WO1994024628A1 - Multi-phase multi-access pipeline memory system - Google Patents
Multi-phase multi-access pipeline memory system Download PDFInfo
- Publication number
- WO1994024628A1 WO1994024628A1 PCT/US1994/004455 US9404455W WO9424628A1 WO 1994024628 A1 WO1994024628 A1 WO 1994024628A1 US 9404455 W US9404455 W US 9404455W WO 9424628 A1 WO9424628 A1 WO 9424628A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- phase
- processor
- memory
- pipeline memory
- processors
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/38—Information transfer, e.g. on bus
- G06F13/42—Bus transfer protocol, e.g. handshake; Synchronisation
- G06F13/4204—Bus transfer protocol, e.g. handshake; Synchronisation on a parallel bus
- G06F13/4208—Bus transfer protocol, e.g. handshake; Synchronisation on a parallel bus being a system bus, e.g. VME bus, Futurebus, Multibus
- G06F13/4217—Bus transfer protocol, e.g. handshake; Synchronisation on a parallel bus being a system bus, e.g. VME bus, Futurebus, Multibus with synchronous protocol
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F13/00—Interconnection of, or transfer of information or other signals between, memories, input/output devices or central processing units
- G06F13/14—Handling requests for interconnection or transfer
- G06F13/16—Handling requests for interconnection or transfer for access to memory bus
- G06F13/1605—Handling requests for interconnection or transfer for access to memory bus based on arbitration
- G06F13/161—Handling requests for interconnection or transfer for access to memory bus based on arbitration with latency improvement
- G06F13/1615—Handling requests for interconnection or transfer for access to memory bus based on arbitration with latency improvement using a concurrent pipeline structrure
Definitions
- This invention relates to a multi-phase, multi-access pipeline memory system.
- the third approach uses a number of separate memory banks so that most of the time each of the microprocessor and I/O processor can be accessing a different bank.
- the user/programmer In order to effect his, however, the user/programmer must superimpose on the programming the goal of minimizing overlap in access demand for the memory banks by the microprocessor and I/O processor. This requires careful scheduling of the I/O and computing tasks so that simultaneous demand by both processors for the same memory bank is avoided or at least reduced. That imposes a burdensome ancillary constraint on the user.
- the fourth approach is to simply run the memory at twice the normal cycle speed. But this is difficult, especially in signal processing systems where memories are already operating at near capacity as a rule.
- the invention results from the realization that a truly effective multi-access memory system can be achieved in which each processor has access to the memory once in each system clock period without interfering with the access of the other processors by phase shifting the operation of the processors and pipelining the memory so that each processor accesses memory during a different phase of the system clock period while maintaining the memory operation at its normal speed.
- this approach is the fundamental realization that in most cases there is no reason why all the processors, be they microprocessors or I/O processors, must be operated in phase so that their operations start and end simultaneously: an overlapped sequence of operation is acceptable and therefore pipelined memory in combination with phase overlapped sequencing can be used to full advantage.
- This invention features a dual phase, dual access pipeline memory system having first and second processors, a pipeline memory including latch means, and a bus means for interconnecting the processor to the pipeline memory.
- a clock circuit responsive to a system clock signal for providing a first clock signal in phase with the system clock signal for operating the first processor, a second clock signal out of phase with the system clock signal for operating the second processor out of phase with the first processor, and a third clock signal at twice the system clock signal rate for clocking the pipeline memory through the latch means to allow an address to be supplied to the pipeline memory by the first processor while accessing data from the address supplied in the previous cycle during one phase. Conversely, it allows an address to be supplied to the pipeline memory by the second processor while accessing data from the address supplied in the previous cycle during the other phase.
- the processors may include an I/O processor and a microprocessor, or the processors may be both microprocessors.
- the processors may also include subprocessors in the same microprocessor.
- the subprocessors may include an instruction fetch unit and a data fetch unit.
- the pipeline memory may include a plurality of memory banks and the bus means may include a plurality of data address bus pairs and there may be third and fourth processors.
