dlhdl.ProcessorConfig.optimizeConfigurationForNetwork - Update network-specific deep learning processor configuration with optimized deep
learning processor configuration - MATLAB ([original](https://www.mathworks.com/help/deep-learning-hdl/ref/dlhdl.processorconfig.optimizeconfigurationfornetwork.html)) ([raw](?raw))Class: dlhdl.ProcessorConfig
Namespace: dlhdl
Update network-specific deep learning processor configuration with optimized deep learning processor configuration
Since R2021b
Syntax
Description
[newprocessorConfigObject](#mw%5Fe4f9d385-26fe-4314-b2d7-24dc08565c91) = optimizeConfigurationForNetwork([processorConfigObject](#mw%5Fa2be16fb-b571-4296-b676-96ba72649e91%5Fsep%5Fmw%5F4aaceff3-0437-4357-96fe-ca0f4d3de873),[network](#mw%5F86b13ef0-a79b-4441-a4b7-c991d1decab8)) updates the processor configuration object specified byprocessorConfigObject with the optimized deep learning processor configuration, optimized for the network and returns a new updated processor configuration object, newprocessorConfigObject.
optimizeConfigurationForNetwork([processorConfigObject](#mw%5Fa2be16fb-b571-4296-b676-96ba72649e91%5Fsep%5Fmw%5F4aaceff3-0437-4357-96fe-ca0f4d3de873),[network](#mw%5F86b13ef0-a79b-4441-a4b7-c991d1decab8)) updates the processor configuration object specified byprocessorConfigObject with the optimized deep learning processor configuration, optimized for the network and overwrites and returns the updated processor configuration object,processorConfigObject.
[newprocessorConfigObject](#mw%5Fe4f9d385-26fe-4314-b2d7-24dc08565c91) = optimizeConfigurationForNetwork(___,[Name,Value](#namevaluepairarguments)) updates the processor configuration, with options specified by one or more name-value arguments.
The optimizeConfigurationForNetwork method optimizes the deep learning processor configuration to achieve your target goal of performance or resource usage optimization. To learn more about the deep learning processor configuration properties and their affects on performance, see Effects of Custom Deep Learning Processor Parameters on Performance and Resource Utilization. You can manually adjust the processor configuration or use the optimizeConfigurationForNetwork method.
Input Arguments
processorConfigObject — Processor configuration
dlhdl.ProcessorConfig object
Processor configuration, specified as adlhdl.ProcessorConfig object.
network — Name of network
SeriesNetwork object | DAGNetwork object | dlnetwork object | quantize object
Name of network for optimized deep learning processor configuration, specified as aSeriesNetwork, DAGNetwork, ordlnetwork.
Example: optimizeConfigurationForNetwork(snet)
Name-Value Arguments
Specify optional pairs of arguments asName1=Value1,...,NameN=ValueN, where Name is the argument name and Value is the corresponding value. Name-value arguments must appear after other arguments, but the order of the pairs does not matter.
Before R2021a, use commas to separate each name and value, and enclose Name in quotes.
FramesPerSecond — Target frames per second
0 (default) | double
Target frames per second for the network, specified as a double data type.
Example: FramesPerSecond=10
Output Arguments
newprocessorConfigObject — Processor configuration
dlhdl.ProcessorConfig object
Processor configuration specified as a dlhdl.ProcessorConfig object.
Examples
Generate Optimized Processor Configuration for MobileNetV2 Network
Retrieve an optimized MobileNetV2 network-specific processor configuration, by using the optimizeConfigurationForNetwork method.
Create a dlhdl.ProcessorConfig object.
net = mobilenetv2; hPC = dlhdl.ProcessorConfig;
To retrieve an optimized processor configuration, call theoptimizeConfigurationForNetwork method. TheoptimizeConfigurationForNetwork method overwrites and returns the deep learning processor configuration hPC.
optimizeConfigurationForNetwork(hPC,net)
Optimizing processor configuration for deep learning network...
