Use a faster way to count 1's (used by Hamming)

diff --git a/modules/ocl/src/opencl/brute_force_match.cl b/modules/ocl/src/opencl/brute_force_match.cl
index edc0bd147ca969270f8bd0105a4b36c3b5c7b0e0..8dcb9d20702733e2aceacc36693013c97ab8c60a 100644 (file)
--- a/modules/ocl/src/opencl/brute_force_match.cl
+++ b/modules/ocl/src/opencl/brute_force_match.cl
@@ -62,15 +62,12 @@
 #define DIST_TYPE 0
 #endif
 
-int bit1Count(int x)
+//http://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
+int bit1Count(int v)
 {
-    int c = 0;
-    for (int i = 0 ; i < 32 ; i++)
-    {
-        c += x & 1;
-        x >>= 1;
-    }
-    return c;
+    v = v - ((v >> 1) & 0x55555555);                    // reuse input as temporary
+    v = (v & 0x33333333) + ((v >> 2) & 0x33333333);     // temp
+    return ((v + (v >> 4) & 0xF0F0F0F) * 0x1010101) >> 24; // count
 }
 
 // dirty fix for non-template support