View DS726_8897117.PDF datasheet online --- IC-ON-LINE

Datasheet File OCR Text:

DS726 april 24, 2009 www.xilinx.com 1 product specification ? 2007- 2009 xilinx, inc. xilinx, the xilinx logo, virtex, spartan, ise and other designated brands included herein are tradema rks of xilinx in the united states and other countries. all other trademarks are the property of their respective owners. introduction the on-chip peripheral bus (opb) to processor local bus (plb v4.6) bridge module translates opb transactions into plbv46 transactions. it functions as a slave on the opb side and a master on the plbv46 side. the opb to plbv46 bridge is necessary in systems where an opb master device, such as a dma engine or an opb based coprocessor, requires access to high performance system resources on the plbv46 bus. the xilinx? opb to plbv46 bridge design allows customers to tailor the bridge to suit their application by setting certain parameters to enable and disable features. the parameterizable features of the design are discussed in design parameters . features the xilinx opb to plbv46 bridge is a soft ip core with the following features: ?bridge functions - uses 16-word deep posted write buffer to decouple opb and plbv46 transactions. - uses 16-word deep read prefetch buffer to eliminate bridge related system lockup issues. ? plbv46 master interface - 32-bit native device width - communicates with 32-bit, 64-bit, and 128-bit plbv46 slaves - non-burst transfers of 1 to 4 bytes - uses fixed length, burst signaling of up to 16, 32-bit words. 0 opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 00 product specification logicore? ip facts core specifics see edk supported device families . version of core opb_plbv46_bridge v1_01_a resources used min max i/o 390 390 luts 467 898 ffs 630 720 block rams 0 0 provided with core documentation product specification design file formats vhdl constraints file n/a verification n/a instantiation te m p l a t e n/a reference designs none design tool requirements xilinx implementation tools see tools for requirements. verification simulation synthesis support provided by xilinx, inc.
opb to plbv46 bridge (v1.01a) 2 www.xilinx.com DS726 april 24, 2009 product specification features (contd) ? opb slave interface - 32-bit opb slave interface that responds to byte enable transfers only. (does not support dynamic bus sizing or non-byte enable transactions.) - decodes up to four separate address ranges - plbv46 and opb clock periods may have a 1:1 or 1:2 synchronous relationship. - utilizes read prefetch and opb retries to elim inate deadlock and increase plb bus performance. ? utilizes post write buffer to improve performance. functional description overview figure 1 provides a high-level overview of the opb to plbv46 bridge. opb transactions are received and decoded in the opb slave and data sent or received to or from the appropriate buffer. the bridge controls the operatio n of the slave and implements the read prefetching and posted writes. as a result the bridge effectiv ely de-couples the opb and plb buses to improve the plb performance and eliminate the typical read lockup potential. clocking the bridge provides for a plbv46:opb clock period ratio of 1:1 or 1:2. the bridge implementation requires that the clocks be generated by one dcm. th is insures that the rising edges of the plbv46 and opb clocks are aligned and that th e necessary and proper period constraint is applied to signals that cross time domain boundaries. figure top x-ref 1 figure 1: opb plbv46 bridge in 1:2 clock ratio configuration pl b v46_m as ter_ bu r s t_v1_00_ a 3 2- b it n a tive d a t a width op b _ s l a ve opb clock dom a in d s 404_01_09170 8 opb plb clock dom a in plbv46 clock dom a in 16 x 3 2 fifo 16 x 3 2 fifo bridge plbv46
opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 www.xilinx.com 3 product specification deadlock prevention deadlock can occur when masters request their bridge s to attempt read transactions at the same time and the opb slave on the opb to plb bridge utilizes timeout suppression. it starts when the plb to opb bridge addracks a read transaction before it knows the opb will be busy, thus tying up the plb read bus. (there is no timeout on the plbv46 once the address phase completes.) simultaneously, an opb master connects to the opb to plb bridge slave for a read and uses timeout suppression to block further access to the opb until its read completes. however, its read will never complete because the plb read bus is locked by the other bridge. the read attempts result in total locking of both buses. the solution used by the opb to plb bridge involv es decoupling the buses through the use of a posted write buffer and a read prefetch buffer. with this solution the system does not require the use of opb timeout suppression and no inter-bridge communication is needed to eliminate the potential for deadlock. plbv46 master burst io signals table 1: plbv46 master burst io signal description signal name interface signal type init status description plb clock and reset mplb_clk plb bus i plb main bus clock. see table note 1 . mplb_rst plb bus i plb main bus reset. see table note 1 . other system signal md_error plb bus o ?0? master detected error status output plb request and qualifier signals m_request plb bus o ?0? see table note 2. m_priority plb bus o ?0? m_buslock plb bus o ?0? m_rnw plb bus o ?0? m_be(0:[ c_mplb_dwidth /8]-1) plb bus o zeros m_msize(0:1) plb bus o "00" m_size(0:3) plb bus o "0000" m_type(0:2) plb bus o "000" m_abus(0: 31 )plb busozeros m_wrburst plb bus o ?0? m_rdburst plb bus o ?0? m_wrdbus(0: c_mplb_dwidth -1) plb bus o zeros plb reply signals