Note: Processing module "conv" property "InputMemorySize" changed from "[227 227 3]" to "[224 224 3]".
Note: Processing module "conv" property "OutputMemorySize" changed from "[227 227 3]" to "[112 112 16]".
Note: Processing module "conv" property "FeatureSizeLimit" changed from "2048" to "1280".
Note: Processing module "conv" property "SegmentationBlockGeneration" changed from "true" to "false".
Note: Processing module "fc" property "InputMemorySize" changed from "25088" to "1280".
Note: Processing module "fc" property "OutputMemorySize" changed from "4096" to "1000".
Note: Processing module "fc" property "SoftmaxBlockGeneration" changed from "false" to "true".
Processing Module "conv"
ModuleGeneration: 'on'
LRNBlockGeneration: 'off'
SegmentationBlockGeneration: 'off'
ConvThreadNumber: 16
InputMemorySize: [224 224 3]
OutputMemorySize: [112 112 16]
FeatureSizeLimit: 1280
Processing Module "fc"
ModuleGeneration: 'on'
SoftmaxBlockGeneration: 'on'
FCThreadNumber: 4
InputMemorySize: 1280
OutputMemorySize: 1000
Processing Module "custom"
ModuleGeneration: 'on'
Addition: 'on'
MishLayer: 'off'
Multiplication: 'on'
Resize2D: 'off'
Sigmoid: 'off'
SwishLayer: 'off'
TanhLayer: 'off'
InputMemorySize: 40
OutputMemorySize: 120
Processor Top Level Properties
RunTimeControl: 'register'
RunTimeStatus: 'register'
InputStreamControl: 'register'
OutputStreamControl: 'register'
SetupControl: 'register'
ProcessorDataType: 'single'
UseVendorLibrary: 'on'
System Level Properties
TargetPlatform: 'Xilinx Zynq UltraScale+ MPSoC ZCU102 Evaluation Kit'
TargetFrequency: 200
SynthesisTool: 'Xilinx Vivado'
ReferenceDesign: 'AXI-Stream DDR Memory Access : 3-AXIM'
SynthesisToolChipFamily: 'Zynq UltraScale+'
SynthesisToolDeviceName: 'xczu9eg-ffvb1156-2-e'
SynthesisToolPackageName: ''
SynthesisToolSpeedValue: ''Optimizing processor configuration for deep learning network complete.
ans = Processing Module "conv" ModuleGeneration: 'on' LRNBlockGeneration: 'off' SegmentationBlockGeneration: 'off' ConvThreadNumber: 16 InputMemorySize: [224 224 3] OutputMemorySize: [112 112 16] FeatureSizeLimit: 1280
Processing Module "fc"
ModuleGeneration: 'on'
SoftmaxBlockGeneration: 'on'
FCThreadNumber: 4
InputMemorySize: 1280
OutputMemorySize: 1000
Processing Module "custom"
ModuleGeneration: 'on'
Addition: 'on'
MishLayer: 'off'
Multiplication: 'on'
Resize2D: 'off'
Sigmoid: 'off'
SwishLayer: 'off'
TanhLayer: 'off'
InputMemorySize: 40
OutputMemorySize: 120
Processor Top Level Properties
RunTimeControl: 'register'
RunTimeStatus: 'register'
InputStreamControl: 'register'
OutputStreamControl: 'register'
SetupControl: 'register'
ProcessorDataType: 'single'
UseVendorLibrary: 'on'
System Level Properties
TargetPlatform: 'Xilinx Zynq UltraScale+ MPSoC ZCU102 Evaluation Kit'
TargetFrequency: 200
SynthesisTool: 'Xilinx Vivado'
ReferenceDesign: 'AXI-Stream DDR Memory Access : 3-AXIM'
SynthesisToolChipFamily: 'Zynq UltraScale+'
SynthesisToolDeviceName: 'xczu9eg-ffvb1156-2-e'
SynthesisToolPackageName: ''
SynthesisToolSpeedValue: ''Optimize Deep Learning Processor Configuration for Network Performance
This example shows how to generate a deep learning processor configuration and estimate the performance of a pretrained network. Generate a deep learning processor configuration optimized for the target frames-per-second value of the network, then generate a custom bitstream by using the optimized processor configuration.
