Intel 80386 manual Bus Cycles with Pipelined Addresses

Models: 80386

1 194
Download 194 pages 57.68 Kb
Page 57
Image 57

HARDWARE IMPLEMENTATION

cannot respond in two clocks can stretch the bus cycle by holding READY inactive, that is, by inserting wait states into the cycle. When running back-to-back 32-bit bus cycles, the 80386's maxi- mum bus bandwidth is 32 megabytes per second at 16 MHz or 25 megabytes per second at 12.5 MHz.

Due to its internal pipelining, the 80386 very often knows the address and definition of the next bus cycle before the external hardware has responded to the current cycle. External hardware can use the 80386's address pipelining

facility to gain early access to the following bus cycle definition when it is available. Address pipelining can give external hardware three clock between address and data while main- taining two-clock bandwidth to the processor.

Address pipelining is best exploited by interleaved memory systems that can respond to accesses in alternate banks in parallel. By asserting Next Address, the external hardware can ask the 80386 to output the next bus cycle definition as soon as it is available in the processor, rather than waiting for READY (see Figure 5-6).

CYCLE 1 (READ)

CLK2 [ (INPUT)

ADDRESS AND [

DEFINITION (OUTPUTS)

ADS [

(OUTPUT)

NA[

(INPUT)

READY [ (INPUT)

00-031

(INPUT [

DURING

READ)

CYCLE 2

CYCLE 3

(READ)

(READ)

Figure 5-6. Bus Cycles with Pipelined Addresses

5-6

Page 57
Image 57
Intel 80386 manual Bus Cycles with Pipelined Addresses