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

#define NEW_SIMD_CODE

#include "inc_vendor.cl"
#include "inc_hash_constants.h"
#include "inc_hash_functions.cl"
#include "inc_types.cl"
#include "inc_common.cl"
#include "inc_simd.cl"

__constant u64a keccakf_rndc[24] =
{
  0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
  0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
  0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
  0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
  0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
  0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
  0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
  0x8000000000008080, 0x0000000080000001, 0x8000000080008008
};

#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] = rotl64 (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;      \
}

void m05000m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset)
{
  /**
   * modifier
   */

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

  /**
   * constants
   */

  const int 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 int 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
  };

  /**
   * loop
   */

  u32 w0l = w0[0];

  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
  {
    const u32x w0r = ix_create_bft (bfs_buf, il_pos);

    const u32x w0lr = w0l | w0r;

    u32x t0[4];
    u32x t1[4];
    u32x t2[4];
    u32x t3[4];

    t0[0] = w0lr;
    t0[1] = w0[1];
    t0[2] = w0[2];
    t0[3] = w0[3];
    t1[0] = w1[0];
    t1[1] = w1[1];
    t1[2] = w1[2];
    t1[3] = w1[3];
    t2[0] = w2[0];
    t2[1] = w2[1];
    t2[2] = w2[2];
    t2[3] = w2[3];
    t3[0] = w3[0];
    t3[1] = w3[1];
    t3[2] = w3[2];
    t3[3] = w3[3];

    /**
     * Keccak
     */

    u64x st[25];

    st[ 0] = hl32_to_64 (t0[1], t0[0]);
    st[ 1] = hl32_to_64 (t0[3], t0[2]);
    st[ 2] = hl32_to_64 (t1[1], t1[0]);
    st[ 3] = hl32_to_64 (t1[3], t1[2]);
    st[ 4] = hl32_to_64 (t2[1], t2[0]);
    st[ 5] = hl32_to_64 (t2[3], t2[2]);
    st[ 6] = hl32_to_64 (t3[1], t3[0]);
    st[ 7] = hl32_to_64 (t3[3], t3[2]);
    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] = 0x8000000000000000;
    st[17] = 0;
    st[18] = 0;
    st[19] = 0;
    st[20] = 0;
    st[21] = 0;
    st[22] = 0;
    st[23] = 0;
    st[24] = 0;

    int round;

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

      u64x bc0 = Theta1 (0);
      u64x bc1 = Theta1 (1);
      u64x bc2 = Theta1 (2);
      u64x bc3 = Theta1 (3);
      u64x bc4 = Theta1 (4);

      u64x t;

      t = bc4 ^ rotl64 (bc1, 1); Theta2 (0);
      t = bc0 ^ rotl64 (bc2, 1); Theta2 (1);
      t = bc1 ^ rotl64 (bc3, 1); Theta2 (2);
      t = bc2 ^ rotl64 (bc4, 1); Theta2 (3);
      t = bc3 ^ rotl64 (bc0, 1); Theta2 (4);

      // Rho Pi

      t = st[1];

      #ifdef _unroll
      #pragma unroll
      #else
      #pragma unroll 1
      #endif
      for (int i = 0; i < 24; i++)
      {
        Rho_Pi (i);
      }

      //  Chi

      #ifdef _unroll
      #pragma unroll
      #else
      #pragma unroll 1
      #endif
      for (int i = 0; i < 25; i += 5)
      {
        Chi (i);
      }

      //  Iota

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

    const u32x r0 = l32_from_64 (st[1]);
    const u32x r1 = h32_from_64 (st[1]);
    const u32x r2 = l32_from_64 (st[2]);
    const u32x r3 = h32_from_64 (st[2]);

    COMPARE_M_SIMD (r0, r1, r2, r3);
  }
}

void m05000s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset)
{
  /**
   * modifier
   */

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

  /**
   * constants
   */

  const int 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 int 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
  };

  /**
   * digest
   */

  const u32 search[4] =
  {
    digests_buf[digests_offset].digest_buf[DGST_R0],
    digests_buf[digests_offset].digest_buf[DGST_R1],
    digests_buf[digests_offset].digest_buf[DGST_R2],
    digests_buf[digests_offset].digest_buf[DGST_R3]
  };

  /**
   * loop
   */

  u32 w0l = w0[0];

  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
  {
    const u32x w0r = ix_create_bft (bfs_buf, il_pos);

    const u32x w0lr = w0l | w0r;

    u32x t0[4];
    u32x t1[4];
    u32x t2[4];
    u32x t3[4];

    t0[0] = w0lr;
    t0[1] = w0[1];
    t0[2] = w0[2];
    t0[3] = w0[3];
    t1[0] = w1[0];
    t1[1] = w1[1];
    t1[2] = w1[2];
    t1[3] = w1[3];
    t2[0] = w2[0];
    t2[1] = w2[1];
    t2[2] = w2[2];
    t2[3] = w2[3];
    t3[0] = w3[0];
    t3[1] = w3[1];
    t3[2] = w3[2];
    t3[3] = w3[3];