Load Pretrained Network and Create Processor Configuration
To load a pretrained ResNet-18 network, enter:
net = imagePretrainedNetwork('resnet18');
Create a custom deep learning processor configuration. For more information, see dlhdl.ProcessorConfig.
hPC = dlhdl.ProcessorConfig;
Estimate Network Performance
Establish the baseline performance of the network, by estimating the performance of the ResNet-18 network. Estimate the performance, by using the estimatePerformance method of the dlhdl.ProcessorConfig object. The method returns the estimated layer latency, network latency, and network performance in frames per second.
estimatePerformance(hPC,net);
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
Optimizing network: Fused 'nnet.cnn.layer.BatchNormalizationLayer' into 'nnet.cnn.layer.Convolution2DLayer'
Notice: The layer 'data' of type 'ImageInputLayer' is split into an image input layer 'data', an addition layer 'data_norm_add', and a multiplication layer 'data_norm' for hardware normalization.
The network includes the following layers:
1 'data' Image Input 224×224×3 images with 'zscore' normalization (SW Layer)
2 'conv1' 2-D Convolution 64 7×7×3 convolutions with stride [2 2] and padding [3 3 3 3] (HW Layer)
3 'conv1_relu' ReLU ReLU (HW Layer)
4 'pool1' 2-D Max Pooling 3×3 max pooling with stride [2 2] and padding [1 1 1 1] (HW Layer)
5 'res2a_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
6 'res2a_branch2a_relu' ReLU ReLU (HW Layer)
7 'res2a_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
8 'res2a' Addition Element-wise addition of 2 inputs (HW Layer)
9 'res2a_relu' ReLU ReLU (HW Layer)
10 'res2b_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
11 'res2b_branch2a_relu' ReLU ReLU (HW Layer)
12 'res2b_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
13 'res2b' Addition Element-wise addition of 2 inputs (HW Layer)
14 'res2b_relu' ReLU ReLU (HW Layer)
15 'res3a_branch2a' 2-D Convolution 128 3×3×64 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
16 'res3a_branch2a_relu' ReLU ReLU (HW Layer)
17 'res3a_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
18 'res3a_branch1' 2-D Convolution 128 1×1×64 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
19 'res3a' Addition Element-wise addition of 2 inputs (HW Layer)
20 'res3a_relu' ReLU ReLU (HW Layer)
21 'res3b_branch2a' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
22 'res3b_branch2a_relu' ReLU ReLU (HW Layer)
23 'res3b_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
24 'res3b' Addition Element-wise addition of 2 inputs (HW Layer)
25 'res3b_relu' ReLU ReLU (HW Layer)
26 'res4a_branch2a' 2-D Convolution 256 3×3×128 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
27 'res4a_branch2a_relu' ReLU ReLU (HW Layer)
28 'res4a_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
29 'res4a_branch1' 2-D Convolution 256 1×1×128 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
30 'res4a' Addition Element-wise addition of 2 inputs (HW Layer)
31 'res4a_relu' ReLU ReLU (HW Layer)
32 'res4b_branch2a' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
33 'res4b_branch2a_relu' ReLU ReLU (HW Layer)
34 'res4b_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
35 'res4b' Addition Element-wise addition of 2 inputs (HW Layer)
36 'res4b_relu' ReLU ReLU (HW Layer)
37 'res5a_branch2a' 2-D Convolution 512 3×3×256 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
38 'res5a_branch2a_relu' ReLU ReLU (HW Layer)
39 'res5a_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
40 'res5a_branch1' 2-D Convolution 512 1×1×256 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
41 'res5a' Addition Element-wise addition of 2 inputs (HW Layer)
42 'res5a_relu' ReLU ReLU (HW Layer)
43 'res5b_branch2a' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
44 'res5b_branch2a_relu' ReLU ReLU (HW Layer)
45 'res5b_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
46 'res5b' Addition Element-wise addition of 2 inputs (HW Layer)
47 'res5b_relu' ReLU ReLU (HW Layer)
48 'pool5' 2-D Global Average Pooling 2-D global average pooling (HW Layer)
49 'fc1000' Fully Connected 1000 fully connected layer (HW Layer)
50 'prob' Softmax softmax (SW Layer)
51 'Output1_prob' Regression Output mean-squared-error (SW Layer)
Notice: The layer 'prob' with type 'nnet.cnn.layer.SoftmaxLayer' is implemented in software.
