diff --git a/OpenCL/m06000_a1.cl b/OpenCL/m06000_a1.cl
index d7d612540..7109ce9eb 100644
--- a/OpenCL/m06000_a1.cl
+++ b/OpenCL/m06000_a1.cl
@@ -343,10 +343,6 @@ __kernel void m06000_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
     w[14] = pw_len * 8;
     w[15] = 0;
 
-    /**
-     * RipeMD160
-     */
-
     u32x dgst[5];
 
     dgst[0] = RIPEMD160M_A;
@@ -507,10 +503,6 @@ __kernel void m06000_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
     w[14] = pw_len * 8;
     w[15] = 0;
 
-    /**
-     * RipeMD160
-     */
-
     u32x dgst[5];
 
     dgst[0] = RIPEMD160M_A;
diff --git a/OpenCL/m08600_a0.cl b/OpenCL/m08600_a0.cl
index 90ca64c7e..3b5ab9aa1 100644
--- a/OpenCL/m08600_a0.cl
+++ b/OpenCL/m08600_a0.cl
@@ -276,14 +276,12 @@ __kernel void m08600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
    */
 
   u32 pw_buf0[4];
+  u32 pw_buf1[4];
 
   pw_buf0[0] = pws[gid].i[ 0];
   pw_buf0[1] = pws[gid].i[ 1];
   pw_buf0[2] = pws[gid].i[ 2];
   pw_buf0[3] = pws[gid].i[ 3];
-
-  u32 pw_buf1[4];
-
   pw_buf1[0] = pws[gid].i[ 4];
   pw_buf1[1] = pws[gid].i[ 5];
   pw_buf1[2] = pws[gid].i[ 6];
@@ -304,6 +302,10 @@ __kernel void m08600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
     const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1);
 
+    /**
+     * domino
+     */
+
     u32x w[16];
 
     w[ 0] = w0[0];
@@ -395,14 +397,12 @@ __kernel void m08600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
    */
 
   u32 pw_buf0[4];
+  u32 pw_buf1[4];
 
   pw_buf0[0] = pws[gid].i[ 0];
   pw_buf0[1] = pws[gid].i[ 1];
   pw_buf0[2] = pws[gid].i[ 2];
   pw_buf0[3] = pws[gid].i[ 3];
-
-  u32 pw_buf1[4];
-
   pw_buf1[0] = pws[gid].i[ 4];
   pw_buf1[1] = pws[gid].i[ 5];
   pw_buf1[2] = pws[gid].i[ 6];
@@ -435,6 +435,10 @@ __kernel void m08600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
     const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1);
 
+    /**
+     * domino
+     */
+
     u32x w[16];
 
     w[ 0] = w0[0];
diff --git a/OpenCL/m08600_a1.cl b/OpenCL/m08600_a1.cl
index d29159c23..b72034896 100644
--- a/OpenCL/m08600_a1.cl
+++ b/OpenCL/m08600_a1.cl
@@ -7,6 +7,9 @@
 
 #define _LOTUS5_
 
+//incompatible
+//#define NEW_SIMD_CODE
+
 #include "include/constants.h"
 #include "include/kernel_vendor.h"
 
@@ -18,9 +21,7 @@
 #include "include/kernel_functions.c"
 #include "OpenCL/types_ocl.c"
 #include "OpenCL/common.c"
-
-#define COMPARE_S "OpenCL/check_single_comp4.c"
-#define COMPARE_M "OpenCL/check_multi_comp4.c"
+#include "OpenCL/simd.c"
 
 __constant u32 lotus_magic_table[256] =
 {
@@ -58,11 +59,21 @@ __constant u32 lotus_magic_table[256] =
   0x29, 0x39, 0xb9, 0xe9, 0x4c, 0xff, 0x43, 0xab,
 };
 
