// Copyright in this code is held by Dr B.R. Gladman but free direct or
// derivative use is permitted subject to acknowledgement of its origin
// and subject to any constraints placed on the use of the algorithm by
// its designers (if such constraints may exist, this will be indicated
// below).
//
// Dr. B. R. Gladman                               . 25th January 2000.
//
// This is an implementation of Serpent, an encryption algorithm designed
// by Anderson, Biham and Knudsen and submitted as a candidate for the
// Advanced Encryption Standard programme of the US National Institute of
// Standards and Technology.
//
// The designers of Serpent have not placed any constraints on the use of
// this algorithm.

#include <sys/types.h>
#include <sys/endian.h>

#include <crypto/serpent/serpent.h>

// Partially optimised Serpent S Box boolean functions derived
// using a recursive descent analyser but without a full search
// of all subtrees. This set of S boxes is the result of work
// by Sam Simpson and Brian Gladman using the spare time on a
// cluster of high capacity servers to search for S boxes with
// this customised search engine.
//
// Copyright:   Dr B. R Gladman (gladman@seven77.demon.co.uk)
//              and Sam Simpson (s.simpson@mia.co.uk)
//              17th December 1998
//
// We hereby give permission for information in this file to be
// used freely subject only to acknowledgement of its origin

// 15 terms

#define rotr(x,n)   (((x) >> ((int)((n) & 0x1f))) | ((x) << ((int)((32 - ((n) & 0x1f))))))
#define rotl(x,n)   (((x) << ((int)((n) & 0x1f))) | ((x) >> ((int)((32 - ((n) & 0x1f))))))

#define sb0(a,b,c,d,e,f,g,h)    \
{   t1 = a ^ d;     \
    t2 = a & d;     \
    t3 = c ^ t1;    \
    t6 = b & t1;    \
    t4 = b ^ t3;    \
    t10 = ~t3;      \
    h = t2 ^ t4;    \
    t7 = a ^ t6;    \
    t14 = ~t7;      \
    t8 = c | t7;    \
    t11 = t3 ^ t7;  \
    g = t4 ^ t8;    \
    t12 = h & t11;  \
    f = t10 ^ t12;  \
    e = t12 ^ t14;  \
}
// 15 terms

#define ib0(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = a ^ b;     \
    t3 = t1 | t2;   \
    t4 = d ^ t3;    \
    t7 = d & t2;    \
    t5 = c ^ t4;    \
    t8 = t1 ^ t7;   \
    g = t2 ^ t5;    \
    t11 = a & t4;   \
    t9 = g & t8;    \
    t14 = t5 ^ t8;  \
    f = t4 ^ t9;    \
    t12 = t5 | f;   \
    h = t11 ^ t12;  \
    e = h ^ t14;    \
}

// 14 terms!

#define sb1(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = b ^ t1;    \
    t3 = a | t2;    \
    t4 = d | t2;    \
    t5 = c ^ t3;    \
    g = d ^ t5;     \
    t7 = b ^ t4;    \
    t8 = t2 ^ g;    \
    t9 = t5 & t7;   \
    h = t8 ^ t9;    \
    t11 = t5 ^ t7;  \
    f = h ^ t11;    \
    t13 = t8 & t11; \
    e = t5 ^ t13;   \
}

// 17 terms

#define ib1(a,b,c,d,e,f,g,h)    \
{   t1 = a ^ d;     \
    t2 = a & b;     \
    t3 = b ^ c;     \
    t4 = a ^ t3;    \
    t5 = b | d;     \
    t7 = c | t1;    \
    h = t4 ^ t5;    \
    t8 = b ^ t7;    \
    t11 = ~t2;      \
    t9 = t4 & t8;   \
    f = t1 ^ t9;    \
    t13 = t9 ^ t11; \
    t12 = h & f;    \
    g = t12 ^ t13;  \
    t15 = a & d;    \
    t16 = c ^ t13;  \
    e = t15 ^ t16;  \
}

