diff --git a/OpenCL/inc_hash_functions.cl b/OpenCL/inc_hash_functions.cl
index fd133b8bd..f87012162 100644
--- a/OpenCL/inc_hash_functions.cl
+++ b/OpenCL/inc_hash_functions.cl
@@ -267,6 +267,19 @@
 #define SHA384_F1o(x,y,z) (SHA384_F1 ((x), (y), (z)))
 #endif
 
+#define SHA384_STEP_S(F0,F1,a,b,c,d,e,f,g,h,x,K)  \
+{                                                 \
+  h += K;                                         \
+  h += x;                                         \
+  h += SHA384_S1_S (e);                           \
+  h += F0 (e, f, g);                              \
+  d += h;                                         \
+  h += SHA384_S0_S (a);                           \
+  h += F1 (a, b, c);                              \
+}
+
+#define SHA384_EXPAND_S(x,y,z,w) (SHA384_S3_S (x) + y + SHA384_S2_S (z) + w)
+
 #define SHA384_STEP(F0,F1,a,b,c,d,e,f,g,h,x,K)  \
 {                                               \
   h += K;                                       \
diff --git a/OpenCL/inc_hash_sha384.cl b/OpenCL/inc_hash_sha384.cl
new file mode 100644
index 000000000..1b2f88acd
--- /dev/null
+++ b/OpenCL/inc_hash_sha384.cl
@@ -0,0 +1,1643 @@
+
+// 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 128 byte aligned when using sha512_update()
+
+__constant u64a k_sha384[80] =
+{
+  SHA512C00, SHA512C01, SHA512C02, SHA512C03,
+  SHA512C04, SHA512C05, SHA512C06, SHA512C07,
+  SHA512C08, SHA512C09, SHA512C0a, SHA512C0b,
+  SHA512C0c, SHA512C0d, SHA512C0e, SHA512C0f,
+  SHA512C10, SHA512C11, SHA512C12, SHA512C13,
+  SHA512C14, SHA512C15, SHA512C16, SHA512C17,
+  SHA512C18, SHA512C19, SHA512C1a, SHA512C1b,
+  SHA512C1c, SHA512C1d, SHA512C1e, SHA512C1f,
+  SHA512C20, SHA512C21, SHA512C22, SHA512C23,
+  SHA512C24, SHA512C25, SHA512C26, SHA512C27,
+  SHA512C28, SHA512C29, SHA512C2a, SHA512C2b,
+  SHA512C2c, SHA512C2d, SHA512C2e, SHA512C2f,
+  SHA512C30, SHA512C31, SHA512C32, SHA512C33,
+  SHA512C34, SHA512C35, SHA512C36, SHA512C37,
+  SHA512C38, SHA512C39, SHA512C3a, SHA512C3b,
+  SHA512C3c, SHA512C3d, SHA512C3e, SHA512C3f,
+  SHA512C40, SHA512C41, SHA512C42, SHA512C43,
+  SHA512C44, SHA512C45, SHA512C46, SHA512C47,
+  SHA512C48, SHA512C49, SHA512C4a, SHA512C4b,
+  SHA512C4c, SHA512C4d, SHA512C4e, SHA512C4f,
+};
+
+typedef struct sha384_ctx
+{
+  u64 h[8];
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[4];
+
+  int len;
+
+} sha384_ctx_t;
+
+void sha384_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], const u32 w4[4], const u32 w5[4], const u32 w6[4], const u32 w7[4], u64 digest[8])
+{
+  u64 a = digest[0];
+  u64 b = digest[1];
+  u64 c = digest[2];
+  u64 d = digest[3];
+  u64 e = digest[4];
+  u64 f = digest[5];
+  u64 g = digest[6];
+  u64 h = digest[7];
+
+  u64 w0_t = hl32_to_64_S (w0[0], w0[1]);
+  u64 w1_t = hl32_to_64_S (w0[2], w0[3]);
+  u64 w2_t = hl32_to_64_S (w1[0], w1[1]);
+  u64 w3_t = hl32_to_64_S (w1[2], w1[3]);
+  u64 w4_t = hl32_to_64_S (w2[0], w2[1]);
+  u64 w5_t = hl32_to_64_S (w2[2], w2[3]);
+  u64 w6_t = hl32_to_64_S (w3[0], w3[1]);
+  u64 w7_t = hl32_to_64_S (w3[2], w3[3]);
+  u64 w8_t = hl32_to_64_S (w4[0], w4[1]);
+  u64 w9_t = hl32_to_64_S (w4[2], w4[3]);
+  u64 wa_t = hl32_to_64_S (w5[0], w5[1]);
+  u64 wb_t = hl32_to_64_S (w5[2], w5[3]);
+  u64 wc_t = hl32_to_64_S (w6[0], w6[1]);
+  u64 wd_t = hl32_to_64_S (w6[2], w6[3]);
+  u64 we_t = hl32_to_64_S (w7[0], w7[1]);
+  u64 wf_t = hl32_to_64_S (w7[2], w7[3]);
+
+  #define ROUND_EXPAND_S()                            \
+  {                                                   \
+    w0_t = SHA384_EXPAND_S (we_t, w9_t, w1_t, w0_t);  \
+    w1_t = SHA384_EXPAND_S (wf_t, wa_t, w2_t, w1_t);  \
+    w2_t = SHA384_EXPAND_S (w0_t, wb_t, w3_t, w2_t);  \
+    w3_t = SHA384_EXPAND_S (w1_t, wc_t, w4_t, w3_t);  \
+    w4_t = SHA384_EXPAND_S (w2_t, wd_t, w5_t, w4_t);  \
+    w5_t = SHA384_EXPAND_S (w3_t, we_t, w6_t, w5_t);  \
+    w6_t = SHA384_EXPAND_S (w4_t, wf_t, w7_t, w6_t);  \
+    w7_t = SHA384_EXPAND_S (w5_t, w0_t, w8_t, w7_t);  \
+    w8_t = SHA384_EXPAND_S (w6_t, w1_t, w9_t, w8_t);  \
+    w9_t = SHA384_EXPAND_S (w7_t, w2_t, wa_t, w9_t);  \
+    wa_t = SHA384_EXPAND_S (w8_t, w3_t, wb_t, wa_t);  \
+    wb_t = SHA384_EXPAND_S (w9_t, w4_t, wc_t, wb_t);  \
+    wc_t = SHA384_EXPAND_S (wa_t, w5_t, wd_t, wc_t);  \
+    wd_t = SHA384_EXPAND_S (wb_t, w6_t, we_t, wd_t);  \
+    we_t = SHA384_EXPAND_S (wc_t, w7_t, wf_t, we_t);  \
