/**
 * Author......: See docs/credits.txt
 * License.....: MIT
 */

#ifdef KERNEL_STATIC
#include M2S(INCLUDE_PATH/inc_vendor.h)
#include M2S(INCLUDE_PATH/inc_types.h)
#include M2S(INCLUDE_PATH/inc_platform.cl)
#include M2S(INCLUDE_PATH/inc_common.cl)
#include M2S(INCLUDE_PATH/inc_hash_sha256.cl)
#endif

#define COMPARE_S M2S(INCLUDE_PATH/inc_comp_single.cl)
#define COMPARE_M M2S(INCLUDE_PATH/inc_comp_multi.cl)

typedef struct
{
  #ifndef SCRYPT_TMP_ELEM
  #define SCRYPT_TMP_ELEM 1
  #endif

  uint4 P[SCRYPT_TMP_ELEM];

} scrypt_tmp_t;

typedef struct ethereum_scrypt
{
  u32 salt_buf[16];
  u32 ciphertext[8];

} ethereum_scrypt_t;

#if defined IS_CUDA || defined IS_HIP

inline __device__ uint4 operator &  (const uint4  a, const u32   b) { return make_uint4 ((a.x &  b  ), (a.y &  b  ), (a.z &  b  ), (a.w &  b  ));  }
inline __device__ uint4 operator << (const uint4  a, const u32   b) { return make_uint4 ((a.x << b  ), (a.y << b  ), (a.z << b  ), (a.w << b  ));  }
inline __device__ uint4 operator >> (const uint4  a, const u32   b) { return make_uint4 ((a.x >> b  ), (a.y >> b  ), (a.z >> b  ), (a.w >> b  ));  }
inline __device__ uint4 operator +  (const uint4  a, const uint4 b) { return make_uint4 ((a.x +  b.x), (a.y +  b.y), (a.z +  b.z), (a.w +  b.w));  }
inline __device__ uint4 operator ^  (const uint4  a, const uint4 b) { return make_uint4 ((a.x ^  b.x), (a.y ^  b.y), (a.z ^  b.z), (a.w ^  b.w));  }
inline __device__ uint4 operator |  (const uint4  a, const uint4 b) { return make_uint4 ((a.x |  b.x), (a.y |  b.y), (a.z |  b.z), (a.w |  b.w));  }
inline __device__ void  operator ^= (      uint4 &a, const uint4 b) {                     a.x ^= b.x;   a.y ^= b.y;   a.z ^= b.z;   a.w ^= b.w;    }

inline __device__ uint4 rotate (const uint4 a, const int n)
{
  return ((a << n) | ((a >> (32 - n))));
}

#endif

DECLSPEC uint4 hc_swap32_4 (uint4 v)
{
  return (rotate ((v & 0x00FF00FF), 24u) | rotate ((v & 0xFF00FF00),  8u));
}

#define GET_SCRYPT_CNT(r,p) (2 * (r) * 16 * (p))
#define GET_SMIX_CNT(r,N)   (2 * (r) * 16 * (N))
#define GET_STATE_CNT(r)    (2 * (r) * 16)

#define SCRYPT_CNT  GET_SCRYPT_CNT (SCRYPT_R, SCRYPT_P)
#define SCRYPT_CNT4 (SCRYPT_CNT / 4)
#define STATE_CNT   GET_STATE_CNT  (SCRYPT_R)
#define STATE_CNT4  (STATE_CNT / 4)

#define ADD_ROTATE_XOR(r,i1,i2,s) (r) ^= rotate ((i1) + (i2), (s));

