Use packv() and unpackv() for slow hash SIMD handling in kernels

diff --git a/OpenCL/m00400.cl b/OpenCL/m00400.cl
index 1610b6a..d8edbcb 100644 (file)
--- a/OpenCL/m00400.cl
+++ b/OpenCL/m00400.cl
@@ -121,7 +121,7 @@ void md5_transform_S (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u
   digest[3] += d;
 }
 
-void md5_transform (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], u32x digest[4])
+void md5_transform_V (const u32x w0[4], const u32x w1[4], const u32x w2[4], const u32x w3[4], u32x digest[4])
 {
   u32x a = digest[0];
   u32x b = digest[1];
@@ -326,217 +326,31 @@ __kernel void m00400_loop (__global pw_t *pws, __global kernel_rule_t *rules_buf
 
   if (gid >= gid_max) return;
 
-  u32x w0[4] = { 0 };
-  u32x w1[4] = { 0 };
-  u32x w2[4] = { 0 };
-
-  u32x pw_len = 0;
-
-  u32x digest[4] = { 0 };
-
-  #if VECT_SIZE == 1
-
-  w0[0] = pws[gid].i[0];
-  w0[1] = pws[gid].i[1];
-  w0[2] = pws[gid].i[2];
-  w0[3] = pws[gid].i[3];
-  w1[0] = pws[gid].i[4];
-  w1[1] = pws[gid].i[5];
-  w1[2] = pws[gid].i[6];
-  w1[3] = pws[gid].i[7];
-  w2[0] = pws[gid].i[8];
-  w2[1] = pws[gid].i[9];
-
-  pw_len = pws[gid].pw_len;
-
-  digest[0] = tmps[gid].digest_buf[0];
-  digest[1] = tmps[gid].digest_buf[1];
-  digest[2] = tmps[gid].digest_buf[2];
-  digest[3] = tmps[gid].digest_buf[3];
-
-  #else
-
-  const u32 gidx = gid * VECT_SIZE;
-
-  #if VECT_SIZE >= 2
-
-  if ((gidx + 0) < gid_max)
-  {
-    w0[0].s0 = pws[gidx + 0].i[0];
-    w0[1].s0 = pws[gidx + 0].i[1];
-    w0[2].s0 = pws[gidx + 0].i[2];
-    w0[3].s0 = pws[gidx + 0].i[3];
-    w1[0].s0 = pws[gidx + 0].i[4];
-    w1[1].s0 = pws[gidx + 0].i[5];
-    w1[2].s0 = pws[gidx + 0].i[6];
-    w1[3].s0 = pws[gidx + 0].i[7];
-    w2[0].s0 = pws[gidx + 0].i[8];
-    w2[1].s0 = pws[gidx + 0].i[9];
-
-    pw_len.s0 = pws[gidx + 0].pw_len;
-
-    digest[0].s0 = tmps[gidx + 0].digest_buf[0];
-    digest[1].s0 = tmps[gidx + 0].digest_buf[1];
-    digest[2].s0 = tmps[gidx + 0].digest_buf[2];
-    digest[3].s0 = tmps[gidx + 0].digest_buf[3];
-  }
-
-  if ((gidx + 1) < gid_max)
-  {
-    w0[0].s1 = pws[gidx + 1].i[0];
-    w0[1].s1 = pws[gidx + 1].i[1];
-    w0[2].s1 = pws[gidx + 1].i[2];
-    w0[3].s1 = pws[gidx + 1].i[3];
-    w1[0].s1 = pws[gidx + 1].i[4];
-    w1[1].s1 = pws[gidx + 1].i[5];
-    w1[2].s1 = pws[gidx + 1].i[6];
-    w1[3].s1 = pws[gidx + 1].i[7];
-    w2[0].s1 = pws[gidx + 1].i[8];
-    w2[1].s1 = pws[gidx + 1].i[9];
-
-    pw_len.s1 = pws[gidx + 1].pw_len;
-
-    digest[0].s1 = tmps[gidx + 1].digest_buf[0];
-    digest[1].s1 = tmps[gidx + 1].digest_buf[1];
-    digest[2].s1 = tmps[gidx + 1].digest_buf[2];
-    digest[3].s1 = tmps[gidx + 1].digest_buf[3];
-  }
-
-  #endif
-
-  #if VECT_SIZE >= 4
-
-  if ((gidx + 2) < gid_max)
-  {
-    w0[0].s2 = pws[gidx + 2].i[0];
-    w0[1].s2 = pws[gidx + 2].i[1];
-    w0[2].s2 = pws[gidx + 2].i[2];
-    w0[3].s2 = pws[gidx + 2].i[3];
-    w1[0].s2 = pws[gidx + 2].i[4];
-    w1[1].s2 = pws[gidx + 2].i[5];
-    w1[2].s2 = pws[gidx + 2].i[6];
-    w1[3].s2 = pws[gidx + 2].i[7];
-    w2[0].s2 = pws[gidx + 2].i[8];
-    w2[1].s2 = pws[gidx + 2].i[9];
-
-    pw_len.s2 = pws[gidx + 2].pw_len;
-
-    digest[0].s2 = tmps[gidx + 2].digest_buf[0];
-    digest[1].s2 = tmps[gidx + 2].digest_buf[1];
-    digest[2].s2 = tmps[gidx + 2].digest_buf[2];
-    digest[3].s2 = tmps[gidx + 2].digest_buf[3];
-  }
-
-  if ((gidx + 3) < gid_max)
-  {
-    w0[0].s3 = pws[gidx + 3].i[0];
-    w0[1].s3 = pws[gidx + 3].i[1];
-    w0[2].s3 = pws[gidx + 3].i[2];
-    w0[3].s3 = pws[gidx + 3].i[3];
-    w1[0].s3 = pws[gidx + 3].i[4];
-    w1[1].s3 = pws[gidx + 3].i[5];
-    w1[2].s3 = pws[gidx + 3].i[6];
-    w1[3].s3 = pws[gidx + 3].i[7];
-    w2[0].s3 = pws[gidx + 3].i[8];
-    w2[1].s3 = pws[gidx + 3].i[9];
-
-    pw_len.s3 = pws[gidx + 3].pw_len;
-
-    digest[0].s3 = tmps[gidx + 3].digest_buf[0];
-    digest[1].s3 = tmps[gidx + 3].digest_buf[1];
-    digest[2].s3 = tmps[gidx + 3].digest_buf[2];
-    digest[3].s3 = tmps[gidx + 3].digest_buf[3];
-  }
-
-  #endif
-  #if VECT_SIZE >= 8
-
-  if ((gidx + 4) < gid_max)
-  {
-    w0[0].s4 = pws[gidx + 4].i[0];
-    w0[1].s4 = pws[gidx + 4].i[1];
-    w0[2].s4 = pws[gidx + 4].i[2];
-    w0[3].s4 = pws[gidx + 4].i[3];
-    w1[0].s4 = pws[gidx + 4].i[4];
-    w1[1].s4 = pws[gidx + 4].i[5];
-    w1[2].s4 = pws[gidx + 4].i[6];
-    w1[3].s4 = pws[gidx + 4].i[7];
-    w2[0].s4 = pws[gidx + 4].i[8];
-    w2[1].s4 = pws[gidx + 4].i[9];
-
-    pw_len.s4 = pws[gidx + 4].pw_len;
-
-    digest[0].s4 = tmps[gidx + 4].digest_buf[0];
-    digest[1].s4 = tmps[gidx + 4].digest_buf[1];
-    digest[2].s4 = tmps[gidx + 4].digest_buf[2];
-    digest[3].s4 = tmps[gidx + 4].digest_buf[3];
-  }
+  u32x w0[4];
+  u32x w1[4];
+  u32x w2[4];
+
+  w0[0] = packv (pws, i, gid, 0);
+  w0[1] = packv (pws, i, gid, 1);
+  w0[2] = packv (pws, i, gid, 2);
+  w0[3] = packv (pws, i, gid, 3);
+  w1[0] = packv (pws, i, gid, 4);
+  w1[1] = packv (pws, i, gid, 5);
+  w1[2] = packv (pws, i, gid, 6);
+  w1[3] = packv (pws, i, gid, 7);
+  w2[0] = packv (pws, i, gid, 8);
+  w2[1] = packv (pws, i, gid, 9);
+  w2[2] = 0;
+  w2[3] = 0;
 
