// important notes on this:
// input buf unused bytes needs to be set to zero
// input buf needs to be in algorithm native byte order (md5 = LE, sha1 = BE, etc)
// input buf needs to be 64 byte aligned when using md5_update()

typedef struct sha1_ctx
{
  u32 h[5];

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

  int len;

} sha1_ctx_t;

void sha1_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], u32 digest[5])
{
  u32 a = digest[0];
  u32 b = digest[1];
  u32 c = digest[2];
  u32 d = digest[3];
  u32 e = digest[4];

  u32 w0_t = w0[0];
  u32 w1_t = w0[1];
  u32 w2_t = w0[2];
  u32 w3_t = w0[3];
  u32 w4_t = w1[0];
  u32 w5_t = w1[1];
  u32 w6_t = w1[2];
  u32 w7_t = w1[3];
  u32 w8_t = w2[0];
  u32 w9_t = w2[1];
  u32 wa_t = w2[2];
  u32 wb_t = w2[3];
  u32 wc_t = w3[0];
  u32 wd_t = w3[1];
  u32 we_t = w3[2];
  u32 wf_t = w3[3];

  #define K SHA1C00

  SHA1_STEP_S (SHA1_F0o, a, b, c, d, e, w0_t);
  SHA1_STEP_S (SHA1_F0o, e, a, b, c, d, w1_t);
  SHA1_STEP_S (SHA1_F0o, d, e, a, b, c, w2_t);
  SHA1_STEP_S (SHA1_F0o, c, d, e, a, b, w3_t);
  SHA1_STEP_S (SHA1_F0o, b, c, d, e, a, w4_t);
  SHA1_STEP_S (SHA1_F0o, a, b, c, d, e, w5_t);
  SHA1_STEP_S (SHA1_F0o, e, a, b, c, d, w6_t);
  SHA1_STEP_S (SHA1_F0o, d, e, a, b, c, w7_t);
  SHA1_STEP_S (SHA1_F0o, c, d, e, a, b, w8_t);
  SHA1_STEP_S (SHA1_F0o, b, c, d, e, a, w9_t);
  SHA1_STEP_S (SHA1_F0o, a, b, c, d, e, wa_t);
  SHA1_STEP_S (SHA1_F0o, e, a, b, c, d, wb_t);
  SHA1_STEP_S (SHA1_F0o, d, e, a, b, c, wc_t);
  SHA1_STEP_S (SHA1_F0o, c, d, e, a, b, wd_t);
  SHA1_STEP_S (SHA1_F0o, b, c, d, e, a, we_t);
  SHA1_STEP_S (SHA1_F0o, a, b, c, d, e, wf_t);
  w0_t = rotl32_S ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP_S (SHA1_F0o, e, a, b, c, d, w0_t);
  w1_t = rotl32_S ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP_S (SHA1_F0o, d, e, a, b, c, w1_t);
  w2_t = rotl32_S ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP_S (SHA1_F0o, c, d, e, a, b, w2_t);
  w3_t = rotl32_S ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP_S (SHA1_F0o, b, c, d, e, a, w3_t);

  #undef K
  #define K SHA1C01

  w4_t = rotl32_S ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, w4_t);
  w5_t = rotl32_S ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, w5_t);
  w6_t = rotl32_S ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, w6_t);
  w7_t = rotl32_S ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, w7_t);
  w8_t = rotl32_S ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, w8_t);
  w9_t = rotl32_S ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, w9_t);
  wa_t = rotl32_S ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, wa_t);
  wb_t = rotl32_S ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, wb_t);
  wc_t = rotl32_S ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, wc_t);
  wd_t = rotl32_S ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, wd_t);
  we_t = rotl32_S ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, we_t);
  wf_t = rotl32_S ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, wf_t);
  w0_t = rotl32_S ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, w0_t);
  w1_t = rotl32_S ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, w1_t);
  w2_t = rotl32_S ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, w2_t);
  w3_t = rotl32_S ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, w3_t);
  w4_t = rotl32_S ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, w4_t);
  w5_t = rotl32_S ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, w5_t);
  w6_t = rotl32_S ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, w6_t);
  w7_t = rotl32_S ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, w7_t);

  #undef K
  #define K SHA1C02

  w8_t = rotl32_S ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP_S (SHA1_F2o, a, b, c, d, e, w8_t);
  w9_t = rotl32_S ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP_S (SHA1_F2o, e, a, b, c, d, w9_t);
  wa_t = rotl32_S ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP_S (SHA1_F2o, d, e, a, b, c, wa_t);
  wb_t = rotl32_S ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP_S (SHA1_F2o, c, d, e, a, b, wb_t);
  wc_t = rotl32_S ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP_S (SHA1_F2o, b, c, d, e, a, wc_t);
  wd_t = rotl32_S ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP_S (SHA1_F2o, a, b, c, d, e, wd_t);
  we_t = rotl32_S ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP_S (SHA1_F2o, e, a, b, c, d, we_t);
  wf_t = rotl32_S ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP_S (SHA1_F2o, d, e, a, b, c, wf_t);
  w0_t = rotl32_S ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP_S (SHA1_F2o, c, d, e, a, b, w0_t);
  w1_t = rotl32_S ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP_S (SHA1_F2o, b, c, d, e, a, w1_t);
  w2_t = rotl32_S ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP_S (SHA1_F2o, a, b, c, d, e, w2_t);
  w3_t = rotl32_S ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP_S (SHA1_F2o, e, a, b, c, d, w3_t);
  w4_t = rotl32_S ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP_S (SHA1_F2o, d, e, a, b, c, w4_t);
  w5_t = rotl32_S ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP_S (SHA1_F2o, c, d, e, a, b, w5_t);
  w6_t = rotl32_S ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP_S (SHA1_F2o, b, c, d, e, a, w6_t);
  w7_t = rotl32_S ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP_S (SHA1_F2o, a, b, c, d, e, w7_t);
  w8_t = rotl32_S ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP_S (SHA1_F2o, e, a, b, c, d, w8_t);
  w9_t = rotl32_S ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP_S (SHA1_F2o, d, e, a, b, c, w9_t);
  wa_t = rotl32_S ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP_S (SHA1_F2o, c, d, e, a, b, wa_t);
  wb_t = rotl32_S ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP_S (SHA1_F2o, b, c, d, e, a, wb_t);