#if defined IS_CUDA || defined IS_HIP

#define SALSA20_2R()                        \
{                                           \
  ADD_ROTATE_XOR (X1, X0, X3,  7);          \
  ADD_ROTATE_XOR (X2, X1, X0,  9);          \
  ADD_ROTATE_XOR (X3, X2, X1, 13);          \
  ADD_ROTATE_XOR (X0, X3, X2, 18);          \
                                            \
  X1 = make_uint4 (X1.w, X1.x, X1.y, X1.z); \
  X2 = make_uint4 (X2.z, X2.w, X2.x, X2.y); \
  X3 = make_uint4 (X3.y, X3.z, X3.w, X3.x); \
                                            \
  ADD_ROTATE_XOR (X3, X0, X1,  7);          \
  ADD_ROTATE_XOR (X2, X3, X0,  9);          \
  ADD_ROTATE_XOR (X1, X2, X3, 13);          \
  ADD_ROTATE_XOR (X0, X1, X2, 18);          \
                                            \
  X1 = make_uint4 (X1.y, X1.z, X1.w, X1.x); \
  X2 = make_uint4 (X2.z, X2.w, X2.x, X2.y); \
  X3 = make_uint4 (X3.w, X3.x, X3.y, X3.z); \
}
#elif defined IS_METAL
#define SALSA20_2R()                        \
{                                           \
  ADD_ROTATE_XOR (X1, X0, X3,  7);          \
  ADD_ROTATE_XOR (X2, X1, X0,  9);          \
  ADD_ROTATE_XOR (X3, X2, X1, 13);          \
  ADD_ROTATE_XOR (X0, X3, X2, 18);          \
                                            \
  X1 = X1.wxyz;                             \
  X2 = X2.zwxy;                             \
  X3 = X3.yzwx;                             \
                                            \
  ADD_ROTATE_XOR (X3, X0, X1,  7);          \
  ADD_ROTATE_XOR (X2, X3, X0,  9);          \
  ADD_ROTATE_XOR (X1, X2, X3, 13);          \
  ADD_ROTATE_XOR (X0, X1, X2, 18);          \
                                            \
  X1 = X1.yzwx;                             \
  X2 = X2.zwxy;                             \
  X3 = X3.wxyz;                             \
}
#else
#define SALSA20_2R()                        \
{                                           \
  ADD_ROTATE_XOR (X1, X0, X3,  7);          \
  ADD_ROTATE_XOR (X2, X1, X0,  9);          \
  ADD_ROTATE_XOR (X3, X2, X1, 13);          \
  ADD_ROTATE_XOR (X0, X3, X2, 18);          \
                                            \
  X1 = X1.s3012;                            \
  X2 = X2.s2301;                            \
  X3 = X3.s1230;                            \
                                            \
  ADD_ROTATE_XOR (X3, X0, X1,  7);          \
  ADD_ROTATE_XOR (X2, X3, X0,  9);          \
  ADD_ROTATE_XOR (X1, X2, X3, 13);          \
  ADD_ROTATE_XOR (X0, X1, X2, 18);          \
                                            \
  X1 = X1.s1230;                            \
  X2 = X2.s2301;                            \
  X3 = X3.s3012;                            \
}
#endif

#define Coord(xd4,y,z) (((xd4) * ySIZE * zSIZE) + ((y) * zSIZE) + (z))
#define CO Coord(xd4,y,z)

DECLSPEC void salsa_r (PRIVATE_AS uint4 *TI)
{
  #if SCRYPT_R > 1

  uint4 TT[STATE_CNT4 / 2];

  for (int dst_off = 0, src_off = 4; src_off < STATE_CNT4; dst_off += 4, src_off += 8)
  {
    TT[dst_off + 0] = TI[src_off + 0];
    TT[dst_off + 1] = TI[src_off + 1];
    TT[dst_off + 2] = TI[src_off + 2];
    TT[dst_off + 3] = TI[src_off + 3];
  }

  for (int dst_off = 4, src_off = 8; src_off < STATE_CNT4; dst_off += 4, src_off += 8)
  {
    TI[dst_off + 0] = TI[src_off + 0];
    TI[dst_off + 1] = TI[src_off + 1];
    TI[dst_off + 2] = TI[src_off + 2];
    TI[dst_off + 3] = TI[src_off + 3];
  }

  for (int dst_off = STATE_CNT4 / 2, src_off = 0; dst_off < STATE_CNT4; dst_off += 4, src_off += 4)
  {
    TI[dst_off + 0] = TT[src_off + 0];
    TI[dst_off + 1] = TT[src_off + 1];
    TI[dst_off + 2] = TT[src_off + 2];
    TI[dst_off + 3] = TT[src_off + 3];
  }

  #endif

  uint4 R0 = TI[STATE_CNT4 - 4];
  uint4 R1 = TI[STATE_CNT4 - 3];
  uint4 R2 = TI[STATE_CNT4 - 2];
  uint4 R3 = TI[STATE_CNT4 - 1];

  for (int i = 0; i < STATE_CNT4; i += 4)
  {
    uint4 Y0 = TI[i + 0];
    uint4 Y1 = TI[i + 1];
    uint4 Y2 = TI[i + 2];
    uint4 Y3 = TI[i + 3];