+    wf_t = SHA384_EXPAND_S (wd_t, w8_t, w0_t, wf_t);  \
+  }
+
+  #define ROUND_STEP_S(i)                                                                   \
+  {                                                                                         \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, a, b, c, d, e, f, g, h, w0_t, k_sha384[i +  0]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, h, a, b, c, d, e, f, g, w1_t, k_sha384[i +  1]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, g, h, a, b, c, d, e, f, w2_t, k_sha384[i +  2]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, f, g, h, a, b, c, d, e, w3_t, k_sha384[i +  3]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, e, f, g, h, a, b, c, d, w4_t, k_sha384[i +  4]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, d, e, f, g, h, a, b, c, w5_t, k_sha384[i +  5]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, c, d, e, f, g, h, a, b, w6_t, k_sha384[i +  6]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, b, c, d, e, f, g, h, a, w7_t, k_sha384[i +  7]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, a, b, c, d, e, f, g, h, w8_t, k_sha384[i +  8]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, h, a, b, c, d, e, f, g, w9_t, k_sha384[i +  9]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, g, h, a, b, c, d, e, f, wa_t, k_sha384[i + 10]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, f, g, h, a, b, c, d, e, wb_t, k_sha384[i + 11]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, e, f, g, h, a, b, c, d, wc_t, k_sha384[i + 12]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, d, e, f, g, h, a, b, c, wd_t, k_sha384[i + 13]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, c, d, e, f, g, h, a, b, we_t, k_sha384[i + 14]); \
+    SHA384_STEP_S (SHA384_F0o, SHA384_F1o, b, c, d, e, f, g, h, a, wf_t, k_sha384[i + 15]); \
+  }
+
+  ROUND_STEP_S (0);
+
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (int i = 16; i < 80; i += 16)
+  {
+    ROUND_EXPAND_S (); ROUND_STEP_S (i);
+  }
+
+  #undef ROUND_EXPAND_S
+  #undef ROUND_STEP_S
+
+  digest[0] += a;
+  digest[1] += b;
+  digest[2] += c;
+  digest[3] += d;
+  digest[4] += e;
+  digest[5] += f;
+  digest[6] += g;
+  digest[7] += h;
+}
+
+void sha384_init (sha384_ctx_t *ctx)
+{
+  ctx->h[0] = SHA384M_A;
+  ctx->h[1] = SHA384M_B;
+  ctx->h[2] = SHA384M_C;
+  ctx->h[3] = SHA384M_D;
+  ctx->h[4] = SHA384M_E;
+  ctx->h[5] = SHA384M_F;
+  ctx->h[6] = SHA384M_G;
+  ctx->h[7] = SHA384M_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->w4[0] = 0;
+  ctx->w4[1] = 0;
+  ctx->w4[2] = 0;
+  ctx->w4[3] = 0;
+  ctx->w5[0] = 0;
+  ctx->w5[1] = 0;
+  ctx->w5[2] = 0;
+  ctx->w5[3] = 0;
+  ctx->w6[0] = 0;
+  ctx->w6[1] = 0;
+  ctx->w6[2] = 0;
+  ctx->w6[3] = 0;
+  ctx->w7[0] = 0;
+  ctx->w7[1] = 0;
+  ctx->w7[2] = 0;
+  ctx->w7[3] = 0;
+
+  ctx->len = 0;
+}
+
+void sha384_update_128 (sha384_ctx_t *ctx, u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 w4[4], u32 w5[4], u32 w6[4], u32 w7[4], const int len)
+{
+  const int pos = ctx->len & 127;
+
+  ctx->len += len;
+
+  if ((pos + len) < 128)
+  {
+    switch_buffer_by_offset_8x4_be_S (w0, w1, w2, w3, w4, w5, w6, w7, 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];
+    ctx->w4[0] |= w4[0];
+    ctx->w4[1] |= w4[1];
+    ctx->w4[2] |= w4[2];
+    ctx->w4[3] |= w4[3];
+    ctx->w5[0] |= w5[0];
+    ctx->w5[1] |= w5[1];
+    ctx->w5[2] |= w5[2];
+    ctx->w5[3] |= w5[3];
+    ctx->w6[0] |= w6[0];
+    ctx->w6[1] |= w6[1];
+    ctx->w6[2] |= w6[2];
+    ctx->w6[3] |= w6[3];
+    ctx->w7[0] |= w7[0];
+    ctx->w7[1] |= w7[1];
+    ctx->w7[2] |= w7[2];
+    ctx->w7[3] |= w7[3];
+  }
+  else
+  {
+    u32 c0[4] = { 0 };
+    u32 c1[4] = { 0 };
+    u32 c2[4] = { 0 };
+    u32 c3[4] = { 0 };
+    u32 c4[4] = { 0 };
+    u32 c5[4] = { 0 };
+    u32 c6[4] = { 0 };
+    u32 c7[4] = { 0 };
+
+    switch_buffer_by_offset_8x4_carry_be_S (w0, w1, w2, w3, w4, w5, w6, w7, c0, c1, c2, c3, c4, c5, c6, c7, 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];
+    ctx->w4[0] |= w4[0];
+    ctx->w4[1] |= w4[1];
+    ctx->w4[2] |= w4[2];
+    ctx->w4[3] |= w4[3];
+    ctx->w5[0] |= w5[0];
+    ctx->w5[1] |= w5[1];
+    ctx->w5[2] |= w5[2];
+    ctx->w5[3] |= w5[3];
+    ctx->w6[0] |= w6[0];
+    ctx->w6[1] |= w6[1];
+    ctx->w6[2] |= w6[2];
+    ctx->w6[3] |= w6[3];
+    ctx->w7[0] |= w7[0];
+    ctx->w7[1] |= w7[1];