  #undef K
  #define K SHA1C03

  wc_t = rotl32_S ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, wc_t);
  wd_t = rotl32_S ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, wd_t);
  we_t = rotl32_S ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, we_t);
  wf_t = rotl32_S ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, wf_t);
  w0_t = rotl32_S ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, w0_t);
  w1_t = rotl32_S ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, w1_t);
  w2_t = rotl32_S ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, w2_t);
  w3_t = rotl32_S ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, w3_t);
  w4_t = rotl32_S ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, w4_t);
  w5_t = rotl32_S ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, w5_t);
  w6_t = rotl32_S ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, w6_t);
  w7_t = rotl32_S ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, w7_t);
  w8_t = rotl32_S ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, w8_t);
  w9_t = rotl32_S ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, w9_t);
  wa_t = rotl32_S ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, wa_t);
  wb_t = rotl32_S ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP_S (SHA1_F1, a, b, c, d, e, wb_t);
  wc_t = rotl32_S ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP_S (SHA1_F1, e, a, b, c, d, wc_t);
  wd_t = rotl32_S ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP_S (SHA1_F1, d, e, a, b, c, wd_t);
  we_t = rotl32_S ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP_S (SHA1_F1, c, d, e, a, b, we_t);
  wf_t = rotl32_S ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP_S (SHA1_F1, b, c, d, e, a, wf_t);

  #undef K

  digest[0] += a;
  digest[1] += b;
  digest[2] += c;
  digest[3] += d;
  digest[4] += e;
}

void sha1_init (sha1_ctx_t *ctx)
{
  ctx->h[0] = SHA1M_A;
  ctx->h[1] = SHA1M_B;
  ctx->h[2] = SHA1M_C;
  ctx->h[3] = SHA1M_D;
  ctx->h[4] = SHA1M_E;

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

  ctx->len = 0;
}

void sha1_update_64 (sha1_ctx_t *ctx, u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const int len)
{
  #ifdef IS_AMD
  volatile const int pos = ctx->len & 63;
  #else
  const int pos = ctx->len & 63;
  #endif

  ctx->len += len;

  if ((pos + len) < 64)
  {
    switch_buffer_by_offset_be_S (w0, w1, w2, w3, pos);

    ctx->w0[0] |= w0[0];
    ctx->w0[1] |= w0[1];
    ctx->w0[2] |= w0[2];
    ctx->w0[3] |= w0[3];
    ctx->w1[0] |= w1[0];
    ctx->w1[1] |= w1[1];
    ctx->w1[2] |= w1[2];
    ctx->w1[3] |= w1[3];
    ctx->w2[0] |= w2[0];
    ctx->w2[1] |= w2[1];
    ctx->w2[2] |= w2[2];
    ctx->w2[3] |= w2[3];
    ctx->w3[0] |= w3[0];
    ctx->w3[1] |= w3[1];
    ctx->w3[2] |= w3[2];
    ctx->w3[3] |= w3[3];
  }
  else
  {
    u32 c0[4] = { 0 };
    u32 c1[4] = { 0 };
    u32 c2[4] = { 0 };
    u32 c3[4] = { 0 };

    switch_buffer_by_offset_carry_be_S (w0, w1, w2, w3, c0, c1, c2, c3, pos);

    ctx->w0[0] |= w0[0];
    ctx->w0[1] |= w0[1];
    ctx->w0[2] |= w0[2];
    ctx->w0[3] |= w0[3];
    ctx->w1[0] |= w1[0];
    ctx->w1[1] |= w1[1];
    ctx->w1[2] |= w1[2];
    ctx->w1[3] |= w1[3];
    ctx->w2[0] |= w2[0];
    ctx->w2[1] |= w2[1];
    ctx->w2[2] |= w2[2];
    ctx->w2[3] |= w2[3];
    ctx->w3[0] |= w3[0];
    ctx->w3[1] |= w3[1];
    ctx->w3[2] |= w3[2];
    ctx->w3[3] |= w3[3];

    sha1_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);

    ctx->w0[0] = c0[0];
    ctx->w0[1] = c0[1];
    ctx->w0[2] = c0[2];
    ctx->w0[3] = c0[3];
    ctx->w1[0] = c1[0];
    ctx->w1[1] = c1[1];
    ctx->w1[2] = c1[2];
    ctx->w1[3] = c1[3];
    ctx->w2[0] = c2[0];
    ctx->w2[1] = c2[1];
    ctx->w2[2] = c2[2];
    ctx->w2[3] = c2[3];
    ctx->w3[0] = c3[0];
    ctx->w3[1] = c3[1];
    ctx->w3[2] = c3[2];
    ctx->w3[3] = c3[3];
  }
}

void sha1_update (sha1_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    sha1_update_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  sha1_update_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_swap (sha1_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    w0[0] = swap32_S (w0[0]);
    w0[1] = swap32_S (w0[1]);
    w0[2] = swap32_S (w0[2]);
    w0[3] = swap32_S (w0[3]);
    w1[0] = swap32_S (w1[0]);
    w1[1] = swap32_S (w1[1]);
    w1[2] = swap32_S (w1[2]);
    w1[3] = swap32_S (w1[3]);
    w2[0] = swap32_S (w2[0]);
    w2[1] = swap32_S (w2[1]);
    w2[2] = swap32_S (w2[2]);
    w2[3] = swap32_S (w2[3]);
    w3[0] = swap32_S (w3[0]);
    w3[1] = swap32_S (w3[1]);
    w3[2] = swap32_S (w3[2]);
    w3[3] = swap32_S (w3[3]);

    sha1_update_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  w0[0] = swap32_S (w0[0]);
  w0[1] = swap32_S (w0[1]);
  w0[2] = swap32_S (w0[2]);
  w0[3] = swap32_S (w0[3]);
  w1[0] = swap32_S (w1[0]);
  w1[1] = swap32_S (w1[1]);
  w1[2] = swap32_S (w1[2]);
  w1[3] = swap32_S (w1[3]);
  w2[0] = swap32_S (w2[0]);
  w2[1] = swap32_S (w2[1]);
  w2[2] = swap32_S (w2[2]);
  w2[3] = swap32_S (w2[3]);
  w3[0] = swap32_S (w3[0]);
  w3[1] = swap32_S (w3[1]);
  w3[2] = swap32_S (w3[2]);
  w3[3] = swap32_S (w3[3]);

  sha1_update_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_utf16le (sha1_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le_S (w1, w2, w3);
    make_utf16le_S (w0, w0, w1);

    sha1_update_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le_S (w1, w2, w3);
  make_utf16le_S (w0, w0, w1);

  sha1_update_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_update_utf16le_swap (sha1_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le_S (w1, w2, w3);
    make_utf16le_S (w0, w0, w1);

    w0[0] = swap32_S (w0[0]);
    w0[1] = swap32_S (w0[1]);
    w0[2] = swap32_S (w0[2]);
    w0[3] = swap32_S (w0[3]);
    w1[0] = swap32_S (w1[0]);
    w1[1] = swap32_S (w1[1]);
    w1[2] = swap32_S (w1[2]);
    w1[3] = swap32_S (w1[3]);
    w2[0] = swap32_S (w2[0]);
    w2[1] = swap32_S (w2[1]);
    w2[2] = swap32_S (w2[2]);
    w2[3] = swap32_S (w2[3]);
    w3[0] = swap32_S (w3[0]);
    w3[1] = swap32_S (w3[1]);
    w3[2] = swap32_S (w3[2]);
    w3[3] = swap32_S (w3[3]);

