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

//incompatible because of branches
//#define NEW_SIMD_CODE

#ifdef KERNEL_STATIC
#include "inc_vendor.h"
#include "inc_types.h"
#include "inc_common.cl"
#include "inc_simd.cl"
#endif

CONSTANT_AS CONSTSPEC u32a crc32tab[0x100] =
{
  0x00000000, 0x77073096, 0xee0e612c, 0x990951ba,
  0x076dc419, 0x706af48f, 0xe963a535, 0x9e6495a3,
  0x0edb8832, 0x79dcb8a4, 0xe0d5e91e, 0x97d2d988,
  0x09b64c2b, 0x7eb17cbd, 0xe7b82d07, 0x90bf1d91,
  0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de,
  0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7,
  0x136c9856, 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec,
  0x14015c4f, 0x63066cd9, 0xfa0f3d63, 0x8d080df5,
  0x3b6e20c8, 0x4c69105e, 0xd56041e4, 0xa2677172,
  0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b,
  0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940,
  0x32d86ce3, 0x45df5c75, 0xdcd60dcf, 0xabd13d59,
  0x26d930ac, 0x51de003a, 0xc8d75180, 0xbfd06116,
  0x21b4f4b5, 0x56b3c423, 0xcfba9599, 0xb8bda50f,
  0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
  0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d,
  0x76dc4190, 0x01db7106, 0x98d220bc, 0xefd5102a,
  0x71b18589, 0x06b6b51f, 0x9fbfe4a5, 0xe8b8d433,
  0x7807c9a2, 0x0f00f934, 0x9609a88e, 0xe10e9818,
  0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01,
  0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e,
  0x6c0695ed, 0x1b01a57b, 0x8208f4c1, 0xf50fc457,
  0x65b0d9c6, 0x12b7e950, 0x8bbeb8ea, 0xfcb9887c,
  0x62dd1ddf, 0x15da2d49, 0x8cd37cf3, 0xfbd44c65,
  0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2,
  0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb,
  0x4369e96a, 0x346ed9fc, 0xad678846, 0xda60b8d0,
  0x44042d73, 0x33031de5, 0xaa0a4c5f, 0xdd0d7cc9,
  0x5005713c, 0x270241aa, 0xbe0b1010, 0xc90c2086,
  0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
  0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4,
  0x59b33d17, 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad,
  0xedb88320, 0x9abfb3b6, 0x03b6e20c, 0x74b1d29a,
  0xead54739, 0x9dd277af, 0x04db2615, 0x73dc1683,
  0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8,
  0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1,
  0xf00f9344, 0x8708a3d2, 0x1e01f268, 0x6906c2fe,
  0xf762575d, 0x806567cb, 0x196c3671, 0x6e6b06e7,
  0xfed41b76, 0x89d32be0, 0x10da7a5a, 0x67dd4acc,
  0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5,
  0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252,
  0xd1bb67f1, 0xa6bc5767, 0x3fb506dd, 0x48b2364b,
  0xd80d2bda, 0xaf0a1b4c, 0x36034af6, 0x41047a60,
  0xdf60efc3, 0xa867df55, 0x316e8eef, 0x4669be79,
  0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
  0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f,
  0xc5ba3bbe, 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04,
  0xc2d7ffa7, 0xb5d0cf31, 0x2cd99e8b, 0x5bdeae1d,
  0x9b64c2b0, 0xec63f226, 0x756aa39c, 0x026d930a,
  0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713,
  0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38,
  0x92d28e9b, 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21,
  0x86d3d2d4, 0xf1d4e242, 0x68ddb3f8, 0x1fda836e,
  0x81be16cd, 0xf6b9265b, 0x6fb077e1, 0x18b74777,
  0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c,
  0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45,
  0xa00ae278, 0xd70dd2ee, 0x4e048354, 0x3903b3c2,
  0xa7672661, 0xd06016f7, 0x4969474d, 0x3e6e77db,
  0xaed16a4a, 0xd9d65adc, 0x40df0b66, 0x37d83bf0,
  0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
  0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6,
  0xbad03605, 0xcdd70693, 0x54de5729, 0x23d967bf,
  0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94,
  0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d
};