-  if ((gidx + 5) < gid_max)
-  {
-    w0[0].s5 = pws[gidx + 5].i[0];
-    w0[1].s5 = pws[gidx + 5].i[1];
-    w0[2].s5 = pws[gidx + 5].i[2];
-    w0[3].s5 = pws[gidx + 5].i[3];
-    w1[0].s5 = pws[gidx + 5].i[4];
-    w1[1].s5 = pws[gidx + 5].i[5];
-    w1[2].s5 = pws[gidx + 5].i[6];
-    w1[3].s5 = pws[gidx + 5].i[7];
-    w2[0].s5 = pws[gidx + 5].i[8];
-    w2[1].s5 = pws[gidx + 5].i[9];
-
-    pw_len.s5 = pws[gidx + 5].pw_len;
-
-    digest[0].s5 = tmps[gidx + 5].digest_buf[0];
-    digest[1].s5 = tmps[gidx + 5].digest_buf[1];
-    digest[2].s5 = tmps[gidx + 5].digest_buf[2];
-    digest[3].s5 = tmps[gidx + 5].digest_buf[3];
-  }
+  u32x pw_len = packvf (pws, pw_len, gid);
 
-  if ((gidx + 6) < gid_max)
-  {
-    w0[0].s6 = pws[gidx + 6].i[0];
-    w0[1].s6 = pws[gidx + 6].i[1];
-    w0[2].s6 = pws[gidx + 6].i[2];
-    w0[3].s6 = pws[gidx + 6].i[3];
-    w1[0].s6 = pws[gidx + 6].i[4];
-    w1[1].s6 = pws[gidx + 6].i[5];
-    w1[2].s6 = pws[gidx + 6].i[6];
-    w1[3].s6 = pws[gidx + 6].i[7];
-    w2[0].s6 = pws[gidx + 6].i[8];
-    w2[1].s6 = pws[gidx + 6].i[9];
-
-    pw_len.s6 = pws[gidx + 6].pw_len;
-
-    digest[0].s6 = tmps[gidx + 6].digest_buf[0];
-    digest[1].s6 = tmps[gidx + 6].digest_buf[1];
-    digest[2].s6 = tmps[gidx + 6].digest_buf[2];
-    digest[3].s6 = tmps[gidx + 6].digest_buf[3];
-  }
-
-  if ((gidx + 7) < gid_max)
-  {
-    w0[0].s7 = pws[gidx + 7].i[0];
-    w0[1].s7 = pws[gidx + 7].i[1];
-    w0[2].s7 = pws[gidx + 7].i[2];
-    w0[3].s7 = pws[gidx + 7].i[3];
-    w1[0].s7 = pws[gidx + 7].i[4];
-    w1[1].s7 = pws[gidx + 7].i[5];
-    w1[2].s7 = pws[gidx + 7].i[6];
-    w1[3].s7 = pws[gidx + 7].i[7];
-    w2[0].s7 = pws[gidx + 7].i[8];
-    w2[1].s7 = pws[gidx + 7].i[9];
-
-    pw_len.s7 = pws[gidx + 7].pw_len;
-
-    digest[0].s7 = tmps[gidx + 7].digest_buf[0];
-    digest[1].s7 = tmps[gidx + 7].digest_buf[1];
-    digest[2].s7 = tmps[gidx + 7].digest_buf[2];
-    digest[3].s7 = tmps[gidx + 7].digest_buf[3];
-  }
+  u32x digest[4];
 