Notice: The layer 'Output1_prob' with type 'nnet.cnn.layer.RegressionOutputLayer' is implemented in software.
Deep Learning Processor Estimator Performance Results
LastFrameLatency(cycles) LastFrameLatency(seconds) FramesNum Total Latency Frames/s
------------- ------------- --------- --------- ---------Network 21627113 0.10814 1 21627113 9.2 data_norm_add 268453 0.00134 data_norm 163081 0.00082 conv1 2164700 0.01082 pool1 515128 0.00258 res2a_branch2a 966477 0.00483 res2a_branch2b 966477 0.00483 res2a 268453 0.00134 res2b_branch2a 966477 0.00483 res2b_branch2b 966477 0.00483 res2b 268453 0.00134 res3a_branch1 541373 0.00271 res3a_branch2a 541261 0.00271 res3a_branch2b 920141 0.00460 res3a 134257 0.00067 res3b_branch2a 920141 0.00460 res3b_branch2b 920141 0.00460 res3b 134257 0.00067 res4a_branch1 505453 0.00253 res4a_branch2a 511309 0.00256 res4a_branch2b 909517 0.00455 res4a 67152 0.00034 res4b_branch2a 909517 0.00455 res4b_branch2b 909517 0.00455 res4b 67152 0.00034 res5a_branch1 750669 0.00375 res5a_branch2a 757837 0.00379 res5a_branch2b 1427661 0.00714 res5a 33582 0.00017 res5b_branch2a 1427661 0.00714 res5b_branch2b 1427661 0.00714 res5b 33582 0.00017 pool5 55746 0.00028 fc1000 207350 0.00104
- The clock frequency of the DL processor is: 200MHz
The estimated frames-per-second performance is 9.4 frames per second. To improve the network performance, you can modify the properties of the custom deep learning processor configuration hPC or use the optimizeConfigurationForNetwork method. In this example, you use the optimizeConfigurationForNetwork method. To learn about modifying the properties manually, see Effects of Custom Deep Learning Processor Parameters on Performance and Resource Utilization.
Generate Optimized Processor Configuration
Optimize the processor configuration by using the optimizeConfigurationForNetwork method. Use the optional FramesPerSecond name-value argument.
hPC_optimized = optimizeConfigurationForNetwork(hPC,net,FramesPerSecond=10);
Optimizing processor configuration for deep learning network...
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
Deep Learning Processor Estimator Resource Results
DSPs Block RAM* LUTs(CLB/ALUT)
------------- ------------- ------------- Available 2520 912 274080 ------------- ------------- ------------- Total 779( 31%) 600( 66%) 270396( 99%) ReferenceDesign 3( 1%) 78( 9%) 35000( 13%) DL_Processor 776( 31%) 522( 58%) 235396( 86%)
- Block RAM represents Block RAM tiles in Xilinx devices and Block RAM bits in Intel devices
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
Optimizing network: Fused 'nnet.cnn.layer.BatchNormalizationLayer' into 'nnet.cnn.layer.Convolution2DLayer'
Notice: The layer 'data' of type 'ImageInputLayer' is split into an image input layer 'data', an addition layer 'data_norm_add', and a multiplication layer 'data_norm' for hardware normalization.