opb to plbv46 bridge (v1.01a) 4 www.xilinx.com DS726 april 24, 2009 product specification plb_mssize(0:1) plb bus i unused see table note 2 plb_maddrack plb bus i see table note 1 . plb_mrearbitrate plb bus i plb_mtimeout plb bus i plb_mrderr plb bus i plb_mwrerr plb bus i plb_mrddbus(0: c_mplb_dwidth -1) plb bus i plb_mrddack plb bus i plb_mwrdack plb bus i plb_rdbterm plb bus i plb_mwrbterm plb bus i plb signal ports included in th e design, but unused internally m_tattribute(0 to 15) plb bus o ?0? unused. see table note 2 . m_lockerr plb bus o ?0? m_abort plb bus o ?0? m_uabus(0: 31) )plb busozeros plb_mbusy plb bus i plb_mirq plb bus i plb_rdwdaddr(0:3) plb bus i opb signals opb_select opb i slave select opb_rnw opb i read=1, write=0 opb_be(0:3) opb i byte enables opb_bexfer opb i unused opb_hwxfer opb i unused opb_fwxfer opb i unused opb_dwxfer opb i unused opb_seqaddr opb i sequential address (burst) indication opb_abus(0:31) opb i transaction address opb_dbus(0:31) opb i write data arriving from the bus. table 1: plbv46 master burst io signal description (contd) signal name interface signal type init status description
opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 www.xilinx.com 5 product specification opb slave interface single transaction bridging the opb slave must complete a transaction before it will accept a new read or write transaction. it responds with the assertion of sl_retry to each master request until the previous transaction is successfully bridged . a new read is a transaction to an address th at is different than the one currently being prefetched. every write is considered to be a new write irrespective of address. successfully bridged means the plbv46_master_burst had an opportunity at delivering the transaction to the plb slave and returned a status of success or failure. sl_toutsup is not used to suppress the opb bus timeout while the bridge attempts to perform a read transaction. for write transactions, a successfully bridged transaction is complete when the posted write buffer has been emptied by the plbv46_master_burst. for read transactions, a successfully bridged transactio n is complete after the read prefetch is satisfied (either with data or an error indication), a new requ est with an address satisf ying the original prefetch address matches, and the prefetch data is used partially or fully to satisfy the request. if the prefetch buffer address does not receive a match in a specified time period, a prefetch match timeout error occurs, which results in a flush of the prefetch buffer (and any error status) and a return to accepting transactions. 1 sl_xferack opb o slave transfer acknowledge sl_beack opb o byte enable transfer acknowledge sl_errack opb o error acknowledge sl_retry opb o asserted high to force master off the bus. sl_dbus(0:31) opb o read data sent back to bus sl_toutsup opb o time out suppress opb system signals opb_clk opb i opb clock domain opb_rst opb i opb slave reset. notes: 1. this function and timing of this signal is defined in the ibm 128-bit processor local bus architecture specification version 4.6. 2. output ports that are not used are dr iven to constant logic levels that are consistent with the inactive state for the subject signal. input ports that are requir ed but not used are internally ignored by the design. 3. for fixed length burst requests, the starting address for the request as specified by the ip2bus_mst_addr(0:31) input must be aligned on an address boundary matching the c_mplb_native_dwidth value. 1. the user must set the c_prefetch_timeout parameter to a value that balances between stalling access to the bridge and thrashing the read prefetch buffer. table 1: plbv46 master burst io signal description (contd) signal name interface signal type init status description