    sha1_update_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le_S (w1, w2, w3);
  make_utf16le_S (w0, w0, w1);

  w0[0] = swap32_S (w0[0]);
  w0[1] = swap32_S (w0[1]);
  w0[2] = swap32_S (w0[2]);
  w0[3] = swap32_S (w0[3]);
  w1[0] = swap32_S (w1[0]);
  w1[1] = swap32_S (w1[1]);
  w1[2] = swap32_S (w1[2]);
  w1[3] = swap32_S (w1[3]);
  w2[0] = swap32_S (w2[0]);
  w2[1] = swap32_S (w2[1]);
  w2[2] = swap32_S (w2[2]);
  w2[3] = swap32_S (w2[3]);
  w3[0] = swap32_S (w3[0]);
  w3[1] = swap32_S (w3[1]);
  w3[2] = swap32_S (w3[2]);
  w3[3] = swap32_S (w3[3]);

  sha1_update_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_update_global (sha1_ctx_t *ctx, const __global u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    sha1_update_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  sha1_update_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_global_swap (sha1_ctx_t *ctx, const __global u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    w0[0] = swap32_S (w0[0]);
    w0[1] = swap32_S (w0[1]);
    w0[2] = swap32_S (w0[2]);
    w0[3] = swap32_S (w0[3]);
    w1[0] = swap32_S (w1[0]);
    w1[1] = swap32_S (w1[1]);
    w1[2] = swap32_S (w1[2]);
    w1[3] = swap32_S (w1[3]);
    w2[0] = swap32_S (w2[0]);
    w2[1] = swap32_S (w2[1]);
    w2[2] = swap32_S (w2[2]);
    w2[3] = swap32_S (w2[3]);
    w3[0] = swap32_S (w3[0]);
    w3[1] = swap32_S (w3[1]);
    w3[2] = swap32_S (w3[2]);
    w3[3] = swap32_S (w3[3]);

    sha1_update_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  w0[0] = swap32_S (w0[0]);
  w0[1] = swap32_S (w0[1]);
  w0[2] = swap32_S (w0[2]);
  w0[3] = swap32_S (w0[3]);
  w1[0] = swap32_S (w1[0]);
  w1[1] = swap32_S (w1[1]);
  w1[2] = swap32_S (w1[2]);
  w1[3] = swap32_S (w1[3]);
  w2[0] = swap32_S (w2[0]);
  w2[1] = swap32_S (w2[1]);
  w2[2] = swap32_S (w2[2]);
  w2[3] = swap32_S (w2[3]);
  w3[0] = swap32_S (w3[0]);
  w3[1] = swap32_S (w3[1]);
  w3[2] = swap32_S (w3[2]);
  w3[3] = swap32_S (w3[3]);

  sha1_update_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_global_utf16le (sha1_ctx_t *ctx, const __global u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le_S (w1, w2, w3);
    make_utf16le_S (w0, w0, w1);

    sha1_update_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le_S (w1, w2, w3);
  make_utf16le_S (w0, w0, w1);

  sha1_update_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_update_global_utf16le_swap (sha1_ctx_t *ctx, const __global u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le_S (w1, w2, w3);
    make_utf16le_S (w0, w0, w1);

    w0[0] = swap32_S (w0[0]);
    w0[1] = swap32_S (w0[1]);
    w0[2] = swap32_S (w0[2]);
    w0[3] = swap32_S (w0[3]);
    w1[0] = swap32_S (w1[0]);
    w1[1] = swap32_S (w1[1]);
    w1[2] = swap32_S (w1[2]);
    w1[3] = swap32_S (w1[3]);
    w2[0] = swap32_S (w2[0]);
    w2[1] = swap32_S (w2[1]);
    w2[2] = swap32_S (w2[2]);
    w2[3] = swap32_S (w2[3]);
    w3[0] = swap32_S (w3[0]);
    w3[1] = swap32_S (w3[1]);
    w3[2] = swap32_S (w3[2]);
    w3[3] = swap32_S (w3[3]);

    sha1_update_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le_S (w1, w2, w3);
  make_utf16le_S (w0, w0, w1);

  w0[0] = swap32_S (w0[0]);
  w0[1] = swap32_S (w0[1]);
  w0[2] = swap32_S (w0[2]);
  w0[3] = swap32_S (w0[3]);
  w1[0] = swap32_S (w1[0]);
  w1[1] = swap32_S (w1[1]);
  w1[2] = swap32_S (w1[2]);
  w1[3] = swap32_S (w1[3]);
  w2[0] = swap32_S (w2[0]);
  w2[1] = swap32_S (w2[1]);
  w2[2] = swap32_S (w2[2]);
  w2[3] = swap32_S (w2[3]);
  w3[0] = swap32_S (w3[0]);
  w3[1] = swap32_S (w3[1]);
  w3[2] = swap32_S (w3[2]);
  w3[3] = swap32_S (w3[3]);

  sha1_update_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_final (sha1_ctx_t *ctx)
{
  const int pos = ctx->len & 63;

  append_0x80_4x4_S (ctx->w0, ctx->w1, ctx->w2, ctx->w3, pos ^ 3);

  if (pos >= 56)
  {
    sha1_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);

    ctx->w0[0] = 0;
    ctx->w0[1] = 0;
    ctx->w0[2] = 0;
    ctx->w0[3] = 0;
    ctx->w1[0] = 0;
    ctx->w1[1] = 0;
    ctx->w1[2] = 0;
    ctx->w1[3] = 0;
    ctx->w2[0] = 0;
    ctx->w2[1] = 0;
    ctx->w2[2] = 0;
    ctx->w2[3] = 0;
    ctx->w3[0] = 0;
    ctx->w3[1] = 0;
    ctx->w3[2] = 0;
    ctx->w3[3] = 0;
  }

  ctx->w3[2] = 0;
  ctx->w3[3] = ctx->len * 8;

  sha1_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);
}

// sha1_hmac

typedef struct sha1_hmac_ctx
{
  sha1_ctx_t ipad;
  sha1_ctx_t opad;

} sha1_hmac_ctx_t;

void sha1_hmac_init_64 (sha1_hmac_ctx_t *ctx, const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4])
{
  u32 t0[4];
  u32 t1[4];
  u32 t2[4];
  u32 t3[4];