- the first and third processors may include data fetch units which access different memory banks from each other in the same phase with each other and the second and fourth processors may include an instruction fetch unit and an I/O processor which access different memory banks from each other in the same phase with each other but out of phase with the first and third processors.
- the invention features a multi-phase, multi-access pipeline memory system which includes a number, n, of processors, a pipeline memory including latch means, arid bus means interconnecting the processors and pipeline memory.
- At least one of the processors may be an I/O processor and one may be a microprocessor, or the processors may both be microprocessors.
- the processors may include subprocessors in the same microprocessor.
- the subprocessors may include a data fetch unit and an instruction fetch unit.
- the latch means may include a plurality of latches to enable pipelining of n accesses of the pipeline memory during each system clock signal period.
- Fig. 1 is a simplified block diagram of a multi-phase, multi-access pipeline memory system according to this invention
- Fig. 2 is an illustration of the clock signals that occur at various points in Fig. 1;
- Fig. 3 is a more detailed block diagram of the system of Fig. 1 showing an additional holding latch which expedites the operation of the pipeline memory;
- Fig. 4 is a timing diagram showing the sequence of occurrences with respect to the operation of the pipeline memory in Figs. 1 and 3;
- Fig. 5 is a more detailed block diagram of the pipeline memory of Figs. 1 and 3;
- Figs. 6A and B are a detailed block diagram of a specific implementation of the invention using a four-bank pipeline memory and a number of subprocessors in a microprocessor sharing dual data and program buses;
- Fig. 7 is a illustration of yet a further improvement according to this invention in the operation of the system of Figs. 6A and 6B.
- a multi-phase, multi-access pipeline memory system 10 including a pipeline memory 12 and two processors 14 and 16, along with clock circuit 18 which is driven by a system clock signal that it receives on line 20.
- Clock circuit 18 divides the system clock signal into two phases: ⁇ delivered on line 22 to processor 14, and ⁇ 2 delivered on line 24 to processor 16.
- Clock circuit 18 also creates a signal at twice the frequency of the incoming system clock signal on line 20 and delivers it on line 26 to latch 28 in pipeline memory 12.
- processor 14 requests an address and provides control signals over line 30 to pipeline memory 12.
- memory 12 provides the data on line 32 back to processor 14 (read) , or the processor 14 supplies data (write) .
- processor 16 on line 34 requests an address and provides control signals and the data is returned (read) or supplied (write) to it in response thereto over line 36 during phase one of the next period of the system clock signal.
- phase one signal 42 is actually a duplication of the system clock signal, in this case where the pipeline memory system is operating as a dual access system and there are but two processors 14 and 16 to deal with.
- the third clock signal delivered to pipeline memory 12 on line 26 appears as shown at 46 as occurring at twice the rate of the system clock signal 40.
- the third clock signal need not be a separately created signal; it can be derived from the rising and falling edges of the original clock.
- the addresses from processors 1 and 2 are driven onto a common address bus.
- the address from processor 1 is driven in ⁇ 2 and is latched in 50, Fig. 3, at the start of ⁇ ,.
- the address from processor 2 is driven in and is latched in 50 at the start of ⁇ j.
- the address decoder can immediately begin decoding the addresses at the start of ⁇ , and ⁇ 2 .
- Latch 50 is clocked by the same 2x clock that drives the pipeline latch 28.
- Fig. 4 is a detailed timing diagram showing the signals that occur at various points in Fig. 3.
- system clock signal 40 has two phases, ⁇ , and ⁇ 2 , which makeup the total clock period 41.
- the next clock period 41' includes phase ⁇ u and ⁇ 12 , and so on.
- the double rate clock signal 46 driving latch 28 triggers latch 50 to store the incoming address request from processor 14 at time 60.
- this address is decoded.
- the decoded address is used to retrieve the data and return it on line 32 to processor 14.
- latch 50 secures the address for processor 16 during the rest of phase ⁇ 2 :
- the use of latch 50 forces the processors to have the address valid at the beginning of each phase, but this address can be removed (by the processors) immediately after the phase begins.
- Latch 50 in essence shifts the timing of address and data such that each processor supplies its address and gets its previous data in the same phase. (See Fig. 4.) This is completed at time 64 at the end of phase ⁇ 2 and the close of clock period 41.
