diff --git a/asmjit/__main__.py b/asmjit/__main__.py
index 9096e98..bd3878d 100644
--- a/asmjit/__main__.py
+++ b/asmjit/__main__.py
@@ -9,9 +9,11 @@ def main():
     # parser.add_argument('mode', metavar='MODE', type=str, choices=['latency', 'throughput'])
     parser.add_argument('instructions', metavar='INSTR', type=op.Instruction.from_string, nargs='+',
                         help='instruction declaration, e.g., "add {src:i32:r} {srcdst:i32:r}"')
+    parser.add_argument('--serialize', action='store_true',
+                        help='Serialize instructions.')
     parser.add_argument('--latency-serial', '-l', type=int, default=8,
                          help='length of serial chain for each instruction in latency benchmark')
-    parser.add_argument('--parallel', '-p',type=int, default=4,
+    parser.add_argument('--parallel', '-p',type=int, default=10,
                         help='number of parallel instances of serial chains in throughput '
                              'benchmark')
     parser.add_argument('--throughput-serial', '-t', type=int, default=8,
@@ -25,6 +27,7 @@ def main():
                                        serial_factor=args.latency_serial,
                                        parallel_factor=args.parallel,
                                        throughput_serial_factor=args.throughput_serial,
+                                       serialize=args.serialize,
                                        verbosity=args.verbose)
     print("Latency: {:.2f} cycle\nThroughput: {:.2f} cycle\n".format(lat, tp))
 
diff --git a/asmjit/bench.py b/asmjit/bench.py
index 164829e..ec1f3a8 100755
--- a/asmjit/bench.py
+++ b/asmjit/bench.py
@@ -238,15 +238,18 @@ class IntegerLoopBenchmark(LoopBenchmark):
 
 
 def bench_instructions(instructions, serial_factor=8, parallel_factor=4, throughput_serial_factor=8,
-                       verbosity=0):
+                       serialize=False, verbosity=0):
     not_serializable = False
     try:
         # Latency Benchmark
         if verbosity > 0:
             print('## Latency Benchmark')
         p_instrs = []
-        for i in instructions:
-            p_instrs.append(op.Serialized([i] * serial_factor))
+        if not serialize:
+            for i in instructions:
+                p_instrs.append(op.Serialized([i] * serial_factor))
+        else:
+            p_instrs = [op.Serialized(instructions * serial_factor)]
         p = op.Parallelized(p_instrs)
         b = IntegerLoopBenchmark(p)
         if verbosity >= 3:
@@ -274,8 +277,11 @@ def bench_instructions(instructions, serial_factor=8, parallel_factor=4, through
     if verbosity > 0:
         print('## Throughput Benchmark')
     p_instrs = []
-    for i in instructions:
-        p_instrs.append(op.Serialized([i] * throughput_serial_factor))
+    if not serialize:
+        for i in instructions:
+            p_instrs.append(op.Serialized([i] * throughput_serial_factor))
+    else:
+        p_instrs = [op.Serialized(instructions * throughput_serial_factor)]
     p = op.Parallelized(p_instrs * parallel_factor)
     b = IntegerLoopBenchmark(p)
     if verbosity >= 3:
diff --git a/asmjit/op.py b/asmjit/op.py
index 0be6120..418da11 100755
--- a/asmjit/op.py
+++ b/asmjit/op.py
@@ -195,12 +195,6 @@ class Synthable:
 class Operation(Synthable):
     """Base class for operations."""
 
-    def __repr__(self):
-        return '{}({})'.format(
-            self.__class__.__name__,
-            ', '.join(['{}={!r}'.format(k, v) for k, v in self.__dict__.items()
-                       if not k.startswith('_')]))
-
 
 class Instruction(Operation):
     def __init__(self, instruction, destination_operand, source_operands):