment: Arithmetic Right Shift

Hello Anne,

On 2014/07/11 18:14, Anne van Kesteren wrote:
> On Sun, Jul 6, 2014 at 9:53 AM, "Martin J. Dürst"
> <duerst@it.aoyama.ac.jp> wrote:
>> I very much agree that we don't want to talk about the *exact*
>> representation of the operand. What I'm proposing is to talk about the
>> *minimum* size of the operands necessary for correct operation.
>
> Seems fair.  Did you mean to make this email public?

Yes. Just fixed a moment ago. Thanks for catching it.

> How about this: "Arithmetic right shifts in this standard require
> operands with at least twenty-two bits precision." Feel free to reply
> on the list.

Well, after a bit more thought, I think that because we are dealing with 
unsigned quantities, using "Logical right shifts" would actually be better.

See http://en.wikipedia.org/wiki/Logical_shift, 
http://upload.wikimedia.org/wikipedia/commons/thumb/6/64/Rotate_right_logically.svg/450px-Rotate_right_logically.svg.png.

That would allow us to shave one more bit off the maximum, so you could 
write "at least twenty-one bits precision" :-). Of course this one bit 
is completely irrelevant, but logical shift is conceptually the right 
thing to use. I'm sorry I didn't get there immediately.

> I see you are using a J. initial. Do you want me to update the
> acknowledgments section?

If I have an initial, it's J., but I don't always use it, so you can 
save that time/commit for something more important. But thanks for 
checking anyway.

Regards,   Martin.

Received on Friday, 11 July 2014 11:11:18 UTC