| Electronic Components Datasheet Search |
|
DP83266 Datasheet(PDF) 34 Page - National Semiconductor (TI) |
|
|
|
|||||||||||||||||||||||||||||
DP83266 Datasheet(HTML) 34 Page - National Semiconductor (TI) |
|
34 / 152 page ![]() 60 Functional Description (Service Engine) (Continued) current transaction Note that Bus Request is guaranteed to be de-asserted for at least two cycles when MR1EAM is enabled It is a requirement of the SBus specification that Bus Request be de-asserted between each transaction If Bus Grant is (re)asserted before the end of the current transaction the MACSI device retains mastership and runs the next transaction This process may be repeated indefi- nitely It is important to note that in default ABus mode (MR1EAM e 0) the MACSI device may take up to two cycles to detect the assertion of Bus Grant Therefore the external interface logic must not remove Bus Grant or latch addresses until a signal such as Address Strobe is asserted indicating that the MACSI device has recognized the Bus Grant If the Bus Arbiter wishes to preempt the MACSI device it deasserts Bus Grant The MACSI device will complete the current bus transaction then release the bus From preemp- tion to bus release is a maximum of (11 bus clocks a (8 times the number of memory wait states)) bus clocks For example in a 1 wait-state system the MACSI device will release the bus within a maximum of 19 bus clocks Parity The state of the FLOW bit in System Interface Mode Regis- ter 0 (SIMR0FLOW) controls two MACSI device modes one for systems using parity the other for systems not using parity Regardless of the state of SIMR0FLOW the MACSI device always provides odd parity on ABus address cycles and Control Bus read cycles The MACSI device System Inter- face does not check data parity on ABus read cycles The User may enable checking of read data parity at the Ring Engine interface by setting the MAC Request Parity bit (MCMR0MRP) For MACSI revision D or later the System Interface will check parity on descriptor reads (when SIMR0FLOW e 1) and report errors via the STARERR bit The ABus data parity (except for Descriptor data parity which is generated internally) flows directly from the Ring Engine interface If parity checking is enabled within the Ring Engine parity flows up from the PLAYERa interface If parity checking is disabled within the Ring Engine good parity is generated at the Ring EngineService Engine inter- face For transmit frame data the parity from the ABus flows di- rectly to the Ring Engine interface when SIMR0FLOW is asserted The data integrity across the ABus and through the Service Engine may be checked by enabling parity checking within the Ring Engine If the SIMR0FLOW bit is deasserted the MACSI device will generate good parity at the transmit data (MR7 – 0) Ring Engine interface When SIMR0FLOW is enabled parity is checked on Con- trol Bus write operations If a parity error is detected the write access is rejected and the STARCPE bit is asserted When SIMR0FLOW is disabled Control Bus write parity is ignored When SIMR0FLOW is enabled parity is checked on MAC Indicate interface (receive data from the Ring Engine) If an error is detected the STARBPE will be asserted and the IDUD for that frame will carry a status of ‘‘parity error’’ When SIMR0FLOW is disabled the parity bit at the MAC Indicate interface is ignored Bandwidth The ABus supports single reads and writes and burst reads and writes With physical addressing back-to-back single readswrites can take place every four clock cycles Burst transactions can transfer 8 32-bit words (32 bytes) every 11 clock cycles With a 25 MHz clock this yields a peak band- width of 727 Mbytessec To allow the bus to operate at high frequency the protocol defines all signals to be both asserted and deasserted by the bus master and slaves Having a bus device actively deassert a signal guarantees a high-speed inactive tran- sition If this were not defined external pull-up resistors would not be able to deassert signals fast enough The pro- tocol also reduces contention by avoiding cases where two bus devices simultaneously drive the same line The MACSI device operates synchronously with the ABus clock In general operations will be asynchronous to the ring since most applications will use a system bus clock that is asynchronous to the ring The MACSI device is de- signed to interface either directly to the host’s main system bus or to external shared memory When interfaced to the host’s bus there are two parameters of critical interest la- tency and bandwidth Data moves between the Request and Indicate Channels and the ABus via four FIFOs two in the receive path (Indi- cate) and two in the transmit path (Request) There are two 1152 x 36-bit data FIFOs for Indicate and Request data (4608 byte of data FIFO in each direction) On the ABus Interface there are two Burst FIFOs each containing two banks of 32 bytes which provide ABus bursting capability The amount of latency covered by the Data FIFO plus one of the banks of the Burst FIFO must meet the average and maximum bus latencies of the external memory With a new byte every 80 ns from the ring a 4608 byte FIFO provides 4608 c 80 ns e 36864 ms maximum latency The two 32 byte burst FIFO in each direction provide an additional 51 ms of latency To assist latency issues the MACSI device can completely empty or fill the Burst FIFO in one bus tenure by asserting Bus Request for multiple transactions Since one bank of the Burst FIFO is 8 words deep if 8-word bursts are en- abled that half of the Burst FIFO can be emptied in one transaction If the second half of the burst FIFO is also full it can be emptied in the same bus tenure by again granting the bus to the MACSI device The MACSI device may be preempted at any time by remov- ing Bus Grant causing the MACSI device to complete the current transaction and release the bus There will be a maximum of 11 clocks (plus any memory wait states) from preemption to bus release (fewer if 8-word bursts are not enabled) 642 Bus States An ABus Master has eight states idle (Ti) bus request (Tbr) virtual address (Tva) MMU translate (Tmmu) physical address (Tpa) data transfer (Td) wait (Tw) and recovery (Tr) An ABus Slave has five states idle (Ti) selected (Ts) data transfer (Td) wait (Tw) and recovery (Tr) 34 |
|
Link URL |
| Does ALLDATASHEET help your business so far? [ DONATE ] |
About Alldatasheet | Advertisement | Contact us | Privacy Policy | Link to Datasheet | Link Exchange | Manufacturer List All Rights Reserved©Alldatasheet.com |
| Russian : Alldatasheetru.com | Korean : Alldatasheet.co.kr | Spanish : Alldatasheet.es | French : Alldatasheet.fr | Italian : Alldatasheetit.com Portuguese : Alldatasheetpt.com | Polish : Alldatasheet.pl | Vietnamese : Alldatasheet.vn Indian : Alldatasheet.in | Mexican : Alldatasheet.com.mx | British : Alldatasheet.co.uk | New Zealand : Alldatasheet.co.nz |
|
Family Site : ic2ic.com |
icmetro.com |