+    ctx->w7[2] |= w7[2];
+    ctx->w7[3] |= w7[3];
+
+    sha384_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, 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];
+    ctx->w4[0] = c4[0];
+    ctx->w4[1] = c4[1];
+    ctx->w4[2] = c4[2];
+    ctx->w4[3] = c4[3];
+    ctx->w5[0] = c5[0];
+    ctx->w5[1] = c5[1];
+    ctx->w5[2] = c5[2];
+    ctx->w5[3] = c5[3];
+    ctx->w6[0] = c6[0];
+    ctx->w6[1] = c6[1];
+    ctx->w6[2] = c6[2];
+    ctx->w6[3] = c6[3];
+    ctx->w7[0] = c7[0];
+    ctx->w7[1] = c7[1];
+    ctx->w7[2] = c7[2];
+    ctx->w7[3] = c7[3];
+  }
+}
+
+void sha384_update (sha384_ctx_t *ctx, const u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[4];
+
+  int pos1;
+  int pos4;
+
+  for (pos1 = 0, pos4 = 0; pos1 < len - 128; pos1 += 128, pos4 += 32)
+  {
+    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];
+    w4[0] = w[pos4 + 16];
+    w4[1] = w[pos4 + 17];
+    w4[2] = w[pos4 + 18];
+    w4[3] = w[pos4 + 19];
+    w5[0] = w[pos4 + 20];
+    w5[1] = w[pos4 + 21];
+    w5[2] = w[pos4 + 22];
+    w5[3] = w[pos4 + 23];
+    w6[0] = w[pos4 + 24];
+    w6[1] = w[pos4 + 25];
+    w6[2] = w[pos4 + 26];
+    w6[3] = w[pos4 + 27];
+    w7[0] = w[pos4 + 28];
+    w7[1] = w[pos4 + 29];
+    w7[2] = w[pos4 + 30];
+    w7[3] = w[pos4 + 31];
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 128);
+  }
+
+  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];
+  w4[0] = w[pos4 + 16];
+  w4[1] = w[pos4 + 17];
+  w4[2] = w[pos4 + 18];
+  w4[3] = w[pos4 + 19];
+  w5[0] = w[pos4 + 20];
+  w5[1] = w[pos4 + 21];
+  w5[2] = w[pos4 + 22];
+  w5[3] = w[pos4 + 23];
+  w6[0] = w[pos4 + 24];
+  w6[1] = w[pos4 + 25];
+  w6[2] = w[pos4 + 26];
+  w6[3] = w[pos4 + 27];
+  w7[0] = w[pos4 + 28];
+  w7[1] = w[pos4 + 29];
+  w7[2] = w[pos4 + 30];
+  w7[3] = w[pos4 + 31];
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, len - pos1);
+}
+
+void sha384_update_swap (sha384_ctx_t *ctx, const u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[4];
+
+  int pos1;
+  int pos4;
+
+  for (pos1 = 0, pos4 = 0; pos1 < len - 128; pos1 += 128, pos4 += 32)
+  {
+    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];
+    w4[0] = w[pos4 + 16];
+    w4[1] = w[pos4 + 17];
+    w4[2] = w[pos4 + 18];
+    w4[3] = w[pos4 + 19];
+    w5[0] = w[pos4 + 20];
+    w5[1] = w[pos4 + 21];
+    w5[2] = w[pos4 + 22];
+    w5[3] = w[pos4 + 23];
+    w6[0] = w[pos4 + 24];
+    w6[1] = w[pos4 + 25];
+    w6[2] = w[pos4 + 26];
+    w6[3] = w[pos4 + 27];
+    w7[0] = w[pos4 + 28];
+    w7[1] = w[pos4 + 29];
+    w7[2] = w[pos4 + 30];
+    w7[3] = w[pos4 + 31];
+
+    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]);
+    w4[0] = swap32_S (w4[0]);
+    w4[1] = swap32_S (w4[1]);
+    w4[2] = swap32_S (w4[2]);
+    w4[3] = swap32_S (w4[3]);
+    w5[0] = swap32_S (w5[0]);
+    w5[1] = swap32_S (w5[1]);
+    w5[2] = swap32_S (w5[2]);
+    w5[3] = swap32_S (w5[3]);
+    w6[0] = swap32_S (w6[0]);
+    w6[1] = swap32_S (w6[1]);
+    w6[2] = swap32_S (w6[2]);
+    w6[3] = swap32_S (w6[3]);
+    w7[0] = swap32_S (w7[0]);
+    w7[1] = swap32_S (w7[1]);
+    w7[2] = swap32_S (w7[2]);
+    w7[3] = swap32_S (w7[3]);
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 128);
+  }
+
+  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];
+  w4[0] = w[pos4 + 16];
+  w4[1] = w[pos4 + 17];
+  w4[2] = w[pos4 + 18];
+  w4[3] = w[pos4 + 19];
+  w5[0] = w[pos4 + 20];
+  w5[1] = w[pos4 + 21];
+  w5[2] = w[pos4 + 22];
+  w5[3] = w[pos4 + 23];
+  w6[0] = w[pos4 + 24];
+  w6[1] = w[pos4 + 25];
+  w6[2] = w[pos4 + 26];
+  w6[3] = w[pos4 + 27];
+  w7[0] = w[pos4 + 28];
+  w7[1] = w[pos4 + 29];
+  w7[2] = w[pos4 + 30];
+  w7[3] = w[pos4 + 31];
+
+  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]);
+  w4[0] = swap32_S (w4[0]);
+  w4[1] = swap32_S (w4[1]);
+  w4[2] = swap32_S (w4[2]);
+  w4[3] = swap32_S (w4[3]);
+  w5[0] = swap32_S (w5[0]);
+  w5[1] = swap32_S (w5[1]);
+  w5[2] = swap32_S (w5[2]);
+  w5[3] = swap32_S (w5[3]);
+  w6[0] = swap32_S (w6[0]);
+  w6[1] = swap32_S (w6[1]);
+  w6[2] = swap32_S (w6[2]);
+  w6[3] = swap32_S (w6[3]);
+  w7[0] = swap32_S (w7[0]);
+  w7[1] = swap32_S (w7[1]);
+  w7[2] = swap32_S (w7[2]);
+  w7[3] = swap32_S (w7[3]);
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, len - pos1);
+}
+
+void sha384_update_global (sha384_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[4];
+
+  int pos1;