- phase ⁇ i the next clock period 41" ending at time 70, the data is retrieved in accordance with the decoded address and returned to processor 16.
- latch 28 is right in the middle of pipeline memory 12, dividing the address decode and the data retrieval functions, this is not a necessary limitation of the invention as one or more latches may be used to implement latch 28 and these may be used at various locations in the pipeline memory 12. Note that at time 62 latch 28 latches the decoded address of processor 14 in response to the address in latch 50. Thus latch 28 affects the pipelining function of memory 12 by permitting the memory to decode the address for one processor request held in latch 50 while simultaneously accessing the memory array with the decoded address held in latch 28.
- the read data waveform 80 indicates that the data to be read or written is valid toward the end of ⁇ 2 for processor 14. If the data is to be read the waveform 80 indicates that the data to be read will be valid when time 64 is reached as indicated at read data 82. If the data is to be written into memory then waveform 84 indicates that the data is valid to write at 86. Similarly, the data valid indications 88 and 90 for microprocessor 16 indicate that the data is valid at time 70.
- pipeline memory 12 shows that it includes an address decoder 100, latch 28, memory array 102, column decoder 104, and latches 106. There is a MUX 108 at the output. Assuming the two processors in this dual-access pipeline memory system include one microprocessor and one I/O processor, the address of each will be fed. in on line 114 during each phase of the clock: drive processor 1 during ⁇ 2 and drive processor 2 during ⁇ ,.
- Latch 112 is a holding latch similar to latch 50 which holds the previous processor address request so that the new one is available for address decoder 100 immediately upon the beginning of the period.
- Latch 28 operates in conjunction with address decoder 100 and memory array 102 in the normal manner of pipeline memories.
- Column decode 104 receives the output data from memory array 102 and delivers it to latches 106 which are also anticipatory latches that maintain the output data at the ready condition for delivery to the requesting processor and the data to be written for delivery to the memory array.
- the data lines 118 and 120 return to the microprocessor and I/O processor, respectively, the data which they have requested. Data lines 118 and 120 may be time-shared in the same way as are the address lines.
- Latch 106 functions to maintain the data at the ready for delivery to the processor. It is this function that produces the overlapping sections 120, 122, 124 and 126 of waveform sections 82, 86, 88 and 90, respectively, indicating that the data is there and ready at the moment the processor is ready to receive or write the data.
- the control signals previously referred to which are delivered by the processor to the pipeline memory and part of the address are delivered over another latch 122 to column decode 104.
- Latch 122 operates to synchronize the operation of column decode 104 with the pipeline
- memory 12 includes four memory banks 150, 152, 154 and 156
- processor 14 is a microprocessor which includes a data fetch unit data address generators 158 and 160, cache memory 162, and an instruction fetch unit, program sequencer 164.
- Microprocessor 14 also includes conventional flags 166, timers 168 and interrupts 170.
- register file 172 floating point fixed multiplier and a fixed point accumulator 174, barrel shifter 176, and a floating point and fixed point arithmetic logic unit 178.
- the other processor, processor 16 is actually an I/O processor which includes I/O processor register 180, serial port 182, nibble port 184, and data memory address control 186.
- the external port 190 includes a pair of MUXes 192, 194 and an external port control 196.
- data address generator 158 and 160 and programmer sequencer 164 each are considered a processor, although in this case they are actually subprocessors of microprocessor 14, so that there are four processors in the eyes of memory 12: data address generators 158 and 160, program sequencer 164, and I/O processor 16.