  // ipad

  t0[0] = w0[0] ^ 0x36363636;
  t0[1] = w0[1] ^ 0x36363636;
  t0[2] = w0[2] ^ 0x36363636;
  t0[3] = w0[3] ^ 0x36363636;
  t1[0] = w1[0] ^ 0x36363636;
  t1[1] = w1[1] ^ 0x36363636;
  t1[2] = w1[2] ^ 0x36363636;
  t1[3] = w1[3] ^ 0x36363636;
  t2[0] = w2[0] ^ 0x36363636;
  t2[1] = w2[1] ^ 0x36363636;
  t2[2] = w2[2] ^ 0x36363636;
  t2[3] = w2[3] ^ 0x36363636;
  t3[0] = w3[0] ^ 0x36363636;
  t3[1] = w3[1] ^ 0x36363636;
  t3[2] = w3[2] ^ 0x36363636;
  t3[3] = w3[3] ^ 0x36363636;

  sha1_init (&ctx->ipad);

  sha1_update_64 (&ctx->ipad, t0, t1, t2, t3, 64);

  // opad

  t0[0] = w0[0] ^ 0x5c5c5c5c;
  t0[1] = w0[1] ^ 0x5c5c5c5c;
  t0[2] = w0[2] ^ 0x5c5c5c5c;
  t0[3] = w0[3] ^ 0x5c5c5c5c;
  t1[0] = w1[0] ^ 0x5c5c5c5c;
  t1[1] = w1[1] ^ 0x5c5c5c5c;
  t1[2] = w1[2] ^ 0x5c5c5c5c;
  t1[3] = w1[3] ^ 0x5c5c5c5c;
  t2[0] = w2[0] ^ 0x5c5c5c5c;
  t2[1] = w2[1] ^ 0x5c5c5c5c;
  t2[2] = w2[2] ^ 0x5c5c5c5c;
  t2[3] = w2[3] ^ 0x5c5c5c5c;
  t3[0] = w3[0] ^ 0x5c5c5c5c;
  t3[1] = w3[1] ^ 0x5c5c5c5c;
  t3[2] = w3[2] ^ 0x5c5c5c5c;
  t3[3] = w3[3] ^ 0x5c5c5c5c;

  sha1_init (&ctx->opad);

  sha1_update_64 (&ctx->opad, t0, t1, t2, t3, 64);
}

void sha1_hmac_init (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  if (len > 64)
  {
    sha1_ctx_t tmp;

    sha1_init (&tmp);

    sha1_update (&tmp, w, len);

    sha1_final (&tmp);

    w0[0] = tmp.h[0];
    w0[1] = tmp.h[1];
    w0[2] = tmp.h[2];
    w0[3] = tmp.h[3];
    w1[0] = tmp.h[4];
    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;
  }
  else
  {
    w0[0] = w[ 0];
    w0[1] = w[ 1];
    w0[2] = w[ 2];
    w0[3] = w[ 3];
    w1[0] = w[ 4];
    w1[1] = w[ 5];
    w1[2] = w[ 6];
    w1[3] = w[ 7];
    w2[0] = w[ 8];
    w2[1] = w[ 9];
    w2[2] = w[10];
    w2[3] = w[11];
    w3[0] = w[12];
    w3[1] = w[13];
    w3[2] = w[14];
    w3[3] = w[15];
  }

  sha1_hmac_init_64 (ctx, w0, w1, w2, w3);
}

void sha1_hmac_init_swap (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  if (len > 64)
  {
    sha1_ctx_t tmp;

    sha1_init (&tmp);

    sha1_update_swap (&tmp, w, len);

    sha1_final (&tmp);

    w0[0] = tmp.h[0];
    w0[1] = tmp.h[1];
    w0[2] = tmp.h[2];
    w0[3] = tmp.h[3];
    w1[0] = tmp.h[4];
    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;
  }
  else
  {
    w0[0] = swap32_S (w[ 0]);
    w0[1] = swap32_S (w[ 1]);
    w0[2] = swap32_S (w[ 2]);
    w0[3] = swap32_S (w[ 3]);
    w1[0] = swap32_S (w[ 4]);
    w1[1] = swap32_S (w[ 5]);
    w1[2] = swap32_S (w[ 6]);
    w1[3] = swap32_S (w[ 7]);
    w2[0] = swap32_S (w[ 8]);
    w2[1] = swap32_S (w[ 9]);
    w2[2] = swap32_S (w[10]);
    w2[3] = swap32_S (w[11]);
    w3[0] = swap32_S (w[12]);
    w3[1] = swap32_S (w[13]);
    w3[2] = swap32_S (w[14]);
    w3[3] = swap32_S (w[15]);
  }

  sha1_hmac_init_64 (ctx, w0, w1, w2, w3);
}

void sha1_hmac_init_global (sha1_hmac_ctx_t *ctx, __global const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  if (len > 64)
  {
    sha1_ctx_t tmp;

    sha1_init (&tmp);

    sha1_update_global (&tmp, w, len);

    sha1_final (&tmp);

    w0[0] = tmp.h[0];
    w0[1] = tmp.h[1];
    w0[2] = tmp.h[2];
    w0[3] = tmp.h[3];
    w1[0] = tmp.h[4];
    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;
  }
  else
  {
    w0[0] = w[ 0];
    w0[1] = w[ 1];
    w0[2] = w[ 2];
    w0[3] = w[ 3];
    w1[0] = w[ 4];
    w1[1] = w[ 5];
    w1[2] = w[ 6];
    w1[3] = w[ 7];
    w2[0] = w[ 8];
    w2[1] = w[ 9];
    w2[2] = w[10];
    w2[3] = w[11];
    w3[0] = w[12];
    w3[1] = w[13];
    w3[2] = w[14];
    w3[3] = w[15];
  }

  sha1_hmac_init_64 (ctx, w0, w1, w2, w3);
}

void sha1_hmac_init_global_swap (sha1_hmac_ctx_t *ctx, __global const u32 *w, const int len)
{
  u32 w0[4];
  u32 w1[4];
  u32 w2[4];
  u32 w3[4];

  if (len > 64)
  {
    sha1_ctx_t tmp;

    sha1_init (&tmp);

    sha1_update_global_swap (&tmp, w, len);

    sha1_final (&tmp);

    w0[0] = tmp.h[0];
    w0[1] = tmp.h[1];
    w0[2] = tmp.h[2];
    w0[3] = tmp.h[3];
    w1[0] = tmp.h[4];
    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;
  }
  else
  {
    w0[0] = swap32_S (w[ 0]);
    w0[1] = swap32_S (w[ 1]);
    w0[2] = swap32_S (w[ 2]);
    w0[3] = swap32_S (w[ 3]);
    w1[0] = swap32_S (w[ 4]);
    w1[1] = swap32_S (w[ 5]);
    w1[2] = swap32_S (w[ 6]);
    w1[3] = swap32_S (w[ 7]);
    w2[0] = swap32_S (w[ 8]);
    w2[1] = swap32_S (w[ 9]);
    w2[2] = swap32_S (w[10]);
    w2[3] = swap32_S (w[11]);
    w3[0] = swap32_S (w[12]);
    w3[1] = swap32_S (w[13]);
    w3[2] = swap32_S (w[14]);
    w3[3] = swap32_S (w[15]);
  }

