Using gpu device 0: GeForce GT 750M (CNMeM is disabled, cuDNN not available)
Starting Reference Theano (gpu) benchmark
Total time: 1.595621s
Single iteration: 0.01595621s
------------------------------
Starting Optimized Theano (gpu) benchmark
Total time: 0.665976s
Single iteration: 0.00665976s
------------------------------
Starting Numpy benchmark
Total time: 0.215270s
Single iteration: 0.00215270s
------------------------------
Theano and Numpy results compatible: True
==============================
Performance comparison:
NN size:  (1000, 1000, 1000, 1000, 1000, 1000, 1000, 1000)
Iterations: 100
theano: 1.00x faster (0.015956s)
numpy: 7.41x faster (0.002153s)
optimized: 2.40x faster (0.006660s)
##############################
Starting Reference Theano (gpu) benchmark
Total time: 5.406381s
Single iteration: 0.05406381s
------------------------------
Starting Optimized Theano (gpu) benchmark
Total time: 1.028230s
Single iteration: 0.01028230s
------------------------------
Starting Numpy benchmark
Total time: 0.697497s
Single iteration: 0.00697497s
------------------------------
Theano and Numpy results compatible: True
==============================
Performance comparison:
NN size:  (2000, 2000, 2000, 2000, 2000, 2000, 2000, 2000)
Iterations: 100
theano: 1.00x faster (0.054064s)
numpy: 7.75x faster (0.006975s)
optimized: 5.26x faster (0.010282s)
##############################
Starting Reference Theano (gpu) benchmark
Total time: 0.678504s
Single iteration: 0.00678504s
------------------------------
Starting Optimized Theano (gpu) benchmark
Total time: 0.425372s
Single iteration: 0.00425372s
------------------------------
Starting Numpy benchmark
Total time: 0.018013s
Single iteration: 0.00018013s
------------------------------
Theano and Numpy results compatible: False
==============================
Performance comparison:
NN size:  [1024, 512, 256, 128, 64, 32, 16, 8]
Iterations: 100
theano: 1.00x faster (0.006785s)
numpy: 37.67x faster (0.000180s)
optimized: 1.60x faster (0.004254s)
##############################
Starting Reference Theano (gpu) benchmark
Total time: 0.117089s
Single iteration: 0.00117089s
------------------------------
Starting Optimized Theano (gpu) benchmark
Total time: 0.131096s
Single iteration: 0.00131096s
------------------------------
Starting Numpy benchmark
Total time: 0.002002s
Single iteration: 0.00002002s
------------------------------
Theano and Numpy results compatible: False
==============================
Performance comparison:
NN size:  (26, 30, 2)
Iterations: 100
theano: 1.00x faster (0.001171s)
numpy: 58.49x faster (0.000020s)
optimized: 0.89x faster (0.001311s)
##############################