diff --git a/OpenCL/m00020_a3-optimized.cl b/OpenCL/m00020_a3-optimized.cl
index 002119f5e..6a8268d27 100644
--- a/OpenCL/m00020_a3-optimized.cl
+++ b/OpenCL/m00020_a3-optimized.cl
@@ -52,63 +52,57 @@ void m00020m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_le (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = pw_salt_len * 8;
+    t3[3]  = 0;
 
     /**
      * md5
@@ -243,63 +237,57 @@ void m00020s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_le (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = pw_salt_len * 8;
+    t3[3]  = 0;
 
     /**
      * md5
diff --git a/OpenCL/m00040_a3-optimized.cl b/OpenCL/m00040_a3-optimized.cl
index 8017e0ca9..063b851a1 100644
--- a/OpenCL/m00040_a3-optimized.cl
+++ b/OpenCL/m00040_a3-optimized.cl
@@ -52,63 +52,57 @@ void m00040m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_le (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = pw_salt_len * 8;
+    t3[3]  = 0;
 
     /**
      * md5
@@ -243,63 +237,57 @@ void m00040s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_le (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = pw_salt_len * 8;
+    t3[3]  = 0;
 
     /**
      * md5
diff --git a/OpenCL/m00120_a3-optimized.cl b/OpenCL/m00120_a3-optimized.cl
index 2170c91f7..011fdbeba 100644
--- a/OpenCL/m00120_a3-optimized.cl
+++ b/OpenCL/m00120_a3-optimized.cl
@@ -52,63 +52,57 @@ void m00120m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha1
@@ -295,63 +289,57 @@ void m00120s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha1
diff --git a/OpenCL/m00140_a3-optimized.cl b/OpenCL/m00140_a3-optimized.cl
index a304aa880..79f98c67d 100644
--- a/OpenCL/m00140_a3-optimized.cl
+++ b/OpenCL/m00140_a3-optimized.cl
@@ -52,63 +52,57 @@ void m00140m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha1
@@ -295,63 +289,57 @@ void m00140s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha1
diff --git a/OpenCL/m01420_a3-optimized.cl b/OpenCL/m01420_a3-optimized.cl
index 21cb7c8cb..70829a8c4 100644
--- a/OpenCL/m01420_a3-optimized.cl
+++ b/OpenCL/m01420_a3-optimized.cl
@@ -66,63 +66,57 @@ void m01420m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha256
@@ -247,6 +241,24 @@ void m01420s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digests_buf[digests_offset].digest_buf[DGST_R3]
   };
 
+  /**
+   * reverse
+   */
+
+  u32 a_rev = digests_buf[digests_offset].digest_buf[0];
+  u32 b_rev = digests_buf[digests_offset].digest_buf[1];
+  u32 c_rev = digests_buf[digests_offset].digest_buf[2];
+  u32 d_rev = digests_buf[digests_offset].digest_buf[3];
+  u32 e_rev = digests_buf[digests_offset].digest_buf[4];
+  u32 f_rev = digests_buf[digests_offset].digest_buf[5];
+  u32 g_rev = digests_buf[digests_offset].digest_buf[6];
+  u32 h_rev = digests_buf[digests_offset].digest_buf[7];
+
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+
   /**
    * salt
    */
@@ -278,81 +290,57 @@ void m01420s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * reverse
-   */
-
-  u32 a_rev = digests_buf[digests_offset].digest_buf[0];
-  u32 b_rev = digests_buf[digests_offset].digest_buf[1];
-  u32 c_rev = digests_buf[digests_offset].digest_buf[2];
-  u32 d_rev = digests_buf[digests_offset].digest_buf[3];
-  u32 e_rev = digests_buf[digests_offset].digest_buf[4];
-  u32 f_rev = digests_buf[digests_offset].digest_buf[5];
-  u32 g_rev = digests_buf[digests_offset].digest_buf[6];
-  u32 h_rev = digests_buf[digests_offset].digest_buf[7];
-
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha256
diff --git a/OpenCL/m01440_a3-optimized.cl b/OpenCL/m01440_a3-optimized.cl
index f142dfdbc..12518e0c0 100644
--- a/OpenCL/m01440_a3-optimized.cl
+++ b/OpenCL/m01440_a3-optimized.cl
@@ -66,63 +66,57 @@ void m01440m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha256
@@ -247,6 +241,24 @@ void m01440s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digests_buf[digests_offset].digest_buf[DGST_R3]
   };
 
