On Thu, 13 Aug 2026 07:12:24 GMT, Shawn Emery <[email protected]> wrote:

>> Isn't the inclusive range permitted? 
>> https://github.com/openjdk/jdk/blob/e1502b52f099c37c9c2fd2aad147dec95ff685b0/src/java.base/share/classes/com/sun/crypto/provider/ML_KEM.java#L1553
>
> Ha, yes, shortly after submitting my response I noticed that the wrong term 
> was used, but didn't go back to edit the comment as I was confident that 
> folks would know what I meant to say.  In regards to inclusive vs. exclusive 
> `q`; yes, I stand corrected, as Barrett reduction would reduce coefficients 
> that are a positive multiple of -3329 to 3329.  I had forgot that the 
> coefficients were signed which would allow for an inclusive `q`.  Please 
> disregard this comment and sorry for the noise.  Note: I surmise that Ferenc 
> used signed coefficients due to better centering (reduction in reductions ;) 
> and support in intrinsics, thereby eliminating normalization instructions.

All good, thanks for checking. May I have a re-approval?

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/32295#discussion_r3773520512

Reply via email to