+  int pos4;
+
+  for (pos1 = 0, pos4 = 0; pos1 < len - 128; pos1 += 128, pos4 += 32)
+  {
+    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];
+    w4[0] = w[pos4 + 16];
+    w4[1] = w[pos4 + 17];
+    w4[2] = w[pos4 + 18];
+    w4[3] = w[pos4 + 19];
+    w5[0] = w[pos4 + 20];
+    w5[1] = w[pos4 + 21];
+    w5[2] = w[pos4 + 22];
+    w5[3] = w[pos4 + 23];
+    w6[0] = w[pos4 + 24];
+    w6[1] = w[pos4 + 25];
+    w6[2] = w[pos4 + 26];
+    w6[3] = w[pos4 + 27];
+    w7[0] = w[pos4 + 28];
+    w7[1] = w[pos4 + 29];
+    w7[2] = w[pos4 + 30];
+    w7[3] = w[pos4 + 31];
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 128);
+  }
+
+  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];
+  w4[0] = w[pos4 + 16];
+  w4[1] = w[pos4 + 17];
+  w4[2] = w[pos4 + 18];
+  w4[3] = w[pos4 + 19];
+  w5[0] = w[pos4 + 20];
+  w5[1] = w[pos4 + 21];
+  w5[2] = w[pos4 + 22];
+  w5[3] = w[pos4 + 23];
+  w6[0] = w[pos4 + 24];
+  w6[1] = w[pos4 + 25];
+  w6[2] = w[pos4 + 26];
+  w6[3] = w[pos4 + 27];
+  w7[0] = w[pos4 + 28];
+  w7[1] = w[pos4 + 29];
+  w7[2] = w[pos4 + 30];
+  w7[3] = w[pos4 + 31];
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, len - pos1);
+}
+
+void sha384_update_global_swap (sha384_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[4];
+
+  int pos1;
+  int pos4;
+
+  for (pos1 = 0, pos4 = 0; pos1 < len - 128; pos1 += 128, pos4 += 32)
+  {
+    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];
+    w4[0] = w[pos4 + 16];
+    w4[1] = w[pos4 + 17];
+    w4[2] = w[pos4 + 18];
+    w4[3] = w[pos4 + 19];
+    w5[0] = w[pos4 + 20];
+    w5[1] = w[pos4 + 21];
+    w5[2] = w[pos4 + 22];
+    w5[3] = w[pos4 + 23];
+    w6[0] = w[pos4 + 24];
+    w6[1] = w[pos4 + 25];
+    w6[2] = w[pos4 + 26];
+    w6[3] = w[pos4 + 27];
+    w7[0] = w[pos4 + 28];
+    w7[1] = w[pos4 + 29];
+    w7[2] = w[pos4 + 30];
+    w7[3] = w[pos4 + 31];
+
+    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]);
+    w4[0] = swap32_S (w4[0]);
+    w4[1] = swap32_S (w4[1]);
+    w4[2] = swap32_S (w4[2]);
+    w4[3] = swap32_S (w4[3]);
+    w5[0] = swap32_S (w5[0]);
+    w5[1] = swap32_S (w5[1]);
+    w5[2] = swap32_S (w5[2]);
+    w5[3] = swap32_S (w5[3]);
+    w6[0] = swap32_S (w6[0]);
+    w6[1] = swap32_S (w6[1]);
+    w6[2] = swap32_S (w6[2]);
+    w6[3] = swap32_S (w6[3]);
+    w7[0] = swap32_S (w7[0]);
+    w7[1] = swap32_S (w7[1]);
+    w7[2] = swap32_S (w7[2]);
+    w7[3] = swap32_S (w7[3]);
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 128);
+  }
+
+  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];
+  w4[0] = w[pos4 + 16];
+  w4[1] = w[pos4 + 17];
+  w4[2] = w[pos4 + 18];
+  w4[3] = w[pos4 + 19];
+  w5[0] = w[pos4 + 20];
+  w5[1] = w[pos4 + 21];
+  w5[2] = w[pos4 + 22];
+  w5[3] = w[pos4 + 23];
+  w6[0] = w[pos4 + 24];
+  w6[1] = w[pos4 + 25];
+  w6[2] = w[pos4 + 26];
+  w6[3] = w[pos4 + 27];
+  w7[0] = w[pos4 + 28];
+  w7[1] = w[pos4 + 29];
+  w7[2] = w[pos4 + 30];
+  w7[3] = w[pos4 + 31];
+
+  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]);
+  w4[0] = swap32_S (w4[0]);
+  w4[1] = swap32_S (w4[1]);
+  w4[2] = swap32_S (w4[2]);
+  w4[3] = swap32_S (w4[3]);
+  w5[0] = swap32_S (w5[0]);
+  w5[1] = swap32_S (w5[1]);
+  w5[2] = swap32_S (w5[2]);
+  w5[3] = swap32_S (w5[3]);
+  w6[0] = swap32_S (w6[0]);
+  w6[1] = swap32_S (w6[1]);
+  w6[2] = swap32_S (w6[2]);
+  w6[3] = swap32_S (w6[3]);
+  w7[0] = swap32_S (w7[0]);
+  w7[1] = swap32_S (w7[1]);
+  w7[2] = swap32_S (w7[2]);
+  w7[3] = swap32_S (w7[3]);
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, len - pos1);
+}
+
+void sha384_update_global_utf16le (sha384_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[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];
+
+    make_utf16le_S (w3, w6, w7);
+    make_utf16le_S (w2, w4, w5);
+    make_utf16le_S (w1, w2, w3);
+    make_utf16le_S (w0, w0, w1);
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 64 * 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];
+  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];
+
+  make_utf16le_S (w3, w6, w7);
+  make_utf16le_S (w2, w4, w5);
+  make_utf16le_S (w1, w2, w3);
+  make_utf16le_S (w0, w0, w1);
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, (len - pos1) * 2);
+}
+
+void sha384_update_global_utf16le_swap (sha384_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+  u32 w4[4];
+  u32 w5[4];
+  u32 w6[4];