    /**
     * Keccak
     */

    u64x st[25];

    st[ 0] = hl32_to_64 (t0[1], t0[0]);
    st[ 1] = hl32_to_64 (t0[3], t0[2]);
    st[ 2] = hl32_to_64 (t1[1], t1[0]);
    st[ 3] = hl32_to_64 (t1[3], t1[2]);
    st[ 4] = hl32_to_64 (t2[1], t2[0]);
    st[ 5] = hl32_to_64 (t2[3], t2[2]);
    st[ 6] = hl32_to_64 (t3[1], t3[0]);
    st[ 7] = hl32_to_64 (t3[3], t3[2]);
    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] = 0x8000000000000000;
    st[17] = 0;
    st[18] = 0;
    st[19] = 0;
    st[20] = 0;
    st[21] = 0;
    st[22] = 0;
    st[23] = 0;
    st[24] = 0;

    int round;

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

      u64x bc0 = Theta1 (0);
      u64x bc1 = Theta1 (1);
      u64x bc2 = Theta1 (2);
      u64x bc3 = Theta1 (3);
      u64x bc4 = Theta1 (4);

      u64x t;

      t = bc4 ^ rotl64 (bc1, 1); Theta2 (0);
      t = bc0 ^ rotl64 (bc2, 1); Theta2 (1);
      t = bc1 ^ rotl64 (bc3, 1); Theta2 (2);
      t = bc2 ^ rotl64 (bc4, 1); Theta2 (3);
      t = bc3 ^ rotl64 (bc0, 1); Theta2 (4);

      // Rho Pi

      t = st[1];

      #ifdef _unroll
      #pragma unroll
      #else
      #pragma unroll 1
      #endif
      for (int i = 0; i < 24; i++)
      {
        Rho_Pi (i);
      }

      //  Chi

      #ifdef _unroll
      #pragma unroll
      #else
      #pragma unroll 1
      #endif
      for (int i = 0; i < 25; i += 5)
      {
        Chi (i);
      }

      //  Iota

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

    const u32x r0 = l32_from_64 (st[1]);
    const u32x r1 = h32_from_64 (st[1]);
    const u32x r2 = l32_from_64 (st[2]);
    const u32x r3 = h32_from_64 (st[2]);

    COMPARE_S_SIMD (r0, r1, r2, r3);
  }
}

__kernel void m05000_m04 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = 0;
  w1[1] = 0;
  w1[2] = 0;
  w1[3] = 0;

  u32 w2[4];

  w2[0] = 0;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;

  u32 w3[4];

  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}

__kernel void m05000_m08 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = pws[gid].i[ 4];
  w1[1] = pws[gid].i[ 5];
  w1[2] = pws[gid].i[ 6];
  w1[3] = pws[gid].i[ 7];

  u32 w2[4];

  w2[0] = 0;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;

  u32 w3[4];

  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}

__kernel void m05000_m16 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = pws[gid].i[ 4];
  w1[1] = pws[gid].i[ 5];
  w1[2] = pws[gid].i[ 6];
  w1[3] = pws[gid].i[ 7];

  u32 w2[4];

  w2[0] = pws[gid].i[ 8];
  w2[1] = pws[gid].i[ 9];
  w2[2] = pws[gid].i[10];
  w2[3] = pws[gid].i[11];

  u32 w3[4];

  w3[0] = pws[gid].i[12];
  w3[1] = pws[gid].i[13];
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000m (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}

__kernel void m05000_s04 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = 0;
  w1[1] = 0;
  w1[2] = 0;
  w1[3] = 0;

  u32 w2[4];

  w2[0] = 0;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;

  u32 w3[4];

  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}

__kernel void m05000_s08 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = pws[gid].i[ 4];
  w1[1] = pws[gid].i[ 5];
  w1[2] = pws[gid].i[ 6];
  w1[3] = pws[gid].i[ 7];

  u32 w2[4];

  w2[0] = 0;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;

  u32 w3[4];

  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}

__kernel void m05000_s16 (__global pw_t *pws, __constant const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u64 gid_max)
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w0[4];

  w0[0] = pws[gid].i[ 0];
  w0[1] = pws[gid].i[ 1];
  w0[2] = pws[gid].i[ 2];
  w0[3] = pws[gid].i[ 3];

  u32 w1[4];

  w1[0] = pws[gid].i[ 4];
  w1[1] = pws[gid].i[ 5];
  w1[2] = pws[gid].i[ 6];
  w1[3] = pws[gid].i[ 7];

  u32 w2[4];

  w2[0] = pws[gid].i[ 8];
  w2[1] = pws[gid].i[ 9];
  w2[2] = pws[gid].i[10];
  w2[3] = pws[gid].i[11];

  u32 w3[4];

  w3[0] = pws[gid].i[12];
  w3[1] = pws[gid].i[13];
  w3[2] = 0;
  w3[3] = 0;

  const u32 pw_len = pws[gid].pw_len;

  /**
   * main
   */

  m05000s (w0, w1, w2, w3, pw_len, pws, rules_buf, combs_buf, bfs_buf, tmps, hooks, bitmaps_buf_s1_a, bitmaps_buf_s1_b, bitmaps_buf_s1_c, bitmaps_buf_s1_d, bitmaps_buf_s2_a, bitmaps_buf_s2_b, bitmaps_buf_s2_c, bitmaps_buf_s2_d, plains_buf, digests_buf, hashes_shown, salt_bufs, esalt_bufs, d_return_buf, d_scryptV0_buf, d_scryptV1_buf, d_scryptV2_buf, d_scryptV3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset);
}