opb to plbv46 bridge (v1.01a) 6 www.xilinx.com DS726 april 24, 2009 product specification the bridge does not support byte or halfword burs ting on the opb bus. bursts must start on a word aligned address (address bits 30 to 31 = ?00), end on a word aligned address, must and contain only full word data. the bridge drives valid byte enables onto the plbv46 only when it detects a single write or single read. it does this by examining opb_seqaddr . when opb_seqaddr =0, it assumes a single transaction with byte enables. xilinx opb masters that use the xi linx ipif are known to follow this assumption. otherwise a plbv46 burst is used and the opb_be() signal is ignored. address decode cycle opb transactions begin with an address decode cycle. a design parameter allows the user to specify the number of address ranges the bridge will respond to. each range has two parameters, c_rngn_baseaddr and c_rngn_highaddr, that specify the 32-bit lower and upper boundary for that range. these parameters define that portio n of the total system address space to which the opb slave will respond. write transactions when an opb master requests write access to the br idge, the slave immediately accepts from 1 to 16 full words of data, then buffers the data. the bridge will not accept write data that is addressed beyond the end of an address range, even if the ranges are back-to-back. for a single word write, indicated by the deassertion by the master of opb_seqaddr , the slave captures the byte enable pattern, opb_abus(0:29) , and sl_xferacks the word. if opb_seqaddr =1, then the slave counts the number of words written to the buffer to provide the plbv46_master_burst with the fixed length for a write burst on the plb bus. byte enables are ignore d for plb burst operations, therefore the opb master should ensure that all byte enables are asserted high when opb_seqaddr is asserted high. after signaling a write request to the plbv46_master_bu rst (of either a single or burst), the opb slave blocks (by issuing retries) until it receives co nfirmation of the plb operation complete status. the nature of posted writes prohibits the return of an plb transaction failure to the opb master that originated the request. any error status is therefore lost.
opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 www.xilinx.com 7 product specification read transactions on receipt of the opb master read transaction in dication, the opb slave captures the address and opb seqaddr qualifier, then asserts sl_retry to force the master off the bus. it then blocks any further requests using sl_retry . simultaneously it makes a fixed size read request to the plbv46_master_burst of 1 or 16 words 1 . the opb only has an indeterminate burst operation, therefore the bridge must over read to fulfill the future transaction retry that will claim the data. when the original master 2 (that initiated the prefetch) returns to bus, it will gain access to the slave by presenting a matching address to the original prefetch request. the slave delivers data to the master until it deasserts the opb_select signal (ending the transaction normally) or the prefetch buffer empties. the buffer might empty early (and contai n < 16 words) if a 16-word prefetch would have accessed data beyond the end of the address range. the result is that all data up to the end of the address range would be transfer-acknowledged to the opb master, but nothing beyond that. if the plbv46_master_burst signals an error at the comp letion of the read prefetch attempt, the slave will capture this status and clear the prefetch buffer . when the original master returns to claim the prefetch data, it will receive an opb_errack assertion. per opb protocol, the opb_errack asserts concurrently with the opb_xferack . upon the selection of the slave, the master will see a continuous stream of opb_xferacks until deselection. multiple read prefetches may be required for long read bursts. opb masters must be careful when performing read bursts at address locations which have read side effects because of the prefetching feature (coherency or destructive read problems). single beat reads should be used when accessing any special memory locations, such as peripherals that destroy the contents of a register when it is read. the prefetch timeout counter (of width determ ined by c_prefetch_timeout) starts counting down as soon as the plbv46_master_burst has returned valid data (or an error) to the bridge. the opb master has until the timer expires to retrieve all of the data. if the timer expires in the middle of a transaction, it will clear the prefet ch buffer without returning an error. plbv46 interface the plbv46_master_burst_v1_00_a pcore services bridge requests for access to the plb. the bridge does not utilize the pl bv46_master_burst bus lock feature. reset the user must ensure that both sides of the brid ge are reset simultaneously with overlapping reset signals. the bridge is not designed to recover from independently applied resets. 1. when opb_seqaddr =0 the bridge knows the explicit size of the read request is 1. in all other cases, the read length is unknown and the bridge resorts to reading in16, 32-bit word chunks. 2. it is important to recognize that the original master may not be the one that gets the re peated transaction. if two masters want to read from the same address, then the first may kick off the read prefetch, but the second may actually receive the data upon retry. the bridge has no way to qualify the address with the master that initiated the request. this condition is extremely unlikely. howeve r, no harm should result because further read attempts by the first master would simply result in a brand new prefetch.