DECLSPEC static u32x round_crc32 (u32x a, const u32x v)
{
  const u32x k = (a ^ v) & 0xff;

  const u32x s = a >> 8;

  #if   VECT_SIZE == 1
  a = (u32x) crc32tab[k];
  #elif VECT_SIZE == 2
  a = (u32x) (crc32tab[k.s0], crc32tab[k.s1]);
  #elif VECT_SIZE == 4
  a = (u32x) (crc32tab[k.s0], crc32tab[k.s1], crc32tab[k.s2], crc32tab[k.s3]);
  #elif VECT_SIZE == 8
  a = (u32x) (crc32tab[k.s0], crc32tab[k.s1], crc32tab[k.s2], crc32tab[k.s3], crc32tab[k.s4], crc32tab[k.s5], crc32tab[k.s6], crc32tab[k.s7]);
  #elif VECT_SIZE == 16
  a = (u32x) (crc32tab[k.s0], crc32tab[k.s1], crc32tab[k.s2], crc32tab[k.s3], crc32tab[k.s4], crc32tab[k.s5], crc32tab[k.s6], crc32tab[k.s7], crc32tab[k.s8], crc32tab[k.s9], crc32tab[k.sa], crc32tab[k.sb], crc32tab[k.sc], crc32tab[k.sd], crc32tab[k.se], crc32tab[k.sf]);
  #endif

  a ^= s;

  return a;
}

DECLSPEC static u32x crc32 (const u32x *w, const u32 pw_len, const u32 iv)
{
  u32x a = iv ^ ~0;

  if (pw_len >=  1) a = round_crc32 (a, w[0] >>  0);
  if (pw_len >=  2) a = round_crc32 (a, w[0] >>  8);
  if (pw_len >=  3) a = round_crc32 (a, w[0] >> 16);
  if (pw_len >=  4) a = round_crc32 (a, w[0] >> 24);

  for (u32 i = 4, j = 1; i < pw_len; i += 4, j += 1)
  {
    if (pw_len >= (i + 1)) a = round_crc32 (a, w[j] >>  0);
    if (pw_len >= (i + 2)) a = round_crc32 (a, w[j] >>  8);
    if (pw_len >= (i + 3)) a = round_crc32 (a, w[j] >> 16);
    if (pw_len >= (i + 4)) a = round_crc32 (a, w[j] >> 24);
  }

  return ~a;
}

DECLSPEC static void m11500m (u32 *w, const u32 pw_len, KERN_ATTR_BASIC ())
{
  /**
   * modifier
   */

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

  /**
   * salt
   */

  const u32 iv = salt_bufs[salt_pos].salt_buf[0];

  /**
   * loop
   */

  u32 w0l = w[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 w0 = w0l | w0r;

    /**
     * crc32
     */

    u32x w_t[16];

    w_t[ 0] = w0;
    w_t[ 1] = w[ 1];
    w_t[ 2] = w[ 2];
    w_t[ 3] = w[ 3];
    w_t[ 4] = w[ 4];
    w_t[ 5] = w[ 5];
    w_t[ 6] = w[ 6];
    w_t[ 7] = w[ 7];
    w_t[ 8] = w[ 8];
    w_t[ 9] = w[ 9];
    w_t[10] = w[10];
    w_t[11] = w[11];
    w_t[12] = w[12];
    w_t[13] = w[13];
    w_t[14] = w[14];
    w_t[15] = w[15];

    u32x a = crc32 (w_t, pw_len, iv);

    u32x z = 0;

    COMPARE_M_SIMD (a, z, z, z);
  }
}

DECLSPEC static void m11500s (u32 *w, const u32 pw_len, KERN_ATTR_BASIC ())
{
  /**
   * modifier
   */

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

  /**
   * salt
   */

  const u32 iv = salt_bufs[salt_pos].salt_buf[0];

  /**
   * digest
   */

  const u32 search[4] =
  {
    digests_buf[digests_offset].digest_buf[DGST_R0],
    0,
    0,
    0
  };

  /**
   * loop
   */

  u32 w0l = w[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 w0 = w0l | w0r;