    R0 = R0 ^ Y0;
    R1 = R1 ^ Y1;
    R2 = R2 ^ Y2;
    R3 = R3 ^ Y3;

    uint4 X0 = R0;
    uint4 X1 = R1;
    uint4 X2 = R2;
    uint4 X3 = R3;

    SALSA20_2R ();
    SALSA20_2R ();
    SALSA20_2R ();
    SALSA20_2R ();

    R0 = R0 + X0;
    R1 = R1 + X1;
    R2 = R2 + X2;
    R3 = R3 + X3;

    TI[i + 0] = R0;
    TI[i + 1] = R1;
    TI[i + 2] = R2;
    TI[i + 3] = R3;
  }
}

DECLSPEC void scrypt_smix_init (PRIVATE_AS uint4 *X, GLOBAL_AS uint4 *V0, GLOBAL_AS uint4 *V1, GLOBAL_AS uint4 *V2, GLOBAL_AS uint4 *V3, const u64 gid)
{
  const u32 ySIZE = SCRYPT_N / SCRYPT_TMTO;
  const u32 zSIZE = STATE_CNT4;

  const u32 x = (u32) gid;

  const u32 xd4 = x / 4;
  const u32 xm4 = x & 3;

  GLOBAL_AS uint4 *V;

  switch (xm4)
  {
    case 0: V = V0; break;
    case 1: V = V1; break;
    case 2: V = V2; break;
    case 3: V = V3; break;
  }

  #if SCRYPT_R > 1

  uint4 TT[STATE_CNT4];

  for (int z = 0; z < zSIZE; z++) TT[z] = X[z];

  for (int dst_off = 8, src_off = 4; dst_off < zSIZE; dst_off += 8, src_off += 4)
  {
    X[dst_off + 0] = TT[src_off + 0];
    X[dst_off + 1] = TT[src_off + 1];
    X[dst_off + 2] = TT[src_off + 2];
    X[dst_off + 3] = TT[src_off + 3];
  }

  for (int dst_off = 4, src_off = zSIZE / 2; dst_off < zSIZE; dst_off += 8, src_off += 4)
  {
    X[dst_off + 0] = TT[src_off + 0];
    X[dst_off + 1] = TT[src_off + 1];
    X[dst_off + 2] = TT[src_off + 2];
    X[dst_off + 3] = TT[src_off + 3];
  }

  #endif

  for (u32 y = 0; y < ySIZE; y++)
  {
    for (u32 z = 0; z < zSIZE; z++) V[CO] = X[z];

    for (u32 i = 0; i < SCRYPT_TMTO; i++) salsa_r (X);
  }
}

DECLSPEC void scrypt_smix_loop (PRIVATE_AS uint4 *X, GLOBAL_AS uint4 *V0, GLOBAL_AS uint4 *V1, GLOBAL_AS uint4 *V2, GLOBAL_AS uint4 *V3, const u64 gid)
{
  const u32 ySIZE = SCRYPT_N / SCRYPT_TMTO;
  const u32 zSIZE = STATE_CNT4;

  const u32 x = (u32) gid;

  const u32 xd4 = x / 4;
  const u32 xm4 = x & 3;

  GLOBAL_AS uint4 *V;

  switch (xm4)
  {
    case 0: V = V0; break;
    case 1: V = V1; break;
    case 2: V = V2; break;
    case 3: V = V3; break;
  }

  // note: fixed 1024 iterations = forced -u 1024

  for (u32 N_pos = 0; N_pos < 1024; N_pos++)
  {
    const u32 k = X[zSIZE - 4].x & (SCRYPT_N - 1);

    const u32 y = k / SCRYPT_TMTO;

    const u32 km = k - (y * SCRYPT_TMTO);

    uint4 T[STATE_CNT4];

    for (u32 z = 0; z < zSIZE; z++) T[z] = V[CO];

    for (u32 i = 0; i < km; i++) salsa_r (T);

    for (u32 z = 0; z < zSIZE; z++) X[z] ^= T[z];

    salsa_r (X);
  }
}

#ifndef KECCAK_ROUNDS
#define KECCAK_ROUNDS 24
#endif

#define Theta1(s) (st[0 + s] ^ st[5 + s] ^ st[10 + s] ^ st[15 + s] ^ st[20 + s])

#define Theta2(s)               \
{                               \
  st[ 0 + s] ^= t;              \
  st[ 5 + s] ^= t;              \
  st[10 + s] ^= t;              \
  st[15 + s] ^= t;              \
  st[20 + s] ^= t;              \
}

