/*
  file LFSR4.C
  created Sun Oct 23 20:24:34 CEST 2005 by Yann GUIDON (whygee@f-cpu.org)
  version Sun Oct 30 05:41:47 CET 2005 : Alpha port

  This file implements a 32-bit LFSR (generates 4 billions bits before wraparound)
  but this adaptation of the Schneier code provides 4 bits each time
  (wraparound after 1 billion calls).

 */

#define U32 unsigned long int

#define LFSR_POLY0 0x80000057
#define LFSR_POLY1 (((LFSR_POLY0>>1)|(LFSR_POLY0<<31)) & 0xFFFFFFFF)
#define LFSR_POLY2 (((LFSR_POLY1>>1)|(LFSR_POLY1<<31)) & 0xFFFFFFFF)
#define LFSR_POLY3 (((LFSR_POLY2>>1)|(LFSR_POLY2<<31)) & 0xFFFFFFFF)

U32 LFSR4_TABLE[16]= {
  0,
  LFSR_POLY0,
             LFSR_POLY1,
  LFSR_POLY0^LFSR_POLY1,
                        LFSR_POLY2,
  LFSR_POLY0^           LFSR_POLY2,
             LFSR_POLY1^LFSR_POLY2,
  LFSR_POLY0^LFSR_POLY1^LFSR_POLY2,
                                   LFSR_POLY3,
  LFSR_POLY0^                      LFSR_POLY3,
             LFSR_POLY1^           LFSR_POLY3,
  LFSR_POLY0^LFSR_POLY1^           LFSR_POLY3,
                        LFSR_POLY2^LFSR_POLY3,
  LFSR_POLY0^           LFSR_POLY2^LFSR_POLY3,
             LFSR_POLY1^LFSR_POLY2^LFSR_POLY3,
  LFSR_POLY0^LFSR_POLY1^LFSR_POLY2^LFSR_POLY3
};

U32 LFSR_reg = 0x12345678; /* anything except zero */

inline U32 lfsr4() {
  U32 t=LFSR_reg & 15;
  U32 u=LFSR_reg ^ LFSR4_TABLE[t];
  LFSR_reg = (u >> 4) | (u << 28);
  return t;
}

#ifdef STANDALONE_TEST_LFSR4

int main() {
  int i, j=0;

  for (i=0; i<1000; i++) {
    j^=lfsr4();
  }
  return j; /* so the LFSR code is not optimised out at all */
}

#endif