-#define BOX(S,i) (S)[(i)]
-
-static void lotus_mix (u32 *in, __local u32 *s_lotus_magic_table)
+#if   VECT_SIZE == 1
+#define BOX1(S,i) (S)[(i)]
+#elif VECT_SIZE == 2
+#define BOX1(S,i) (u32x) ((S)[(i).s0], (S)[(i).s1])
+#elif VECT_SIZE == 4
+#define BOX1(S,i) (u32x) ((S)[(i).s0], (S)[(i).s1], (S)[(i).s2], (S)[(i).s3])
+#elif VECT_SIZE == 8
+#define BOX1(S,i) (u32x) ((S)[(i).s0], (S)[(i).s1], (S)[(i).s2], (S)[(i).s3], (S)[(i).s4], (S)[(i).s5], (S)[(i).s6], (S)[(i).s7])
+#elif VECT_SIZE == 16
+#define BOX1(S,i) (u32x) ((S)[(i).s0], (S)[(i).s1], (S)[(i).s2], (S)[(i).s3], (S)[(i).s4], (S)[(i).s5], (S)[(i).s6], (S)[(i).s7], (S)[(i).s8], (S)[(i).s9], (S)[(i).sa], (S)[(i).sb], (S)[(i).sc], (S)[(i).sd], (S)[(i).se], (S)[(i).sf])
+#endif
+
+static void lotus_mix (u32x *in, __local u32 *s_lotus_magic_table)
 {
-  u32 p = 0;
+  u32x p = 0;
 
   for (int i = 0; i < 18; i++)
   {
@@ -71,49 +82,49 @@ static void lotus_mix (u32 *in, __local u32 *s_lotus_magic_table)
     #pragma unroll 12
     for (int j = 0; j < 12; j++)
     {
-      u32 tmp_in = in[j];
-      u32 tmp_out = 0;
+      u32x tmp_in = in[j];
+      u32x tmp_out = 0;
 
-      p = (p + s--) & 0xff; p = ((tmp_in >>  0) & 0xff) ^ BOX (s_lotus_magic_table, p); tmp_out |= p <<  0;
-      p = (p + s--) & 0xff; p = ((tmp_in >>  8) & 0xff) ^ BOX (s_lotus_magic_table, p); tmp_out |= p <<  8;
-      p = (p + s--) & 0xff; p = ((tmp_in >> 16) & 0xff) ^ BOX (s_lotus_magic_table, p); tmp_out |= p << 16;
-      p = (p + s--) & 0xff; p = ((tmp_in >> 24) & 0xff) ^ BOX (s_lotus_magic_table, p); tmp_out |= p << 24;
+      p = (p + s--) & 0xff; p = ((tmp_in >>  0) & 0xff) ^ BOX1 (s_lotus_magic_table, p); tmp_out |= p <<  0;
+      p = (p + s--) & 0xff; p = ((tmp_in >>  8) & 0xff) ^ BOX1 (s_lotus_magic_table, p); tmp_out |= p <<  8;
+      p = (p + s--) & 0xff; p = ((tmp_in >> 16) & 0xff) ^ BOX1 (s_lotus_magic_table, p); tmp_out |= p << 16;
+      p = (p + s--) & 0xff; p = ((tmp_in >> 24) & 0xff) ^ BOX1 (s_lotus_magic_table, p); tmp_out |= p << 24;
 
       in[j] = tmp_out;
     }
   }
 }
 
-static void lotus_transform_password (u32 in[4], u32 out[4], __local u32 *s_lotus_magic_table)
+static void lotus_transform_password (u32x in[4], u32x out[4], __local u32 *s_lotus_magic_table)
 {
-  u32 t = out[3] >> 24;
+  u32x t = out[3] >> 24;
 
-  u32 c;
+  u32x c;
 
