![](http://datasheet.mmic.net.cn/330000/PEF20321_datasheet_16444041/PEF20321_146.png)
PEB 20321
PEF 20321
Local Bus Interface (LBI)
Data Sheet
146
2001-02-14
6.1.1
Standard PCI Slave transactions are used when the PCI host system communicates with
non-intelligent LBI peripherals. The MUNICH32X handles 8-, 16- or 32-bit PCI
transactions and translates them into the corresponding local bus transactions, e.g. a 32-
bit PCI transaction results in two 16-bit local bus transactions when the local bus
interface is configured for 16-bit operations.
For reads, a PCI Retry sequence of operations is performed, in which the MUNICH32X
will immediately terminate the PCI transaction (and request a retry) until it terminates the
transaction to the LBI. The MUNICH32X uses the retry procedure because the time to
complete the data phase will require more than the maximum allowed 16 PCI clocks
(from the assertion of FRAME to the completion of the first data phase). Data transfer
will be successfully completed within a PCI retry cycle. The number of necessary PCI
retry cycles depend on PCI arbitration behavior and the time it needs to terminate the
transaction on the local bus; PCI TRDY wait states will not be added for the sequential
retry read cycles unless the LBI arbitration time is excessive.
For write transactions, the MUNICH32X will store a single data DWORD and then
immediately terminate the PCI transaction successfully. It will then arbitrate the local bus
and perform the write transaction after being granted depending on the selected number
of wait states and LRDY bus control signal.
Thus write accesses to LBI are performed as
‘
posted write
’
transactions from the PCI
view. A consecutive write transaction results in PCI retry cycles in the case that the
preceding write transaction is not yet finished on LBI.
Note that the MUNICH32X performs single word PCI Slave read or write transactions
only; Slave burst transactions to LBI are not supported.
Transactions with Non-intelligent Peripherals
6.1.2
The MUNICH32X uses an
‘
exclusive-access
’
Mailbox Command Register MBCMD to
control the transfer of information between the PCI host system and an intelligent LBI
peripheral (e.g., a CPU). The PCI host system always reads the contents that was
written to Mailbox Command Register by the LBI peripheral, while the intelligent LBI
peripheral always reads the contents that was written to Mailbox Command Register by
the PCI host system.
As an
example
, consider when the PCI host system wants to transfer data to an
intelligent LBI peripheral. First, assuming it has
‘
ownership
’
of the Mailbox registers, it
loads data into the Mailbox Data Registers, and then writes a
‘
1
’
to INPCI bit field of the
Mailbox Command Register. This last action causes the LINTO output signal to become
asserted, indicating to the intelligent LBI peripheral that data is ready.
The intelligent LBI peripheral will read Mailbox Command Register (which deasserts the
LINTO output signal and resets the INPCI bit field of Mailbox Command Register), and
then reads the data from the Mailbox Data Registers. Finally, it writes a
‘
1
’
to the INLBI
Transactions with Intelligent Peripherals