diff --git a/LeopardCommon.h b/LeopardCommon.h
index e7ccb10..1644fbf 100644
--- a/LeopardCommon.h
+++ b/LeopardCommon.h
@@ -30,11 +30,9 @@
 
 /*
     TODO:
-    + New 8-bit Muladd inner loops
-        + Benchmarks for smaller data!
+    + Fixes for all different input sizes
     + New 16-bit Muladd inner loops
         + Benchmarks for large data!
-    + Use parallel row ops
     + Add multi-threading to split up long parallelizable calculations
         + Write detailed comments for all the routines
         + Final benchmarks!
@@ -55,13 +53,13 @@
 // Constants
 
 // Unroll inner loops 4 times
-//#define LEO_USE_VECTOR4_OPT
+#define LEO_USE_VECTOR4_OPT
 
 // Define this to enable the optimized version of FWHT()
-//#define LEO_FWHT_OPT
+#define LEO_FWHT_OPT
 
 // Avoid scheduling reduced FFT operations that are unneeded
-//#define LEO_SCHEDULE_OPT
+#define LEO_SCHEDULE_OPT
 
 
 //------------------------------------------------------------------------------
diff --git a/tests/benchmark.cpp b/tests/benchmark.cpp
index 0e1c5c5..d0a8bd1 100644
--- a/tests/benchmark.cpp
+++ b/tests/benchmark.cpp
@@ -389,11 +389,11 @@ struct TestParameters
     unsigned original_count = 1000; // under 65536
     unsigned recovery_count = 100; // under 65536 - original_count
 #else
-    unsigned original_count = 100; // under 65536
-    unsigned recovery_count = 10; // under 65536 - original_count
+    unsigned original_count = 128; // under 65536
+    unsigned recovery_count = 128; // under 65536 - original_count
 #endif
     unsigned buffer_bytes = 64000; // multiple of 64 bytes
-    unsigned loss_count = 10; // some fraction of original_count
+    unsigned loss_count = 128; // some fraction of original_count
     unsigned seed = 0;
     bool multithreaded = true;
 };