// 16 terms

#define sb2(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = b ^ d;     \
    t3 = c & t1;    \
    t13 = d | t1;   \
    e = t2 ^ t3;    \
    t5 = c ^ t1;    \
    t6 = c ^ e;     \
    t7 = b & t6;    \
    t10 = e | t5;   \
    h = t5 ^ t7;    \
    t9 = d | t7;    \
    t11 = t9 & t10; \
    t14 = t2 ^ h;   \
    g = a ^ t11;    \
    t15 = g ^ t13;  \
    f = t14 ^ t15;  \
}

// 16 terms

#define ib2(a,b,c,d,e,f,g,h)    \
{   t1 = b ^ d;     \
    t2 = ~t1;       \
    t3 = a ^ c;     \
    t4 = c ^ t1;    \
    t7 = a | t2;    \
    t5 = b & t4;    \
    t8 = d ^ t7;    \
    t11 = ~t4;      \
    e = t3 ^ t5;    \
    t9 = t3 | t8;   \
    t14 = d & t11;  \
    h = t1 ^ t9;    \
    t12 = e | h;    \
    f = t11 ^ t12;  \
    t15 = t3 ^ t12; \
    g = t14 ^ t15;  \
}

// 17 terms

#define sb3(a,b,c,d,e,f,g,h)    \
{   t1 = a ^ c;     \
    t2 = d ^ t1;    \
    t3 = a & t2;    \
    t4 = d ^ t3;    \
    t5 = b & t4;    \
    g = t2 ^ t5;    \
    t7 = a | g;     \
    t8 = b | d;     \
    t11 = a | d;    \
    t9 = t4 & t7;   \
    f = t8 ^ t9;    \
    t12 = b ^ t11;  \
    t13 = g ^ t9;   \
    t15 = t3 ^ t8;  \
    h = t12 ^ t13;  \
    t16 = c & t15;  \
    e = t12 ^ t16;  \
}

// 16 term solution that performs less well than 17 term one
// in my environment (PPro/PII)

#if 0

#define sb3(a,b,c,d,e,f,g,h)    \
{   t1 = a ^ b;     \
    t2 = a & c;     \
    t3 = a | d;     \
    t4 = c ^ d;     \
    t5 = t1 & t3;   \
    t6 = t2 | t5;   \
    g = t4 ^ t6;    \
    t8 = b ^ t3;    \
    t9 = t6 ^ t8;   \
    t10 = t4 & t9;  \
    e = t1 ^ t10;   \
    t12 = g & e;    \
    f = t9 ^ t12;   \
    t14 = b | d;    \
    t15 = t4 ^ t12; \
    h = t14 ^ t15;  \
}

#endif

// 17 terms

#define ib3(a,b,c,d,e,f,g,h)    \
{   t1 = b ^ c;     \
    t2 = b | c;     \
    t3 = a ^ c;     \
    t7 = a ^ d;     \
    t4 = t2 ^ t3;   \
    t5 = d | t4;    \
    t9 = t2 ^ t7;   \
    e = t1 ^ t5;    \
    t8 = t1 | t5;   \
    t11 = a & t4;   \
    g = t8 ^ t9;    \
    t12 = e | t9;   \
    f = t11 ^ t12;  \
    t14 = a & g;    \
    t15 = t2 ^ t14; \
    t16 = e & t15;  \
    h = t4 ^ t16;   \
}

// 15 terms

#define sb4(a,b,c,d,e,f,g,h)    \
{   t1 = a ^ d;     \
    t2 = d & t1;    \
    t3 = c ^ t2;    \
    t4 = b | t3;    \
    h = t1 ^ t4;    \
    t6 = ~b;        \
    t7 = t1 | t6;   \
    e = t3 ^ t7;    \
    t9 = a & e;     \
    t10 = t1 ^ t6;  \
    t11 = t4 & t10; \
    g = t9 ^ t11;   \
    t13 = a ^ t3;   \
    t14 = t10 & g;  \
    f = t13 ^ t14;  \
}