  sha1_hmac_init_64 (ctx, w0, w1, w2, w3);
}

void sha1_hmac_update_64 (sha1_hmac_ctx_t *ctx, u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const int len)
{
  sha1_update_64 (&ctx->ipad, w0, w1, w2, w3, len);
}

void sha1_hmac_update (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  sha1_update (&ctx->ipad, w, len);
}

void sha1_hmac_update_swap (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  sha1_update_swap (&ctx->ipad, w, len);
}

void sha1_hmac_update_utf16le (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  sha1_update_utf16le (&ctx->ipad, w, len);
}

void sha1_hmac_update_utf16le_swap (sha1_hmac_ctx_t *ctx, const u32 *w, const int len)
{
  sha1_update_utf16le_swap (&ctx->ipad, w, len);
}

void sha1_hmac_update_global (sha1_hmac_ctx_t *ctx, const __global u32 *w, const int len)
{
  sha1_update_global (&ctx->ipad, w, len);
}

void sha1_hmac_update_global_swap (sha1_hmac_ctx_t *ctx, const __global u32 *w, const int len)
{
  sha1_update_global_swap (&ctx->ipad, w, len);
}

void sha1_hmac_update_global_utf16le (sha1_hmac_ctx_t *ctx, const __global u32 *w, const int len)
{
  sha1_update_global_utf16le (&ctx->ipad, w, len);
}

void sha1_hmac_update_global_utf16le_swap (sha1_hmac_ctx_t *ctx, const __global u32 *w, const int len)
{
  sha1_update_global_utf16le_swap (&ctx->ipad, w, len);
}

void sha1_hmac_final (sha1_hmac_ctx_t *ctx)
{
  sha1_final (&ctx->ipad);

  u32 t0[4];
  u32 t1[4];
  u32 t2[4];
  u32 t3[4];

  t0[0] = ctx->ipad.h[0];
  t0[1] = ctx->ipad.h[1];
  t0[2] = ctx->ipad.h[2];
  t0[3] = ctx->ipad.h[3];
  t1[0] = ctx->ipad.h[4];
  t1[1] = 0;
  t1[2] = 0;
  t1[3] = 0;
  t2[0] = 0;
  t2[1] = 0;
  t2[2] = 0;
  t2[3] = 0;
  t3[0] = 0;
  t3[1] = 0;
  t3[2] = 0;
  t3[3] = 0;

  sha1_update_64 (&ctx->opad, t0, t1, t2, t3, 20);

  sha1_final (&ctx->opad);
}

// while input buf can be a vector datatype, the length of the different elements can not

typedef struct sha1_ctx_vector
{
  u32x h[5];

  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int  len;

} sha1_ctx_vector_t;

void sha1_transform_vector (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], u32x digest[5])
{
  u32x a = digest[0];
  u32x b = digest[1];
  u32x c = digest[2];
  u32x d = digest[3];
  u32x e = digest[4];

  u32x w0_t = w0[0];
  u32x w1_t = w0[1];
  u32x w2_t = w0[2];
  u32x w3_t = w0[3];
  u32x w4_t = w1[0];
  u32x w5_t = w1[1];
  u32x w6_t = w1[2];
  u32x w7_t = w1[3];
  u32x w8_t = w2[0];
  u32x w9_t = w2[1];
  u32x wa_t = w2[2];
  u32x wb_t = w2[3];
  u32x wc_t = w3[0];
  u32x wd_t = w3[1];
  u32x we_t = w3[2];
  u32x wf_t = w3[3];

  #define K SHA1C00

  SHA1_STEP (SHA1_F0o, a, b, c, d, e, w0_t);
  SHA1_STEP (SHA1_F0o, e, a, b, c, d, w1_t);
  SHA1_STEP (SHA1_F0o, d, e, a, b, c, w2_t);
  SHA1_STEP (SHA1_F0o, c, d, e, a, b, w3_t);
  SHA1_STEP (SHA1_F0o, b, c, d, e, a, w4_t);
  SHA1_STEP (SHA1_F0o, a, b, c, d, e, w5_t);
  SHA1_STEP (SHA1_F0o, e, a, b, c, d, w6_t);
  SHA1_STEP (SHA1_F0o, d, e, a, b, c, w7_t);
  SHA1_STEP (SHA1_F0o, c, d, e, a, b, w8_t);
  SHA1_STEP (SHA1_F0o, b, c, d, e, a, w9_t);
  SHA1_STEP (SHA1_F0o, a, b, c, d, e, wa_t);
  SHA1_STEP (SHA1_F0o, e, a, b, c, d, wb_t);
  SHA1_STEP (SHA1_F0o, d, e, a, b, c, wc_t);
  SHA1_STEP (SHA1_F0o, c, d, e, a, b, wd_t);
  SHA1_STEP (SHA1_F0o, b, c, d, e, a, we_t);
  SHA1_STEP (SHA1_F0o, a, b, c, d, e, wf_t);
  w0_t = rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F0o, e, a, b, c, d, w0_t);
  w1_t = rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F0o, d, e, a, b, c, w1_t);
  w2_t = rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F0o, c, d, e, a, b, w2_t);
  w3_t = rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F0o, b, c, d, e, a, w3_t);

  #undef K
  #define K SHA1C01

  w4_t = rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w4_t);
  w5_t = rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w5_t);
  w6_t = rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w6_t);
  w7_t = rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w7_t);
  w8_t = rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w8_t);
  w9_t = rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w9_t);
  wa_t = rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, wa_t);
  wb_t = rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, wb_t);
  wc_t = rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, wc_t);
  wd_t = rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wd_t);
  we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, we_t);
  wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, wf_t);
  w0_t = rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w0_t);
  w1_t = rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w1_t);
  w2_t = rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w2_t);
  w3_t = rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w3_t);
  w4_t = rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w4_t);
  w5_t = rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w5_t);
  w6_t = rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w6_t);
  w7_t = rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w7_t);

  #undef K
  #define K SHA1C02

  w8_t = rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F2o, a, b, c, d, e, w8_t);
  w9_t = rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F2o, e, a, b, c, d, w9_t);
  wa_t = rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F2o, d, e, a, b, c, wa_t);
  wb_t = rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F2o, c, d, e, a, b, wb_t);
  wc_t = rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F2o, b, c, d, e, a, wc_t);
  wd_t = rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F2o, a, b, c, d, e, wd_t);
  we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F2o, e, a, b, c, d, we_t);
  wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F2o, d, e, a, b, c, wf_t);
  w0_t = rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F2o, c, d, e, a, b, w0_t);
  w1_t = rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F2o, b, c, d, e, a, w1_t);
  w2_t = rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F2o, a, b, c, d, e, w2_t);
  w3_t = rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F2o, e, a, b, c, d, w3_t);
  w4_t = rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F2o, d, e, a, b, c, w4_t);
  w5_t = rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F2o, c, d, e, a, b, w5_t);
  w6_t = rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F2o, b, c, d, e, a, w6_t);
  w7_t = rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F2o, a, b, c, d, e, w7_t);
  w8_t = rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F2o, e, a, b, c, d, w8_t);
  w9_t = rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F2o, d, e, a, b, c, w9_t);
  wa_t = rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F2o, c, d, e, a, b, wa_t);
  wb_t = rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F2o, b, c, d, e, a, wb_t);