+  u32 w7[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];
+
+    make_utf16le_S (w3, w6, w7);
+    make_utf16le_S (w2, w4, w5);
+    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]);
+    w4[0] = swap32_S (w4[0]);
+    w4[1] = swap32_S (w4[1]);
+    w4[2] = swap32_S (w4[2]);
+    w4[3] = swap32_S (w4[3]);
+    w5[0] = swap32_S (w5[0]);
+    w5[1] = swap32_S (w5[1]);
+    w5[2] = swap32_S (w5[2]);
+    w5[3] = swap32_S (w5[3]);
+    w6[0] = swap32_S (w6[0]);
+    w6[1] = swap32_S (w6[1]);
+    w6[2] = swap32_S (w6[2]);
+    w6[3] = swap32_S (w6[3]);
+    w7[0] = swap32_S (w7[0]);
+    w7[1] = swap32_S (w7[1]);
+    w7[2] = swap32_S (w7[2]);
+    w7[3] = swap32_S (w7[3]);
+
+    sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 64 * 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];
+  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];
+
+  make_utf16le_S (w3, w6, w7);
+  make_utf16le_S (w2, w4, w5);
+  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]);
+  w4[0] = swap32_S (w4[0]);
+  w4[1] = swap32_S (w4[1]);
+  w4[2] = swap32_S (w4[2]);
+  w4[3] = swap32_S (w4[3]);
+  w5[0] = swap32_S (w5[0]);
+  w5[1] = swap32_S (w5[1]);
+  w5[2] = swap32_S (w5[2]);
+  w5[3] = swap32_S (w5[3]);
+  w6[0] = swap32_S (w6[0]);
+  w6[1] = swap32_S (w6[1]);
+  w6[2] = swap32_S (w6[2]);
+  w6[3] = swap32_S (w6[3]);
+  w7[0] = swap32_S (w7[0]);
+  w7[1] = swap32_S (w7[1]);
+  w7[2] = swap32_S (w7[2]);
+  w7[3] = swap32_S (w7[3]);
+
+  sha384_update_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, (len - pos1) * 2);
+}
+
+void sha384_final (sha384_ctx_t *ctx)
+{
+  int pos = ctx->len & 127;
+
+  append_0x80_8x4_S (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, pos ^ 3);
+
+  if (pos >= 112)
+  {
+    sha384_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, 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->w4[0] = 0;
+    ctx->w4[1] = 0;
+    ctx->w4[2] = 0;
+    ctx->w4[3] = 0;
+    ctx->w5[0] = 0;
+    ctx->w5[1] = 0;
+    ctx->w5[2] = 0;
+    ctx->w5[3] = 0;
+    ctx->w6[0] = 0;
+    ctx->w6[1] = 0;
+    ctx->w6[2] = 0;
+    ctx->w6[3] = 0;
+    ctx->w7[0] = 0;
+    ctx->w7[1] = 0;
+    ctx->w7[2] = 0;
+    ctx->w7[3] = 0;
+  }
+
+  ctx->w7[2] = 0;
+  ctx->w7[3] = ctx->len * 8;
+
+  sha384_transform (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, ctx->h);
+}
+
+// sha384_hmac
+
+typedef struct sha384_hmac_ctx
+{
+  sha384_ctx_t ipad;
+  sha384_ctx_t opad;
+
+} sha384_hmac_ctx_t;
+
+void sha384_hmac_init (sha384_hmac_ctx_t *ctx, const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], const u32 w4[4], const u32 w5[4], const u32 w6[4], const u32 w7[4])
+{
+  u32 t0[4];
+  u32 t1[4];
+  u32 t2[4];
+  u32 t3[4];
+  u32 t4[4];
+  u32 t5[4];
+  u32 t6[4];
+  u32 t7[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;
+  t4[0] = w4[0] ^ 0x36363636;
+  t4[1] = w4[1] ^ 0x36363636;
+  t4[2] = w4[2] ^ 0x36363636;
+  t4[3] = w4[3] ^ 0x36363636;
+  t5[0] = w5[0] ^ 0x36363636;
+  t5[1] = w5[1] ^ 0x36363636;
+  t5[2] = w5[2] ^ 0x36363636;
+  t5[3] = w5[3] ^ 0x36363636;
+  t6[0] = w6[0] ^ 0x36363636;
+  t6[1] = w6[1] ^ 0x36363636;
+  t6[2] = w6[2] ^ 0x36363636;
+  t6[3] = w6[3] ^ 0x36363636;
+  t7[0] = w7[0] ^ 0x36363636;
+  t7[1] = w7[1] ^ 0x36363636;
+  t7[2] = w7[2] ^ 0x36363636;
+  t7[3] = w7[3] ^ 0x36363636;
+
+  sha384_init (&ctx->ipad);
+
+  sha384_update_128 (&ctx->ipad, t0, t1, t2, t3, t4, t5, t6, t7, 128);
+
+  // 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;
+  t4[0] = w4[0] ^ 0x5c5c5c5c;
+  t4[1] = w4[1] ^ 0x5c5c5c5c;
+  t4[2] = w4[2] ^ 0x5c5c5c5c;
+  t4[3] = w4[3] ^ 0x5c5c5c5c;
+  t5[0] = w5[0] ^ 0x5c5c5c5c;
+  t5[1] = w5[1] ^ 0x5c5c5c5c;
+  t5[2] = w5[2] ^ 0x5c5c5c5c;
+  t5[3] = w5[3] ^ 0x5c5c5c5c;
+  t6[0] = w6[0] ^ 0x5c5c5c5c;
+  t6[1] = w6[1] ^ 0x5c5c5c5c;
+  t6[2] = w6[2] ^ 0x5c5c5c5c;
+  t6[3] = w6[3] ^ 0x5c5c5c5c;
+  t7[0] = w7[0] ^ 0x5c5c5c5c;
+  t7[1] = w7[1] ^ 0x5c5c5c5c;
+  t7[2] = w7[2] ^ 0x5c5c5c5c;
+  t7[3] = w7[3] ^ 0x5c5c5c5c;
+
+  sha384_init (&ctx->opad);
+
+  sha384_update_128 (&ctx->opad, t0, t1, t2, t3, t4, t5, t6, t7, 128);
+}
+
+void sha384_hmac_update_128 (sha384_hmac_ctx_t *ctx, u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], u32 w4[4], u32 w5[4], u32 w6[4], u32 w7[4], const int len)
+{
+  sha384_update_128 (&ctx->ipad, w0, w1, w2, w3, w4, w5, w6, w7, len);
+}
+
+void sha384_hmac_update (sha384_hmac_ctx_t *ctx, const u32 *w, const int len)