   #pragma unroll 4
   for (int i = 0; i < 4; i++)
   {
-    t ^= (in[i] >>  0) & 0xff; c = BOX (s_lotus_magic_table, t); out[i] ^= c <<  0; t = ((out[i] >>  0) & 0xff);
-    t ^= (in[i] >>  8) & 0xff; c = BOX (s_lotus_magic_table, t); out[i] ^= c <<  8; t = ((out[i] >>  8) & 0xff);
-    t ^= (in[i] >> 16) & 0xff; c = BOX (s_lotus_magic_table, t); out[i] ^= c << 16; t = ((out[i] >> 16) & 0xff);
-    t ^= (in[i] >> 24) & 0xff; c = BOX (s_lotus_magic_table, t); out[i] ^= c << 24; t = ((out[i] >> 24) & 0xff);
+    t ^= (in[i] >>  0) & 0xff; c = BOX1 (s_lotus_magic_table, t); out[i] ^= c <<  0; t = ((out[i] >>  0) & 0xff);
+    t ^= (in[i] >>  8) & 0xff; c = BOX1 (s_lotus_magic_table, t); out[i] ^= c <<  8; t = ((out[i] >>  8) & 0xff);
+    t ^= (in[i] >> 16) & 0xff; c = BOX1 (s_lotus_magic_table, t); out[i] ^= c << 16; t = ((out[i] >> 16) & 0xff);
+    t ^= (in[i] >> 24) & 0xff; c = BOX1 (s_lotus_magic_table, t); out[i] ^= c << 24; t = ((out[i] >> 24) & 0xff);
   }
 }
 
-static void pad (u32 w[4], const u32 len)
+static void pad (u32x w[4], const u32 len)
 {
   const u32 val = 16 - len;
 
-  const u32 mask1 = val << 24;
+  const u32x mask1 = val << 24;
 
-  const u32 mask2 = val << 16
+  const u32x mask2 = val << 16
                    | val << 24;
 
-  const u32 mask3 = val <<  8
+  const u32x mask3 = val <<  8
                    | val << 16
                    | val << 24;
 
-  const u32 mask4 = val <<  0
+  const u32x mask4 = val <<  0
                    | val <<  8
                    | val << 16
                    | val << 24;
@@ -179,9 +190,9 @@ static void pad (u32 w[4], const u32 len)
   }
 }
 
-static void mdtransform_norecalc (u32 state[4], u32 block[4], __local u32 *s_lotus_magic_table)
+static void mdtransform_norecalc (u32x state[4], u32x block[4], __local u32 *s_lotus_magic_table)
 {
-  u32 x[12];
+  u32x x[12];
 
   x[ 0] = state[0];
   x[ 1] = state[1];
@@ -204,23 +215,23 @@ static void mdtransform_norecalc (u32 state[4], u32 block[4], __local u32 *s_lot
   state[3] = x[3];
 }
 
-static void mdtransform (u32 state[4], u32 checksum[4], u32 block[4], __local u32 *s_lotus_magic_table)
+static void mdtransform (u32x state[4], u32x checksum[4], u32x block[4], __local u32 *s_lotus_magic_table)
 {
   mdtransform_norecalc (state, block, s_lotus_magic_table);
 
   lotus_transform_password (block, checksum, s_lotus_magic_table);
 }
 
-static void domino_big_md (const u32 saved_key[16], const u32 size, u32 state[4], __local u32 *s_lotus_magic_table)
+static void domino_big_md (const u32x saved_key[16], const u32x size, u32x state[4], __local u32 *s_lotus_magic_table)
 {
-  u32 checksum[4];
+  u32x checksum[4];
 
   checksum[0] = 0;
   checksum[1] = 0;
   checksum[2] = 0;
   checksum[3] = 0;
 
-  u32 block[4];
+  u32x block[4];
 
   block[0] = saved_key[0];
   block[1] = saved_key[1];
@@ -261,104 +272,107 @@ __kernel void m08600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
    * base
    */
 
-  u32 wordl0[4];
-
-  wordl0[0] = pws[gid].i[ 0];
-  wordl0[1] = pws[gid].i[ 1];
-  wordl0[2] = pws[gid].i[ 2];
-  wordl0[3] = pws[gid].i[ 3];
-
-  u32 wordl1[4];
-
-  wordl1[0] = pws[gid].i[ 4];
-  wordl1[1] = pws[gid].i[ 5];
-  wordl1[2] = pws[gid].i[ 6];
-  wordl1[3] = pws[gid].i[ 7];
-
-  u32 wordl2[4];
+  u32 pw_buf0[4];
+  u32 pw_buf1[4];
 