#define Rho_Pi(s)               \
{                               \
  u32 j = keccakf_piln[s];      \
  u32 k = keccakf_rotc[s];      \
  bc0 = st[j];                  \
  st[j] = hc_rotl64_S (t, k);   \
  t = bc0;                      \
}

#define Chi(s)                  \
{                               \
  bc0 = st[0 + s];              \
  bc1 = st[1 + s];              \
  bc2 = st[2 + s];              \
  bc3 = st[3 + s];              \
  bc4 = st[4 + s];              \
  st[0 + s] ^= ~bc1 & bc2;      \
  st[1 + s] ^= ~bc2 & bc3;      \
  st[2 + s] ^= ~bc3 & bc4;      \
  st[3 + s] ^= ~bc4 & bc0;      \
  st[4 + s] ^= ~bc0 & bc1;      \
}

CONSTANT_VK u64a keccakf_rndc[24] =
{
  KECCAK_RNDC_00, KECCAK_RNDC_01, KECCAK_RNDC_02, KECCAK_RNDC_03,
  KECCAK_RNDC_04, KECCAK_RNDC_05, KECCAK_RNDC_06, KECCAK_RNDC_07,
  KECCAK_RNDC_08, KECCAK_RNDC_09, KECCAK_RNDC_10, KECCAK_RNDC_11,
  KECCAK_RNDC_12, KECCAK_RNDC_13, KECCAK_RNDC_14, KECCAK_RNDC_15,
  KECCAK_RNDC_16, KECCAK_RNDC_17, KECCAK_RNDC_18, KECCAK_RNDC_19,
  KECCAK_RNDC_20, KECCAK_RNDC_21, KECCAK_RNDC_22, KECCAK_RNDC_23
};

DECLSPEC void keccak_transform_S (PRIVATE_AS u64 *st)
{
  const u8 keccakf_rotc[24] =
  {
     1,  3,  6, 10, 15, 21, 28, 36, 45, 55,  2, 14,
    27, 41, 56,  8, 25, 43, 62, 18, 39, 61, 20, 44
  };

  const u8 keccakf_piln[24] =
  {
    10,  7, 11, 17, 18,  3,  5, 16,  8, 21, 24,  4,
    15, 23, 19, 13, 12,  2, 20, 14, 22,  9,  6,  1
  };

  /**
   * Keccak
   */

  int round;

  for (round = 0; round < KECCAK_ROUNDS; round++)
  {
    // Theta

    u64 bc0 = Theta1 (0);
    u64 bc1 = Theta1 (1);
    u64 bc2 = Theta1 (2);
    u64 bc3 = Theta1 (3);
    u64 bc4 = Theta1 (4);

    u64 t;

    t = bc4 ^ hc_rotl64_S (bc1, 1); Theta2 (0);
    t = bc0 ^ hc_rotl64_S (bc2, 1); Theta2 (1);
    t = bc1 ^ hc_rotl64_S (bc3, 1); Theta2 (2);
    t = bc2 ^ hc_rotl64_S (bc4, 1); Theta2 (3);
    t = bc3 ^ hc_rotl64_S (bc0, 1); Theta2 (4);

    // Rho Pi

    t = st[1];

    Rho_Pi (0);
    Rho_Pi (1);
    Rho_Pi (2);
    Rho_Pi (3);
    Rho_Pi (4);
    Rho_Pi (5);
    Rho_Pi (6);
    Rho_Pi (7);
    Rho_Pi (8);
    Rho_Pi (9);
    Rho_Pi (10);
    Rho_Pi (11);
    Rho_Pi (12);
    Rho_Pi (13);
    Rho_Pi (14);
    Rho_Pi (15);
    Rho_Pi (16);
    Rho_Pi (17);
    Rho_Pi (18);
    Rho_Pi (19);
    Rho_Pi (20);
    Rho_Pi (21);
    Rho_Pi (22);
    Rho_Pi (23);

    //  Chi

    Chi (0);
    Chi (5);
    Chi (10);
    Chi (15);
    Chi (20);

