Add pure kernel for -m 6300

OpenCL/m06300-pure.cl

@@ -0,0 +1,253 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+//#define NEW_SIMD_CODE
+
+#include "inc_vendor.cl"
+#include "inc_hash_constants.h"
+#include "inc_hash_functions.cl"
+#include "inc_types.cl"
+#include "inc_common.cl"
+#include "inc_hash_md5.cl"
+
+#define COMPARE_S "inc_comp_single.cl"
+#define COMPARE_M "inc_comp_multi.cl"
+
+__kernel void m06300_init (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global md5crypt_tmp_t *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * base
+   */
+
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * init
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32 w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  const u32 salt_len = salt_bufs[salt_pos].salt_len;
+
+  const u32 salt_lenv = ceil ((float) salt_len / 4);
+
+  u32 s[64] = { 0 };
+
+  for (int idx = 0; idx < salt_lenv; idx++)
+  {
+    s[idx] = salt_bufs[salt_pos].salt_buf[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  /**
+   * prepare
+   */
+
+  md5_ctx_t md5_ctx1;
+
+  md5_init (&md5_ctx1);
+
+  md5_update (&md5_ctx1, w, pw_len);
+
+  md5_update (&md5_ctx1, s, salt_len);
+
+  md5_update (&md5_ctx1, w, pw_len);
+
+  md5_final (&md5_ctx1);
+
+  u32 final[16] = { 0 };
+
+  final[0] = md5_ctx1.h[0];
+  final[1] = md5_ctx1.h[1];
+  final[2] = md5_ctx1.h[2];
+  final[3] = md5_ctx1.h[3];
+
+  md5_ctx_t md5_ctx;
+
+  md5_init (&md5_ctx);
+
+  md5_update (&md5_ctx, w, pw_len);
+
+  md5_update (&md5_ctx, s, salt_len);
+
+  int pl;
+
+	for (pl = pw_len; pl > 16; pl -= 16)
+  {
+    md5_update (&md5_ctx, final, 16);
+  }
+
+  truncate_block_4x4_le (final, pl);
+
+  md5_update (&md5_ctx, final, pl);
+
+  /* Then something really weird... */
+
+	for (int i = pw_len; i != 0; i >>= 1)
+  {
+    u32 t[16] = { 0 };
+
+		if (i & 1)
+    {
+      t[0] = 0;
+    }
+    else
+    {
+      t[0] = w[0] & 0xff;
+    }
+
+    md5_update (&md5_ctx, t, 1);
+  }
+
+  md5_final (&md5_ctx);
+
+  tmps[gid].digest_buf[0] = md5_ctx.h[0];
+  tmps[gid].digest_buf[1] = md5_ctx.h[1];
+  tmps[gid].digest_buf[2] = md5_ctx.h[2];
+  tmps[gid].digest_buf[3] = md5_ctx.h[3];
+}
+
+__kernel void m06300_loop (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global md5crypt_tmp_t *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * base
+   */
+
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * init
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32 w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  const u32 salt_len = salt_bufs[salt_pos].salt_len;
+
+  const u32 salt_lenv = ceil ((float) salt_len / 4);
+
+  u32 s[64] = { 0 };
+
+  for (int idx = 0; idx < salt_lenv; idx++)
+  {
+    s[idx] = salt_bufs[salt_pos].salt_buf[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  /**
+   * digest
+   */
+
+  u32 digest[16] = { 0 };
+
+  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];
+
+  /**
+   * loop
+   */
+
+  for (u32 i = 0, j = loop_pos; i < loop_cnt; i++, j++)
+  {
+    md5_ctx_t md5_ctx;
+
+    md5_init (&md5_ctx);
+
+		if (j & 1)
+    {
+			md5_update (&md5_ctx, w, pw_len);
+    }
+		else
+    {
+			md5_update (&md5_ctx, digest, 16);
+    }
+
+		if (j % 3)
+    {
+			md5_update (&md5_ctx, s, salt_len);
+    }
+
+		if (j % 7)
+    {
+			md5_update (&md5_ctx, w, pw_len);
+    }
+
+		if (j & 1)
+    {
+			md5_update (&md5_ctx, digest, 16);
+    }
+		else
+    {
+			md5_update (&md5_ctx, w, pw_len);
+    }
+
+    md5_final (&md5_ctx);
+
+    digest[0] = md5_ctx.h[0];
+    digest[1] = md5_ctx.h[1];
+    digest[2] = md5_ctx.h[2];
+    digest[3] = md5_ctx.h[3];
+  }
+
+  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];
+}
+
+__kernel void m06300_comp (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global md5crypt_tmp_t *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  const u32 lid = get_local_id (0);
+
+  /**
+   * digest
+   */
+
+  const u32 r0 = tmps[gid].digest_buf[DGST_R0];
+  const u32 r1 = tmps[gid].digest_buf[DGST_R1];
+  const u32 r2 = tmps[gid].digest_buf[DGST_R2];
+  const u32 r3 = tmps[gid].digest_buf[DGST_R3];
+
+  #define il_pos 0
+
+  #include COMPARE_M
+}

OpenCL/m06300.cl

@@ -781,7 +781,7 @@ __kernel void m06300_init (__global pw_t *pws, __global const kernel_rule_t *rul
 
   block_len += salt_len;
 
-  truncate_block (digest, pw_len);
+  truncate_block_4x4_le (digest, pw_len);
 
   memcat16 (block0, block1, block2, block3, block_len, digest);