+  /**
+   * reverse
+   */
+
+  u32 a_rev = digests_buf[digests_offset].digest_buf[0];
+  u32 b_rev = digests_buf[digests_offset].digest_buf[1];
+  u32 c_rev = digests_buf[digests_offset].digest_buf[2];
+  u32 d_rev = digests_buf[digests_offset].digest_buf[3];
+  u32 e_rev = digests_buf[digests_offset].digest_buf[4];
+  u32 f_rev = digests_buf[digests_offset].digest_buf[5];
+  u32 g_rev = digests_buf[digests_offset].digest_buf[6];
+  u32 h_rev = digests_buf[digests_offset].digest_buf[7];
+
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
+
   /**
    * salt
    */
@@ -278,81 +290,57 @@ void m01440s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * reverse
-   */
-
-  u32 a_rev = digests_buf[digests_offset].digest_buf[0];
-  u32 b_rev = digests_buf[digests_offset].digest_buf[1];
-  u32 c_rev = digests_buf[digests_offset].digest_buf[2];
-  u32 d_rev = digests_buf[digests_offset].digest_buf[3];
-  u32 e_rev = digests_buf[digests_offset].digest_buf[4];
-  u32 f_rev = digests_buf[digests_offset].digest_buf[5];
-  u32 g_rev = digests_buf[digests_offset].digest_buf[6];
-  u32 h_rev = digests_buf[digests_offset].digest_buf[7];
-
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-  SHA256_STEP_REV (a_rev, b_rev, c_rev, d_rev, e_rev, f_rev, g_rev, h_rev);
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha256
diff --git a/OpenCL/m01720_a3-optimized.cl b/OpenCL/m01720_a3-optimized.cl
index 4e57bfcf2..3ebbea0d8 100644
--- a/OpenCL/m01720_a3-optimized.cl
+++ b/OpenCL/m01720_a3-optimized.cl
@@ -175,90 +175,62 @@ void m01720m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha512
      */
 
-    u32x w0_t[4];
-    u32x w1_t[4];
-    u32x w2_t[4];
-    u32x w3_t[4];
-
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = 0;
-    w3_t[3] = pw_salt_len * 8;
-
     u64x digest[8];
 
     digest[0] = SHA512M_A;
@@ -270,7 +242,7 @@ void m01720m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digest[6] = SHA512M_G;
     digest[7] = SHA512M_H;
 
-    sha512_transform (w0_t, w1_t, w2_t, w3_t, digest);
+    sha512_transform (t0, t1, t2, t3, digest);
 
     const u32x r0 = l32_from_64 (digest[7]);
     const u32x r1 = h32_from_64 (digest[7]);
@@ -333,90 +305,62 @@ void m01720s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha512
      */
 
-    u32x w0_t[4];
-    u32x w1_t[4];
-    u32x w2_t[4];
-    u32x w3_t[4];
-
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = 0;
-    w3_t[3] = pw_salt_len * 8;
-
     u64x digest[8];
 
     digest[0] = SHA512M_A;
@@ -428,7 +372,7 @@ void m01720s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digest[6] = SHA512M_G;
     digest[7] = SHA512M_H;
 
-    sha512_transform (w0_t, w1_t, w2_t, w3_t, digest);
+    sha512_transform (t0, t1, t2, t3, digest);
 
     const u32x r0 = l32_from_64 (digest[7]);
     const u32x r1 = h32_from_64 (digest[7]);
diff --git a/OpenCL/m01740_a3-optimized.cl b/OpenCL/m01740_a3-optimized.cl
index e78a02ae0..f1811cc7c 100644
--- a/OpenCL/m01740_a3-optimized.cl
+++ b/OpenCL/m01740_a3-optimized.cl
@@ -175,90 +175,62 @@ void m01740m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha512
      */
 
-    u32x w0_t[4];
-    u32x w1_t[4];
-    u32x w2_t[4];
-    u32x w3_t[4];
-
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = 0;
-    w3_t[3] = pw_salt_len * 8;
-
     u64x digest[8];
 
     digest[0] = SHA512M_A;
@@ -270,7 +242,7 @@ void m01740m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digest[6] = SHA512M_G;
     digest[7] = SHA512M_H;
 