opb to plbv46 bridge (v1.01a) 8 www.xilinx.com DS726 april 24, 2009 product specification design parameters allowable parameter combinations the current implementation of the plbv46 master burst has the following restrictions which apply to parameter value settings: ? the assigned value for c_mplb_awidth is currently restricted to 32. ? the assigned value for c_mplb_native_dwidth is currently restricted to 32. table 2: bridge design parameters feature/description parameter name allowable values default values vhdl type decoder address range definition number of address ranges c_ num_addr_rng 1-4 1 integer address range definition base address c_rngn_baseaddr (0 <= n <= 3) 0x 00000000 to 0x ffffffff x" ffffffff " std_logic_ vector address range definition high address c_rngn_highaddr (0<=n<=3) 0x 00000000 to 0x ffffffff x" 00000000 " std_logic_ vector bridge configuration establishes the ratio of plb to opb bus clock periods. the clocks must be synchronous with minimal phase difference. c_bus_clock_ period_ratio 1=1:1, 2=1:2 1 integer specifies the width of the timeout counter that deter mines the amount of time (in plbv46 clocks) the bridge waits for a master to retrieve all the read prefetch data before the prefetch buffer is flushed and new transactions are accepted again. c_prefetch _timeout 5-32 10 integer plb i/o specification specifies the number of used address bits out of the available 64 bits of plbv46 addressing c_mplb_awidth 32 32 integer width of the plb data bus to which the master is attached c_mplb_dwidth 32, 64, 128 32 integer specifies the internal native data width of the master c_mplb_native _dwidth 32 32 integer fpga family type xilinx fpga family c_family spartan3, virtex4, virtex5 "virtex4" string
opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 www.xilinx.com 9 product specification parameter-port dependencies n/a device utilization and performance benchmarks core performance because the opb_plbv46_bridge is a module that will be used with other design pieces in the fpga, the resource utilization and timing numbers reported in this section are estimates only. when the opb_plbv46_bridge is combined with other pieces of the fpga design, the utilization of fpga resources and timing of the design will vary from the results reported here. for spartan?-3e fpga systems the performance of th e plbv46 interface in 1:2 clock ratio mode should meet or exceed 90 mhz. similarly, for virtex?-5 fp ga systems the performance should meet or exceed 120 mhz. in some system configurat ions (in either 1:1 or 1:2 clock ra tio mode) the opb bus could be the limiting factor thus preventing the plbv46 interface from reaching full speed. use of the core in 1:1 clock ratio mode is offered only as an option and no clock frequency performance numbers are provided for it. the plbv46_opb_bridge resource utilization benchmar ks for an xc5vlx50-1-ff676 fpga for a variety of generic parameter combinations applied on top of a base parameter set are shown in table 3 . table 3: fpga resource utilization benchmarks parameter values (for example) device resources c_num_addr_rng rng0 size rng1 size rng2 size rng3 size c_bus_clock_ period_ratio c_prefetch_ timeout slice registers slice luts occupied slices 12 32 x x x 1 10 414 571 346 20x 20000000 0x 20000000 x x 1 10 408 567 314 30x 20000000 0x 20000000 0x 20000000 x 1 10 408 559 298 40x 20000000 0x 20000000 0x 20000000 0x 20000000 1 10 408 564 367 4 0x200 0x200 0x200 0x200 1 10 368 496 320 4 0x200 0x200 0x200 0x200 2 20 380 517 305 notes: generic parameters used: 1. c_mplb_awidth=32 2. c_mplb_dwidth=32 3. c_mplb_native_dwidth=32 4. c_family="virtex5"
opb to plbv46 bridge (v1.01a) 10 www.xilinx.com DS726 april 24, 2009 product specification system performance to measure the system performance (f max ) of this core, this core was added as the device under test (dut) to a virtex-4 fpga system as shown in figure 2 , to a virtex-5 fpga system as shown in figure 3 , and to a spartan-3a fpga system as shown in figure 4 . the dut in this core is the plbv46 opb bridge and the opb plbv46 bridge with the opb buses connected. because the opb to plbv46 bridge core will be used with other design modules in the fpga, the utilization and timing numbers reported in this sect ion are estimates only. when this core is combined with other designs in the system, the utilization of fpga resources and timing of the core design will vary from the results reported here.