-  #endif
-  #endif
+  digest[0] = packv (tmps, digest_buf, gid, 0);
+  digest[1] = packv (tmps, digest_buf, gid, 1);
+  digest[2] = packv (tmps, digest_buf, gid, 2);
+  digest[3] = packv (tmps, digest_buf, gid, 3);
 
   /**
    * loop
@@ -584,94 +398,13 @@ __kernel void m00400_loop (__global pw_t *pws, __global kernel_rule_t *rules_buf
     digest[2] = MD5M_C;
     digest[3] = MD5M_D;
 
-    md5_transform (block0, block1, block2, block3, digest);
-  }
-
-  #if VECT_SIZE == 1
-
-  tmps[gid].digest_buf[0] = digest[0];
-  tmps[gid].digest_buf[1] = digest[1];
-  tmps[gid].digest_buf[2] = digest[2];
-  tmps[gid].digest_buf[3] = digest[3];
-
-  #else
-
-  #if VECT_SIZE >= 2
-
-  if ((gidx + 0) < gid_max)
-  {
-    tmps[gidx + 0].digest_buf[0] = digest[0].s0;
-    tmps[gidx + 0].digest_buf[1] = digest[1].s0;
-    tmps[gidx + 0].digest_buf[2] = digest[2].s0;
-    tmps[gidx + 0].digest_buf[3] = digest[3].s0;
-  }
-
-  if ((gidx + 1) < gid_max)
-  {
-    tmps[gidx + 1].digest_buf[0] = digest[0].s1;
-    tmps[gidx + 1].digest_buf[1] = digest[1].s1;
-    tmps[gidx + 1].digest_buf[2] = digest[2].s1;
-    tmps[gidx + 1].digest_buf[3] = digest[3].s1;
-  }
-
-  #endif
-
-  #if VECT_SIZE >= 4
-
-  if ((gidx + 2) < gid_max)
-  {
-    tmps[gidx + 2].digest_buf[0] = digest[0].s2;
-    tmps[gidx + 2].digest_buf[1] = digest[1].s2;
-    tmps[gidx + 2].digest_buf[2] = digest[2].s2;
-    tmps[gidx + 2].digest_buf[3] = digest[3].s2;
-  }
-
-  if ((gidx + 3) < gid_max)
-  {
-    tmps[gidx + 3].digest_buf[0] = digest[0].s3;
-    tmps[gidx + 3].digest_buf[1] = digest[1].s3;
-    tmps[gidx + 3].digest_buf[2] = digest[2].s3;
-    tmps[gidx + 3].digest_buf[3] = digest[3].s3;
-  }
-
-  #endif
-
-  #if VECT_SIZE >= 8
-
-  if ((gidx + 4) < gid_max)
-  {
-    tmps[gidx + 4].digest_buf[0] = digest[0].s4;
-    tmps[gidx + 4].digest_buf[1] = digest[1].s4;
-    tmps[gidx + 4].digest_buf[2] = digest[2].s4;
-    tmps[gidx + 4].digest_buf[3] = digest[3].s4;
-  }
-
-  if ((gidx + 5) < gid_max)
-  {
-    tmps[gidx + 5].digest_buf[0] = digest[0].s5;
-    tmps[gidx + 5].digest_buf[1] = digest[1].s5;
-    tmps[gidx + 5].digest_buf[2] = digest[2].s5;
-    tmps[gidx + 5].digest_buf[3] = digest[3].s5;
-  }
-
-  if ((gidx + 6) < gid_max)
-  {
-    tmps[gidx + 6].digest_buf[0] = digest[0].s6;
-    tmps[gidx + 6].digest_buf[1] = digest[1].s6;
-    tmps[gidx + 6].digest_buf[2] = digest[2].s6;
-    tmps[gidx + 6].digest_buf[3] = digest[3].s6;
-  }
-
-  if ((gidx + 7) < gid_max)
-  {
-    tmps[gidx + 7].digest_buf[0] = digest[0].s7;
-    tmps[gidx + 7].digest_buf[1] = digest[1].s7;
-    tmps[gidx + 7].digest_buf[2] = digest[2].s7;
-    tmps[gidx + 7].digest_buf[3] = digest[3].s7;
+    md5_transform_V (block0, block1, block2, block3, digest);
   }
 
-  #endif
-  #endif
+  unpackv (tmps, digest_buf, gid, 0, digest[0]);
+  unpackv (tmps, digest_buf, gid, 1, digest[1]);
+  unpackv (tmps, digest_buf, gid, 2, digest[2]);
+  unpackv (tmps, digest_buf, gid, 3, digest[3]);
 }
 
 __kernel void m00400_comp (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global phpass_tmp_t *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)

diff --git a/OpenCL/m02500.cl b/OpenCL/m02500.cl
index 7b62c69..d36993a 100644 (file)
--- a/OpenCL/m02500.cl
+++ b/OpenCL/m02500.cl
@@ -18,6 +18,7 @@
 #include "include/kernel_functions.c"
 #include "OpenCL/types_ocl.c"
 #include "OpenCL/common.c"
+#include "OpenCL/simd.c"
 
 #define COMPARE_S "OpenCL/check_single_comp4.c"
 #define COMPARE_M "OpenCL/check_multi_comp4.c"
@@ -750,312 +751,34 @@ __kernel void m02500_loop (__global pw_t *pws, __global kernel_rule_t *rules_buf
   u32x ipad[5];
   u32x opad[5];
 
