Unlock kernel_threads > 64

[hashcat.git] / OpenCL / m03710_a3.cl

diff --git a/OpenCL/m03710_a3.cl b/OpenCL/m03710_a3.cl
index 721b6ab..7d3a1ab 100644 (file)
--- a/OpenCL/m03710_a3.cl
+++ b/OpenCL/m03710_a3.cl
@@ -35,7 +35,7 @@
 #define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i).s0], l_bin2asc[(i).s1], l_bin2asc[(i).s2], l_bin2asc[(i).s3], l_bin2asc[(i).s4], l_bin2asc[(i).s5], l_bin2asc[(i).s6], l_bin2asc[(i).s7], l_bin2asc[(i).s8], l_bin2asc[(i).s9], l_bin2asc[(i).sa], l_bin2asc[(i).sb], l_bin2asc[(i).sc], l_bin2asc[(i).sd], l_bin2asc[(i).se], l_bin2asc[(i).sf])
 #endif
 
-static void m03710m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *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, __local u32 *l_bin2asc)
+void m03710m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *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, __local u32 *l_bin2asc)
 {
   /**
    * modifier
@@ -322,7 +322,7 @@ static void m03710m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le
   }
 }
 
-static void m03710s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *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, __local u32 *l_bin2asc)
+void m03710s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_len, __global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *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, __local u32 *l_bin2asc)
 {
   /**
    * modifier
@@ -744,30 +744,14 @@ __kernel void m03710_m08 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
   __local u32 l_bin2asc[256];
 
-  const u32 lid4 = lid * 4;
-
-  const u32 lid40 = lid4 + 0;
-  const u32 lid41 = lid4 + 1;
-  const u32 lid42 = lid4 + 2;
-  const u32 lid43 = lid4 + 3;
-
-  const u32 v400 = (lid40 >> 0) & 15;
-  const u32 v401 = (lid40 >> 4) & 15;
-  const u32 v410 = (lid41 >> 0) & 15;
-  const u32 v411 = (lid41 >> 4) & 15;
-  const u32 v420 = (lid42 >> 0) & 15;
-  const u32 v421 = (lid42 >> 4) & 15;
-  const u32 v430 = (lid43 >> 0) & 15;
-  const u32 v431 = (lid43 >> 4) & 15;
-
-  l_bin2asc[lid40] = ((v400 < 10) ? '0' + v400 : 'a' - 10 + v400) << 8
-                   | ((v401 < 10) ? '0' + v401 : 'a' - 10 + v401) << 0;
-  l_bin2asc[lid41] = ((v410 < 10) ? '0' + v410 : 'a' - 10 + v410) << 8
-                   | ((v411 < 10) ? '0' + v411 : 'a' - 10 + v411) << 0;
-  l_bin2asc[lid42] = ((v420 < 10) ? '0' + v420 : 'a' - 10 + v420) << 8
-                   | ((v421 < 10) ? '0' + v421 : 'a' - 10 + v421) << 0;
-  l_bin2asc[lid43] = ((v430 < 10) ? '0' + v430 : 'a' - 10 + v430) << 8
-                   | ((v431 < 10) ? '0' + v431 : 'a' - 10 + v431) << 0;
+  for (u32 i = lid; i < 256; i += lsz)
+  {
+    const u32 i0 = (i >> 0) & 15;
+    const u32 i1 = (i >> 4) & 15;
+
+    l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
+                 | ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
+  }
 
   barrier (CLK_LOCAL_MEM_FENCE);
 
@@ -830,30 +814,14 @@ __kernel void m03710_m16 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
   __local u32 l_bin2asc[256];
 
-  const u32 lid4 = lid * 4;
-
-  const u32 lid40 = lid4 + 0;
-  const u32 lid41 = lid4 + 1;
-  const u32 lid42 = lid4 + 2;
-  const u32 lid43 = lid4 + 3;
-
-  const u32 v400 = (lid40 >> 0) & 15;
-  const u32 v401 = (lid40 >> 4) & 15;
-  const u32 v410 = (lid41 >> 0) & 15;
-  const u32 v411 = (lid41 >> 4) & 15;
-  const u32 v420 = (lid42 >> 0) & 15;
-  const u32 v421 = (lid42 >> 4) & 15;
-  const u32 v430 = (lid43 >> 0) & 15;
-  const u32 v431 = (lid43 >> 4) & 15;
-
-  l_bin2asc[lid40] = ((v400 < 10) ? '0' + v400 : 'a' - 10 + v400) << 8
-                   | ((v401 < 10) ? '0' + v401 : 'a' - 10 + v401) << 0;
-  l_bin2asc[lid41] = ((v410 < 10) ? '0' + v410 : 'a' - 10 + v410) << 8
-                   | ((v411 < 10) ? '0' + v411 : 'a' - 10 + v411) << 0;
-  l_bin2asc[lid42] = ((v420 < 10) ? '0' + v420 : 'a' - 10 + v420) << 8
-                   | ((v421 < 10) ? '0' + v421 : 'a' - 10 + v421) << 0;
-  l_bin2asc[lid43] = ((v430 < 10) ? '0' + v430 : 'a' - 10 + v430) << 8
-                   | ((v431 < 10) ? '0' + v431 : 'a' - 10 + v431) << 0;
+  for (u32 i = lid; i < 256; i += lsz)
+  {
+    const u32 i0 = (i >> 0) & 15;
+    const u32 i1 = (i >> 4) & 15;
+
+    l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
+                 | ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
+  }
 
   barrier (CLK_LOCAL_MEM_FENCE);
 
