Interfacing Processors and Peripherals - PDF document

Interfacing Processors and Peripherals Mainõmemory I/Oõcontroller I/Oõcontroller I/Oõcontroller Disk Graphicsõoutput Network MemoryÐI/O busProcessor Cache Interrupts Disk • Bus lines Control lines  Data lines (data, commands, addresses) Bus transactions Read (output): memory to I/O device Write (input): I/O device to memory Types of buses Processor-memory (specific) I/O buses (standard) Backplane bases (standard)  Request a read on control lines and supply data at the data lines Memory access Memory signals data availability and transfer them on the data line to I/O Memory Processor Control linesData linesDisks Memory Processor Control linesData linesDisks Processor Control linesData linesDisks Memory  Write request for memory on the control lines and address on the data lines Memory signal the I/O device that is ready and transfers starts Memory Processor Control linesData linesDisks Processor Control linesData linesDisksMemory Types of buses Processor Memory Backplane bus I/O devices Processor MemoryProcessor-memory bus Busõadapter Busõadapter I/OõbusI/Oõbus Busõadapter I/Oõbus Processor MemoryProcessor-memory bus Busõadapter Backplaneõbus Busõadapter I/O bus Busõadapter I/O bus  Synchronous buses: Using a clock synchronized  All devices use a single clock Cannot be long Asynchronous buses: handshaking protocol Reading a word from memory: Read request ( Data are ready to read ( Acknowledge the ReadReq and DataRdy signals of the other party (Ack DataRdyAckData ReadReq 34576422 Implementing handshaking Record fromõdata linesõand assertõAck ReadReqReadReq________ ReadReqReadReq3, 4õDrop Ack;õput memoryõdata on dataõlines; assertõDataRdy AckAckRelease dataõlines andõDataRdy ___________Memory Release dataõlines; deassertõReadReq Ack DataRdyDataRdyRead memoryõdata from dataõlines;õassert Ack DataRdyDataRdyDeassert Ack I/O devicePut addressõon dataõlines; assertõReadReq________Ack___________ New I/O request N e w I / O r e q u e s t Increasing the bus bandwidth Increasing the width of the data bus lines Using separate address and data lines Transferring multiple words (blocks) Bus access: master and slave Memory Processor Bus request linesBusDisks Bus request linesBusDisks Processor Bus request linesBusDisks Processor MemoryMemory  I/O device generates a request to the processor to use the bus The processor responds and generates the corresponding bus control signals The processor notifies the device and the device places the address on the bus Daisy chain arbitration: A single bus grant line is run through the devices from highestpriority to lowest. A higher priority device intercepts the bus grant signal, not allowing alower priority device to see it.Centralized, parallel arbitration: Using multiple request lines managed by acentralized arbiter.Distributed arbitration by self-selection: Multiple request lirequesting the bus access determine who will be granted access. Each bus requestingdevice places its identity code on the bus.Distributed arbitration by collision detection: Each device indepebus. Multiple simultaneous requests result in a collision, which is detected and solved. Device n Lowest priorit y Device 2 Device 1 Highest priority Busõarbiter GrantGrantGrant ReleaseRequest CharacteristicPCISCSI TypeBackplaneI/O Data bus width32-648-32 Address/data multiplexed?YesYes Number of bus mastersmultiplemultiple ArbitrationCentralized, parallelSelf-selection ClockingSynchronous 33-66MHzAsynchronous orsynchronous 5-10MHz Theoretical peak bandwidth133-512 MB/sec5-40 MB/sec Estimated achievablebandwidth80 MB/sec2.5-40 MB/sec Maximum number of devices10247-31 Maximum bus length0.5 meter25 meters