-  #if VECT_SIZE == 1
+  ipad[0] = packv (tmps, ipad, gid, 0);
+  ipad[1] = packv (tmps, ipad, gid, 1);
+  ipad[2] = packv (tmps, ipad, gid, 2);
+  ipad[3] = packv (tmps, ipad, gid, 3);
+  ipad[4] = packv (tmps, ipad, gid, 4);
 
-  ipad[0] = tmps[gid].ipad[0];
-  ipad[1] = tmps[gid].ipad[1];
-  ipad[2] = tmps[gid].ipad[2];
-  ipad[3] = tmps[gid].ipad[3];
-  ipad[4] = tmps[gid].ipad[4];
-
-  opad[0] = tmps[gid].opad[0];
-  opad[1] = tmps[gid].opad[1];
-  opad[2] = tmps[gid].opad[2];
-  opad[3] = tmps[gid].opad[3];
-  opad[4] = tmps[gid].opad[4];
-
-  #else
-
-  const u32 gidx = gid * VECT_SIZE;
-
-  #if VECT_SIZE >= 2
-
-  if ((gidx + 0) < gid_max)
-  {
-    ipad[0].s0 = tmps[gidx + 0].ipad[0];
-    ipad[1].s0 = tmps[gidx + 0].ipad[1];
-    ipad[2].s0 = tmps[gidx + 0].ipad[2];
-    ipad[3].s0 = tmps[gidx + 0].ipad[3];
-    ipad[4].s0 = tmps[gidx + 0].ipad[4];
-
-    opad[0].s0 = tmps[gidx + 0].opad[0];
-    opad[1].s0 = tmps[gidx + 0].opad[1];
-    opad[2].s0 = tmps[gidx + 0].opad[2];
-    opad[3].s0 = tmps[gidx + 0].opad[3];
-    opad[4].s0 = tmps[gidx + 0].opad[4];
-  }
-
-  if ((gidx + 1) < gid_max)
-  {
-    ipad[0].s1 = tmps[gidx + 1].ipad[0];
-    ipad[1].s1 = tmps[gidx + 1].ipad[1];
-    ipad[2].s1 = tmps[gidx + 1].ipad[2];
-    ipad[3].s1 = tmps[gidx + 1].ipad[3];
-    ipad[4].s1 = tmps[gidx + 1].ipad[4];
-
-    opad[0].s1 = tmps[gidx + 1].opad[0];
-    opad[1].s1 = tmps[gidx + 1].opad[1];
-    opad[2].s1 = tmps[gidx + 1].opad[2];
-    opad[3].s1 = tmps[gidx + 1].opad[3];
-    opad[4].s1 = tmps[gidx + 1].opad[4];
-  }
-
-  #endif
-
-  #if VECT_SIZE >= 4
-
-  if ((gidx + 2) < gid_max)
-  {
-    ipad[0].s2 = tmps[gidx + 2].ipad[0];
-    ipad[1].s2 = tmps[gidx + 2].ipad[1];
-    ipad[2].s2 = tmps[gidx + 2].ipad[2];
-    ipad[3].s2 = tmps[gidx + 2].ipad[3];
-    ipad[4].s2 = tmps[gidx + 2].ipad[4];
-
-    opad[0].s2 = tmps[gidx + 2].opad[0];
-    opad[1].s2 = tmps[gidx + 2].opad[1];
-    opad[2].s2 = tmps[gidx + 2].opad[2];
-    opad[3].s2 = tmps[gidx + 2].opad[3];
-    opad[4].s2 = tmps[gidx + 2].opad[4];
-  }
-
-  if ((gidx + 3) < gid_max)
-  {
-    ipad[0].s3 = tmps[gidx + 3].ipad[0];
-    ipad[1].s3 = tmps[gidx + 3].ipad[1];
-    ipad[2].s3 = tmps[gidx + 3].ipad[2];
-    ipad[3].s3 = tmps[gidx + 3].ipad[3];
-    ipad[4].s3 = tmps[gidx + 3].ipad[4];
-
-    opad[0].s3 = tmps[gidx + 3].opad[0];
-    opad[1].s3 = tmps[gidx + 3].opad[1];
-    opad[2].s3 = tmps[gidx + 3].opad[2];
-    opad[3].s3 = tmps[gidx + 3].opad[3];
-    opad[4].s3 = tmps[gidx + 3].opad[4];
-  }
-
-  #endif
-
-  #if VECT_SIZE >= 8
-
-  if ((gidx + 4) < gid_max)
-  {
-    ipad[0].s4 = tmps[gidx + 4].ipad[0];
-    ipad[1].s4 = tmps[gidx + 4].ipad[1];
-    ipad[2].s4 = tmps[gidx + 4].ipad[2];
-    ipad[3].s4 = tmps[gidx + 4].ipad[3];
-    ipad[4].s4 = tmps[gidx + 4].ipad[4];
-
-    opad[0].s4 = tmps[gidx + 4].opad[0];
-    opad[1].s4 = tmps[gidx + 4].opad[1];
-    opad[2].s4 = tmps[gidx + 4].opad[2];
-    opad[3].s4 = tmps[gidx + 4].opad[3];
-    opad[4].s4 = tmps[gidx + 4].opad[4];
-  }
-
-  if ((gidx + 5) < gid_max)
-  {
-    ipad[0].s5 = tmps[gidx + 5].ipad[0];
-    ipad[1].s5 = tmps[gidx + 5].ipad[1];
-    ipad[2].s5 = tmps[gidx + 5].ipad[2];
-    ipad[3].s5 = tmps[gidx + 5].ipad[3];
-    ipad[4].s5 = tmps[gidx + 5].ipad[4];
-
-    opad[0].s5 = tmps[gidx + 5].opad[0];
-    opad[1].s5 = tmps[gidx + 5].opad[1];
-    opad[2].s5 = tmps[gidx + 5].opad[2];
-    opad[3].s5 = tmps[gidx + 5].opad[3];
-    opad[4].s5 = tmps[gidx + 5].opad[4];
-  }
-
-  if ((gidx + 6) < gid_max)
-  {
-    ipad[0].s6 = tmps[gidx + 6].ipad[0];
-    ipad[1].s6 = tmps[gidx + 6].ipad[1];
-    ipad[2].s6 = tmps[gidx + 6].ipad[2];
-    ipad[3].s6 = tmps[gidx + 6].ipad[3];
-    ipad[4].s6 = tmps[gidx + 6].ipad[4];
-
-    opad[0].s6 = tmps[gidx + 6].opad[0];
-    opad[1].s6 = tmps[gidx + 6].opad[1];
-    opad[2].s6 = tmps[gidx + 6].opad[2];
-    opad[3].s6 = tmps[gidx + 6].opad[3];
-    opad[4].s6 = tmps[gidx + 6].opad[4];
-  }
-
-  if ((gidx + 7) < gid_max)
-  {
-    ipad[0].s7 = tmps[gidx + 7].ipad[0];
-    ipad[1].s7 = tmps[gidx + 7].ipad[1];
-    ipad[2].s7 = tmps[gidx + 7].ipad[2];
-    ipad[3].s7 = tmps[gidx + 7].ipad[3];
-    ipad[4].s7 = tmps[gidx + 7].ipad[4];
-
-    opad[0].s7 = tmps[gidx + 7].opad[0];
-    opad[1].s7 = tmps[gidx + 7].opad[1];
-    opad[2].s7 = tmps[gidx + 7].opad[2];
-    opad[3].s7 = tmps[gidx + 7].opad[3];
-    opad[4].s7 = tmps[gidx + 7].opad[4];
-  }
-
-  #endif
-
-  #endif
+  opad[0] = packv (tmps, opad, gid, 0);
+  opad[1] = packv (tmps, opad, gid, 1);
+  opad[2] = packv (tmps, opad, gid, 2);
+  opad[3] = packv (tmps, opad, gid, 3);
+  opad[4] = packv (tmps, opad, gid, 4);
 
