Main Content

registerDeepLearningMemoryAddressSpace

Class: hdlcoder.ReferenceDesign
Package: hdlcoder

Add memory address space to reference design

Syntax

registerDeepLearningMemoryAddressSpace(baseAddr,addrRange)

Description

registerDeepLearningMemoryAddressSpace(baseAddr,addrRange) registers memory address space accessible by the deep learning processor IP core to your reference design. The deep learning processor IP core uses this registered memory space to store the inputs to the deep learning network, network weights, and intermediate computations performed by the deep learning processor.

Input Arguments

expand all

Base address of the deep learning memory address space, in bytes as uint32.

Example: 0x80000000

Address range of the deep learning memory address space, in bytes as uint32.

Example: 0x40000000

Requirements

The registered memory address space must be accessible by the AXI4 Master interfaces in your reference design.

Tips

  • A minimum of 32 MB of memory address space is required to run any deep learning network. To run a variety of deep learning networks, it is recommended to register a minimum of 512 MB of memory address space.

  • The compile method of the dlhdl.Workflow object generates the memory address space for a deep learning network. See compile. Determine the memory address space for your deep learning network by using the compile method of the dlhdl.Workflow object.

Introduced in R2021b