// 17 terms

#define ib4(a,b,c,d,e,f,g,h)    \
{   t1 = c ^ d;     \
    t2 = c | d;     \
    t3 = b ^ t2;    \
    t4 = a & t3;    \
    f = t1 ^ t4;    \
    t6 = a ^ d;     \
    t7 = b | d;     \
    t8 = t6 & t7;   \
    h = t3 ^ t8;    \
    t10 = ~a;       \
    t11 = c ^ h;    \
    t12 = t10 | t11;\
    e = t3 ^ t12;   \
    t14 = c | t4;   \
    t15 = t7 ^ t14; \
    t16 = h | t10;  \
    g = t15 ^ t16;  \
}

// 16 terms

#define sb5(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = a ^ b;     \
    t3 = a ^ d;     \
    t4 = c ^ t1;    \
    t5 = t2 | t3;   \
    e = t4 ^ t5;    \
    t7 = d & e;     \
    t8 = t2 ^ e;    \
    t10 = t1 | e;   \
    f = t7 ^ t8;    \
    t11 = t2 | t7;  \
    t12 = t3 ^ t10; \
    t14 = b ^ t7;   \
    g = t11 ^ t12;  \
    t15 = f & t12;  \
    h = t14 ^ t15;  \
}

// 16 terms

#define ib5(a,b,c,d,e,f,g,h)    \
{   t1 = ~c;        \
    t2 = b & t1;    \
    t3 = d ^ t2;    \
    t4 = a & t3;    \
    t5 = b ^ t1;    \
    h = t4 ^ t5;    \
    t7 = b | h;     \
    t8 = a & t7;    \
    f = t3 ^ t8;    \
    t10 = a | d;    \
    t11 = t1 ^ t7;  \
    e = t10 ^ t11;  \
    t13 = a ^ c;    \
    t14 = b & t10;  \
    t15 = t4 | t13; \
    g = t14 ^ t15;  \
}

// 15 terms

#define sb6(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = a ^ d;     \
    t3 = b ^ t2;    \
    t4 = t1 | t2;   \
    t5 = c ^ t4;    \
    f = b ^ t5;     \
    t13 = ~t5;      \
    t7 = t2 | f;    \
    t8 = d ^ t7;    \
    t9 = t5 & t8;   \
    g = t3 ^ t9;    \
    t11 = t5 ^ t8;  \
    e = g ^ t11;    \
    t14 = t3 & t11; \
    h = t13 ^ t14;  \
}

// 15 terms

#define ib6(a,b,c,d,e,f,g,h)    \
{   t1 = ~a;        \
    t2 = a ^ b;     \
    t3 = c ^ t2;    \
    t4 = c | t1;    \
    t5 = d ^ t4;    \
    t13 = d & t1;   \
    f = t3 ^ t5;    \
    t7 = t3 & t5;   \
    t8 = t2 ^ t7;   \
    t9 = b | t8;    \
    h = t5 ^ t9;    \
    t11 = b | h;    \
    e = t8 ^ t11;   \
    t14 = t3 ^ t11; \
    g = t13 ^ t14;  \
}

// 17 terms

#define sb7(a,b,c,d,e,f,g,h)    \
{   t1 = ~c;        \
    t2 = b ^ c;     \
    t3 = b | t1;    \
    t4 = d ^ t3;    \
    t5 = a & t4;    \
    t7 = a ^ d;     \
    h = t2 ^ t5;    \
    t8 = b ^ t5;    \
    t9 = t2 | t8;   \
    t11 = d & t3;   \
    f = t7 ^ t9;    \
    t12 = t5 ^ f;   \
    t15 = t1 | t4;  \
    t13 = h & t12;  \
    g = t11 ^ t13;  \
    t16 = t12 ^ g;  \
    e = t15 ^ t16;  \
}