@@ -916,30 +884,14 @@ __kernel void m03710_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
   __local u32 l_bin2asc[256];
 
-  const u32 lid4 = lid * 4;
-
-  const u32 lid40 = lid4 + 0;
-  const u32 lid41 = lid4 + 1;
-  const u32 lid42 = lid4 + 2;
-  const u32 lid43 = lid4 + 3;
-
-  const u32 v400 = (lid40 >> 0) & 15;
-  const u32 v401 = (lid40 >> 4) & 15;
-  const u32 v410 = (lid41 >> 0) & 15;
-  const u32 v411 = (lid41 >> 4) & 15;
-  const u32 v420 = (lid42 >> 0) & 15;
-  const u32 v421 = (lid42 >> 4) & 15;
-  const u32 v430 = (lid43 >> 0) & 15;
-  const u32 v431 = (lid43 >> 4) & 15;
-
-  l_bin2asc[lid40] = ((v400 < 10) ? '0' + v400 : 'a' - 10 + v400) << 8
-                   | ((v401 < 10) ? '0' + v401 : 'a' - 10 + v401) << 0;
-  l_bin2asc[lid41] = ((v410 < 10) ? '0' + v410 : 'a' - 10 + v410) << 8
-                   | ((v411 < 10) ? '0' + v411 : 'a' - 10 + v411) << 0;
-  l_bin2asc[lid42] = ((v420 < 10) ? '0' + v420 : 'a' - 10 + v420) << 8
-                   | ((v421 < 10) ? '0' + v421 : 'a' - 10 + v421) << 0;
-  l_bin2asc[lid43] = ((v430 < 10) ? '0' + v430 : 'a' - 10 + v430) << 8
-                   | ((v431 < 10) ? '0' + v431 : 'a' - 10 + v431) << 0;
+  for (u32 i = lid; i < 256; i += lsz)
+  {
+    const u32 i0 = (i >> 0) & 15;
+    const u32 i1 = (i >> 4) & 15;
+
+    l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
+                 | ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
+  }
 
   barrier (CLK_LOCAL_MEM_FENCE);
 
@@ -1002,30 +954,14 @@ __kernel void m03710_s08 (__global pw_t *pws, __global kernel_rule_t *rules_buf,
 
   __local u32 l_bin2asc[256];
 
-  const u32 lid4 = lid * 4;
-
-  const u32 lid40 = lid4 + 0;
-  const u32 lid41 = lid4 + 1;
-  const u32 lid42 = lid4 + 2;
-  const u32 lid43 = lid4 + 3;
-
-  const u32 v400 = (lid40 >> 0) & 15;
-  const u32 v401 = (lid40 >> 4) & 15;
-  const u32 v410 = (lid41 >> 0) & 15;
-  const u32 v411 = (lid41 >> 4) & 15;
-  const u32 v420 = (lid42 >> 0) & 15;
-  const u32 v421 = (lid42 >> 4) & 15;
-  const u32 v430 = (lid43 >> 0) & 15;
-  const u32 v431 = (lid43 >> 4) & 15;
-
-  l_bin2asc[lid40] = ((v400 < 10) ? '0' + v400 : 'a' - 10 + v400) << 8
-                   | ((v401 < 10) ? '0' + v401 : 'a' - 10 + v401) << 0;
-  l_bin2asc[lid41] = ((v410 < 10) ? '0' + v410 : 'a' - 10 + v410) << 8
-                   | ((v411 < 10) ? '0' + v411 : 'a' - 10 + v411) << 0;
-  l_bin2asc[lid42] = ((v420 < 10) ? '0' + v420 : 'a' - 10 + v420) << 8
-                   | ((v421 < 10) ? '0' + v421 : 'a' - 10 + v421) << 0;
-  l_bin2asc[lid43] = ((v430 < 10) ? '0' + v430 : 'a' - 10 + v430) << 8
-                   | ((v431 < 10) ? '0' + v431 : 'a' - 10 + v431) << 0;
+  for (u32 i = lid; i < 256; i += lsz)
+  {
+    const u32 i0 = (i >> 0) & 15;
+    const u32 i1 = (i >> 4) & 15;
+
+    l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
+                 | ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
+  }
 
   barrier (CLK_LOCAL_MEM_FENCE);