-  wordl2[0] = 0;
-  wordl2[1] = 0;
-  wordl2[2] = 0;
-  wordl2[3] = 0;
-
-  u32 wordl3[4];
-
-  wordl3[0] = 0;
-  wordl3[1] = 0;
-  wordl3[2] = 0;
-  wordl3[3] = 0;
+  pw_buf0[0] = pws[gid].i[0];
+  pw_buf0[1] = pws[gid].i[1];
+  pw_buf0[2] = pws[gid].i[2];
+  pw_buf0[3] = pws[gid].i[3];
+  pw_buf1[0] = pws[gid].i[4];
+  pw_buf1[1] = pws[gid].i[5];
+  pw_buf1[2] = pws[gid].i[6];
+  pw_buf1[3] = pws[gid].i[7];
 
   const u32 pw_l_len = pws[gid].pw_len;
 
-  if (combs_mode == COMBINATOR_MODE_BASE_RIGHT)
-  {
-    switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len);
-  }
-
   /**
    * loop
    */
 
-  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
-    const u32 pw_r_len = combs_buf[il_pos].pw_len;
-
-    u32 pw_len = pw_l_len + pw_r_len;
-
-    u32 wordr0[4];
+    const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos);
 
-    wordr0[0] = combs_buf[il_pos].i[0];
-    wordr0[1] = combs_buf[il_pos].i[1];
-    wordr0[2] = combs_buf[il_pos].i[2];
-    wordr0[3] = combs_buf[il_pos].i[3];
+    const u32x pw_len = pw_l_len + pw_r_len;
 
-    u32 wordr1[4];
-
-    wordr1[0] = combs_buf[il_pos].i[4];
-    wordr1[1] = combs_buf[il_pos].i[5];
-    wordr1[2] = combs_buf[il_pos].i[6];
-    wordr1[3] = combs_buf[il_pos].i[7];
-
-    u32 wordr2[4];
-
-    wordr2[0] = 0;
-    wordr2[1] = 0;
-    wordr2[2] = 0;
-    wordr2[3] = 0;
-
-    u32 wordr3[4];
+    /**
+     * concat password candidate
+     */
 
-    wordr3[0] = 0;
-    wordr3[1] = 0;
-    wordr3[2] = 0;
-    wordr3[3] = 0;
+    u32x wordl0[4] = { 0 };
+    u32x wordl1[4] = { 0 };
+    u32x wordl2[4] = { 0 };
+    u32x wordl3[4] = { 0 };
+
+    wordl0[0] = pw_buf0[0];
+    wordl0[1] = pw_buf0[1];
+    wordl0[2] = pw_buf0[2];
+    wordl0[3] = pw_buf0[3];
+    wordl1[0] = pw_buf1[0];
+    wordl1[1] = pw_buf1[1];
+    wordl1[2] = pw_buf1[2];
+    wordl1[3] = pw_buf1[3];
+
+    u32x wordr0[4] = { 0 };
+    u32x wordr1[4] = { 0 };
+    u32x wordr2[4] = { 0 };
+    u32x wordr3[4] = { 0 };
+
+    wordr0[0] = ix_create_combt (combs_buf, il_pos, 0);
+    wordr0[1] = ix_create_combt (combs_buf, il_pos, 1);
+    wordr0[2] = ix_create_combt (combs_buf, il_pos, 2);
+    wordr0[3] = ix_create_combt (combs_buf, il_pos, 3);
+    wordr1[0] = ix_create_combt (combs_buf, il_pos, 4);
+    wordr1[1] = ix_create_combt (combs_buf, il_pos, 5);
+    wordr1[2] = ix_create_combt (combs_buf, il_pos, 6);
+    wordr1[3] = ix_create_combt (combs_buf, il_pos, 7);
 