// 17 terms

#define ib7(a,b,c,d,e,f,g,h)    \
{   t1 = a & b;     \
    t2 = a | b;     \
    t3 = c | t1;    \
    t4 = d & t2;    \
    h = t3 ^ t4;    \
    t6 = ~d;        \
    t7 = b ^ t4;    \
    t8 = h ^ t6;    \
    t11 = c ^ t7;   \
    t9 = t7 | t8;   \
    f = a ^ t9;     \
    t12 = d | f;    \
    e = t11 ^ t12;  \
    t14 = a & h;    \
    t15 = t3 ^ f;   \
    t16 = e ^ t14;  \
    g = t15 ^ t16;  \
}

#define k_xor(r,a,b,c,d)             \
{   a ^= ctx->l_key[4 * r +  8]; \
    b ^= ctx->l_key[4 * r +  9]; \
    c ^= ctx->l_key[4 * r + 10]; \
    d ^= ctx->l_key[4 * r + 11]; \
}

#define k_set(r,a,b,c,d)             \
{   a = ctx->l_key[4 * r +  8];  \
    b = ctx->l_key[4 * r +  9];  \
    c = ctx->l_key[4 * r + 10];  \
    d = ctx->l_key[4 * r + 11];  \
}

#define k_get(r,a,b,c,d)    \
{   ctx->l_key[4 * r +  8] = a;  \
    ctx->l_key[4 * r +  9] = b;  \
    ctx->l_key[4 * r + 10] = c;  \
    ctx->l_key[4 * r + 11] = d;  \
}

// the linear transformation and its inverse

#define rot(a,b,c,d)    \
{   a = rotl(a, 13);    \
    c = rotl(c, 3);     \
    d ^= c ^ (a << 3);  \
    b ^= a ^ c;         \
    d = rotl(d, 7);     \
    b = rotl(b, 1);     \
    a ^= b ^ d;         \
    c ^= d ^ (b << 7);  \
    a = rotl(a, 5);     \
    c = rotl(c, 22);    \
}

#define irot(a,b,c,d)   \
{   c = rotr(c, 22);    \
    a = rotr(a, 5);     \
    c ^= d ^ (b << 7);  \
    a ^= b ^ d;         \
    d = rotr(d, 7);     \
    b = rotr(b, 1);     \
    d ^= c ^ (a << 3);  \
    b ^= a ^ c;         \
    c = rotr(c, 3);     \
    a = rotr(a, 13);    \
}

// initialise the key schedule from the user supplied key

void serpent_set_key(serpent_ctx *ctx, const u_int8_t in_key[], int key_len)
{
    u_int32_t  i,lk,a,b,c,d,e,f,g,h;
    u_int32_t  t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,t11,t12,t13,t14,t15,t16;

    if(key_len > 256)
        return;

    i = 0; lk = (key_len + 31) / 32;

    while(i < lk)
    {
        ctx->l_key[i] = htole32(*((const u_int32_t *)(in_key + 4 * i))); i++;
    }

    if(key_len < 256)
    {
        while(i < 8)

            ctx->l_key[i++] = 0;

        i = key_len / 32; lk = 1 << key_len % 32;

        ctx->l_key[i] = (ctx->l_key[i] & (lk - 1)) | lk;
    }

    t1 = ctx->l_key[2] ^ ctx->l_key[4] ^ ctx->l_key[6] ^ 0x9e3779b9;
    t2 = ctx->l_key[3] ^ ctx->l_key[5] ^ ctx->l_key[7] ^ 0x9e3779b9;

    for(i = 0; i < 132; i += 2)
    {
        ctx->l_key[i + 8] = rotr(i ^ ctx->l_key[i] ^ t2, 21);

        t1 ^= ctx->l_key[i + 2] ^ ctx->l_key[i + 8];

        ctx->l_key[i + 9] = rotr((i + 1) ^ ctx->l_key[i + 1] ^ t1, 21);