  #undef K
  #define K SHA1C03

  wc_t = rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, wc_t);
  wd_t = rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, wd_t);
  we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, we_t);
  wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, wf_t);
  w0_t = rotl32 ((wd_t ^ w8_t ^ w2_t ^ w0_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w0_t);
  w1_t = rotl32 ((we_t ^ w9_t ^ w3_t ^ w1_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w1_t);
  w2_t = rotl32 ((wf_t ^ wa_t ^ w4_t ^ w2_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w2_t);
  w3_t = rotl32 ((w0_t ^ wb_t ^ w5_t ^ w3_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w3_t);
  w4_t = rotl32 ((w1_t ^ wc_t ^ w6_t ^ w4_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w4_t);
  w5_t = rotl32 ((w2_t ^ wd_t ^ w7_t ^ w5_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, w5_t);
  w6_t = rotl32 ((w3_t ^ we_t ^ w8_t ^ w6_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, w6_t);
  w7_t = rotl32 ((w4_t ^ wf_t ^ w9_t ^ w7_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, w7_t);
  w8_t = rotl32 ((w5_t ^ w0_t ^ wa_t ^ w8_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, w8_t);
  w9_t = rotl32 ((w6_t ^ w1_t ^ wb_t ^ w9_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, w9_t);
  wa_t = rotl32 ((w7_t ^ w2_t ^ wc_t ^ wa_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wa_t);
  wb_t = rotl32 ((w8_t ^ w3_t ^ wd_t ^ wb_t), 1u); SHA1_STEP (SHA1_F1, a, b, c, d, e, wb_t);
  wc_t = rotl32 ((w9_t ^ w4_t ^ we_t ^ wc_t), 1u); SHA1_STEP (SHA1_F1, e, a, b, c, d, wc_t);
  wd_t = rotl32 ((wa_t ^ w5_t ^ wf_t ^ wd_t), 1u); SHA1_STEP (SHA1_F1, d, e, a, b, c, wd_t);
  we_t = rotl32 ((wb_t ^ w6_t ^ w0_t ^ we_t), 1u); SHA1_STEP (SHA1_F1, c, d, e, a, b, we_t);
  wf_t = rotl32 ((wc_t ^ w7_t ^ w1_t ^ wf_t), 1u); SHA1_STEP (SHA1_F1, b, c, d, e, a, wf_t);

  #undef K

  digest[0] += a;
  digest[1] += b;
  digest[2] += c;
  digest[3] += d;
  digest[4] += e;
}

void sha1_init_vector (sha1_ctx_vector_t *ctx)
{
  ctx->h[0] = SHA1M_A;
  ctx->h[1] = SHA1M_B;
  ctx->h[2] = SHA1M_C;
  ctx->h[3] = SHA1M_D;
  ctx->h[4] = SHA1M_E;

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

  ctx->len = 0;
}

void sha1_init_vector_from_scalar (sha1_ctx_vector_t *ctx, sha1_ctx_t *ctx0)
{
  ctx->h[0] = ctx0->h[0];
  ctx->h[1] = ctx0->h[1];
  ctx->h[2] = ctx0->h[2];
  ctx->h[3] = ctx0->h[3];
  ctx->h[4] = ctx0->h[4];

  ctx->w0[0] = ctx0->w0[0];
  ctx->w0[1] = ctx0->w0[1];
  ctx->w0[2] = ctx0->w0[2];
  ctx->w0[3] = ctx0->w0[3];
  ctx->w1[0] = ctx0->w1[0];
  ctx->w1[1] = ctx0->w1[1];
  ctx->w1[2] = ctx0->w1[2];
  ctx->w1[3] = ctx0->w1[3];
  ctx->w2[0] = ctx0->w2[0];
  ctx->w2[1] = ctx0->w2[1];
  ctx->w2[2] = ctx0->w2[2];
  ctx->w2[3] = ctx0->w2[3];
  ctx->w3[0] = ctx0->w3[0];
  ctx->w3[1] = ctx0->w3[1];
  ctx->w3[2] = ctx0->w3[2];
  ctx->w3[3] = ctx0->w3[3];

  ctx->len = ctx0->len;
}

void sha1_update_vector_64 (sha1_ctx_vector_t *ctx, u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], const int len)
{
  #ifdef IS_AMD
  volatile const int pos = ctx->len & 63;
  #else
  const int pos = ctx->len & 63;
  #endif

  ctx->len += len;

  if ((pos + len) < 64)
  {
    switch_buffer_by_offset_be (w0, w1, w2, w3, pos);

    ctx->w0[0] |= w0[0];
    ctx->w0[1] |= w0[1];
    ctx->w0[2] |= w0[2];
    ctx->w0[3] |= w0[3];
    ctx->w1[0] |= w1[0];
    ctx->w1[1] |= w1[1];
    ctx->w1[2] |= w1[2];
    ctx->w1[3] |= w1[3];
    ctx->w2[0] |= w2[0];
    ctx->w2[1] |= w2[1];
    ctx->w2[2] |= w2[2];
    ctx->w2[3] |= w2[3];
    ctx->w3[0] |= w3[0];
    ctx->w3[1] |= w3[1];
    ctx->w3[2] |= w3[2];
    ctx->w3[3] |= w3[3];
  }
  else
  {
    u32x c0[4] = { 0 };
    u32x c1[4] = { 0 };
    u32x c2[4] = { 0 };
    u32x c3[4] = { 0 };

    switch_buffer_by_offset_carry_be (w0, w1, w2, w3, c0, c1, c2, c3, pos);

    ctx->w0[0] |= w0[0];
    ctx->w0[1] |= w0[1];
    ctx->w0[2] |= w0[2];
    ctx->w0[3] |= w0[3];
    ctx->w1[0] |= w1[0];
    ctx->w1[1] |= w1[1];
    ctx->w1[2] |= w1[2];
    ctx->w1[3] |= w1[3];
    ctx->w2[0] |= w2[0];
    ctx->w2[1] |= w2[1];
    ctx->w2[2] |= w2[2];
    ctx->w2[3] |= w2[3];
    ctx->w3[0] |= w3[0];
    ctx->w3[1] |= w3[1];
    ctx->w3[2] |= w3[2];
    ctx->w3[3] |= w3[3];

    sha1_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);