    //  Iota

    st[0] ^= keccakf_rndc[round];
  }
}

KERNEL_FQ void m15700_init (KERN_ATTR_TMPS_ESALT (scrypt_tmp_t, ethereum_scrypt_t))
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= GID_CNT) return;

  sha256_hmac_ctx_t sha256_hmac_ctx;

  sha256_hmac_init_global_swap (&sha256_hmac_ctx, pws[gid].i, pws[gid].pw_len);

  sha256_hmac_update_global_swap (&sha256_hmac_ctx, salt_bufs[SALT_POS_HOST].salt_buf, salt_bufs[SALT_POS_HOST].salt_len);

  for (u32 i = 0, j = 1, k = 0; i < SCRYPT_CNT; i += 8, j += 1, k += 2)
  {
    sha256_hmac_ctx_t sha256_hmac_ctx2 = sha256_hmac_ctx;

    u32 w0[4];
    u32 w1[4];
    u32 w2[4];
    u32 w3[4];

    w0[0] = j;
    w0[1] = 0;
    w0[2] = 0;
    w0[3] = 0;
    w1[0] = 0;
    w1[1] = 0;
    w1[2] = 0;
    w1[3] = 0;
    w2[0] = 0;
    w2[1] = 0;
    w2[2] = 0;
    w2[3] = 0;
    w3[0] = 0;
    w3[1] = 0;
    w3[2] = 0;
    w3[3] = 0;

    sha256_hmac_update_64 (&sha256_hmac_ctx2, w0, w1, w2, w3, 4);

    sha256_hmac_final (&sha256_hmac_ctx2);

    u32 digest[8];

    digest[0] = sha256_hmac_ctx2.opad.h[0];
    digest[1] = sha256_hmac_ctx2.opad.h[1];
    digest[2] = sha256_hmac_ctx2.opad.h[2];
    digest[3] = sha256_hmac_ctx2.opad.h[3];
    digest[4] = sha256_hmac_ctx2.opad.h[4];
    digest[5] = sha256_hmac_ctx2.opad.h[5];
    digest[6] = sha256_hmac_ctx2.opad.h[6];
    digest[7] = sha256_hmac_ctx2.opad.h[7];

    #if defined IS_CUDA || defined IS_HIP
    const uint4 tmp0 = make_uint4 (digest[0], digest[1], digest[2], digest[3]);
    const uint4 tmp1 = make_uint4 (digest[4], digest[5], digest[6], digest[7]);
    #elif defined IS_METAL
    const uint4 tmp0 = uint4 (digest[0], digest[1], digest[2], digest[3]);
    const uint4 tmp1 = uint4 (digest[4], digest[5], digest[6], digest[7]);
    #else
    const uint4 tmp0 = (uint4) (digest[0], digest[1], digest[2], digest[3]);
    const uint4 tmp1 = (uint4) (digest[4], digest[5], digest[6], digest[7]);
    #endif

    tmps[gid].P[k + 0] = tmp0;
    tmps[gid].P[k + 1] = tmp1;
  }

  for (u32 l = 0; l < SCRYPT_CNT4; l += 4)
  {
    uint4 T[4];

    T[0] = tmps[gid].P[l + 0];
    T[1] = tmps[gid].P[l + 1];
    T[2] = tmps[gid].P[l + 2];
    T[3] = tmps[gid].P[l + 3];

    T[0] = hc_swap32_4 (T[0]);
    T[1] = hc_swap32_4 (T[1]);
    T[2] = hc_swap32_4 (T[2]);
    T[3] = hc_swap32_4 (T[3]);

    uint4 X[4];

    #if defined IS_CUDA || defined IS_HIP
    X[0] = make_uint4 (T[0].x, T[1].y, T[2].z, T[3].w);
    X[1] = make_uint4 (T[1].x, T[2].y, T[3].z, T[0].w);
    X[2] = make_uint4 (T[2].x, T[3].y, T[0].z, T[1].w);
    X[3] = make_uint4 (T[3].x, T[0].y, T[1].z, T[2].w);
    #elif defined IS_METAL
    X[0] = uint4 (T[0].x, T[1].y, T[2].z, T[3].w);
    X[1] = uint4 (T[1].x, T[2].y, T[3].z, T[0].w);
    X[2] = uint4 (T[2].x, T[3].y, T[0].z, T[1].w);
    X[3] = uint4 (T[3].x, T[0].y, T[1].z, T[2].w);
    #else
    X[0] = (uint4) (T[0].x, T[1].y, T[2].z, T[3].w);
    X[1] = (uint4) (T[1].x, T[2].y, T[3].z, T[0].w);
    X[2] = (uint4) (T[2].x, T[3].y, T[0].z, T[1].w);
    X[3] = (uint4) (T[3].x, T[0].y, T[1].z, T[2].w);
    #endif

    tmps[gid].P[l + 0] = X[0];
    tmps[gid].P[l + 1] = X[1];
    tmps[gid].P[l + 2] = X[2];
    tmps[gid].P[l + 3] = X[3];
  }
}