The network includes the following layers:
1 'data' Image Input 224×224×3 images with 'zscore' normalization (SW Layer)
2 'conv1' 2-D Convolution 64 7×7×3 convolutions with stride [2 2] and padding [3 3 3 3] (HW Layer)
3 'conv1_relu' ReLU ReLU (HW Layer)
4 'pool1' 2-D Max Pooling 3×3 max pooling with stride [2 2] and padding [1 1 1 1] (HW Layer)
5 'res2a_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
6 'res2a_branch2a_relu' ReLU ReLU (HW Layer)
7 'res2a_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
8 'res2a' Addition Element-wise addition of 2 inputs (HW Layer)
9 'res2a_relu' ReLU ReLU (HW Layer)
10 'res2b_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
11 'res2b_branch2a_relu' ReLU ReLU (HW Layer)
12 'res2b_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
13 'res2b' Addition Element-wise addition of 2 inputs (HW Layer)
14 'res2b_relu' ReLU ReLU (HW Layer)
15 'res3a_branch2a' 2-D Convolution 128 3×3×64 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
16 'res3a_branch2a_relu' ReLU ReLU (HW Layer)
17 'res3a_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
18 'res3a_branch1' 2-D Convolution 128 1×1×64 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
19 'res3a' Addition Element-wise addition of 2 inputs (HW Layer)
20 'res3a_relu' ReLU ReLU (HW Layer)
21 'res3b_branch2a' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
22 'res3b_branch2a_relu' ReLU ReLU (HW Layer)
23 'res3b_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
24 'res3b' Addition Element-wise addition of 2 inputs (HW Layer)
25 'res3b_relu' ReLU ReLU (HW Layer)
26 'res4a_branch2a' 2-D Convolution 256 3×3×128 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
27 'res4a_branch2a_relu' ReLU ReLU (HW Layer)
28 'res4a_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
29 'res4a_branch1' 2-D Convolution 256 1×1×128 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
30 'res4a' Addition Element-wise addition of 2 inputs (HW Layer)
31 'res4a_relu' ReLU ReLU (HW Layer)
32 'res4b_branch2a' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
33 'res4b_branch2a_relu' ReLU ReLU (HW Layer)
34 'res4b_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
35 'res4b' Addition Element-wise addition of 2 inputs (HW Layer)
36 'res4b_relu' ReLU ReLU (HW Layer)
37 'res5a_branch2a' 2-D Convolution 512 3×3×256 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
38 'res5a_branch2a_relu' ReLU ReLU (HW Layer)
39 'res5a_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
40 'res5a_branch1' 2-D Convolution 512 1×1×256 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
41 'res5a' Addition Element-wise addition of 2 inputs (HW Layer)
42 'res5a_relu' ReLU ReLU (HW Layer)
43 'res5b_branch2a' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
44 'res5b_branch2a_relu' ReLU ReLU (HW Layer)
45 'res5b_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
46 'res5b' Addition Element-wise addition of 2 inputs (HW Layer)
47 'res5b_relu' ReLU ReLU (HW Layer)
48 'pool5' 2-D Global Average Pooling 2-D global average pooling (HW Layer)
49 'fc1000' Fully Connected 1000 fully connected layer (HW Layer)
50 'prob' Softmax softmax (HW Layer)
51 'Output1_prob' Regression Output mean-squared-error (SW Layer)
Notice: The layer 'Output1_prob' with type 'nnet.cnn.layer.RegressionOutputLayer' is implemented in software.