+{
+  sha384_update (&ctx->ipad, w, len);
+}
+
+void sha384_hmac_update_global (sha384_hmac_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  sha384_update_global (&ctx->ipad, w, len);
+}
+
+void sha384_hmac_update_global_swap (sha384_hmac_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  sha384_update_global_swap (&ctx->ipad, w, len);
+}
+
+void sha384_hmac_update_global_utf16le (sha384_hmac_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  sha384_update_global_utf16le (&ctx->ipad, w, len);
+}
+
+void sha384_hmac_update_global_utf16le_swap (sha384_hmac_ctx_t *ctx, const __global u32 *w, const int len)
+{
+  sha384_update_global_utf16le_swap (&ctx->ipad, w, len);
+}
+
+void sha384_hmac_final (sha384_hmac_ctx_t *ctx)
+{
+  sha384_final (&ctx->ipad);
+
+  u32 t0[4];
+  u32 t1[4];
+  u32 t2[4];
+  u32 t3[4];
+  u32 t4[4];
+  u32 t5[4];
+  u32 t6[4];
+  u32 t7[4];
+
+  t0[0] = h32_from_64_S (ctx->ipad.h[0]);
+  t0[1] = l32_from_64_S (ctx->ipad.h[0]);
+  t0[2] = h32_from_64_S (ctx->ipad.h[1]);
+  t0[3] = l32_from_64_S (ctx->ipad.h[1]);
+  t1[0] = h32_from_64_S (ctx->ipad.h[2]);
+  t1[1] = l32_from_64_S (ctx->ipad.h[2]);
+  t1[2] = h32_from_64_S (ctx->ipad.h[3]);
+  t1[3] = l32_from_64_S (ctx->ipad.h[3]);
+  t2[0] = h32_from_64_S (ctx->ipad.h[4]);
+  t2[1] = l32_from_64_S (ctx->ipad.h[4]);
+  t2[2] = h32_from_64_S (ctx->ipad.h[5]);
+  t2[3] = l32_from_64_S (ctx->ipad.h[5]);
+  t3[0] = h32_from_64_S (ctx->ipad.h[6]);
+  t3[1] = l32_from_64_S (ctx->ipad.h[6]);
+  t3[2] = h32_from_64_S (ctx->ipad.h[7]);
+  t3[3] = l32_from_64_S (ctx->ipad.h[7]);
+  t4[0] = 0;
+  t4[1] = 0;
+  t4[2] = 0;
+  t4[3] = 0;
+  t5[0] = 0;
+  t5[1] = 0;
+  t5[2] = 0;
+  t5[3] = 0;
+  t6[0] = 0;
+  t6[1] = 0;
+  t6[2] = 0;
+  t6[3] = 0;
+  t7[0] = 0;
+  t7[1] = 0;
+  t7[2] = 0;
+  t7[3] = 0;
+
+  sha384_update_128 (&ctx->opad, t0, t1, t2, t3, t4, t5, t6, t7, 64);
+
+  sha384_final (&ctx->opad);
+}
+
+// while input buf can be a vector datatype, the length of the different elements can not
+
+typedef struct sha384_ctx_vector
+{
+  u64x h[8];
+
+  u32x w0[4];
+  u32x w1[4];
+  u32x w2[4];
+  u32x w3[4];
+  u32x w4[4];
+  u32x w5[4];
+  u32x w6[4];
+  u32x w7[4];
+
+  int  len;
+
+} sha384_ctx_vector_t;
+
+void sha384_transform_vector (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], const u32x w4[4], const u32x w5[4], const u32x w6[4], const u32x w7[4], u64x digest[8])
+{
+  u64x a = digest[0];
+  u64x b = digest[1];
+  u64x c = digest[2];
+  u64x d = digest[3];
+  u64x e = digest[4];
+  u64x f = digest[5];
+  u64x g = digest[6];
+  u64x h = digest[7];
+
+  u64x w0_t = hl32_to_64 (w0[0], w0[1]);
+  u64x w1_t = hl32_to_64 (w0[2], w0[3]);
+  u64x w2_t = hl32_to_64 (w1[0], w1[1]);
+  u64x w3_t = hl32_to_64 (w1[2], w1[3]);
+  u64x w4_t = hl32_to_64 (w2[0], w2[1]);
+  u64x w5_t = hl32_to_64 (w2[2], w2[3]);
+  u64x w6_t = hl32_to_64 (w3[0], w3[1]);
+  u64x w7_t = hl32_to_64 (w3[2], w3[3]);
+  u64x w8_t = hl32_to_64 (w4[0], w4[1]);
+  u64x w9_t = hl32_to_64 (w4[2], w4[3]);
+  u64x wa_t = hl32_to_64 (w5[0], w5[1]);
+  u64x wb_t = hl32_to_64 (w5[2], w5[3]);
+  u64x wc_t = hl32_to_64 (w6[0], w6[1]);
+  u64x wd_t = hl32_to_64 (w6[2], w6[3]);
+  u64x we_t = hl32_to_64 (w7[0], w7[1]);
+  u64x wf_t = hl32_to_64 (w7[2], w7[3]);
+
+  #define ROUND_EXPAND()                            \
+  {                                                 \
+    w0_t = SHA384_EXPAND (we_t, w9_t, w1_t, w0_t);  \
+    w1_t = SHA384_EXPAND (wf_t, wa_t, w2_t, w1_t);  \
+    w2_t = SHA384_EXPAND (w0_t, wb_t, w3_t, w2_t);  \
+    w3_t = SHA384_EXPAND (w1_t, wc_t, w4_t, w3_t);  \
+    w4_t = SHA384_EXPAND (w2_t, wd_t, w5_t, w4_t);  \
+    w5_t = SHA384_EXPAND (w3_t, we_t, w6_t, w5_t);  \
+    w6_t = SHA384_EXPAND (w4_t, wf_t, w7_t, w6_t);  \
+    w7_t = SHA384_EXPAND (w5_t, w0_t, w8_t, w7_t);  \
+    w8_t = SHA384_EXPAND (w6_t, w1_t, w9_t, w8_t);  \
+    w9_t = SHA384_EXPAND (w7_t, w2_t, wa_t, w9_t);  \
+    wa_t = SHA384_EXPAND (w8_t, w3_t, wb_t, wa_t);  \
+    wb_t = SHA384_EXPAND (w9_t, w4_t, wc_t, wb_t);  \
+    wc_t = SHA384_EXPAND (wa_t, w5_t, wd_t, wc_t);  \
+    wd_t = SHA384_EXPAND (wb_t, w6_t, we_t, wd_t);  \
+    we_t = SHA384_EXPAND (wc_t, w7_t, wf_t, we_t);  \
+    wf_t = SHA384_EXPAND (wd_t, w8_t, w0_t, wf_t);  \
+  }
+
+  #define ROUND_STEP(i)                                                                   \