     if (combs_mode == COMBINATOR_MODE_BASE_LEFT)
     {
-      switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len);
+      switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len);
     }
+    else
+    {
+      switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len);
+    }
+
+    u32x w0[4];
+    u32x w1[4];
+
+    w0[0] = wordl0[0] | wordr0[0];
+    w0[1] = wordl0[1] | wordr0[1];
+    w0[2] = wordl0[2] | wordr0[2];
+    w0[3] = wordl0[3] | wordr0[3];
+    w1[0] = wordl1[0] | wordr1[0];
+    w1[1] = wordl1[1] | wordr1[1];
+    w1[2] = wordl1[2] | wordr1[2];
+    w1[3] = wordl1[3] | wordr1[3];
+
+    /**
+     * domino
+     */
 
-    u32 w[16];
-
-    w[ 0] = wordl0[0] | wordr0[0];
-    w[ 1] = wordl0[1] | wordr0[1];
-    w[ 2] = wordl0[2] | wordr0[2];
-    w[ 3] = wordl0[3] | wordr0[3];
-    w[ 4] = wordl1[0] | wordr1[0];
-    w[ 5] = wordl1[1] | wordr1[1];
-    w[ 6] = wordl1[2] | wordr1[2];
-    w[ 7] = wordl1[3] | wordr1[3];
-    w[ 8] = wordl2[0] | wordr2[0];
-    w[ 9] = wordl2[1] | wordr2[1];
-    w[10] = wordl2[2] | wordr2[2];
-    w[11] = wordl2[3] | wordr2[3];
-    w[12] = wordl3[0] | wordr3[0];
-    w[13] = wordl3[1] | wordr3[1];
-    w[14] = wordl3[2] | wordr3[2];
-    w[15] = wordl3[3] | wordr3[3];
-
-    u32 state[4];
+    u32x w[16];
+
+    w[ 0] = w0[0];
+    w[ 1] = w0[1];
+    w[ 2] = w0[2];
+    w[ 3] = w0[3];
+    w[ 4] = w1[0];
+    w[ 5] = w1[1];
+    w[ 6] = w1[2];
+    w[ 7] = w1[3];
+    w[ 8] = 0;
+    w[ 9] = 0;
+    w[10] = 0;
+    w[11] = 0;
+    w[12] = 0;
+    w[13] = 0;
+    w[14] = 0;
+    w[15] = 0;
+
+    u32x state[4];
 
     state[0] = 0;
     state[1] = 0;
@@ -388,12 +402,7 @@ __kernel void m08600_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
     domino_big_md (w, pw_len, state, s_lotus_magic_table);
 
-    const u32 r0 = state[0];
-    const u32 r1 = state[1];
-    const u32 r2 = state[2];
-    const u32 r3 = state[3];
-
-    #include COMPARE_M
+    COMPARE_M_SIMD (state[0], state[1], state[2], state[3]);
   }
 }
 
@@ -434,41 +443,20 @@ __kernel void m08600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
    * base
    */
 
-  u32 wordl0[4];
-
-  wordl0[0] = pws[gid].i[ 0];
-  wordl0[1] = pws[gid].i[ 1];
-  wordl0[2] = pws[gid].i[ 2];
-  wordl0[3] = pws[gid].i[ 3];
-
-  u32 wordl1[4];
-
-  wordl1[0] = pws[gid].i[ 4];
-  wordl1[1] = pws[gid].i[ 5];
-  wordl1[2] = pws[gid].i[ 6];
-  wordl1[3] = pws[gid].i[ 7];
-
-  u32 wordl2[4];
-
-  wordl2[0] = 0;
-  wordl2[1] = 0;
-  wordl2[2] = 0;
-  wordl2[3] = 0;
-
-  u32 wordl3[4];
+  u32 pw_buf0[4];
+  u32 pw_buf1[4];
 