-    sha512_transform (w0_t, w1_t, w2_t, w3_t, digest);
+    sha512_transform (t0, t1, t2, t3, digest);
 
     const u32x r0 = l32_from_64 (digest[7]);
     const u32x r1 = h32_from_64 (digest[7]);
@@ -333,90 +305,62 @@ void m01740s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 pw_salt_len = pw_len + salt_len;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_be_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = w3[2];
-  t3[3] = w3[3];
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_be_4x4 (t0, t1, t2, t3, w0lr, salt_len);
+    u32x t0[4];
+    u32x t1[4];
+    u32x t2[4];
+    u32x t3[4];
+
+    t0[0] = w0lr;
+    t0[1] = w0[1];
+    t0[2] = w0[2];
+    t0[3] = w0[3];
+    t1[0] = w1[0];
+    t1[1] = w1[1];
+    t1[2] = w1[2];
+    t1[3] = w1[3];
+    t2[0] = w2[0];
+    t2[1] = w2[1];
+    t2[2] = w2[2];
+    t2[3] = w2[3];
+    t3[0] = w3[0];
+    t3[1] = w3[1];
+    t3[2] = w3[2];
+    t3[3] = w3[3];
+
+    switch_buffer_by_offset_be (t0, t1, t2, t3, salt_len);
+
+    t0[0] |= salt_buf0[0];
+    t0[1] |= salt_buf0[1];
+    t0[2] |= salt_buf0[2];
+    t0[3] |= salt_buf0[3];
+    t1[0] |= salt_buf1[0];
+    t1[1] |= salt_buf1[1];
+    t1[2] |= salt_buf1[2];
+    t1[3] |= salt_buf1[3];
+    t2[0] |= salt_buf2[0];
+    t2[1] |= salt_buf2[1];
+    t2[2] |= salt_buf2[2];
+    t2[3] |= salt_buf2[3];
+    t3[0] |= salt_buf3[0];
+    t3[1] |= salt_buf3[1];
+    t3[2]  = 0;
+    t3[3]  = pw_salt_len * 8;
 
     /**
      * sha512
      */
 
-    u32x w0_t[4];
-    u32x w1_t[4];
-    u32x w2_t[4];
-    u32x w3_t[4];
-
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = 0;
-    w3_t[3] = pw_salt_len * 8;
-
     u64x digest[8];
 
     digest[0] = SHA512M_A;
@@ -428,7 +372,7 @@ void m01740s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
     digest[6] = SHA512M_G;
     digest[7] = SHA512M_H;
 
-    sha512_transform (w0_t, w1_t, w2_t, w3_t, digest);
+    sha512_transform (t0, t1, t2, t3, digest);
 
     const u32x r0 = l32_from_64 (digest[7]);
     const u32x r1 = h32_from_64 (digest[7]);
diff --git a/OpenCL/m03100_a3-optimized.cl b/OpenCL/m03100_a3-optimized.cl
index 54460d8b7..77a4c7e9c 100644
--- a/OpenCL/m03100_a3-optimized.cl
+++ b/OpenCL/m03100_a3-optimized.cl
@@ -513,62 +513,6 @@ void m03100m (__local u32 (*s_SPtrans)[64], __local u32 (*s_skb)[64], u32 w[16],
 
   const u32 salt_word_len = (salt_len + pw_len) * 2;
 