   for (u32 i = 0; i < 8; i += 5)
   {
     u32x dgst[5];
     u32x out[5];
 
-    #if VECT_SIZE == 1
-
-    dgst[0] = tmps[gid].dgst[i + 0];
-    dgst[1] = tmps[gid].dgst[i + 1];
-    dgst[2] = tmps[gid].dgst[i + 2];
-    dgst[3] = tmps[gid].dgst[i + 3];
-    dgst[4] = tmps[gid].dgst[i + 4];
-
-    out[0] = tmps[gid].out[i + 0];
-    out[1] = tmps[gid].out[i + 1];
-    out[2] = tmps[gid].out[i + 2];
-    out[3] = tmps[gid].out[i + 3];
-    out[4] = tmps[gid].out[i + 4];
-
-    #else
-
-    #if VECT_SIZE >= 2
-
-    if ((gidx + 0) < gid_max)
-    {
-      dgst[0].s0 = tmps[gidx + 0].dgst[i + 0];
-      dgst[1].s0 = tmps[gidx + 0].dgst[i + 1];
-      dgst[2].s0 = tmps[gidx + 0].dgst[i + 2];
-      dgst[3].s0 = tmps[gidx + 0].dgst[i + 3];
-      dgst[4].s0 = tmps[gidx + 0].dgst[i + 4];
-
-      out[0].s0 = tmps[gidx + 0].out[i + 0];
-      out[1].s0 = tmps[gidx + 0].out[i + 1];
-      out[2].s0 = tmps[gidx + 0].out[i + 2];
-      out[3].s0 = tmps[gidx + 0].out[i + 3];
-      out[4].s0 = tmps[gidx + 0].out[i + 4];
-    }
-
-    if ((gidx + 1) < gid_max)
-    {
-      dgst[0].s1 = tmps[gidx + 1].dgst[i + 0];
-      dgst[1].s1 = tmps[gidx + 1].dgst[i + 1];
-      dgst[2].s1 = tmps[gidx + 1].dgst[i + 2];
-      dgst[3].s1 = tmps[gidx + 1].dgst[i + 3];
-      dgst[4].s1 = tmps[gidx + 1].dgst[i + 4];
-
-      out[0].s1 = tmps[gidx + 1].out[i + 0];
-      out[1].s1 = tmps[gidx + 1].out[i + 1];
-      out[2].s1 = tmps[gidx + 1].out[i + 2];
-      out[3].s1 = tmps[gidx + 1].out[i + 3];
-      out[4].s1 = tmps[gidx + 1].out[i + 4];
-    }
-
-    #endif
-
-    #if VECT_SIZE >= 4
-
-    if ((gidx + 2) < gid_max)
-    {
-      dgst[0].s2 = tmps[gidx + 2].dgst[i + 0];
-      dgst[1].s2 = tmps[gidx + 2].dgst[i + 1];
-      dgst[2].s2 = tmps[gidx + 2].dgst[i + 2];
-      dgst[3].s2 = tmps[gidx + 2].dgst[i + 3];
-      dgst[4].s2 = tmps[gidx + 2].dgst[i + 4];
-
-      out[0].s2 = tmps[gidx + 2].out[i + 0];
-      out[1].s2 = tmps[gidx + 2].out[i + 1];
-      out[2].s2 = tmps[gidx + 2].out[i + 2];
-      out[3].s2 = tmps[gidx + 2].out[i + 3];
-      out[4].s2 = tmps[gidx + 2].out[i + 4];
-    }
-
-    if ((gidx + 3) < gid_max)
-    {
-      dgst[0].s3 = tmps[gidx + 3].dgst[i + 0];
-      dgst[1].s3 = tmps[gidx + 3].dgst[i + 1];
-      dgst[2].s3 = tmps[gidx + 3].dgst[i + 2];
-      dgst[3].s3 = tmps[gidx + 3].dgst[i + 3];
-      dgst[4].s3 = tmps[gidx + 3].dgst[i + 4];
-
-      out[0].s3 = tmps[gidx + 3].out[i + 0];
-      out[1].s3 = tmps[gidx + 3].out[i + 1];
-      out[2].s3 = tmps[gidx + 3].out[i + 2];
-      out[3].s3 = tmps[gidx + 3].out[i + 3];
-      out[4].s3 = tmps[gidx + 3].out[i + 4];
-    }
-
-    #endif
+    dgst[0] = packv (tmps, dgst, gid, 0);
+    dgst[1] = packv (tmps, dgst, gid, 1);
+    dgst[2] = packv (tmps, dgst, gid, 2);
+    dgst[3] = packv (tmps, dgst, gid, 3);
+    dgst[4] = packv (tmps, dgst, gid, 4);
 