-  wordl3[0] = 0;
-  wordl3[1] = 0;
-  wordl3[2] = 0;
-  wordl3[3] = 0;
+  pw_buf0[0] = pws[gid].i[0];
+  pw_buf0[1] = pws[gid].i[1];
+  pw_buf0[2] = pws[gid].i[2];
+  pw_buf0[3] = pws[gid].i[3];
+  pw_buf1[0] = pws[gid].i[4];
+  pw_buf1[1] = pws[gid].i[5];
+  pw_buf1[2] = pws[gid].i[6];
+  pw_buf1[3] = pws[gid].i[7];
 
   const u32 pw_l_len = pws[gid].pw_len;
 
-  if (combs_mode == COMBINATOR_MODE_BASE_RIGHT)
-  {
-    switch_buffer_by_offset_le (wordl0, wordl1, wordl2, wordl3, combs_buf[0].pw_len);
-  }
-
   /**
    * digest
    */
@@ -485,65 +473,89 @@ __kernel void m08600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
    * loop
    */
 
-  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
   {
-    const u32 pw_r_len = combs_buf[il_pos].pw_len;
-
-    u32 pw_len = pw_l_len + pw_r_len;
-
-    u32 wordr0[4];
-
-    wordr0[0] = combs_buf[il_pos].i[0];
-    wordr0[1] = combs_buf[il_pos].i[1];
-    wordr0[2] = combs_buf[il_pos].i[2];
-    wordr0[3] = combs_buf[il_pos].i[3];
+    const u32x pw_r_len = pwlenx_create_combt (combs_buf, il_pos);
 
-    u32 wordr1[4];
+    const u32x pw_len = pw_l_len + pw_r_len;
 
-    wordr1[0] = combs_buf[il_pos].i[4];
-    wordr1[1] = combs_buf[il_pos].i[5];
-    wordr1[2] = combs_buf[il_pos].i[6];
-    wordr1[3] = combs_buf[il_pos].i[7];
-
-    u32 wordr2[4];
-
-    wordr2[0] = 0;
-    wordr2[1] = 0;
-    wordr2[2] = 0;
-    wordr2[3] = 0;
-
-    u32 wordr3[4];
+    /**
+     * concat password candidate
+     */
 
-    wordr3[0] = 0;
-    wordr3[1] = 0;
-    wordr3[2] = 0;
-    wordr3[3] = 0;
+    u32x wordl0[4] = { 0 };
+    u32x wordl1[4] = { 0 };
+    u32x wordl2[4] = { 0 };
+    u32x wordl3[4] = { 0 };
+
+    wordl0[0] = pw_buf0[0];
+    wordl0[1] = pw_buf0[1];
+    wordl0[2] = pw_buf0[2];
+    wordl0[3] = pw_buf0[3];
+    wordl1[0] = pw_buf1[0];
+    wordl1[1] = pw_buf1[1];
+    wordl1[2] = pw_buf1[2];
+    wordl1[3] = pw_buf1[3];
+
+    u32x wordr0[4] = { 0 };
+    u32x wordr1[4] = { 0 };
+    u32x wordr2[4] = { 0 };
+    u32x wordr3[4] = { 0 };
+
+    wordr0[0] = ix_create_combt (combs_buf, il_pos, 0);
+    wordr0[1] = ix_create_combt (combs_buf, il_pos, 1);
+    wordr0[2] = ix_create_combt (combs_buf, il_pos, 2);
+    wordr0[3] = ix_create_combt (combs_buf, il_pos, 3);
+    wordr1[0] = ix_create_combt (combs_buf, il_pos, 4);
+    wordr1[1] = ix_create_combt (combs_buf, il_pos, 5);
+    wordr1[2] = ix_create_combt (combs_buf, il_pos, 6);
+    wordr1[3] = ix_create_combt (combs_buf, il_pos, 7);
 
     if (combs_mode == COMBINATOR_MODE_BASE_LEFT)
     {
-      switch_buffer_by_offset_le (wordr0, wordr1, wordr2, wordr3, pw_l_len);
+      switch_buffer_by_offset_le_VV (wordr0, wordr1, wordr2, wordr3, pw_l_len);
+    }
+    else
+    {
+      switch_buffer_by_offset_le_VV (wordl0, wordl1, wordl2, wordl3, pw_r_len);
     }
 
