New AXI Scatter-Gather DMA Core Transfers Streaming Data to/from System Memory

Data processing accelerators are often designed with streaming interfaces and are meant for continuous flow through a custom hardware processing pipeline. Such cores include engines for compression, cryptography, or packet processing. A custom hardware pipeline built of such cores would be similar to this:

Data Processing architecture with Streaming

Sometimes, though, designers want the data processing accelerator to store or retrieve its data to or from the systems memory rather than directly transferring data to or from the next or preceding pipeline stage. So, our example custom hardware pipeline now transforms to something like this:  

Data processing memory storage architecture

 We’ve often received requests from customers to help them build such an architecture. We’re now happy to provide a solution with our new AXI4-SGDMA AXI4 to/from Stream Scatter-Gather DMA IP Core.

Two modes of the CAST AXI4-SGDMA IP Core

The AXI4-SGDMA gives you a quick way to implement a Direct Memory Access (DMA) engine that connects a core’s AXI4-Stream data interfaces to system memory through an AMBA® AXI4 bus matrix. It works in two ways:

  • As a Host-to-Peripherals (H2P) interface, it transfers data for processing from memory through the subject core’s AXI4-Stream Slave port, or
  • As a Peripheral-to-Host (P2H) interface, it moves processed data to system memory through the subject core’s AXI4-Stream Master port.

It is typically the host CPU that needs to orchestrate these data transfers, causing an extra load on system operation. The AXI4-SGDMA minimizes this load with its Scatter-Gather (SG) mode, in which it reads transfer descriptors from a memory-mapped location defined at run time. (It can also operate in Direct Mode, reading descriptors from local registers.)

The AXI4-SGDMA is remarkably configurable—both before synthesis and at run time—and includes numerous features that make it easy to integrate into a variety of systems (see the product page for details). 

An obvious choice for companion use with our GZIP data compression engines or our image and video codecs, the AXI4-SGDMA will also work well with our IP stacks and encryption cores. See below for an example system diagram.

Call Sales (+1 201.391.8300) or submit a Request Info form to talk with us about how the AXI4-SGDMA might save you time and headaches on your next system design.

CAST AXI4-SGDMA IP Cores in an Example System
CAST AXI4-SGDMA Streaming to.from memory scatter-gather DMA IP core