KERNEL_FQ void m15700_loop_prepare (KERN_ATTR_TMPS (scrypt_tmp_t))
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);
  const u64 lid = get_local_id (0);

  if (gid >= GID_CNT) return;

  // SCRYPT part, init V

  GLOBAL_AS uint4 *d_scrypt0_buf = (GLOBAL_AS uint4 *) d_extra0_buf;
  GLOBAL_AS uint4 *d_scrypt1_buf = (GLOBAL_AS uint4 *) d_extra1_buf;
  GLOBAL_AS uint4 *d_scrypt2_buf = (GLOBAL_AS uint4 *) d_extra2_buf;
  GLOBAL_AS uint4 *d_scrypt3_buf = (GLOBAL_AS uint4 *) d_extra3_buf;

  uint4 X[STATE_CNT4];

  const u32 P_offset = SALT_REPEAT * STATE_CNT4;

  GLOBAL_AS uint4 *P = tmps[gid].P + P_offset;

  for (int z = 0; z < STATE_CNT4; z++) X[z] = P[z];

  scrypt_smix_init (X, d_scrypt0_buf, d_scrypt1_buf, d_scrypt2_buf, d_scrypt3_buf, gid);

  for (int z = 0; z < STATE_CNT4; z++) P[z] = X[z];
}

KERNEL_FQ void m15700_loop (KERN_ATTR_TMPS (scrypt_tmp_t))
{
  const u64 gid = get_global_id (0);
  const u64 lid = get_local_id (0);

  if (gid >= GID_CNT) return;

  GLOBAL_AS uint4 *d_scrypt0_buf = (GLOBAL_AS uint4 *) d_extra0_buf;
  GLOBAL_AS uint4 *d_scrypt1_buf = (GLOBAL_AS uint4 *) d_extra1_buf;
  GLOBAL_AS uint4 *d_scrypt2_buf = (GLOBAL_AS uint4 *) d_extra2_buf;
  GLOBAL_AS uint4 *d_scrypt3_buf = (GLOBAL_AS uint4 *) d_extra3_buf;

  uint4 X[STATE_CNT4];

  const u32 P_offset = SALT_REPEAT * STATE_CNT4;

  GLOBAL_AS uint4 *P = tmps[gid].P + P_offset;

  for (int z = 0; z < STATE_CNT4; z++) X[z] = P[z];

  scrypt_smix_loop (X, d_scrypt0_buf, d_scrypt1_buf, d_scrypt2_buf, d_scrypt3_buf, gid);

  for (int z = 0; z < STATE_CNT4; z++) P[z] = X[z];
}

KERNEL_FQ void m15700_comp (KERN_ATTR_TMPS_ESALT (scrypt_tmp_t, ethereum_scrypt_t))
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);
  const u64 lid = get_local_id (0);

  if (gid >= GID_CNT) return;

  /**
   * 2nd pbkdf2, creates B
   */

  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  sha256_hmac_ctx_t ctx;

  sha256_hmac_init_global_swap (&ctx, pws[gid].i, pws[gid].pw_len);


  for (u32 i = 0; i < SCRYPT_CNT4; i += STATE_CNT4)
  {
   for (u32 j = 0; j < (STATE_CNT4 * 2); j += 8)
   {
    uint4 X[4];

    const u32 l =  i + j + ((j >= STATE_CNT4) ? (4 - STATE_CNT4) : 0);

    X[0] = tmps[gid].P[l + 0];
    X[1] = tmps[gid].P[l + 1];
    X[2] = tmps[gid].P[l + 2];
    X[3] = tmps[gid].P[l + 3];

    uint4 T[4];

