Author stutzbach
Recipients belopolsky, draghuram, mark.dickinson, rhettinger, stutzbach
Date 2010-05-12.16:00:30
SpamBayes Score 0.00240094
Marked as misclassified No
Message-id <AANLkTinC_5l5402K-kpj4hdTFsEM0in_4D-FyF-KDYWG@mail.gmail.com>
In-reply-to <AANLkTim61n5Wp8ttjtKSPzUFzjVfaOgWzeWIGwhRRiTJ@mail.gmail.com>
Content
On Wed, May 12, 2010 at 10:31 AM, Alexander Belopolsky
<report@bugs.python.org> wrote:
> if ((k & 1) != 1)
>          k = k - 1;
>
> looks odd to me. Maybe k += (k & 1) - 1?

If we change the logic slightly to put the odd entry on the right
instead of the left, we can do:

        k = (n + m) / 2;
        k |= 1; /* Ensure that k is odd */
        left = factorial_part_product(n, k-2);
        if (left == NULL) goto done;
        right = factorial_part_product(k, m);
        if (right == NULL) goto done;

That will split 1*3*5*7*9*11 into (1*3*5) * (7*9*11), just like the
old code.  It will split 1*3*5*7*9 into (1*3) * (5*7*9) while the old
code did (1*3*5) * (7*9), which is fine.

It's easier to read and fewer operations. :-)
History
Date User Action Args
2010-05-12 16:03:08stutzbachsetrecipients: + stutzbach, rhettinger, mark.dickinson, belopolsky, draghuram
2010-05-12 16:01:01stutzbachlinkissue8692 messages
2010-05-12 16:00:30stutzbachcreate