opb to plbv46 bridge (v1.01a) DS726 april 24, 2009 www.xilinx.com 11 product specification figure top x-ref 2 figure 2: virtex-4 fx fpga system figure top x-ref 3 figure 3: virtex-5 lx fpga system figure top x-ref 4 figure 4: spartan-3a fpga system ppc405 mpmc 3 xp s cdma dut xp s uart lite xp s gpio xp s intc xp s bram dplb1 iplb1 dplb0 iplb0 xp s cdma plbv46 plbv46 plbv46 virtex-4 fx fpga s y s tem d s 404_02_09170 8 microbl a ze mpmc 3 xp s cdma dut xp s uart lite xp s gpio xp s intc xp s bram virtex-5 lxt fpga s y s tem xp s cdma mdm xcl xcl plbv46 d s 404_0 3 _09170 8 microbl a ze mpmc 3 xp s cdma dut xp s uart lite xp s gpio xp s intc xp s bram xp s cdma mdm plbv46 virtex-4 fx fpga s y s tem d s 404_04_09170 8
opb to plbv46 bridge (v1.01a) 12 www.xilinx.com DS726 april 24, 2009 product specification the target fpga was then filled with logic to drive the lut and block ram utilization to approximately 70% and the i/o utilization to appr oximately 80%. using the default tool options and the slowest speed grade for the target fpga, the resulting target f max numbers are shown in table 4 . the target f max is influenced by the exact system and is provided solely for guidance. it is not a guaranteed value across all systems. reference documents the following documents contain reference information important to understanding the opb to plbv46 bridge design: 1. ibm coreconnect 128-bit processor local bus: architecture specification 2. ibm coreconnect 64-bit on-chip peripher al bus: architecture specifications 3. xilinx plbv46 interconnect and interfaces simplifications an d feature subset specification revision history the following table shows the revision history for this document. notice of disclaimer .xilinx is providing this product documentation, herein after ?information,? to you ?as is? with no warranty of any kind, express or implied. xilinx makes no repr esentation that the information, or any particular implementation thereof, is free from any claims of infringement. you are re sponsible for obtaining any rights you may require for any implementation based on the informat ion. all specifications ar e subject to change without notice. xilinx expressly disclaims any warranty whatsoever with respect to the adequacy of the information or any implementation based thereon, including but not limited to any warranties or representa tions that this implementation is free from claims of infringement and any im plied warranties of merchantability or fitness for a particular purpose. except as stated herein, none of the information may be copied, reproduced, distributed, republished, downloaded, displayed , posted, or transmitted in any form or by any means including, but not limited to, electronic, mechanical, ph otocopying, recording, or otherwise, without the prior written consent of xilinx. table 4: opb to plbv46 bridge co re system performance target fpga target f max (mhz) s3a700 -4 90 v4fx60 -10 100 v5lxt50 -1 120 date version revision 6/11/07 1.0 initial xilinx release. 10/3/07 1.1 added fmax margin system performance section. 12/13/07 1.2 added virtex-ii pro fpga support. 9/17/08 1.3 updated for edk11.1 release; removed virtex-ii support. 04/24/09 1.4 replaced references to supported device families and tool name(s) with hyperlinks to pdf files; updated trademark information. assigned a new doc id - DS726 to replace old doc id - ds404. another dat a sheet already had the ds404 number.

▲Up To Search▲

Price & Availability of DS726

	To Download DS726 Datasheet File
If you can't view the Datasheet, Please click here to try to view without PDF Reader .