+  {                                                                                       \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, a, b, c, d, e, f, g, h, w0_t, k_sha384[i +  0]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, h, a, b, c, d, e, f, g, w1_t, k_sha384[i +  1]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, g, h, a, b, c, d, e, f, w2_t, k_sha384[i +  2]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, f, g, h, a, b, c, d, e, w3_t, k_sha384[i +  3]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, e, f, g, h, a, b, c, d, w4_t, k_sha384[i +  4]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, d, e, f, g, h, a, b, c, w5_t, k_sha384[i +  5]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, c, d, e, f, g, h, a, b, w6_t, k_sha384[i +  6]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, b, c, d, e, f, g, h, a, w7_t, k_sha384[i +  7]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, a, b, c, d, e, f, g, h, w8_t, k_sha384[i +  8]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, h, a, b, c, d, e, f, g, w9_t, k_sha384[i +  9]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, g, h, a, b, c, d, e, f, wa_t, k_sha384[i + 10]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, f, g, h, a, b, c, d, e, wb_t, k_sha384[i + 11]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, e, f, g, h, a, b, c, d, wc_t, k_sha384[i + 12]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, d, e, f, g, h, a, b, c, wd_t, k_sha384[i + 13]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, c, d, e, f, g, h, a, b, we_t, k_sha384[i + 14]); \
+    SHA384_STEP (SHA384_F0o, SHA384_F1o, b, c, d, e, f, g, h, a, wf_t, k_sha384[i + 15]); \
+  }
+
+  ROUND_STEP (0);
+
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (int i = 16; i < 80; i += 16)
+  {
+    ROUND_EXPAND (); ROUND_STEP (i);
+  }
+
+  #undef ROUND_EXPAND
+  #undef ROUND_STEP
+
+  digest[0] += a;
+  digest[1] += b;
+  digest[2] += c;
+  digest[3] += d;
+  digest[4] += e;
+  digest[5] += f;
+  digest[6] += g;
+  digest[7] += h;
+}
+
+void sha384_init_vector (sha384_ctx_vector_t *ctx)
+{
+  ctx->h[0] = SHA384M_A;
+  ctx->h[1] = SHA384M_B;
+  ctx->h[2] = SHA384M_C;
+  ctx->h[3] = SHA384M_D;
+  ctx->h[4] = SHA384M_E;
+  ctx->h[5] = SHA384M_F;
+  ctx->h[6] = SHA384M_G;
+  ctx->h[7] = SHA384M_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->w4[0] = 0;
+  ctx->w4[1] = 0;
+  ctx->w4[2] = 0;
+  ctx->w4[3] = 0;
+  ctx->w5[0] = 0;
+  ctx->w5[1] = 0;
+  ctx->w5[2] = 0;
+  ctx->w5[3] = 0;
+  ctx->w6[0] = 0;
+  ctx->w6[1] = 0;
+  ctx->w6[2] = 0;
+  ctx->w6[3] = 0;
+  ctx->w7[0] = 0;
+  ctx->w7[1] = 0;
+  ctx->w7[2] = 0;
+  ctx->w7[3] = 0;
+
+  ctx->len = 0;
+}
+
+void sha384_update_vector_128 (sha384_ctx_vector_t *ctx, u32x w0[4], u32x w1[4], u32x w2[4], u32x w3[4], u32x w4[4], u32x w5[4], u32x w6[4], u32x w7[4], const int len)
+{
+  const int pos = ctx->len & 127;
+
+  ctx->len += len;
+
+  if ((pos + len) < 128)
+  {
+    switch_buffer_by_offset_8x4_be (w0, w1, w2, w3, w4, w5, w6, w7, 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];
+    ctx->w4[0] |= w4[0];
+    ctx->w4[1] |= w4[1];
+    ctx->w4[2] |= w4[2];
+    ctx->w4[3] |= w4[3];
+    ctx->w5[0] |= w5[0];
+    ctx->w5[1] |= w5[1];
+    ctx->w5[2] |= w5[2];
+    ctx->w5[3] |= w5[3];
+    ctx->w6[0] |= w6[0];
+    ctx->w6[1] |= w6[1];
+    ctx->w6[2] |= w6[2];
+    ctx->w6[3] |= w6[3];
+    ctx->w7[0] |= w7[0];
+    ctx->w7[1] |= w7[1];
+    ctx->w7[2] |= w7[2];
+    ctx->w7[3] |= w7[3];
+  }
+  else
+  {
+    u32x c0[4] = { 0 };
+    u32x c1[4] = { 0 };
+    u32x c2[4] = { 0 };
+    u32x c3[4] = { 0 };
+    u32x c4[4] = { 0 };
+    u32x c5[4] = { 0 };
+    u32x c6[4] = { 0 };
+    u32x c7[4] = { 0 };
+
+    switch_buffer_by_offset_8x4_carry_be (w0, w1, w2, w3, w4, w5, w6, w7, c0, c1, c2, c3, c4, c5, c6, c7, 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];
+    ctx->w4[0] |= w4[0];
+    ctx->w4[1] |= w4[1];
+    ctx->w4[2] |= w4[2];
+    ctx->w4[3] |= w4[3];
+    ctx->w5[0] |= w5[0];
+    ctx->w5[1] |= w5[1];
+    ctx->w5[2] |= w5[2];
+    ctx->w5[3] |= w5[3];
+    ctx->w6[0] |= w6[0];
+    ctx->w6[1] |= w6[1];
+    ctx->w6[2] |= w6[2];
+    ctx->w6[3] |= w6[3];
+    ctx->w7[0] |= w7[0];
+    ctx->w7[1] |= w7[1];
+    ctx->w7[2] |= w7[2];
+    ctx->w7[3] |= w7[3];
+
+    sha384_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, 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];