-    u32 w[16];
-
-    w[ 0] = wordl0[0] | wordr0[0];
-    w[ 1] = wordl0[1] | wordr0[1];
-    w[ 2] = wordl0[2] | wordr0[2];
-    w[ 3] = wordl0[3] | wordr0[3];
-    w[ 4] = wordl1[0] | wordr1[0];
-    w[ 5] = wordl1[1] | wordr1[1];
-    w[ 6] = wordl1[2] | wordr1[2];
-    w[ 7] = wordl1[3] | wordr1[3];
-    w[ 8] = wordl2[0] | wordr2[0];
-    w[ 9] = wordl2[1] | wordr2[1];
-    w[10] = wordl2[2] | wordr2[2];
-    w[11] = wordl2[3] | wordr2[3];
-    w[12] = wordl3[0] | wordr3[0];
-    w[13] = wordl3[1] | wordr3[1];
-    w[14] = wordl3[2] | wordr3[2];
-    w[15] = wordl3[3] | wordr3[3];
-
-    u32 state[4];
+    u32x w0[4];
+    u32x w1[4];
+
+    w0[0] = wordl0[0] | wordr0[0];
+    w0[1] = wordl0[1] | wordr0[1];
+    w0[2] = wordl0[2] | wordr0[2];
+    w0[3] = wordl0[3] | wordr0[3];
+    w1[0] = wordl1[0] | wordr1[0];
+    w1[1] = wordl1[1] | wordr1[1];
+    w1[2] = wordl1[2] | wordr1[2];
+    w1[3] = wordl1[3] | wordr1[3];
+
+    /**
+     * domino
+     */
+
+    u32x w[16];
+
+    w[ 0] = w0[0];
+    w[ 1] = w0[1];
+    w[ 2] = w0[2];
+    w[ 3] = w0[3];
+    w[ 4] = w1[0];
+    w[ 5] = w1[1];
+    w[ 6] = w1[2];
+    w[ 7] = w1[3];
+    w[ 8] = 0;
+    w[ 9] = 0;
+    w[10] = 0;
+    w[11] = 0;
+    w[12] = 0;
+    w[13] = 0;
+    w[14] = 0;
+    w[15] = 0;
+
+    u32x state[4];
 
     state[0] = 0;
     state[1] = 0;
@@ -573,12 +585,7 @@ __kernel void m08600_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
     domino_big_md (w, pw_len, state, s_lotus_magic_table);
 
-    const u32 r0 = state[0];
-    const u32 r1 = state[1];
-    const u32 r2 = state[2];
-    const u32 r3 = state[3];
-
-    #include COMPARE_S
+    COMPARE_S_SIMD (state[0], state[1], state[2], state[3]);
   }
 }
 
diff --git a/OpenCL/m08600_a3.cl b/OpenCL/m08600_a3.cl
index 6a307a478..4ee8d400a 100644
--- a/OpenCL/m08600_a3.cl
+++ b/OpenCL/m08600_a3.cl
@@ -253,7 +253,7 @@ static void m08600m (__local u32 *s_lotus_magic_table, u32 w[16], const u32 pw_l
   const u32 lid = get_local_id (0);
 
   /**
-   * padding
+   * base
    */
 
   if (pw_len < 16)
@@ -283,26 +283,26 @@ static void m08600m (__local u32 *s_lotus_magic_table, u32 w[16], const u32 pw_l
   {
     const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
 
