libgcj/1615

[Per Bothner]

The following reply was made to PR libgcj/1615; it has been noted by GNATS.

From: Per Bothner <per@bothner.com>
To: gcc-gnats@gcc.gnu.org, mjr@statesmean.com, warrenl@gcc.gnu.org
Cc: per@bothner.com
Subject: Re: libgcj/1615
Date: 11 Feb 2001 15:36:59 -0800

 I took a look at the code for BigInteger.modPow, and noticed something
 rather suspicious: The variable s is never used.  Perhaps
 double-checking the referenced algorithm would make sense.
 
 Another problem: if exponent.isNegative() then modInverse(m) is
 called.  That seems unlikely to be correct if exponent < -1.
 
 One trivial performance improvement that should be done: Replace
 u.and(ONE).isOne() by u.isOdd().  Even better would be to drop u
 entirely, and just use bitValue(exponent, i) (where i starts at 0 is
 is increment on each iteration), where:
 
   /** Return the value of a specified bit in an BigInteger. */
   public static boolean bitValue (BigInteger x, int bitno)
   {
     int i = x.ival;
     if (x.words == null)
       {
         return bitno >= 32 ? i < 0 : ((i >> bitno) & 1) != 0;
       }
     else
       {
         int wordno = bitno >> 5;
         return wordno >= i ? x.words[i-1] < 0
           : (((x.words[wordno]) >> bitno) & 1) != 0;
       }
   }
 
 /* This is cut-and-pasted from Kawa, so may need soem tweaking. */
 -- 
 	--Per Bothner
 per@bothner.com   http://www.bothner.com/~per/