-    #if VECT_SIZE >= 8
-
-    if ((gidx + 4) < gid_max)
-    {
-      dgst[0].s4 = tmps[gidx + 4].dgst[i + 0];
-      dgst[1].s4 = tmps[gidx + 4].dgst[i + 1];
-      dgst[2].s4 = tmps[gidx + 4].dgst[i + 2];
-      dgst[3].s4 = tmps[gidx + 4].dgst[i + 3];
-      dgst[4].s4 = tmps[gidx + 4].dgst[i + 4];
-
-      out[0].s4 = tmps[gidx + 4].out[i + 0];
-      out[1].s4 = tmps[gidx + 4].out[i + 1];
-      out[2].s4 = tmps[gidx + 4].out[i + 2];
-      out[3].s4 = tmps[gidx + 4].out[i + 3];
-      out[4].s4 = tmps[gidx + 4].out[i + 4];
-    }
-
-    if ((gidx + 5) < gid_max)
-    {
-      dgst[0].s5 = tmps[gidx + 5].dgst[i + 0];
-      dgst[1].s5 = tmps[gidx + 5].dgst[i + 1];
-      dgst[2].s5 = tmps[gidx + 5].dgst[i + 2];
-      dgst[3].s5 = tmps[gidx + 5].dgst[i + 3];
-      dgst[4].s5 = tmps[gidx + 5].dgst[i + 4];
-
-      out[0].s5 = tmps[gidx + 5].out[i + 0];
-      out[1].s5 = tmps[gidx + 5].out[i + 1];
-      out[2].s5 = tmps[gidx + 5].out[i + 2];
-      out[3].s5 = tmps[gidx + 5].out[i + 3];
-      out[4].s5 = tmps[gidx + 5].out[i + 4];
-    }
-
-    if ((gidx + 6) < gid_max)
-    {
-      dgst[0].s6 = tmps[gidx + 6].dgst[i + 0];
-      dgst[1].s6 = tmps[gidx + 6].dgst[i + 1];
-      dgst[2].s6 = tmps[gidx + 6].dgst[i + 2];
-      dgst[3].s6 = tmps[gidx + 6].dgst[i + 3];
-      dgst[4].s6 = tmps[gidx + 6].dgst[i + 4];
-
-      out[0].s6 = tmps[gidx + 6].out[i + 0];
-      out[1].s6 = tmps[gidx + 6].out[i + 1];
-      out[2].s6 = tmps[gidx + 6].out[i + 2];
-      out[3].s6 = tmps[gidx + 6].out[i + 3];
-      out[4].s6 = tmps[gidx + 6].out[i + 4];
-    }
-
-    if ((gidx + 7) < gid_max)
-    {
-      dgst[0].s7 = tmps[gidx + 7].dgst[i + 0];
-      dgst[1].s7 = tmps[gidx + 7].dgst[i + 1];
-      dgst[2].s7 = tmps[gidx + 7].dgst[i + 2];
-      dgst[3].s7 = tmps[gidx + 7].dgst[i + 3];
-      dgst[4].s7 = tmps[gidx + 7].dgst[i + 4];
-
-      out[0].s7 = tmps[gidx + 7].out[i + 0];
-      out[1].s7 = tmps[gidx + 7].out[i + 1];
-      out[2].s7 = tmps[gidx + 7].out[i + 2];
-      out[3].s7 = tmps[gidx + 7].out[i + 3];
-      out[4].s7 = tmps[gidx + 7].out[i + 4];
-    }
-
-    #endif
-
-    #endif
+    out[0] = packv (tmps, out, gid, 0);
+    out[1] = packv (tmps, out, gid, 1);
+    out[2] = packv (tmps, out, gid, 2);
+    out[3] = packv (tmps, out, gid, 3);
+    out[4] = packv (tmps, out, gid, 4);
 
     for (u32 j = 0; j < loop_cnt; j++)
     {
@@ -1090,155 +813,17 @@ __kernel void m02500_loop (__global pw_t *pws, __global kernel_rule_t *rules_buf
       out[4] ^= dgst[4];
     }
 
