User manual

Table Of Contents
Zynq-7000 AP SoC Technical Reference Manual www.xilinx.com 91
UG585 (v1.11) September 27, 2016
Chapter 3: Application Processing Unit
The SCU can also copy clean data from one processor cache to another and eliminate the need for
main memory accesses to perform this task. Furthermore, it can move dirty data between the
processors, skipping the shared state and avoiding the latency associated with the write-back.
IMPORTANT: It is important to note that the Cortex-A9 does not guarantee coherency between the L1
instructions caches as the processor is not capable of modifying the L1 contents directly.
3.3.2 Address Filtering
One of the functions of the SCU is to filter transactions that are generated by the processors and the
ACP based on their addresses and route them accordingly to the OCM or L2 controller. The
granularity of the address filtering within the SCU is 1 MB; therefore, all accesses by the processors
or through the ACP whose addresses are within a 1 MB window can only target the OCM or L2
controller. The default setting of the address filtering within the SCU routes all the upper and lower
1M addresses within the 4G address space to the OCM and the rest of the addresses are routed to
the L2 controller. Refer to the SCU Address Filtering section of Chapter 29, On-Chip Memory (OCM)
for more information on the SCU address filtering.
3.3.3 SCU Master Ports
Each of the SCU AXI master ports to the L2 or OCM has the following write and read issuing
capabilities:
Write issuing capability:
°
10 write transactions per processor:
- 8 non-cacheable writes
- 2 evictions from L1
°
2 additional writes for eviction traffic from the SCU
°
3 more write transactions from the ACP
Read issuing capability:
°
14 read transactions per processor:
- 4 instruction reads
- 6 linefill reads
- 4 non-cacheable read
°
7 more read transactions from the ACP