Optimize out the swap32() of sha256 in 7zip

[hashcat.git] / OpenCL / m05100_a3.cl

diff --git a/OpenCL/m05100_a3.cl b/OpenCL/m05100_a3.cl
index 956157f..5a41ae2 100644 (file)
--- a/OpenCL/m05100_a3.cl
+++ b/OpenCL/m05100_a3.cl
@@ -20,7 +20,7 @@
 #include "OpenCL/common.c"
 #include "OpenCL/simd.c"
 
-static void m05100m (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)
+void m05100m (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)
 {
   /**
    * modifier
@@ -145,18 +145,15 @@ static void m05100m (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le
     c += MD5M_C;
     d += MD5M_D;
 
-    u32x e = 0;
-    u32x f = 0;
+    u32x z = 0;
 
-    COMPARE_M_SIMD (a, b, e, f);
-
-    COMPARE_M_SIMD (b, c, e, f);
-
-    COMPARE_M_SIMD (c, d, e, f);
+    COMPARE_M_SIMD (a, b, z, z);
+    COMPARE_M_SIMD (b, c, z, z);
+    COMPARE_M_SIMD (c, d, z, z);
   }
 }
 
-static void m05100s (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)
+void m05100s (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)
 {
   /**
    * modifier
@@ -173,8 +170,8 @@ static void m05100s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le
   {
     digests_buf[digests_offset].digest_buf[DGST_R0],
     digests_buf[digests_offset].digest_buf[DGST_R1],
-    digests_buf[digests_offset].digest_buf[DGST_R2],
-    digests_buf[digests_offset].digest_buf[DGST_R3]
+    0,
+    0
   };
 
   /**
@@ -293,14 +290,11 @@ static void m05100s (u32 w0[4], u32 w1[4], u32 w2[4], u32 w3[4], const u32 pw_le
     c += MD5M_C;
     d += MD5M_D;
 
-    u32x e = 0;
-    u32x f = 0;
-
-    COMPARE_S_SIMD (a, b, e, f);
-
-    COMPARE_S_SIMD (b, c, e, f);
+    u32x z = 0;
 
-    COMPARE_S_SIMD (c, d, e, f);
+    COMPARE_S_SIMD (a, b, z, z);
+    COMPARE_S_SIMD (b, c, z, z);
+    COMPARE_S_SIMD (c, d, z, z);
   }
 }