-    #if VECT_SIZE == 1
-
-    tmps[gid].dgst[i + 0] = dgst[0];
-    tmps[gid].dgst[i + 1] = dgst[1];
-    tmps[gid].dgst[i + 2] = dgst[2];
-    tmps[gid].dgst[i + 3] = dgst[3];
-    tmps[gid].dgst[i + 4] = dgst[4];
-
-    tmps[gid].out[i + 0] = out[0];
-    tmps[gid].out[i + 1] = out[1];
-    tmps[gid].out[i + 2] = out[2];
-    tmps[gid].out[i + 3] = out[3];
-    tmps[gid].out[i + 4] = out[4];
-
-    #else
-
-    #if VECT_SIZE >= 2
-
-    if ((gidx + 0) < gid_max)
-    {
-      tmps[gidx + 0].dgst[i + 0] = dgst[0].s0;
-      tmps[gidx + 0].dgst[i + 1] = dgst[1].s0;
-      tmps[gidx + 0].dgst[i + 2] = dgst[2].s0;
-      tmps[gidx + 0].dgst[i + 3] = dgst[3].s0;
-      tmps[gidx + 0].dgst[i + 4] = dgst[4].s0;
-
-      tmps[gidx + 0].out[i + 0] = out[0].s0;
-      tmps[gidx + 0].out[i + 1] = out[1].s0;
-      tmps[gidx + 0].out[i + 2] = out[2].s0;
-      tmps[gidx + 0].out[i + 3] = out[3].s0;
-      tmps[gidx + 0].out[i + 4] = out[4].s0;
-    }
-
-    if ((gidx + 1) < gid_max)
-    {
-      tmps[gidx + 1].dgst[i + 0] = dgst[0].s1;
-      tmps[gidx + 1].dgst[i + 1] = dgst[1].s1;
-      tmps[gidx + 1].dgst[i + 2] = dgst[2].s1;
-      tmps[gidx + 1].dgst[i + 3] = dgst[3].s1;
-      tmps[gidx + 1].dgst[i + 4] = dgst[4].s1;
-
-      tmps[gidx + 1].out[i + 0] = out[0].s1;
-      tmps[gidx + 1].out[i + 1] = out[1].s1;
-      tmps[gidx + 1].out[i + 2] = out[2].s1;
-      tmps[gidx + 1].out[i + 3] = out[3].s1;
-      tmps[gidx + 1].out[i + 4] = out[4].s1;
-    }
-
-    #endif
-
-    #if VECT_SIZE >= 4
-
-    if ((gidx + 2) < gid_max)
-    {
-      tmps[gidx + 2].dgst[i + 0] = dgst[0].s2;
-      tmps[gidx + 2].dgst[i + 1] = dgst[1].s2;
-      tmps[gidx + 2].dgst[i + 2] = dgst[2].s2;
-      tmps[gidx + 2].dgst[i + 3] = dgst[3].s2;
-      tmps[gidx + 2].dgst[i + 4] = dgst[4].s2;
-
-      tmps[gidx + 2].out[i + 0] = out[0].s2;
-      tmps[gidx + 2].out[i + 1] = out[1].s2;
-      tmps[gidx + 2].out[i + 2] = out[2].s2;
-      tmps[gidx + 2].out[i + 3] = out[3].s2;
-      tmps[gidx + 2].out[i + 4] = out[4].s2;
-    }
-
-    if ((gidx + 3) < gid_max)
-    {
-      tmps[gidx + 3].dgst[i + 0] = dgst[0].s3;
-      tmps[gidx + 3].dgst[i + 1] = dgst[1].s3;
-      tmps[gidx + 3].dgst[i + 2] = dgst[2].s3;
-      tmps[gidx + 3].dgst[i + 3] = dgst[3].s3;
-      tmps[gidx + 3].dgst[i + 4] = dgst[4].s3;
-
-      tmps[gidx + 3].out[i + 0] = out[0].s3;
-      tmps[gidx + 3].out[i + 1] = out[1].s3;
-      tmps[gidx + 3].out[i + 2] = out[2].s3;
-      tmps[gidx + 3].out[i + 3] = out[3].s3;
-      tmps[gidx + 3].out[i + 4] = out[4].s3;
-    }
-
-    #endif
-
-    #if VECT_SIZE >= 8
-
-    if ((gidx + 4) < gid_max)
-    {
-      tmps[gidx + 4].dgst[i + 0] = dgst[0].s4;
-      tmps[gidx + 4].dgst[i + 1] = dgst[1].s4;
-      tmps[gidx + 4].dgst[i + 2] = dgst[2].s4;
-      tmps[gidx + 4].dgst[i + 3] = dgst[3].s4;
-      tmps[gidx + 4].dgst[i + 4] = dgst[4].s4;
-
-      tmps[gidx + 4].out[i + 0] = out[0].s4;
-      tmps[gidx + 4].out[i + 1] = out[1].s4;
-      tmps[gidx + 4].out[i + 2] = out[2].s4;
-      tmps[gidx + 4].out[i + 3] = out[3].s4;
-      tmps[gidx + 4].out[i + 4] = out[4].s4;
-    }
-
-    if ((gidx + 5) < gid_max)
-    {
-      tmps[gidx + 5].dgst[i + 0] = dgst[0].s5;
-      tmps[gidx + 5].dgst[i + 1] = dgst[1].s5;
-      tmps[gidx + 5].dgst[i + 2] = dgst[2].s5;
-      tmps[gidx + 5].dgst[i + 3] = dgst[3].s5;
-      tmps[gidx + 5].dgst[i + 4] = dgst[4].s5;
-
-      tmps[gidx + 5].out[i + 0] = out[0].s5;
-      tmps[gidx + 5].out[i + 1] = out[1].s5;
-      tmps[gidx + 5].out[i + 2] = out[2].s5;
-      tmps[gidx + 5].out[i + 3] = out[3].s5;
-      tmps[gidx + 5].out[i + 4] = out[4].s5;
-    }
-
-    if ((gidx + 6) < gid_max)
-    {
-      tmps[gidx + 6].dgst[i + 0] = dgst[0].s6;
-      tmps[gidx + 6].dgst[i + 1] = dgst[1].s6;
-      tmps[gidx + 6].dgst[i + 2] = dgst[2].s6;
-      tmps[gidx + 6].dgst[i + 3] = dgst[3].s6;
-      tmps[gidx + 6].dgst[i + 4] = dgst[4].s6;
-
-      tmps[gidx + 6].out[i + 0] = out[0].s6;
-      tmps[gidx + 6].out[i + 1] = out[1].s6;
-      tmps[gidx + 6].out[i + 2] = out[2].s6;
-      tmps[gidx + 6].out[i + 3] = out[3].s6;
-      tmps[gidx + 6].out[i + 4] = out[4].s6;
-    }
-
-    if ((gidx + 7) < gid_max)
-    {
-      tmps[gidx + 7].dgst[i + 0] = dgst[0].s7;
-      tmps[gidx + 7].dgst[i + 1] = dgst[1].s7;
-      tmps[gidx + 7].dgst[i + 2] = dgst[2].s7;
-      tmps[gidx + 7].dgst[i + 3] = dgst[3].s7;
-      tmps[gidx + 7].dgst[i + 4] = dgst[4].s7;
-
-      tmps[gidx + 7].out[i + 0] = out[0].s7;
-      tmps[gidx + 7].out[i + 1] = out[1].s7;
-      tmps[gidx + 7].out[i + 2] = out[2].s7;
-      tmps[gidx + 7].out[i + 3] = out[3].s7;
-      tmps[gidx + 7].out[i + 4] = out[4].s7;
-    }
-
-    #endif
-
-    #endif
+    unpackv (tmps, dgst, gid, 0, dgst[0]);
+    unpackv (tmps, dgst, gid, 1, dgst[1]);
+    unpackv (tmps, dgst, gid, 2, dgst[2]);
+    unpackv (tmps, dgst, gid, 3, dgst[3]);
+    unpackv (tmps, dgst, gid, 4, dgst[4]);
+
+    unpackv (tmps, out, gid, 0, out[0]);
+    unpackv (tmps, out, gid, 1, out[1]);
+    unpackv (tmps, out, gid, 2, out[2]);
+    unpackv (tmps, out, gid, 3, out[3]);
+    unpackv (tmps, out, gid, 4, out[4]);
   }
 }
 

