3. Processor Bus Interface
95
PowerSpan II User Manual
80A1010_MA001_09
Integrated Device Technology
www.idt.com
3.3.2.3
Cache Line Size
The supported embedded PowerPC processors implement a 32-byte cache line size. Cache wrap reads
are supported by the PB slave for burst and extended transactions.
3.3.2.4
Reads
Address Retry Enable
The PB slave supports up to eight concurrent delayed reads when the Address Retry Enable
(ARTRY_EN) bit in the
“Processor Bus Miscellaneous Control and Status Register” on page 304
is set
for more information on read pipelining in PowerSpan II.
When an external master makes an initial read
request, the PowerSpan II PB slave latches the address.
This initiates a read on the destination bus. The destination bus is specified by the Destination Bus
(DEST) bit in the
“Processor Bus Slave Image x Control Register” on page 287
Delayed Reads
The outstanding read is referred to as a delayed read. Delayed reads consist of the following phases:
1.
Delayed Read Request
— PowerSpan II PB Slave latches transaction parameters and issues a retry
2.
Delayed Read Completion
— The PB Slave obtains the requested data and completion status on the destination bus
3.
Read Completion
— The master repeats the transaction with the same parameters used for the initial request
Any attempt by a processor bus master to complete the read
transaction is retried by the PowerSpan II
PB Slave until the following byte quantities are available in the line buffer:
•
32 bytes
•
8 bytes if the RD_AMT=0 (see
“Processor Bus Slave Image x Control Register” on page 287
•
16 bytes if the RD_AMT=1
Read Amount
All PowerSpan II PB slave reads destined for PCI Memory space are considered prefetchable to 8-byte
boundaries by default. Setting the MEM_IO bit in the
“Processor Bus Slave Image x Control Register”
enables 1,2,3, or 4 byte reads from the PCI bus(es).
PowerPC processors do not generate cache wrap writes.