another small optimization

@@ -748,14 +748,16 @@ TR_malloc(size_t size)
 		if (size < 8) {
 			size = 8;
 		} else {
-			size_t check = size;
-			size_t mask  = 1;
+			size_t check = size >> 4;
+			size_t mask  = 0x1F;
 
 			while (check >>= 1) {
 				mask = (mask << 1) | 1;
 			}
 
-			size = (size << 1) & ~mask;
+			if (size != (size & ~(mask >> 1))) {
+				size = (size << 1) & ~mask;
+			}
 		}
 	}