hgcd2_jacobi

[Marco Bodrato]

Ciao,

Il 2020-02-01 16:43 nisse at lysator.liu.se ha scritto:
> Jacob Glickman <jhg023 at bucknell.edu> writes:
> 
>> Do these improvements warrant any changes to the mpn_jacobi_base 
>> function
>> (with JACOBI_BASE_METHOD = 4) in mpn/generic/jacbase.c? I use this 
>> function

Me too, when writing a (partial) jacobi function for mini-gmp, got 
inspiration from jacbase.c, the JACOBI_BASE_METHOD == 4 variant.

I just pushed a small optimisation to mini-gmp:
https://gmplib.org/repo/gmp/rev/2a398323c9c7

And maybe a similar optimisation for mpn/generic/jacbase.c should be 
committed.

I propose:
diff -r 2624648aa8e1 mpn/generic/jacbase.c
--- a/mpn/generic/jacbase.c     Sun Feb 02 16:06:52 2020 +0100
+++ b/mpn/generic/jacbase.c     Sun Feb 02 16:10:28 2020 +0100
@@ -235,7 +235,7 @@
        bit ^= c & (b ^ (b >> 1));
        a >>= c;
      }
-  while (b > 0);
+  while (a > 0);

    return 1-2*(bit & 1);
  }

It saves an unneeded round of the loop.

Ĝis,
m