-  /**
-   * prepend salt
-   */
-
-  u32 w0_t[4];
-  u32 w1_t[4];
-  u32 w2_t[4];
-  u32 w3_t[4];
-
-  w0_t[0] = w[ 0];
-  w0_t[1] = w[ 1];
-  w0_t[2] = w[ 2];
-  w0_t[3] = w[ 3];
-  w1_t[0] = w[ 4];
-  w1_t[1] = w[ 5];
-  w1_t[2] = w[ 6];
-  w1_t[3] = w[ 7];
-  w2_t[0] = w[ 8];
-  w2_t[1] = w[ 9];
-  w2_t[2] = w[10];
-  w2_t[3] = w[11];
-  w3_t[0] = w[12];
-  w3_t[1] = w[13];
-  w3_t[2] = w[14];
-  w3_t[3] = w[15];
-
-  switch_buffer_by_offset_le_S (w0_t, w1_t, w2_t, w3_t, salt_len);
-
-  w0_t[0] |= salt_buf0[0];
-  w0_t[1] |= salt_buf0[1];
-  w0_t[2] |= salt_buf0[2];
-  w0_t[3] |= salt_buf0[3];
-  w1_t[0] |= salt_buf1[0];
-  w1_t[1] |= salt_buf1[1];
-  w1_t[2] |= salt_buf1[2];
-  w1_t[3] |= salt_buf1[3];
-
-  u32x dst[16];
-
-  dst[ 0] = w0_t[0];
-  dst[ 1] = w0_t[1];
-  dst[ 2] = w0_t[2];
-  dst[ 3] = w0_t[3];
-  dst[ 4] = w1_t[0];
-  dst[ 5] = w1_t[1];
-  dst[ 6] = w1_t[2];
-  dst[ 7] = w1_t[3];
-  dst[ 8] = w2_t[0];
-  dst[ 9] = w2_t[1];
-  dst[10] = w2_t[2];
-  dst[11] = w2_t[3];
-  dst[12] = w3_t[0];
-  dst[13] = w3_t[1];
-  dst[14] = w3_t[2];
-  dst[15] = w3_t[3];
-
   /**
    * loop
    */
@@ -579,9 +523,50 @@ void m03100m (__local u32 (*s_SPtrans)[64], __local u32 (*s_skb)[64], u32 w[16],
   {
     const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
 
-    const u32x w0 = w0l | w0r;
+    const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le (dst, w0, salt_len);
+    u32x w0[4];
+    u32x w1[4];
+    u32x w2[4];
+    u32x w3[4];
+
+    w0[0] = w0lr;
+    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];
+
+    switch_buffer_by_offset_le (w0, w1, w2, w3, salt_len);
+
+    u32x dst[16];
+
+    dst[ 0] = w0[0] | salt_buf0[0];
+    dst[ 1] = w0[1] | salt_buf0[1];
+    dst[ 2] = w0[2] | salt_buf0[2];
+    dst[ 3] = w0[3] | salt_buf0[3];
+    dst[ 4] = w1[0] | salt_buf1[0];
+    dst[ 5] = w1[1] | salt_buf1[1];
+    dst[ 6] = w1[2] | salt_buf1[2];
+    dst[ 7] = w1[3] | salt_buf1[3];
+    dst[ 8] = w2[0];
+    dst[ 9] = w2[1];
+    dst[10] = w2[2];
+    dst[11] = w2[3];
+    dst[12] = w3[0];
+    dst[13] = w3[1];
+    dst[14] = w3[2];
+    dst[15] = w3[3];
 
     /**
      * precompute key1 since key is static: 0x0123456789abcdef
@@ -709,62 +694,6 @@ void m03100s (__local u32 (*s_SPtrans)[64], __local u32 (*s_skb)[64], u32 w[16],
 
   const u32 salt_word_len = (salt_len + pw_len) * 2;
 
-  /**
-   * prepend salt
-   */
-
-  u32 w0_t[4];
-  u32 w1_t[4];
-  u32 w2_t[4];
-  u32 w3_t[4];
-
-  w0_t[0] = w[ 0];
-  w0_t[1] = w[ 1];
-  w0_t[2] = w[ 2];
-  w0_t[3] = w[ 3];
-  w1_t[0] = w[ 4];
-  w1_t[1] = w[ 5];
-  w1_t[2] = w[ 6];
-  w1_t[3] = w[ 7];
-  w2_t[0] = w[ 8];
-  w2_t[1] = w[ 9];
-  w2_t[2] = w[10];
-  w2_t[3] = w[11];
-  w3_t[0] = w[12];
-  w3_t[1] = w[13];
-  w3_t[2] = w[14];
-  w3_t[3] = w[15];
-
-  switch_buffer_by_offset_le_S (w0_t, w1_t, w2_t, w3_t, salt_len);
-
-  w0_t[0] |= salt_buf0[0];
-  w0_t[1] |= salt_buf0[1];
-  w0_t[2] |= salt_buf0[2];
-  w0_t[3] |= salt_buf0[3];
-  w1_t[0] |= salt_buf1[0];
-  w1_t[1] |= salt_buf1[1];
-  w1_t[2] |= salt_buf1[2];
-  w1_t[3] |= salt_buf1[3];
-
-  u32x dst[16];
-
-  dst[ 0] = w0_t[0];
-  dst[ 1] = w0_t[1];
-  dst[ 2] = w0_t[2];
-  dst[ 3] = w0_t[3];
-  dst[ 4] = w1_t[0];
-  dst[ 5] = w1_t[1];
-  dst[ 6] = w1_t[2];
-  dst[ 7] = w1_t[3];
-  dst[ 8] = w2_t[0];
-  dst[ 9] = w2_t[1];
-  dst[10] = w2_t[2];
-  dst[11] = w2_t[3];
-  dst[12] = w3_t[0];
-  dst[13] = w3_t[1];
-  dst[14] = w3_t[2];
-  dst[15] = w3_t[3];
-
   /**
    * digest
    */
@@ -787,9 +716,50 @@ void m03100s (__local u32 (*s_SPtrans)[64], __local u32 (*s_skb)[64], u32 w[16],
   {
     const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
 