Deep Learning Processor Estimator Performance Results
LastFrameLatency(cycles) LastFrameLatency(seconds) FramesNum Total Latency Frames/s
------------- ------------- --------- --------- ---------Network 19640680 0.09820 1 19640680 10.2 data_norm_add 268453 0.00134 data_norm 163081 0.00082 conv1 2164700 0.01082 pool1 515128 0.00258 res2a_branch2a 966477 0.00483 res2a_branch2b 966477 0.00483 res2a 268453 0.00134 res2b_branch2a 966477 0.00483 res2b_branch2b 966477 0.00483 res2b 268453 0.00134 res3a_branch1 541373 0.00271 res3a_branch2a 541261 0.00271 res3a_branch2b 920141 0.00460 res3a 134257 0.00067 res3b_branch2a 920141 0.00460 res3b_branch2b 920141 0.00460 res3b 134257 0.00067 res4a_branch1 505453 0.00253 res4a_branch2a 511309 0.00256 res4a_branch2b 909517 0.00455 res4a 67152 0.00034 res4b_branch2a 909517 0.00455 res4b_branch2b 909517 0.00455 res4b 67152 0.00034 res5a_branch1 515149 0.00258 res5a_branch2a 522317 0.00261 res5a_branch2b 956621 0.00478 res5a 33582 0.00017 res5b_branch2a 956621 0.00478 res5b_branch2b 956621 0.00478 res5b 33582 0.00017 pool5 55746 0.00028 fc1000 103850 0.00052 prob 1227 0.00001
- The clock frequency of the DL processor is: 200MHz
Note: Processing module "conv" property "SegmentationBlockGeneration" changed from "true" to "false".
Note: Processing module "fc" property "FCThreadNumber" changed from "4" to "8".
Note: Processing module "fc" property "WeightAXIDataBitwidth" changed from "128" to "256".
Note: Processing module "fc" property "SoftmaxBlockGeneration" changed from "false" to "true".
Processing Module "conv"
ModuleGeneration: 'on'
LRNBlockGeneration: 'off'
SegmentationBlockGeneration: 'off'
ConvThreadNumber: 16
InputMemorySize: [227 227 3]
OutputMemorySize: [227 227 3]
FeatureSizeLimit: 2048
Processing Module "fc"
ModuleGeneration: 'on'
SoftmaxBlockGeneration: 'on'
FCThreadNumber: 8
InputMemorySize: 25088
OutputMemorySize: 4096
Processing Module "custom"
ModuleGeneration: 'on'
Addition: 'on'
MishLayer: 'off'
Multiplication: 'on'
Resize2D: 'off'
Sigmoid: 'off'
SwishLayer: 'off'
TanhLayer: 'off'
InputMemorySize: 40
OutputMemorySize: 120
Processor Top Level Properties
RunTimeControl: 'register'
RunTimeStatus: 'register'
InputStreamControl: 'register'
OutputStreamControl: 'register'
SetupControl: 'register'
ProcessorDataType: 'single'
UseVendorLibrary: 'on'
System Level Properties
TargetPlatform: 'Xilinx Zynq UltraScale+ MPSoC ZCU102 Evaluation Kit'
TargetFrequency: 200
SynthesisTool: 'Xilinx Vivado'
ReferenceDesign: 'AXI-Stream DDR Memory Access : 3-AXIM'
SynthesisToolChipFamily: 'Zynq UltraScale+'
SynthesisToolDeviceName: 'xczu9eg-ffvb1156-2-e'
SynthesisToolPackageName: ''
SynthesisToolSpeedValue: ''Optimizing processor configuration for deep learning network complete.
Estimate performance of the ResNet-18 network by using the new optimized deep learning processor configuration.
estimatePerformance(hPC_optimized,net);
An output layer called 'Output1_prob' of type 'nnet.cnn.layer.RegressionOutputLayer' has been added to the provided network. This layer performs no operation during prediction and thus does not affect the output of the network.
Optimizing network: Fused 'nnet.cnn.layer.BatchNormalizationLayer' into 'nnet.cnn.layer.Convolution2DLayer'
Notice: The layer 'data' of type 'ImageInputLayer' is split into an image input layer 'data', an addition layer 'data_norm_add', and a multiplication layer 'data_norm' for hardware normalization.