- data address generator 160 or program sequencer 164 can access the program memory bus pair 200 and 202. Also during phase one the data address generator 158 can access the data memory address pair 206, 208. During phase two, only I/O processor 16 accesses the memory banks of memory 12 via program memory address pair 200 and 202. If both the program sequencer 164 and data address generator 160 need to address the program memory address bus 200, 202, then one must access it in phase one of a first clock period and the other obtain its access during phase one of the next clock period.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multi Processors (AREA)
- Advance Control (AREA)
- Static Random-Access Memory (AREA)
- Dram (AREA)
- Memory System Of A Hierarchy Structure (AREA)
Abstract
A multi-phase, multi-access pipeline memory system includes a numer, n, of processors (14, 16); a pipeline memory (12) including a latch (28); and a bus (30, 32, 34, 36) for interconnecting the processors and pipeline memory; a clock circuit (18) responsive to a system clock signal (20) into n phases for providing multiple clock signals (22, 24, 26) corresponding to the n phases of the system clock signal (20) for operating each processor (14, 16) to allow data and address to be transferred only during its assigned phase thereby enabling the memory (12) and each processor (14, 16) to operate at the system clock rate while allowing n accesses to the memory (12) during each system clock signal period, one access for each processor (14, 16).
Description
MULTI-PHASE MULTI-ACCESS PIPELINE MEMORY SYSTEM
FIELD OF INVENTION This invention relates to a multi-phase, multi-access pipeline memory system.
BACKGROUND OF INVENTION In conventional computers and microcomputers there is a constant contention between the microprocessor and the I/O processor for memory access. This is a particularly serious problem in signal processing applications requiring high speed processing of massive amounts of data and high I/O rates. There are four approaches commonly used to enable fair sharing of the memory. One is to use dual port memories: these are generally complex and expensive and have small capacity. A second approach uses "cycle stealing" wherein one of the microprocessor and I/O processor has priority over the other and "steals" memory access when it needs it, thereby interrupting the operation of the other. This slows down the response of the interrupted processor and of the whole system. The third approach uses a number of separate memory banks so that most of the time each of the microprocessor and I/O processor can be accessing a different bank. In order to effect his, however, the user/programmer must superimpose on the programming the goal of minimizing overlap in access demand for the memory banks by the microprocessor and I/O processor. This requires careful scheduling of the I/O and computing tasks so that simultaneous demand by both processors
for the same memory bank is avoided or at least reduced. That imposes a burdensome ancillary constraint on the user. The fourth approach is to simply run the memory at twice the normal cycle speed. But this is difficult, especially in signal processing systems where memories are already operating at near capacity as a rule.
SUMMARY OF INVENTION
It is therefore an object of this invention to provide an improved, multi-phase multi-access pipeline memory system.
It is a further object of this invention to provide such a multi-phase multi-access pipeline memory system which does not increase, expand or limit memory capacity.
It is a further object of this invention to provide such a multi-phase, multi-access pipeline memory system which does not interrupt other processors' access to the memory and does not slow down the system operation as a whole.
It is a further object of this invention to provide such a multi-phase, multi-access pipeline memory system which requires no special programming or scheduling of processor-memory interaction.
It is a further object of this invention to provide such a multi-phase, multi-access pipeline memory system which does not require operating the memory above normal speeds.
The invention results from the realization that a truly effective multi-access memory system can be achieved in which each processor has access to the memory once in each system clock
period without interfering with the access of the other processors by phase shifting the operation of the processors and pipelining the memory so that each processor accesses memory during a different phase of the system clock period while maintaining the memory operation at its normal speed. Underlying this approach is the fundamental realization that in most cases there is no reason why all the processors, be they microprocessors or I/O processors, must be operated in phase so that their operations start and end simultaneously: an overlapped sequence of operation is acceptable and therefore pipelined memory in combination with phase overlapped sequencing can be used to full advantage.
This invention features a dual phase, dual access pipeline memory system having first and second processors, a pipeline memory including latch means, and a bus means for interconnecting the processor to the pipeline memory. There is a clock circuit responsive to a system clock signal for providing a first clock signal in phase with the system clock signal for operating the first processor, a second clock signal out of phase with the system clock signal for operating the second processor out of phase with the first processor, and a third clock signal at twice the system clock signal rate for clocking the pipeline memory through the latch means to allow an address to be supplied to the pipeline memory by the first processor while accessing data from the address supplied in the previous cycle during one phase. Conversely, it allows an address to be supplied to the pipeline memory by the second processor while accessing data from the
address supplied in the previous cycle during the other phase.