diff --git a/OpenCL/simd.c b/OpenCL/simd.c
index d4a8fab..865819d 100644 (file)
--- a/OpenCL/simd.c
+++ b/OpenCL/simd.c
@@ -1239,6 +1239,18 @@ inline u32x ix_create_combt (__global comb_t *combs_buf, const u32 il_pos, const
 #define packv(arr,var,gid,idx) (u32x) ((arr)[((gid) * 16) + 0].var[(idx)], (arr)[((gid) * 16) + 1].var[(idx)], (arr)[((gid) * 16) + 2].var[(idx)], (arr)[((gid) * 16) + 3].var[(idx)], (arr)[((gid) * 16) + 4].var[(idx)], (arr)[((gid) * 16) + 5].var[(idx)], (arr)[((gid) * 16) + 6].var[(idx)], (arr)[((gid) * 16) + 7].var[(idx)], (arr)[((gid) * 16) + 8].var[(idx)], (arr)[((gid) * 16) + 9].var[(idx)], (arr)[((gid) * 16) + 10].var[(idx)], (arr)[((gid) * 16) + 11].var[(idx)], (arr)[((gid) * 16) + 12].var[(idx)], (arr)[((gid) * 16) + 13].var[(idx)], (arr)[((gid) * 16) + 14].var[(idx)], (arr)[((gid) * 16) + 15].var[(idx)])
 #endif
 
+#if   VECT_SIZE == 1
+#define packvf(arr,var,gid) (u32x) ((arr)[((gid) *  1) + 0].var)
+#elif VECT_SIZE == 2
+#define packvf(arr,var,gid) (u32x) ((arr)[((gid) *  2) + 0].var, (arr)[((gid) *  2) + 1].var)
+#elif VECT_SIZE == 4
+#define packvf(arr,var,gid) (u32x) ((arr)[((gid) *  4) + 0].var, (arr)[((gid) *  4) + 1].var, (arr)[((gid) *  4) + 2].var, (arr)[((gid) *  4) + 3].var)
+#elif VECT_SIZE == 8
+#define packvf(arr,var,gid) (u32x) ((arr)[((gid) *  8) + 0].var, (arr)[((gid) *  8) + 1].var, (arr)[((gid) *  8) + 2].var, (arr)[((gid) *  8) + 3].var, (arr)[((gid) *  8) + 4].var, (arr)[((gid) *  8) + 5].var, (arr)[((gid) *  8) + 6].var, (arr)[((gid) *  8) + 7].var)
+#elif VECT_SIZE == 16
+#define packvf(arr,var,gid) (u32x) ((arr)[((gid) * 16) + 0].var, (arr)[((gid) * 16) + 1].var, (arr)[((gid) * 16) + 2].var, (arr)[((gid) * 16) + 3].var, (arr)[((gid) * 16) + 4].var, (arr)[((gid) * 16) + 5].var, (arr)[((gid) * 16) + 6].var, (arr)[((gid) * 16) + 7].var, (arr)[((gid) * 16) + 8].var, (arr)[((gid) * 16) + 9].var, (arr)[((gid) * 16) + 10].var, (arr)[((gid) * 16) + 11].var, (arr)[((gid) * 16) + 12].var, (arr)[((gid) * 16) + 13].var, (arr)[((gid) * 16) + 14].var, (arr)[((gid) * 16) + 15].var)
+#endif
+
 #if   VECT_SIZE == 1
 #define unpackv(arr,var,gid,idx,val) (arr)[((gid) *  1) + 0].var[(idx)] = val;
 #elif VECT_SIZE == 2