The network includes the following layers:
1 'data' Image Input 224×224×3 images with 'zscore' normalization (SW Layer)
2 'conv1' 2-D Convolution 64 7×7×3 convolutions with stride [2 2] and padding [3 3 3 3] (HW Layer)
3 'conv1_relu' ReLU ReLU (HW Layer)
4 'pool1' 2-D Max Pooling 3×3 max pooling with stride [2 2] and padding [1 1 1 1] (HW Layer)
5 'res2a_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
6 'res2a_branch2a_relu' ReLU ReLU (HW Layer)
7 'res2a_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
8 'res2a' Addition Element-wise addition of 2 inputs (HW Layer)
9 'res2a_relu' ReLU ReLU (HW Layer)
10 'res2b_branch2a' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
11 'res2b_branch2a_relu' ReLU ReLU (HW Layer)
12 'res2b_branch2b' 2-D Convolution 64 3×3×64 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
13 'res2b' Addition Element-wise addition of 2 inputs (HW Layer)
14 'res2b_relu' ReLU ReLU (HW Layer)
15 'res3a_branch2a' 2-D Convolution 128 3×3×64 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
16 'res3a_branch2a_relu' ReLU ReLU (HW Layer)
17 'res3a_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
18 'res3a_branch1' 2-D Convolution 128 1×1×64 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
19 'res3a' Addition Element-wise addition of 2 inputs (HW Layer)
20 'res3a_relu' ReLU ReLU (HW Layer)
21 'res3b_branch2a' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
22 'res3b_branch2a_relu' ReLU ReLU (HW Layer)
23 'res3b_branch2b' 2-D Convolution 128 3×3×128 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
24 'res3b' Addition Element-wise addition of 2 inputs (HW Layer)
25 'res3b_relu' ReLU ReLU (HW Layer)
26 'res4a_branch2a' 2-D Convolution 256 3×3×128 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
27 'res4a_branch2a_relu' ReLU ReLU (HW Layer)
28 'res4a_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
29 'res4a_branch1' 2-D Convolution 256 1×1×128 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
30 'res4a' Addition Element-wise addition of 2 inputs (HW Layer)
31 'res4a_relu' ReLU ReLU (HW Layer)
32 'res4b_branch2a' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
33 'res4b_branch2a_relu' ReLU ReLU (HW Layer)
34 'res4b_branch2b' 2-D Convolution 256 3×3×256 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
35 'res4b' Addition Element-wise addition of 2 inputs (HW Layer)
36 'res4b_relu' ReLU ReLU (HW Layer)
37 'res5a_branch2a' 2-D Convolution 512 3×3×256 convolutions with stride [2 2] and padding [1 1 1 1] (HW Layer)
38 'res5a_branch2a_relu' ReLU ReLU (HW Layer)
39 'res5a_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
40 'res5a_branch1' 2-D Convolution 512 1×1×256 convolutions with stride [2 2] and padding [0 0 0 0] (HW Layer)
41 'res5a' Addition Element-wise addition of 2 inputs (HW Layer)
42 'res5a_relu' ReLU ReLU (HW Layer)
43 'res5b_branch2a' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
44 'res5b_branch2a_relu' ReLU ReLU (HW Layer)
45 'res5b_branch2b' 2-D Convolution 512 3×3×512 convolutions with stride [1 1] and padding [1 1 1 1] (HW Layer)
46 'res5b' Addition Element-wise addition of 2 inputs (HW Layer)
47 'res5b_relu' ReLU ReLU (HW Layer)
48 'pool5' 2-D Global Average Pooling 2-D global average pooling (HW Layer)
49 'fc1000' Fully Connected 1000 fully connected layer (HW Layer)
50 'prob' Softmax softmax (HW Layer)
51 'Output1_prob' Regression Output mean-squared-error (SW Layer)
Notice: The layer 'Output1_prob' with type 'nnet.cnn.layer.RegressionOutputLayer' is implemented in software.
