Create a custom processor configuration object of class
dlhdl.ProcessorConfig.
Create an object of class workflow by using the
dlhdl.Workflow class.
Set the deep learning network and processor configuration for the workflow object.
Call the estimate function for the workflow object.
The speed and latency is stored in a structure struct and
displayed on the screen.
For example:
hPC = dlhdl.ProcessorConfig; snet = vgg19; hW = dlhdl.Workflow('Network', snet, 'ProcessorConfig',hPC); result = hW.estimate('Performance');
The result of the estimation is:
Deep Learning Processor Estimator Performance Results
LastLayerLatency(cycles) LastLayerLatency(seconds) FramesNum Total Latency Frames/s
------------- ------------- --------- --------- ---------
Network 202770372 1.01385 1 202770372 1.0
conv_module 158812469 0.79406
conv1_1 2022004 0.01011
conv1_2 15855549 0.07928
pool1 2334753 0.01167
conv2_1 7536365 0.03768
conv2_2 14837392 0.07419
pool2 1446960 0.00723
conv3_1 7950445 0.03975
conv3_2 14365933 0.07183
conv3_3 14365933 0.07183
conv3_4 14365933 0.07183
pool3 930145 0.00465
conv4_1 7073684 0.03537
conv4_2 13761300 0.06881
conv4_3 13761300 0.06881
conv4_4 13761300 0.06881
pool4 572644 0.00286
conv5_1 3432645 0.01716
conv5_2 3432645 0.01716
conv5_3 3432645 0.01716
conv5_4 3432645 0.01716
pool5 140249 0.00070
fc_module 43957903 0.21979
fc6 36535923 0.18268
fc7 5965299 0.02983
fc8 1456681 0.00728
* The clock frequency of the DL processor is: 200MHz