        t2 ^= ctx->l_key[i + 3] ^ ctx->l_key[i + 9];
    }

    k_set( 0,a,b,c,d);sb3(a,b,c,d,e,f,g,h);k_get( 0,e,f,g,h);
    k_set( 1,a,b,c,d);sb2(a,b,c,d,e,f,g,h);k_get( 1,e,f,g,h);
    k_set( 2,a,b,c,d);sb1(a,b,c,d,e,f,g,h);k_get( 2,e,f,g,h);
    k_set( 3,a,b,c,d);sb0(a,b,c,d,e,f,g,h);k_get( 3,e,f,g,h);
    k_set( 4,a,b,c,d);sb7(a,b,c,d,e,f,g,h);k_get( 4,e,f,g,h);
    k_set( 5,a,b,c,d);sb6(a,b,c,d,e,f,g,h);k_get( 5,e,f,g,h);
    k_set( 6,a,b,c,d);sb5(a,b,c,d,e,f,g,h);k_get( 6,e,f,g,h);
    k_set( 7,a,b,c,d);sb4(a,b,c,d,e,f,g,h);k_get( 7,e,f,g,h);
    k_set( 8,a,b,c,d);sb3(a,b,c,d,e,f,g,h);k_get( 8,e,f,g,h);
    k_set( 9,a,b,c,d);sb2(a,b,c,d,e,f,g,h);k_get( 9,e,f,g,h);
    k_set(10,a,b,c,d);sb1(a,b,c,d,e,f,g,h);k_get(10,e,f,g,h);
    k_set(11,a,b,c,d);sb0(a,b,c,d,e,f,g,h);k_get(11,e,f,g,h);
    k_set(12,a,b,c,d);sb7(a,b,c,d,e,f,g,h);k_get(12,e,f,g,h);
    k_set(13,a,b,c,d);sb6(a,b,c,d,e,f,g,h);k_get(13,e,f,g,h);
    k_set(14,a,b,c,d);sb5(a,b,c,d,e,f,g,h);k_get(14,e,f,g,h);
    k_set(15,a,b,c,d);sb4(a,b,c,d,e,f,g,h);k_get(15,e,f,g,h);
    k_set(16,a,b,c,d);sb3(a,b,c,d,e,f,g,h);k_get(16,e,f,g,h);
    k_set(17,a,b,c,d);sb2(a,b,c,d,e,f,g,h);k_get(17,e,f,g,h);
    k_set(18,a,b,c,d);sb1(a,b,c,d,e,f,g,h);k_get(18,e,f,g,h);
    k_set(19,a,b,c,d);sb0(a,b,c,d,e,f,g,h);k_get(19,e,f,g,h);
    k_set(20,a,b,c,d);sb7(a,b,c,d,e,f,g,h);k_get(20,e,f,g,h);
    k_set(21,a,b,c,d);sb6(a,b,c,d,e,f,g,h);k_get(21,e,f,g,h);
    k_set(22,a,b,c,d);sb5(a,b,c,d,e,f,g,h);k_get(22,e,f,g,h);
    k_set(23,a,b,c,d);sb4(a,b,c,d,e,f,g,h);k_get(23,e,f,g,h);
    k_set(24,a,b,c,d);sb3(a,b,c,d,e,f,g,h);k_get(24,e,f,g,h);
    k_set(25,a,b,c,d);sb2(a,b,c,d,e,f,g,h);k_get(25,e,f,g,h);
    k_set(26,a,b,c,d);sb1(a,b,c,d,e,f,g,h);k_get(26,e,f,g,h);
    k_set(27,a,b,c,d);sb0(a,b,c,d,e,f,g,h);k_get(27,e,f,g,h);
    k_set(28,a,b,c,d);sb7(a,b,c,d,e,f,g,h);k_get(28,e,f,g,h);
    k_set(29,a,b,c,d);sb6(a,b,c,d,e,f,g,h);k_get(29,e,f,g,h);
    k_set(30,a,b,c,d);sb5(a,b,c,d,e,f,g,h);k_get(30,e,f,g,h);
    k_set(31,a,b,c,d);sb4(a,b,c,d,e,f,g,h);k_get(31,e,f,g,h);
    k_set(32,a,b,c,d);sb3(a,b,c,d,e,f,g,h);k_get(32,e,f,g,h);

