Deep Learning Processor Customization and IP Generation

Configure, build, and generate custom bitstreams and processor IP cores, estimate and benchmark custom deep learning processor performance

Deep Learning HDL Toolbox™ provides functions to configure, build, and generate custom bitstreams and a custom processor IP. Obtain performance and resource utilization of a pretrained series network on the custom processor. Optimize the custom processor by using the estimation results.

Classes

expand all

Processor Configuration

dlhdl.ProcessorConfig Configure custom deep learning processor (Since R2020b)

Objects

`dlhdl.layer.mishLayer`	Mish layer (Since R2024a)
`dlhdl.layer.reshapeLayer`	Layer that reshapes activation data (Since R2024b)
`dlhdl.layer.sliceLayer`	Slice layer (Since R2024a)

Functions

expand all

Build Bitstream

dlhdl.buildProcessor Build and generate custom processor IP (Since R2020b)

Custom Board Calibration

`buildCalibrationBitstream`	Generate calibration bitstream to obtain double data rate (DDR) RAM response time (Since R2022b)
`deployCalibrationBitstream`	Deploy calibration bitstream and generate calibration data file (Since R2022b)

Performance and Resource Estimation

`estimatePerformance`	Retrieve layer-level latencies and performance by using `estimatePerformance` method (Since R2021a)
`estimateResources`	Return estimated resources used by custom bitstream configuration (Since R2021a)
`optimizeConfigurationForNetwork`	Update network-specific deep learning processor configuration with optimized deep learning processor configuration (Since R2021b)

Processor Configuration

`getModuleProperty`	Use the `getModuleProperty` method to get values of module properties within the `dlhdl.ProcessorConfig` object (Since R2020b)
`setModuleProperty`	Use the `setModuleProperty` method to set properties of modules within the `dlhdl.ProcessorConfig` object (Since R2020b)

Custom Layer Support

`openCustomLayerModel`	Open a generated custom layer verification model to verify your custom layers (Since R2022a)
`registerCustomLayer`	Register the custom layer definition and Simulink model representation of the custom layer (Since R2022a)
`verifyCustomLayerModel`	Verify the functionality and accuracy of the custom layer by using the generated custom layer verification model (Since R2022a)

Topics

Custom Processor Configuration

Custom Processor Configuration Workflow
Accelerate the estimation and optimization of custom deep learning processor by configuring parameters of the conv processor and fc processor, created by using the dlhdl.ProcessorConfig object workflow.
Deep Learning Processor IP Core Architecture
Learn about the FPGA architecture based custom deep learning processor architecture and using it to create a MATLAB^® controlled deep learning processor.
Estimate Performance of Deep Learning Network
Analyze the deep learning network layer level latencies and overall performance before deployment.
Estimate Resource Utilization for Custom Processor Configuration
Expedite the time to identify a target hardware board that meets resource utilization budgets before deployment.
Effects of Custom Deep Learning Processor Parameters on Performance and Resource Utilization
Rapidly prototype custom processor configuration and networks by understanding how deep learning processor parameters affect resource utilization and network performance.
Generate Custom Bitstream to Meet Custom Deep Learning Network Requirements
Deploy your custom network that only has layers with the convolution module output format or only layers with the fully connected module output format by generating a resource optimized custom bitstream that satisfies your performance and resource requirements.
Create Deep Learning Processor Configuration for Custom Layers
Create a deep learning processor configuration that includes your custom layers.

Custom Processor Code Generation

Generate Custom Bitstream
Rapidly prototype and iterate custom deep learning networks performance by configuring, building and generating custom bitstreams which can then be deployed to target FPGA and SoC boards.
Generate Custom Processor IP
Build and generate IP for the dlhdl.ProcessorConfig.

Featured Examples

Custom Deep Learning Processor Generation to Meet Performance Requirements

Create a custom processor configuration and estimate the performance of a pretrained series network. You can then modify parameters of the custom processor configuration and re-estimate the performance. Once you have achieved your performance requirements you can generate a custom bitstream by using the custom processor configuration.

Open Live Script

Customize Bitstream Configuration to Meet Resource Use Requirements

Deploy a digit recognition network with a target performance of 500 frames per second (FPS) to a Xilinx™ ZCU102 ZU4CG device. The target device resource counts are:

Open Live Script

Estimate Resource Utilization for Custom Board and Reference Design

Rapidly prototype the deployment of deep learning networks to your custom board by using the estimateResources function. Estimate the resource utilization of the deep learning processor configuration for your custom board. Optimize the integration of custom IP cores and reference design into your system by using the estimateResources function to estimate the resource utilization of your reference design. The synthesis tool that you use must be in the list of tools supported by the SynthesisTool property of the dlhdl.ProcessorConfig object. For a list of supported tools and device families, see SynthesisTool and SynthesisToolChipFamily.

Open Live Script

Register, Validate, and Deploy Custom Natural Logarithm Layer Network to FPGA

Register, validate, and deploy a custom natural logarithm (log) layer network by using Deep Learning HDL Toolbox™. To deploy the network with the custom natural logarithm (log) layer:

Open Live Script

Optimize Deep Learning Processor Configuration for Network Performance

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.

Open Live Script

Deep Learning Processor IP Core Generation for Custom Board

Create custom board and generate a deep learning processor IP core for the custom board. In this example you:

Open Live Script

Use AXI-Based BRAM IP to Remove DDR Requirement

Deploy a deep learning network to an FPGA which does not have external DDR memory.

Since R2024a
Open Live Script