-    const u32x w0 = w0l | w0r;
+    const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le (dst, w0, salt_len);
+    u32x w0[4];
+    u32x w1[4];
+    u32x w2[4];
+    u32x w3[4];
+
+    w0[0] = w0lr;
+    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];
+
+    switch_buffer_by_offset_le (w0, w1, w2, w3, salt_len);
+
+    u32x dst[16];
+
+    dst[ 0] = w0[0] | salt_buf0[0];
+    dst[ 1] = w0[1] | salt_buf0[1];
+    dst[ 2] = w0[2] | salt_buf0[2];
+    dst[ 3] = w0[3] | salt_buf0[3];
+    dst[ 4] = w1[0] | salt_buf1[0];
+    dst[ 5] = w1[1] | salt_buf1[1];
+    dst[ 6] = w1[2] | salt_buf1[2];
+    dst[ 7] = w1[3] | salt_buf1[3];
+    dst[ 8] = w2[0];
+    dst[ 9] = w2[1];
+    dst[10] = w2[2];
+    dst[11] = w2[3];
+    dst[12] = w3[0];
+    dst[13] = w3[1];
+    dst[14] = w3[2];
+    dst[15] = w3[3];
 
     /**
      * precompute key1 since key is static: 0x0123456789abcdef
diff --git a/OpenCL/m04010_a3-optimized.cl b/OpenCL/m04010_a3-optimized.cl
index 54c2faca8..ff7abfe05 100644
--- a/OpenCL/m04010_a3-optimized.cl
+++ b/OpenCL/m04010_a3-optimized.cl
@@ -65,85 +65,57 @@ void m04010m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 final_len   = salt_len + 32;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
-
     u32x w0_t[4];
     u32x w1_t[4];
     u32x w2_t[4];
     u32x w3_t[4];
 
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = t3[2];
-    w3_t[3] = t3[3];
+    w0_t[0] = w0lr;
+    w0_t[1] = w0[1];
+    w0_t[2] = w0[2];
+    w0_t[3] = w0[3];
+    w1_t[0] = w1[0];
+    w1_t[1] = w1[1];
+    w1_t[2] = w1[2];
+    w1_t[3] = w1[3];
+    w2_t[0] = w2[0];
+    w2_t[1] = w2[1];
+    w2_t[2] = w2[2];
+    w2_t[3] = w2[3];
+    w3_t[0] = w3[0];
+    w3_t[1] = w3[1];
+    w3_t[2] = w3[2];
+    w3_t[3] = w3[3];
+
+    switch_buffer_by_offset_le (w0_t, w1_t, w2_t, w3_t, salt_len);
+
+    w0_t[0] |= salt_buf0[0];
+    w0_t[1] |= salt_buf0[1];
+    w0_t[2] |= salt_buf0[2];
+    w0_t[3] |= salt_buf0[3];
+    w1_t[0] |= salt_buf1[0];
+    w1_t[1] |= salt_buf1[1];
+    w1_t[2] |= salt_buf1[2];
+    w1_t[3] |= salt_buf1[3];
+    w2_t[0] |= salt_buf2[0];
+    w2_t[1] |= salt_buf2[1];
+    w2_t[2] |= salt_buf2[2];
+    w2_t[3] |= salt_buf2[3];
+    w3_t[0] |= salt_buf3[0];
+    w3_t[1] |= salt_buf3[1];
+    w3_t[2]  = pw_salt_len * 8;
+    w3_t[3]  = 0;
 
     /**
      * md5
@@ -412,85 +384,57 @@ void m04010s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __gl
   const u32 final_len   = salt_len + 32;
 