-    const u32x w0 = w0l | w0r;
-
-    u32x w_tmp[16];
-
-    w_tmp[ 0] = w0;
-    w_tmp[ 1] = w[ 1];
-    w_tmp[ 2] = w[ 2];
-    w_tmp[ 3] = w[ 3];
-    w_tmp[ 4] = w[ 4];
-    w_tmp[ 5] = w[ 5];
-    w_tmp[ 6] = w[ 6];
-    w_tmp[ 7] = w[ 7];
-    w_tmp[ 8] = w[ 8];
-    w_tmp[ 9] = w[ 9];
-    w_tmp[10] = w[10];
-    w_tmp[11] = w[11];
-    w_tmp[12] = w[12];
-    w_tmp[13] = w[13];
-    w_tmp[14] = w[14];
-    w_tmp[15] = w[15];
+    const u32x w0lr = w0l | w0r;
+
+    u32x w_t[16];
+
+    w_t[ 0] = w0lr;
+    w_t[ 1] = w[ 1];
+    w_t[ 2] = w[ 2];
+    w_t[ 3] = w[ 3];
+    w_t[ 4] = w[ 4];
+    w_t[ 5] = w[ 5];
+    w_t[ 6] = w[ 6];
+    w_t[ 7] = w[ 7];
+    w_t[ 8] = w[ 8];
+    w_t[ 9] = w[ 9];
+    w_t[10] = w[10];
+    w_t[11] = w[11];
+    w_t[12] = w[12];
+    w_t[13] = w[13];
+    w_t[14] = w[14];
+    w_t[15] = w[15];
 
     u32x state[4];
 
@@ -311,7 +311,7 @@ static void m08600m (__local u32 *s_lotus_magic_table, u32 w[16], const u32 pw_l
     state[2] = 0;
     state[3] = 0;
 
-    domino_big_md (w_tmp, pw_len, state, s_lotus_magic_table);
+    domino_big_md (w_t, pw_len, state, s_lotus_magic_table);
 
     COMPARE_M_SIMD (state[0], state[1], state[2], state[3]);
   }
@@ -369,26 +369,26 @@ static void m08600s (__local u32 *s_lotus_magic_table, u32 w[16], const u32 pw_l
   {
     const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
 
-    const u32x w0 = w0l | w0r;
-
-    u32x w_tmp[16];
-
-    w_tmp[ 0] = w0;
-    w_tmp[ 1] = w[ 1];
-    w_tmp[ 2] = w[ 2];
-    w_tmp[ 3] = w[ 3];
-    w_tmp[ 4] = w[ 4];
-    w_tmp[ 5] = w[ 5];
-    w_tmp[ 6] = w[ 6];
-    w_tmp[ 7] = w[ 7];
-    w_tmp[ 8] = w[ 8];
-    w_tmp[ 9] = w[ 9];
-    w_tmp[10] = w[10];
-    w_tmp[11] = w[11];
-    w_tmp[12] = w[12];
-    w_tmp[13] = w[13];
-    w_tmp[14] = w[14];
-    w_tmp[15] = w[15];
+    const u32x w0lr = w0l | w0r;
+
+    u32x w_t[16];
+
+    w_t[ 0] = w0lr;
+    w_t[ 1] = w[ 1];
+    w_t[ 2] = w[ 2];
+    w_t[ 3] = w[ 3];
+    w_t[ 4] = w[ 4];
+    w_t[ 5] = w[ 5];
+    w_t[ 6] = w[ 6];
+    w_t[ 7] = w[ 7];
+    w_t[ 8] = w[ 8];
+    w_t[ 9] = w[ 9];
+    w_t[10] = w[10];
+    w_t[11] = w[11];
+    w_t[12] = w[12];
+    w_t[13] = w[13];
+    w_t[14] = w[14];
+    w_t[15] = w[15];
 
     u32x state[4];
 
@@ -397,7 +397,7 @@ static void m08600s (__local u32 *s_lotus_magic_table, u32 w[16], const u32 pw_l
     state[2] = 0;
     state[3] = 0;
 
-    domino_big_md (w_tmp, pw_len, state, s_lotus_magic_table);
+    domino_big_md (w_t, pw_len, state, s_lotus_magic_table);
 
     COMPARE_S_SIMD (state[0], state[1], state[2], state[3]);
   }