Deep Learning Processor Estimator Performance Results
LastFrameLatency(cycles) LastFrameLatency(seconds) FramesNum Total Latency Frames/s
------------- ------------- --------- --------- ---------Network 19640680 0.09820 1 19640680 10.2 data_norm_add 268453 0.00134 data_norm 163081 0.00082 conv1 2164700 0.01082 pool1 515128 0.00258 res2a_branch2a 966477 0.00483 res2a_branch2b 966477 0.00483 res2a 268453 0.00134 res2b_branch2a 966477 0.00483 res2b_branch2b 966477 0.00483 res2b 268453 0.00134 res3a_branch1 541373 0.00271 res3a_branch2a 541261 0.00271 res3a_branch2b 920141 0.00460 res3a 134257 0.00067 res3b_branch2a 920141 0.00460 res3b_branch2b 920141 0.00460 res3b 134257 0.00067 res4a_branch1 505453 0.00253 res4a_branch2a 511309 0.00256 res4a_branch2b 909517 0.00455 res4a 67152 0.00034 res4b_branch2a 909517 0.00455 res4b_branch2b 909517 0.00455 res4b 67152 0.00034 res5a_branch1 515149 0.00258 res5a_branch2a 522317 0.00261 res5a_branch2b 956621 0.00478 res5a 33582 0.00017 res5b_branch2a 956621 0.00478 res5b_branch2b 956621 0.00478 res5b 33582 0.00017 pool5 55746 0.00028 fc1000 103850 0.00052 prob 1227 0.00001
- The clock frequency of the DL processor is: 200MHz
The new estimated frames per second performance is 10 frames per second.
This image shows the comparison between the original processor configuration and the optimized processor configuration:
The optimized processor configuration has:
SegmentationBlockGenerationturned off.SoftMaxBlockGenerationturned on.FCThreadNumberincreased to 8.
Generate Optimized Custom Bitstream
Use the optimized custom deep learning processor configuration to build and generate a custom bitstream. Use the custom bitstream to deploy the pretrained ResNet-18 network to your target FPGA board.
hdlsetuptoolpath('ToolName', 'Xilinx Vivado', 'ToolPath', 'C:\Xilinx\Vivado\2023.1\bin\vivado.bat'); dlhdl.buildProcessor(hPC_optimized);
Algorithms
InputMemorySize and OutputMemorySize Optimization
The optimizeConfigurationForNetwork method optimizes the deep learning processor configuration Conv module input and output memory size, by:
- Checking all the convolution layers in the network and scanning their input activation size. Because the deep learning processor uses the same
Convmodule on the FPGA to execute all the convolution layers in the network, theConvmoduleInputMemorySizeandOutputMemorySizemust have the memory to cover all the convolution layers in the network. - If the activation size of the largest convolution layer is less than the default value of 227-by-227-by-3,
optimizeConfigurationForNetworklowers theInputMemorySizeandOutputMemorySizeof theConvmodule. - If the activation size of the largest convolution layer is greater than the default value of 227-by-227-by-3,
optimizeConfigurationForNetworkdoes not modify theInputMemorySizeandOutputMemorySizeof theConvmodule.
The optimizeConfigurationForNetwork method optimizes the deep learning processor configuration FC module input and output memory size by:
- Checking all the fully connected layers in the network and scanning their activation size. Because Deep Learning HDL Toolbox™ does not support tile cutting on the FC layer, the activation size must fit within the
InputMemorySizeandOutputMemorySize. - If the input activation size of the largest fully connected layer is less than the default value of 25088,
optimizeConfigurationForNetworklowers theInputMemorySizeof theFCmodule. If the output activation size is less than the default value of 4096,optimizeConfigurationForNetworklowers theOutputMemorySizeof theFCmodule. - If the input activation size of the largest fully connected layer is greater than the default value of 25088,
optimizeConfigurationForNetworkincreases theInputMemorySizeof theFCmodule. If the output activation size is greater than the default value of 4096,optimizeConfigurationForNetworkincreases theOutputMemorySizeof theFCmodule.
The optimizeConfigurationForNetwork method does not modify theInputMemorySize and OutputMemorySize of the custom module of the deep learning processor configuration.
Version History
Introduced in R2021b
R2022b: Added FramesPerSecond name-value argument
You can specify the target frames per second for the network by using the optionalFramesPerSecond name-value argument.