   /**
-   * prepend salt
+   * loop
    */
 
   const u32 w0l = w0[0];
 
-  switch_buffer_by_offset_le_S (w0, w1, w2, w3, salt_len);
-
-  w0[0] |= salt_buf0[0];
-  w0[1] |= salt_buf0[1];
-  w0[2] |= salt_buf0[2];
-  w0[3] |= salt_buf0[3];
-  w1[0] |= salt_buf1[0];
-  w1[1] |= salt_buf1[1];
-  w1[2] |= salt_buf1[2];
-  w1[3] |= salt_buf1[3];
-  w2[0] |= salt_buf2[0];
-  w2[1] |= salt_buf2[1];
-  w2[2] |= salt_buf2[2];
-  w2[3] |= salt_buf2[3];
-  w3[0] |= salt_buf3[0];
-  w3[1] |= salt_buf3[1];
-  w3[2] |= salt_buf3[2];
-  w3[3] |= salt_buf3[3];
-
-  u32x t0[4];
-  u32x t1[4];
-  u32x t2[4];
-  u32x t3[4];
-
-  t0[0] = w0[0];
-  t0[1] = w0[1];
-  t0[2] = w0[2];
-  t0[3] = w0[3];
-  t1[0] = w1[0];
-  t1[1] = w1[1];
-  t1[2] = w1[2];
-  t1[3] = w1[3];
-  t2[0] = w2[0];
-  t2[1] = w2[1];
-  t2[2] = w2[2];
-  t2[3] = w2[3];
-  t3[0] = w3[0];
-  t3[1] = w3[1];
-  t3[2] = pw_salt_len * 8;
-  t3[3] = 0;
-
-  /**
-   * loop
-   */
-
   for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
     const u32x w0r = ix_create_bft (bfs_buf, il_pos);
 
     const u32x w0lr = w0l | w0r;
 
-    overwrite_at_le_4x4 (t0, t1, t2, t3, w0lr, salt_len);
-
     u32x w0_t[4];
     u32x w1_t[4];
     u32x w2_t[4];
     u32x w3_t[4];
 
-    w0_t[0] = t0[0];
-    w0_t[1] = t0[1];
-    w0_t[2] = t0[2];
-    w0_t[3] = t0[3];
-    w1_t[0] = t1[0];
-    w1_t[1] = t1[1];
-    w1_t[2] = t1[2];
-    w1_t[3] = t1[3];
-    w2_t[0] = t2[0];
-    w2_t[1] = t2[1];
-    w2_t[2] = t2[2];
-    w2_t[3] = t2[3];
-    w3_t[0] = t3[0];
-    w3_t[1] = t3[1];
-    w3_t[2] = t3[2];
-    w3_t[3] = t3[3];
+    w0_t[0] = w0lr;
+    w0_t[1] = w0[1];
+    w0_t[2] = w0[2];
+    w0_t[3] = w0[3];
+    w1_t[0] = w1[0];
+    w1_t[1] = w1[1];
+    w1_t[2] = w1[2];
+    w1_t[3] = w1[3];
+    w2_t[0] = w2[0];
+    w2_t[1] = w2[1];
+    w2_t[2] = w2[2];
+    w2_t[3] = w2[3];
+    w3_t[0] = w3[0];
+    w3_t[1] = w3[1];
+    w3_t[2] = w3[2];
+    w3_t[3] = w3[3];
+
+    switch_buffer_by_offset_le (w0_t, w1_t, w2_t, w3_t, salt_len);
+
+    w0_t[0] |= salt_buf0[0];
+    w0_t[1] |= salt_buf0[1];
+    w0_t[2] |= salt_buf0[2];
+    w0_t[3] |= salt_buf0[3];
+    w1_t[0] |= salt_buf1[0];
+    w1_t[1] |= salt_buf1[1];
+    w1_t[2] |= salt_buf1[2];
+    w1_t[3] |= salt_buf1[3];
+    w2_t[0] |= salt_buf2[0];
+    w2_t[1] |= salt_buf2[1];
+    w2_t[2] |= salt_buf2[2];
+    w2_t[3] |= salt_buf2[3];
+    w3_t[0] |= salt_buf3[0];
+    w3_t[1] |= salt_buf3[1];
+    w3_t[2]  = pw_salt_len * 8;
+    w3_t[3]  = 0;
 
     /**
      * md5