In the preferred embodiment the processors may include an I/O processor and a microprocessor, or the processors may be both microprocessors. The processors may also include subprocessors in the same microprocessor. The subprocessors may include an instruction fetch unit and a data fetch unit. The pipeline memory may include a plurality of memory banks and the bus means may include a plurality of data address bus pairs and there may be third and fourth processors. The first and third processors may include data fetch units which access different memory banks from each other in the same phase with each other and the second and fourth processors may include an instruction fetch unit and an I/O processor which access different memory banks from each other in the same phase with each other but out of phase with the first and third processors.
In a more comprehensive sense, the invention features a multi-phase, multi-access pipeline memory system which includes a number, n, of processors, a pipeline memory including latch means, arid bus means interconnecting the processors and pipeline memory. There is a clock circuit responsive to a system clock signal for dividing the system clock signal into n phases for providing multiple clock signals corresponding to the n phases of the system clock signal for operating each processor to allow data and addresses to be transferred only during the assigned phase, thereby enabling the memory and each processor to operate at the system clock rate while allowing n accesses to the memory during each system clock signal period, one access for each
processor .
In a preferred embodiment at least one of the processors may be an I/O processor and one may be a microprocessor, or the processors may both be microprocessors. The processors may include subprocessors in the same microprocessor. The subprocessors may include a data fetch unit and an instruction fetch unit. The latch means may include a plurality of latches to enable pipelining of n accesses of the pipeline memory during each system clock signal period.
DISCLOSURE OF PREFERRED EMBODIMENT
Other objects, features and advantages will occur to those skilled in the art from the following description of a preferred embodiment and the accompanying drawings, in which:
Fig. 1 is a simplified block diagram of a multi-phase, multi-access pipeline memory system according to this invention;
Fig. 2 is an illustration of the clock signals that occur at various points in Fig. 1;
Fig. 3 is a more detailed block diagram of the system of Fig. 1 showing an additional holding latch which expedites the operation of the pipeline memory;
Fig. 4 is a timing diagram showing the sequence of occurrences with respect to the operation of the pipeline memory in Figs. 1 and 3;
Fig. 5 is a more detailed block diagram of the pipeline memory of Figs. 1 and 3;
Figs. 6A and B are a detailed block diagram of a specific
implementation of the invention using a four-bank pipeline memory and a number of subprocessors in a microprocessor sharing dual data and program buses; and
Fig. 7 is a illustration of yet a further improvement according to this invention in the operation of the system of Figs. 6A and 6B.
There is shown in Fig. 1 a multi-phase, multi-access pipeline memory system 10 according to this invention including a pipeline memory 12 and two processors 14 and 16, along with clock circuit 18 which is driven by a system clock signal that it receives on line 20. Clock circuit 18 divides the system clock signal into two phases: λ delivered on line 22 to processor 14, and ø2 delivered on line 24 to processor 16. Clock circuit 18 also creates a signal at twice the frequency of the incoming system clock signal on line 20 and delivers it on line 26 to latch 28 in pipeline memory 12. During phase one, processor 14 requests an address and provides control signals over line 30 to pipeline memory 12. During phase two, memory 12 provides the data on line 32 back to processor 14 (read) , or the processor 14 supplies data (write) . Also during phase two, processor 16 on line 34 requests an address and provides control signals and the data is returned (read) or supplied (write) to it in response thereto over line 36 during phase one of the next period of the system clock signal.
This can be seen more readily with respect to the waveform shown in Fig. 2. There the system clock 40 which appears on line 20 is shown as having two phases: φx and φ2. In clock circuit 18
there are two clock signals developed from that: phase one signal 42 and the phase two signal 44, which is the inverse of the phase one signal 42. Signal 42 is actually a duplication of the system clock signal, in this case where the pipeline memory system is operating as a dual access system and there are but two processors 14 and 16 to deal with. The third clock signal delivered to pipeline memory 12 on line 26 appears as shown at 46 as occurring at twice the rate of the system clock signal 40. However, it should be understood that this is not operating the memory at twice the normal memory speed; rather, it is simply operating latch 28 to perform the pipelining function of pipeline memory 12. As the number of processors increases, the number of latches must also increase in order to adequately pipeline the requests and data inputs and outputs. The third clock signal need not be a separately created signal; it can be derived from the rising and falling edges of the original clock.