    #if defined IS_CUDA || defined IS_HIP
    T[0] = make_uint4 (X[0].x, X[3].y, X[2].z, X[1].w);
    T[1] = make_uint4 (X[1].x, X[0].y, X[3].z, X[2].w);
    T[2] = make_uint4 (X[2].x, X[1].y, X[0].z, X[3].w);
    T[3] = make_uint4 (X[3].x, X[2].y, X[1].z, X[0].w);
    #elif defined IS_METAL
    T[0] = uint4 (X[0].x, X[3].y, X[2].z, X[1].w);
    T[1] = uint4 (X[1].x, X[0].y, X[3].z, X[2].w);
    T[2] = uint4 (X[2].x, X[1].y, X[0].z, X[3].w);
    T[3] = uint4 (X[3].x, X[2].y, X[1].z, X[0].w);
    #else
    T[0] = (uint4) (X[0].x, X[3].y, X[2].z, X[1].w);
    T[1] = (uint4) (X[1].x, X[0].y, X[3].z, X[2].w);
    T[2] = (uint4) (X[2].x, X[1].y, X[0].z, X[3].w);
    T[3] = (uint4) (X[3].x, X[2].y, X[1].z, X[0].w);
    #endif

    T[0] = hc_swap32_4 (T[0]);
    T[1] = hc_swap32_4 (T[1]);
    T[2] = hc_swap32_4 (T[2]);
    T[3] = hc_swap32_4 (T[3]);

    w0[0] = T[0].x;
    w0[1] = T[0].y;
    w0[2] = T[0].z;
    w0[3] = T[0].w;
    w1[0] = T[1].x;
    w1[1] = T[1].y;
    w1[2] = T[1].z;
    w1[3] = T[1].w;
    w2[0] = T[2].x;
    w2[1] = T[2].y;
    w2[2] = T[2].z;
    w2[3] = T[2].w;
    w3[0] = T[3].x;
    w3[1] = T[3].y;
    w3[2] = T[3].z;
    w3[3] = T[3].w;

    sha256_hmac_update_64 (&ctx, w0, w1, w2, w3, 64);
   }
  }

  w0[0] = 1;
  w0[1] = 0;
  w0[2] = 0;
  w0[3] = 0;
  w1[0] = 0;
  w1[1] = 0;
  w1[2] = 0;
  w1[3] = 0;
  w2[0] = 0;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;
  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = 0;

  sha256_hmac_update_64 (&ctx, w0, w1, w2, w3, 4);

  sha256_hmac_final (&ctx);

  /**
   * keccak
   */

  u32 ciphertext[8];

  ciphertext[0] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[0];
  ciphertext[1] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[1];
  ciphertext[2] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[2];
  ciphertext[3] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[3];
  ciphertext[4] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[4];
  ciphertext[5] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[5];
  ciphertext[6] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[6];
  ciphertext[7] = esalt_bufs[DIGESTS_OFFSET_HOST].ciphertext[7];

  u32 key[4];

  key[0] = hc_swap32_S (ctx.opad.h[4]);
  key[1] = hc_swap32_S (ctx.opad.h[5]);
  key[2] = hc_swap32_S (ctx.opad.h[6]);
  key[3] = hc_swap32_S (ctx.opad.h[7]);

  u64 st[25];

  st[ 0] = hl32_to_64_S (key[1], key[0]);
  st[ 1] = hl32_to_64_S (key[3], key[2]);
  st[ 2] = hl32_to_64_S (ciphertext[1], ciphertext[0]);
  st[ 3] = hl32_to_64_S (ciphertext[3], ciphertext[2]);
  st[ 4] = hl32_to_64_S (ciphertext[5], ciphertext[4]);
  st[ 5] = hl32_to_64_S (ciphertext[7], ciphertext[6]);
  st[ 6] = 0x01;
  st[ 7] = 0;
  st[ 8] = 0;
  st[ 9] = 0;
  st[10] = 0;
  st[11] = 0;
  st[12] = 0;
  st[13] = 0;
  st[14] = 0;
  st[15] = 0;
  st[16] = 0;
  st[17] = 0;
  st[18] = 0;
  st[19] = 0;
  st[20] = 0;
  st[21] = 0;
  st[22] = 0;
  st[23] = 0;
  st[24] = 0;

  const u32 mdlen = 32;

  const u32 rsiz = 200 - (2 * mdlen);

  const u32 add80w = (rsiz - 1) / 8;

  st[add80w] |= 0x8000000000000000UL;

  keccak_transform_S (st);

  const u32 r0 = l32_from_64_S (st[0]);
  const u32 r1 = h32_from_64_S (st[0]);
  const u32 r2 = l32_from_64_S (st[1]);
  const u32 r3 = h32_from_64_S (st[1]);

  #define il_pos 0

  #ifdef KERNEL_STATIC
  #include COMPARE_M
  #endif
}