    return;
}

// encrypt a block of text

void serpent_encrypt(serpent_ctx *ctx, const u_int8_t in_blk[],
		     u_int8_t out_blk[])
{
    u_int32_t  a,b,c,d,e,f,g,h;
    u_int32_t  t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,t11,t12,t13,t14,t15,t16;

    a = htole32(*((const u_int32_t *)(in_blk     )));
    b = htole32(*((const u_int32_t *)(in_blk + 4 )));
    c = htole32(*((const u_int32_t *)(in_blk + 8 )));
    d = htole32(*((const u_int32_t *)(in_blk + 12)));

    k_xor( 0,a,b,c,d); sb0(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor( 1,e,f,g,h); sb1(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor( 2,a,b,c,d); sb2(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor( 3,e,f,g,h); sb3(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor( 4,a,b,c,d); sb4(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor( 5,e,f,g,h); sb5(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor( 6,a,b,c,d); sb6(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor( 7,e,f,g,h); sb7(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor( 8,a,b,c,d); sb0(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor( 9,e,f,g,h); sb1(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(10,a,b,c,d); sb2(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(11,e,f,g,h); sb3(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(12,a,b,c,d); sb4(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(13,e,f,g,h); sb5(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(14,a,b,c,d); sb6(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(15,e,f,g,h); sb7(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(16,a,b,c,d); sb0(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(17,e,f,g,h); sb1(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(18,a,b,c,d); sb2(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(19,e,f,g,h); sb3(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(20,a,b,c,d); sb4(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(21,e,f,g,h); sb5(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(22,a,b,c,d); sb6(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(23,e,f,g,h); sb7(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(24,a,b,c,d); sb0(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(25,e,f,g,h); sb1(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(26,a,b,c,d); sb2(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(27,e,f,g,h); sb3(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(28,a,b,c,d); sb4(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(29,e,f,g,h); sb5(e,f,g,h,a,b,c,d); rot(a,b,c,d);
    k_xor(30,a,b,c,d); sb6(a,b,c,d,e,f,g,h); rot(e,f,g,h);
    k_xor(31,e,f,g,h); sb7(e,f,g,h,a,b,c,d); k_xor(32,a,b,c,d);

    *((u_int32_t *)(out_blk     )) = le32toh(a);
    *((u_int32_t *)(out_blk +  4)) = le32toh(b);
    *((u_int32_t *)(out_blk +  8)) = le32toh(c);
    *((u_int32_t *)(out_blk + 12)) = le32toh(d);
}

// decrypt a block of text

void serpent_decrypt(serpent_ctx *ctx, const u_int8_t in_blk[],
		     u_int8_t out_blk[])
{
    u_int32_t  a,b,c,d,e,f,g,h;
    u_int32_t  t1,t2,t3,t4,t5,t6,t7,t8,t9,t10,t11,t12,t13,t14,t15,t16;

    a = htole32(*((const u_int32_t *)(in_blk     )));
    b = htole32(*((const u_int32_t *)(in_blk + 4 )));
    c = htole32(*((const u_int32_t *)(in_blk + 8 )));
    d = htole32(*((const u_int32_t *)(in_blk + 12)));