    ctx->w0[0] = c0[0];
    ctx->w0[1] = c0[1];
    ctx->w0[2] = c0[2];
    ctx->w0[3] = c0[3];
    ctx->w1[0] = c1[0];
    ctx->w1[1] = c1[1];
    ctx->w1[2] = c1[2];
    ctx->w1[3] = c1[3];
    ctx->w2[0] = c2[0];
    ctx->w2[1] = c2[1];
    ctx->w2[2] = c2[2];
    ctx->w2[3] = c2[3];
    ctx->w3[0] = c3[0];
    ctx->w3[1] = c3[1];
    ctx->w3[2] = c3[2];
    ctx->w3[3] = c3[3];
  }
}

void sha1_update_vector (sha1_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    sha1_update_vector_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  sha1_update_vector_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_vector_swap (sha1_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 64; pos1 += 64, pos4 += 16)
  {
    w0[0] = w[pos4 +  0];
    w0[1] = w[pos4 +  1];
    w0[2] = w[pos4 +  2];
    w0[3] = w[pos4 +  3];
    w1[0] = w[pos4 +  4];
    w1[1] = w[pos4 +  5];
    w1[2] = w[pos4 +  6];
    w1[3] = w[pos4 +  7];
    w2[0] = w[pos4 +  8];
    w2[1] = w[pos4 +  9];
    w2[2] = w[pos4 + 10];
    w2[3] = w[pos4 + 11];
    w3[0] = w[pos4 + 12];
    w3[1] = w[pos4 + 13];
    w3[2] = w[pos4 + 14];
    w3[3] = w[pos4 + 15];

    w0[0] = swap32 (w0[0]);
    w0[1] = swap32 (w0[1]);
    w0[2] = swap32 (w0[2]);
    w0[3] = swap32 (w0[3]);
    w1[0] = swap32 (w1[0]);
    w1[1] = swap32 (w1[1]);
    w1[2] = swap32 (w1[2]);
    w1[3] = swap32 (w1[3]);
    w2[0] = swap32 (w2[0]);
    w2[1] = swap32 (w2[1]);
    w2[2] = swap32 (w2[2]);
    w2[3] = swap32 (w2[3]);
    w3[0] = swap32 (w3[0]);
    w3[1] = swap32 (w3[1]);
    w3[2] = swap32 (w3[2]);
    w3[3] = swap32 (w3[3]);

    sha1_update_vector_64 (ctx, w0, w1, w2, w3, 64);
  }

  w0[0] = w[pos4 +  0];
  w0[1] = w[pos4 +  1];
  w0[2] = w[pos4 +  2];
  w0[3] = w[pos4 +  3];
  w1[0] = w[pos4 +  4];
  w1[1] = w[pos4 +  5];
  w1[2] = w[pos4 +  6];
  w1[3] = w[pos4 +  7];
  w2[0] = w[pos4 +  8];
  w2[1] = w[pos4 +  9];
  w2[2] = w[pos4 + 10];
  w2[3] = w[pos4 + 11];
  w3[0] = w[pos4 + 12];
  w3[1] = w[pos4 + 13];
  w3[2] = w[pos4 + 14];
  w3[3] = w[pos4 + 15];

  w0[0] = swap32 (w0[0]);
  w0[1] = swap32 (w0[1]);
  w0[2] = swap32 (w0[2]);
  w0[3] = swap32 (w0[3]);
  w1[0] = swap32 (w1[0]);
  w1[1] = swap32 (w1[1]);
  w1[2] = swap32 (w1[2]);
  w1[3] = swap32 (w1[3]);
  w2[0] = swap32 (w2[0]);
  w2[1] = swap32 (w2[1]);
  w2[2] = swap32 (w2[2]);
  w2[3] = swap32 (w2[3]);
  w3[0] = swap32 (w3[0]);
  w3[1] = swap32 (w3[1]);
  w3[2] = swap32 (w3[2]);
  w3[3] = swap32 (w3[3]);

  sha1_update_vector_64 (ctx, w0, w1, w2, w3, len - pos1);
}

void sha1_update_vector_utf16le (sha1_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le (w1, w2, w3);
    make_utf16le (w0, w0, w1);

    sha1_update_vector_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le (w1, w2, w3);
  make_utf16le (w0, w0, w1);

  sha1_update_vector_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_update_vector_utf16le_swap (sha1_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16le (w1, w2, w3);
    make_utf16le (w0, w0, w1);

    w0[0] = swap32 (w0[0]);
    w0[1] = swap32 (w0[1]);
    w0[2] = swap32 (w0[2]);
    w0[3] = swap32 (w0[3]);
    w1[0] = swap32 (w1[0]);
    w1[1] = swap32 (w1[1]);
    w1[2] = swap32 (w1[2]);
    w1[3] = swap32 (w1[3]);
    w2[0] = swap32 (w2[0]);
    w2[1] = swap32 (w2[1]);
    w2[2] = swap32 (w2[2]);
    w2[3] = swap32 (w2[3]);
    w3[0] = swap32 (w3[0]);
    w3[1] = swap32 (w3[1]);
    w3[2] = swap32 (w3[2]);
    w3[3] = swap32 (w3[3]);

    sha1_update_vector_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16le (w1, w2, w3);
  make_utf16le (w0, w0, w1);

  w0[0] = swap32 (w0[0]);
  w0[1] = swap32 (w0[1]);
  w0[2] = swap32 (w0[2]);
  w0[3] = swap32 (w0[3]);
  w1[0] = swap32 (w1[0]);
  w1[1] = swap32 (w1[1]);
  w1[2] = swap32 (w1[2]);
  w1[3] = swap32 (w1[3]);
  w2[0] = swap32 (w2[0]);
  w2[1] = swap32 (w2[1]);
  w2[2] = swap32 (w2[2]);
  w2[3] = swap32 (w2[3]);
  w3[0] = swap32 (w3[0]);
  w3[1] = swap32 (w3[1]);
  w3[2] = swap32 (w3[2]);
  w3[3] = swap32 (w3[3]);

  sha1_update_vector_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_update_vector_utf16beN (sha1_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  int pos1;
  int pos4;

  for (pos1 = 0, pos4 = 0; pos1 < len - 32; pos1 += 32, pos4 += 8)
  {
    w0[0] = w[pos4 + 0];
    w0[1] = w[pos4 + 1];
    w0[2] = w[pos4 + 2];
    w0[3] = w[pos4 + 3];
    w1[0] = w[pos4 + 4];
    w1[1] = w[pos4 + 5];
    w1[2] = w[pos4 + 6];
    w1[3] = w[pos4 + 7];

    make_utf16beN (w1, w2, w3);
    make_utf16beN (w0, w0, w1);