+    ctx->w4[0] = c4[0];
+    ctx->w4[1] = c4[1];
+    ctx->w4[2] = c4[2];
+    ctx->w4[3] = c4[3];
+    ctx->w5[0] = c5[0];
+    ctx->w5[1] = c5[1];
+    ctx->w5[2] = c5[2];
+    ctx->w5[3] = c5[3];
+    ctx->w6[0] = c6[0];
+    ctx->w6[1] = c6[1];
+    ctx->w6[2] = c6[2];
+    ctx->w6[3] = c6[3];
+    ctx->w7[0] = c7[0];
+    ctx->w7[1] = c7[1];
+    ctx->w7[2] = c7[2];
+    ctx->w7[3] = c7[3];
+  }
+}
+
+void sha384_update_vector (sha384_ctx_vector_t *ctx, const u32x *w, const int len)
+{
+  u32x w0[4];
+  u32x w1[4];
+  u32x w2[4];
+  u32x w3[4];
+  u32x w4[4];
+  u32x w5[4];
+  u32x w6[4];
+  u32x w7[4];
+
+  int pos1;
+  int pos4;
+
+  for (pos1 = 0, pos4 = 0; pos1 < len - 128; pos1 += 128, pos4 += 32)
+  {
+    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];
+    w4[0] = w[pos4 + 16];
+    w4[1] = w[pos4 + 17];
+    w4[2] = w[pos4 + 18];
+    w4[3] = w[pos4 + 19];
+    w5[0] = w[pos4 + 20];
+    w5[1] = w[pos4 + 21];
+    w5[2] = w[pos4 + 22];
+    w5[3] = w[pos4 + 23];
+    w6[0] = w[pos4 + 24];
+    w6[1] = w[pos4 + 25];
+    w6[2] = w[pos4 + 26];
+    w6[3] = w[pos4 + 27];
+    w7[0] = w[pos4 + 28];
+    w7[1] = w[pos4 + 29];
+    w7[2] = w[pos4 + 30];
+    w7[3] = w[pos4 + 31];
+
+    sha384_update_vector_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, 128);
+  }
+
+  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];
+  w4[0] = w[pos4 + 16];
+  w4[1] = w[pos4 + 17];
+  w4[2] = w[pos4 + 18];
+  w4[3] = w[pos4 + 19];
+  w5[0] = w[pos4 + 20];
+  w5[1] = w[pos4 + 21];
+  w5[2] = w[pos4 + 22];
+  w5[3] = w[pos4 + 23];
+  w6[0] = w[pos4 + 24];
+  w6[1] = w[pos4 + 25];
+  w6[2] = w[pos4 + 26];
+  w6[3] = w[pos4 + 27];
+  w7[0] = w[pos4 + 28];
+  w7[1] = w[pos4 + 29];
+  w7[2] = w[pos4 + 30];
+  w7[3] = w[pos4 + 31];
+
+  sha384_update_vector_128 (ctx, w0, w1, w2, w3, w4, w5, w6, w7, len - pos1);
+}
+
+void sha384_final_vector (sha384_ctx_vector_t *ctx)
+{
+  int pos = ctx->len & 127;
+
+  append_0x80_8x4 (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, pos ^ 3);
+
+  if (pos >= 112)
+  {
+    sha384_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, 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->w4[0] = 0;
+    ctx->w4[1] = 0;
+    ctx->w4[2] = 0;
+    ctx->w4[3] = 0;
+    ctx->w5[0] = 0;
+    ctx->w5[1] = 0;
+    ctx->w5[2] = 0;
+    ctx->w5[3] = 0;
+    ctx->w6[0] = 0;
+    ctx->w6[1] = 0;
+    ctx->w6[2] = 0;
+    ctx->w6[3] = 0;
+    ctx->w7[0] = 0;
+    ctx->w7[1] = 0;
+    ctx->w7[2] = 0;
+    ctx->w7[3] = 0;
+  }
+
+  ctx->w7[2] = 0;
+  ctx->w7[3] = ctx->len * 8;
+
+  sha384_transform_vector (ctx->w0, ctx->w1, ctx->w2, ctx->w3, ctx->w4, ctx->w5, ctx->w6, ctx->w7, ctx->h);
+}
diff --git a/OpenCL/m10800_a3-pure.cl b/OpenCL/m10800_a3-pure.cl
new file mode 100644
index 000000000..ea2143d52
--- /dev/null
+++ b/OpenCL/m10800_a3-pure.cl
@@ -0,0 +1,144 @@
+/**
+ * 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"
+#include "inc_hash_sha384.cl"
+
+__kernel void m10800_mxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const u32x *words_buf_r, __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 u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32x w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    sha384_ctx_vector_t ctx;
+
+    sha384_init_vector (&ctx);
+
+    sha384_update_vector (&ctx, w, pw_len);
+
+    sha384_final_vector (&ctx);
+
+    const u32x r0 = l32_from_64 (ctx.h[3]);
+    const u32x r1 = h32_from_64 (ctx.h[3]);
+    const u32x r2 = l32_from_64 (ctx.h[2]);
+    const u32x r3 = h32_from_64 (ctx.h[2]);
+
+    COMPARE_M_SIMD (r0, r1, r2, r3);
+  }
+}
+
+__kernel void m10800_sxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const u32x *words_buf_r, __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 u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * 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]
+  };
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32x w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    sha384_ctx_vector_t ctx;
+
+    sha384_init_vector (&ctx);
+
+    sha384_update_vector (&ctx, w, pw_len);
+
+    sha384_final_vector (&ctx);
+
+    const u32x r0 = l32_from_64 (ctx.h[3]);
+    const u32x r1 = h32_from_64 (ctx.h[3]);
+    const u32x r2 = l32_from_64 (ctx.h[2]);
+    const u32x r3 = h32_from_64 (ctx.h[2]);
+
+    COMPARE_S_SIMD (r0, r1, r2, r3);
+  }
+}