    /**
     * crc32
     */

    u32x w_t[16];

    w_t[ 0] = w0;
    w_t[ 1] = w[ 1];
    w_t[ 2] = w[ 2];
    w_t[ 3] = w[ 3];
    w_t[ 4] = w[ 4];
    w_t[ 5] = w[ 5];
    w_t[ 6] = w[ 6];
    w_t[ 7] = w[ 7];
    w_t[ 8] = w[ 8];
    w_t[ 9] = w[ 9];
    w_t[10] = w[10];
    w_t[11] = w[11];
    w_t[12] = w[12];
    w_t[13] = w[13];
    w_t[14] = w[14];
    w_t[15] = w[15];

    u32x a = crc32 (w_t, pw_len, iv);

    u32x z = 0;

    COMPARE_S_SIMD (a, z, z, z);
  }
}

KERNEL_FQ void m11500_m04 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = 0;
  w[ 5] = 0;
  w[ 6] = 0;
  w[ 7] = 0;
  w[ 8] = 0;
  w[ 9] = 0;
  w[10] = 0;
  w[11] = 0;
  w[12] = 0;
  w[13] = 0;
  w[14] = 0;
  w[15] = 0;

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500m (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}

KERNEL_FQ void m11500_m08 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = pws[gid].i[ 4];
  w[ 5] = pws[gid].i[ 5];
  w[ 6] = pws[gid].i[ 6];
  w[ 7] = pws[gid].i[ 7];
  w[ 8] = 0;
  w[ 9] = 0;
  w[10] = 0;
  w[11] = 0;
  w[12] = 0;
  w[13] = 0;
  w[14] = 0;
  w[15] = 0;

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500m (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}

KERNEL_FQ void m11500_m16 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = pws[gid].i[ 4];
  w[ 5] = pws[gid].i[ 5];
  w[ 6] = pws[gid].i[ 6];
  w[ 7] = pws[gid].i[ 7];
  w[ 8] = pws[gid].i[ 8];
  w[ 9] = pws[gid].i[ 9];
  w[10] = pws[gid].i[10];
  w[11] = pws[gid].i[11];
  w[12] = pws[gid].i[12];
  w[13] = pws[gid].i[13];
  w[14] = pws[gid].i[14];
  w[15] = pws[gid].i[15];

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500m (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}

KERNEL_FQ void m11500_s04 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = 0;
  w[ 5] = 0;
  w[ 6] = 0;
  w[ 7] = 0;
  w[ 8] = 0;
  w[ 9] = 0;
  w[10] = 0;
  w[11] = 0;
  w[12] = 0;
  w[13] = 0;
  w[14] = 0;
  w[15] = 0;

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500s (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}

KERNEL_FQ void m11500_s08 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = pws[gid].i[ 4];
  w[ 5] = pws[gid].i[ 5];
  w[ 6] = pws[gid].i[ 6];
  w[ 7] = pws[gid].i[ 7];
  w[ 8] = 0;
  w[ 9] = 0;
  w[10] = 0;
  w[11] = 0;
  w[12] = 0;
  w[13] = 0;
  w[14] = 0;
  w[15] = 0;

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500s (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}

KERNEL_FQ void m11500_s16 (KERN_ATTR_BASIC ())
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= gid_max) return;

  u32 w[16];

  w[ 0] = pws[gid].i[ 0];
  w[ 1] = pws[gid].i[ 1];
  w[ 2] = pws[gid].i[ 2];
  w[ 3] = pws[gid].i[ 3];
  w[ 4] = pws[gid].i[ 4];
  w[ 5] = pws[gid].i[ 5];
  w[ 6] = pws[gid].i[ 6];
  w[ 7] = pws[gid].i[ 7];
  w[ 8] = pws[gid].i[ 8];
  w[ 9] = pws[gid].i[ 9];
  w[10] = pws[gid].i[10];
  w[11] = pws[gid].i[11];
  w[12] = pws[gid].i[12];
  w[13] = pws[gid].i[13];
  w[14] = pws[gid].i[14];
  w[15] = pws[gid].i[15];

  const u32 pw_len = pws[gid].pw_len & 63;

  /**
   * main
   */

  m11500s (w, 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_extra0_buf, d_extra1_buf, d_extra2_buf, d_extra3_buf, bitmap_mask, bitmap_shift1, bitmap_shift2, salt_pos, loop_pos, loop_cnt, il_cnt, digests_cnt, digests_offset, combs_mode, gid_max);
}