    k_xor(32,a,b,c,d); ib7(a,b,c,d,e,f,g,h); k_xor(31,e,f,g,h);
    irot(e,f,g,h); ib6(e,f,g,h,a,b,c,d); k_xor(30,a,b,c,d);
    irot(a,b,c,d); ib5(a,b,c,d,e,f,g,h); k_xor(29,e,f,g,h);
    irot(e,f,g,h); ib4(e,f,g,h,a,b,c,d); k_xor(28,a,b,c,d);
    irot(a,b,c,d); ib3(a,b,c,d,e,f,g,h); k_xor(27,e,f,g,h);
    irot(e,f,g,h); ib2(e,f,g,h,a,b,c,d); k_xor(26,a,b,c,d);
    irot(a,b,c,d); ib1(a,b,c,d,e,f,g,h); k_xor(25,e,f,g,h);
    irot(e,f,g,h); ib0(e,f,g,h,a,b,c,d); k_xor(24,a,b,c,d);
    irot(a,b,c,d); ib7(a,b,c,d,e,f,g,h); k_xor(23,e,f,g,h);
    irot(e,f,g,h); ib6(e,f,g,h,a,b,c,d); k_xor(22,a,b,c,d);
    irot(a,b,c,d); ib5(a,b,c,d,e,f,g,h); k_xor(21,e,f,g,h);
    irot(e,f,g,h); ib4(e,f,g,h,a,b,c,d); k_xor(20,a,b,c,d);
    irot(a,b,c,d); ib3(a,b,c,d,e,f,g,h); k_xor(19,e,f,g,h);
    irot(e,f,g,h); ib2(e,f,g,h,a,b,c,d); k_xor(18,a,b,c,d);
    irot(a,b,c,d); ib1(a,b,c,d,e,f,g,h); k_xor(17,e,f,g,h);
    irot(e,f,g,h); ib0(e,f,g,h,a,b,c,d); k_xor(16,a,b,c,d);
    irot(a,b,c,d); ib7(a,b,c,d,e,f,g,h); k_xor(15,e,f,g,h);
    irot(e,f,g,h); ib6(e,f,g,h,a,b,c,d); k_xor(14,a,b,c,d);
    irot(a,b,c,d); ib5(a,b,c,d,e,f,g,h); k_xor(13,e,f,g,h);
    irot(e,f,g,h); ib4(e,f,g,h,a,b,c,d); k_xor(12,a,b,c,d);
    irot(a,b,c,d); ib3(a,b,c,d,e,f,g,h); k_xor(11,e,f,g,h);
    irot(e,f,g,h); ib2(e,f,g,h,a,b,c,d); k_xor(10,a,b,c,d);
    irot(a,b,c,d); ib1(a,b,c,d,e,f,g,h); k_xor( 9,e,f,g,h);
    irot(e,f,g,h); ib0(e,f,g,h,a,b,c,d); k_xor( 8,a,b,c,d);
    irot(a,b,c,d); ib7(a,b,c,d,e,f,g,h); k_xor( 7,e,f,g,h);
    irot(e,f,g,h); ib6(e,f,g,h,a,b,c,d); k_xor( 6,a,b,c,d);
    irot(a,b,c,d); ib5(a,b,c,d,e,f,g,h); k_xor( 5,e,f,g,h);
    irot(e,f,g,h); ib4(e,f,g,h,a,b,c,d); k_xor( 4,a,b,c,d);
    irot(a,b,c,d); ib3(a,b,c,d,e,f,g,h); k_xor( 3,e,f,g,h);
    irot(e,f,g,h); ib2(e,f,g,h,a,b,c,d); k_xor( 2,a,b,c,d);
    irot(a,b,c,d); ib1(a,b,c,d,e,f,g,h); k_xor( 1,e,f,g,h);
    irot(e,f,g,h); ib0(e,f,g,h,a,b,c,d); k_xor( 0,a,b,c,d);

    *((u_int32_t *)(out_blk     )) = le32toh(a);
    *((u_int32_t *)(out_blk +  4)) = le32toh(b);
    *((u_int32_t *)(out_blk +  8)) = le32toh(c);
    *((u_int32_t *)(out_blk + 12)) = le32toh(d);
}