switch from clz() to ffz() for bitsliced algorithms

OpenCL/inc_common.cl

@@ -7,6 +7,18 @@
  * pure scalar functions
  */
 
+inline int ffz (const u32 v)
+{
+  for (int i = 0; i < 32; i++)
+  {
+    if ((v >> i) & 1) continue;
+
+    return i;
+  }
+
+  return -1;
+}
+
 inline int hash_comp (const u32 d1[4], __global u32 *d2)
 {
   if (d1[3] > d2[DGST_R3]) return ( 1);

OpenCL/m01500_a3.cl

@@ -2214,7 +2214,7 @@ void m01500m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
       if (tmpResult == 0xffffffff) continue;
 
-      const u32 slice = 31 - clz (~tmpResult);
+      const u32 slice = ffz (tmpResult);
 
       const u32 r0 = search[0];
       const u32 r1 = search[1];
@@ -2234,8 +2234,8 @@ void m01500m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int i = 0; i < 32; i++)
     {
-      out0[i] = out[ 0 + 31 - i];
-      out1[i] = out[32 + 31 - i];
+      out0[i] = out[ 0 + i];
+      out1[i] = out[32 + i];
     }
 
     transpose32c (out0);
@@ -2246,8 +2246,8 @@ void m01500m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int slice = 0; slice < 32; slice++)
     {
-      const u32 r0 = out0[31 - slice];
-      const u32 r1 = out1[31 - slice];
+      const u32 r0 = out0[slice];
+      const u32 r1 = out1[slice];
       const u32 r2 = 0;
       const u32 r3 = 0;
 
@@ -2640,7 +2640,7 @@ void m01500s (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
   if (tmpResult == 0xffffffff) return;
 
-  const u32 slice = 31 - clz (~tmpResult);
+  const u32 slice = ffz (tmpResult);
 
   #include COMPARE_S
 }

OpenCL/m03000_a3.cl

@@ -2056,7 +2056,7 @@ void m03000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
       if (tmpResult == 0xffffffff) continue;
 
-      const u32 slice = 31 - clz (~tmpResult);
+      const u32 slice = ffz (tmpResult);
 
       const u32 r0 = search[0];
       const u32 r1 = search[1];
@@ -2076,8 +2076,8 @@ void m03000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int i = 0; i < 32; i++)
     {
-      out0[i] = out[ 0 + 31 - i];
-      out1[i] = out[32 + 31 - i];
+      out0[i] = out[ 0 + i];
+      out1[i] = out[32 + i];
     }
 
     transpose32c (out0);
@@ -2088,8 +2088,8 @@ void m03000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int slice = 0; slice < 32; slice++)
     {
-      const u32 r0 = out0[31 - slice];
-      const u32 r1 = out1[31 - slice];
+      const u32 r0 = out0[slice];
+      const u32 r1 = out1[slice];
       const u32 r2 = 0;
       const u32 r3 = 0;
 
@@ -2482,7 +2482,7 @@ void m03000s (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
   if (tmpResult == 0xffffffff) return;
 
-  const u32 slice = 31 - clz (~tmpResult);
+  const u32 slice = ffz (tmpResult);
 
   #include COMPARE_S
 }

OpenCL/m14000_a3.cl

@@ -2194,7 +2194,7 @@ void m14000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
       if (tmpResult == 0xffffffff) continue;
 
-      const u32 slice = 31 - clz (~tmpResult);
+      const u32 slice = ffz (tmpResult);
 
       const u32 r0 = search[0];
       const u32 r1 = search[1];
@@ -2214,8 +2214,8 @@ void m14000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int i = 0; i < 32; i++)
     {
-      out0[i] = out[ 0 + 31 - i];
-      out1[i] = out[32 + 31 - i];
+      out0[i] = out[ 0 + i];
+      out1[i] = out[32 + i];
     }
 
     transpose32c (out0);
@@ -2226,8 +2226,8 @@ void m14000m (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
     #endif
     for (int slice = 0; slice < 32; slice++)
     {
-      const u32 r0 = out0[31 - slice];
-      const u32 r1 = out1[31 - slice];
+      const u32 r0 = out0[slice];
+      const u32 r1 = out1[slice];
       const u32 r2 = 0;
       const u32 r3 = 0;
 
@@ -2684,7 +2684,7 @@ void m14000s (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global co
 
   if (tmpResult == 0xffffffff) return;
 
-  const u32 slice = 31 - clz (~tmpResult);
+  const u32 slice = ffz (tmpResult);
 
   #include COMPARE_S
 }