    sha1_update_vector_64 (ctx, w0, w1, w2, w3, 32 * 2);
  }

  w0[0] = w[pos4 + 0];
  w0[1] = w[pos4 + 1];
  w0[2] = w[pos4 + 2];
  w0[3] = w[pos4 + 3];
  w1[0] = w[pos4 + 4];
  w1[1] = w[pos4 + 5];
  w1[2] = w[pos4 + 6];
  w1[3] = w[pos4 + 7];

  make_utf16beN (w1, w2, w3);
  make_utf16beN (w0, w0, w1);

  sha1_update_vector_64 (ctx, w0, w1, w2, w3, (len - pos1) * 2);
}

void sha1_final_vector (sha1_ctx_vector_t *ctx)
{
  const int pos = ctx->len & 63;

  append_0x80_4x4 (ctx->w0, ctx->w1, ctx->w2, ctx->w3, pos ^ 3);

  if (pos >= 56)
  {
    sha1_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);

    ctx->w0[0] = 0;
    ctx->w0[1] = 0;
    ctx->w0[2] = 0;
    ctx->w0[3] = 0;
    ctx->w1[0] = 0;
    ctx->w1[1] = 0;
    ctx->w1[2] = 0;
    ctx->w1[3] = 0;
    ctx->w2[0] = 0;
    ctx->w2[1] = 0;
    ctx->w2[2] = 0;
    ctx->w2[3] = 0;
    ctx->w3[0] = 0;
    ctx->w3[1] = 0;
    ctx->w3[2] = 0;
    ctx->w3[3] = 0;
  }

  ctx->w3[2] = 0;
  ctx->w3[3] = ctx->len * 8;

  sha1_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->h);
}

// HMAC + Vector

typedef struct sha1_hmac_ctx_vector
{
  sha1_ctx_vector_t ipad;
  sha1_ctx_vector_t opad;

} sha1_hmac_ctx_vector_t;

void sha1_hmac_init_vector_64 (sha1_hmac_ctx_vector_t *ctx, const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4])
{
  u32x t0[4];
  u32x t1[4];
  u32x t2[4];
  u32x t3[4];

  // ipad

  t0[0] = w0[0] ^ 0x36363636;
  t0[1] = w0[1] ^ 0x36363636;
  t0[2] = w0[2] ^ 0x36363636;
  t0[3] = w0[3] ^ 0x36363636;
  t1[0] = w1[0] ^ 0x36363636;
  t1[1] = w1[1] ^ 0x36363636;
  t1[2] = w1[2] ^ 0x36363636;
  t1[3] = w1[3] ^ 0x36363636;
  t2[0] = w2[0] ^ 0x36363636;
  t2[1] = w2[1] ^ 0x36363636;
  t2[2] = w2[2] ^ 0x36363636;
  t2[3] = w2[3] ^ 0x36363636;
  t3[0] = w3[0] ^ 0x36363636;
  t3[1] = w3[1] ^ 0x36363636;
  t3[2] = w3[2] ^ 0x36363636;
  t3[3] = w3[3] ^ 0x36363636;

  sha1_init_vector (&ctx->ipad);

  sha1_update_vector_64 (&ctx->ipad, t0, t1, t2, t3, 64);

  // opad

  t0[0] = w0[0] ^ 0x5c5c5c5c;
  t0[1] = w0[1] ^ 0x5c5c5c5c;
  t0[2] = w0[2] ^ 0x5c5c5c5c;
  t0[3] = w0[3] ^ 0x5c5c5c5c;
  t1[0] = w1[0] ^ 0x5c5c5c5c;
  t1[1] = w1[1] ^ 0x5c5c5c5c;
  t1[2] = w1[2] ^ 0x5c5c5c5c;
  t1[3] = w1[3] ^ 0x5c5c5c5c;
  t2[0] = w2[0] ^ 0x5c5c5c5c;
  t2[1] = w2[1] ^ 0x5c5c5c5c;
  t2[2] = w2[2] ^ 0x5c5c5c5c;
  t2[3] = w2[3] ^ 0x5c5c5c5c;
  t3[0] = w3[0] ^ 0x5c5c5c5c;
  t3[1] = w3[1] ^ 0x5c5c5c5c;
  t3[2] = w3[2] ^ 0x5c5c5c5c;
  t3[3] = w3[3] ^ 0x5c5c5c5c;

  sha1_init_vector (&ctx->opad);

  sha1_update_vector_64 (&ctx->opad, t0, t1, t2, t3, 64);
}

void sha1_hmac_init_vector (sha1_hmac_ctx_vector_t *ctx, const u32x *w, const int len)
{
  u32x w0[4];
  u32x w1[4];
  u32x w2[4];
  u32x w3[4];

  if (len > 64)
  {
    sha1_ctx_vector_t tmp;

    sha1_init_vector (&tmp);

    sha1_update_vector (&tmp, w, len);

    sha1_final_vector (&tmp);

    w0[0] = tmp.h[0];
    w0[1] = tmp.h[1];
    w0[2] = tmp.h[2];
    w0[3] = tmp.h[3];
    w1[0] = tmp.h[4];
    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;
  }
  else
  {
    w0[0] = w[ 0];
    w0[1] = w[ 1];
    w0[2] = w[ 2];
    w0[3] = w[ 3];
    w1[0] = w[ 4];
    w1[1] = w[ 5];
    w1[2] = w[ 6];
    w1[3] = w[ 7];
    w2[0] = w[ 8];
    w2[1] = w[ 9];
    w2[2] = w[10];
    w2[3] = w[11];
    w3[0] = w[12];
    w3[1] = w[13];
    w3[2] = w[14];
    w3[3] = w[15];
  }

  sha1_hmac_init_vector_64 (ctx, w0, w1, w2, w3);
}

void sha1_hmac_update_vector_64 (sha1_hmac_ctx_vector_t *ctx, u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], const int len)
{
  sha1_update_vector_64 (&ctx->ipad, w0, w1, w2, w3, len);
}

void sha1_hmac_update_vector (sha1_hmac_ctx_vector_t *ctx, const u32x *w, const int len)
{
  sha1_update_vector (&ctx->ipad, w, len);
}

void sha1_hmac_final_vector (sha1_hmac_ctx_vector_t *ctx)
{
  sha1_final_vector (&ctx->ipad);

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

  t0[0] = ctx->ipad.h[0];
  t0[1] = ctx->ipad.h[1];
  t0[2] = ctx->ipad.h[2];
  t0[3] = ctx->ipad.h[3];
  t1[0] = ctx->ipad.h[4];
  t1[1] = 0;
  t1[2] = 0;
  t1[3] = 0;
  t2[0] = 0;
  t2[1] = 0;
  t2[2] = 0;
  t2[3] = 0;
  t3[0] = 0;
  t3[1] = 0;
  t3[2] = 0;
  t3[3] = 0;

  sha1_update_vector_64 (&ctx->opad, t0, t1, t2, t3, 20);

  sha1_final_vector (&ctx->opad);
}