User Guide

Table Of Contents
www.ti.com
Switch
Switch
Endpoint
Endpoint
C0
C0
B0
B0
B2
B2
A1
A1
B1
B1
A0
A0
Open
Open
Open
Open
Open
Open
Open
Full
Open
Open
Full
Full
Retry
Retry
Retry
Retry
Retry
Retry
Accept
Retry
Retry
Retry
Action
Action
Retry
Retry
Scenario A - Default
Scenario B - In order mode
Data flow destined for the
same Rx Queue
Rx Queue Status when
packet arrives
Rx Queue Status when
packet arrives
Records SourceID/letter of
first retry packet
SRIO Functional Description
Figure 20. RX CPPI Mode Explanation
In addition, multiple messages can be interleaved at the receive port due to ordering within a connected
switch’s output queue. This can occur when using a single or multiple priorities. The RX CPPI block must
handle simultaneous interleaved multi-segment messages. This implies that state information (write
pointers and srcID) must be maintained on each simultaneous message to properly store the segments in
memory. The number of simultaneous transactions supported directly impacts the number of states to be
stored, and the size of the buffer descriptor memory outside the peripheral. With this in mind, the
peripheral’s supported buffer descriptor SRAM is parameterizable. A minimum size of 1.25KB is
recommended, which will allow up to 64 buffer descriptors to be stored at any given time for one core.
These buffer descriptors can be configured to support any combination of single and multi-segment
messages. For example, if the application only handles single-segment messages, all 64 buffers can be
allotted to that queue. Note that a given RX queue can contain packets of all priorities which have been
directed from any of the receive ports.
A CPU may wish to stop receiving messages and reclaim buffers belonging to a specific queue. This is
called queue teardown. The CPU initiates a RX queue teardown by writing to the RX Queue Teardown
command register.
SPRU976 March 2006 Serial RapidIO (SRIO) 47
Submit Documentation Feedback