User manual

Table Of Contents
Zynq-7000 AP SoC Technical Reference Manual www.xilinx.com 655
UG585 (v1.11) September 27, 2016
Chapter 22: Programmable Logic Design Guide
22.4 Choosing a Programmable Logic Interface
This section discusses various options to connecting Programmable Logic (PL) to the Processing
System (PS). The main emphasis is on data movement tasks such as direct memory access (DMA).
22.4.1 PL Interface Comparison Summary
Table 22-8 presents a qualitative overview of data transfer use cases. The estimated throughput
column reflects suggested maximum throughput in a single direction (read/write).
22.4.2 Cortex-A9 CPU via General Purpose Masters
The least intrusive method from a software perspective is to use the Cortex-A9 to move data
between the PS and PL (see Figure 22-1). Data flow is directly moved by a CPU, removing the need
to handle events from a separate DMA. Access to the PL is provided through the two M_AXI_GP
master ports, which each have a memory address range to originate PL AXI transactions. The PL
design is also simplified since as little as a single AXI slave can be implemented to service the CPU
requests.
Table 22-8: Data Movement Method Comparison Summary
Method Benefits Drawbacks Suggested Uses
Estimated
Throughput
CPU Programmed I/O Simple Software
Least PL Resources
Simple PL Slaves
Lowest Throughput Control Functions <25 MB/s
PS DMAC Least PL Resources
Medium Throughput
Multiple Channels
Simple PL Slaves
•Somewhat complex
DMA programming
Limited PL
Resource DMAs
600 MB/s
PL AXI_HP DMA Highest Throughput
•Multiple Interfaces
Command/Data FIFOs
OCM/DDR access only
•More complex PL
Master design
High Performance
DMA for large
datasets
1,200 MB/s
(per interface)
PL AXI_ACP DMA Highest Throughput
Lowest Latency
Optional Cache
Coherency
•Large burst might cause
cache thrashing
•Shares CPU
Interconnect bandwidth
•More complex PL
Master design
High Performance
DMA for smaller,
coherent datasets
•Medium
granularity CPU
offload
1,200 MB/s
PL AXI_GP DMA Medium Throughput More complex PL
Master design
•PL to PS Control
Functions
•PS I/O Peripheral
Access
600 MB/s