In order to expedite processing of the request by pipeline memory 12, the addresses from processors 1 and 2 are driven onto a common address bus. The address from processor 1 is driven in φ2 and is latched in 50, Fig. 3, at the start of ø,. The address from processor 2 is driven in and is latched in 50 at the start of øj. Thus the address decoder can immediately begin decoding the addresses at the start of ø, and ø2. Latch 50 is clocked by the same 2x clock that drives the pipeline latch 28.
Fig. 4 is a detailed timing diagram showing the signals that occur at various points in Fig. 3. As can be seen, system clock signal 40 has two phases, ø, and ø2, which makeup the total clock
period 41. The next clock period 41' includes phase φu and ø12, and so on. The double rate clock signal 46 driving latch 28 triggers latch 50 to store the incoming address request from processor 14 at time 60. During the first phase, ø., ending at time 62, this address is decoded. In the second phase, ending at time 6 , the decoded address is used to retrieve the data and return it on line 32 to processor 14. Simultaneously during phase ø2, beginning at time 62, latch 50 secures the address for processor 16 during the rest of phase ø2: The use of latch 50 forces the processors to have the address valid at the beginning of each phase, but this address can be removed (by the processors) immediately after the phase begins. Latch 50 in essence shifts the timing of address and data such that each processor supplies its address and gets its previous data in the same phase. (See Fig. 4.) This is completed at time 64 at the end of phase ø2 and the close of clock period 41. During phase øi,, of the next clock period 41" ending at time 70, the data is retrieved in accordance with the decoded address and returned to processor 16. Although in this particular explanation latch 28 is right in the middle of pipeline memory 12, dividing the address decode and the data retrieval functions, this is not a necessary limitation of the invention as one or more latches may be used to implement latch 28 and these may be used at various locations in the pipeline memory 12. Note that at time 62 latch 28 latches the decoded address of processor 14 in response to the address in latch 50. Thus latch 28 affects the pipelining function of memory 12 by permitting the memory to decode the
address for one processor request held in latch 50 while simultaneously accessing the memory array with the decoded address held in latch 28.
Included in the address request on lines 30 and 34 from processors 14 and 16 are control signals which indicate whether there is information that has to be written into or read out of, or neither, of the decoded address. The read data waveform 80, Fig. 4, indicates that the data to be read or written is valid toward the end of ø2 for processor 14. If the data is to be read the waveform 80 indicates that the data to be read will be valid when time 64 is reached as indicated at read data 82. If the data is to be written into memory then waveform 84 indicates that the data is valid to write at 86. Similarly, the data valid indications 88 and 90 for microprocessor 16 indicate that the data is valid at time 70.
A more detailed implementation of pipeline memory 12, Fig. 5, shows that it includes an address decoder 100, latch 28, memory array 102, column decoder 104, and latches 106. There is a MUX 108 at the output. Assuming the two processors in this dual-access pipeline memory system include one microprocessor and one I/O processor, the address of each will be fed. in on line 114 during each phase of the clock: drive processor 1 during ø2 and drive processor 2 during ø,. Latch 112 is a holding latch similar to latch 50 which holds the previous processor address request so that the new one is available for address decoder 100 immediately upon the beginning of the period. Latch 28 operates in conjunction with address decoder 100 and memory array 102 in
the normal manner of pipeline memories. Column decode 104 receives the output data from memory array 102 and delivers it to latches 106 which are also anticipatory latches that maintain the output data at the ready condition for delivery to the requesting processor and the data to be written for delivery to the memory array. The data lines 118 and 120 return to the microprocessor and I/O processor, respectively, the data which they have requested. Data lines 118 and 120 may be time-shared in the same way as are the address lines. Latch 106 functions to maintain the data at the ready for delivery to the processor. It is this function that produces the overlapping sections 120, 122, 124 and 126 of waveform sections 82, 86, 88 and 90, respectively, indicating that the data is there and ready at the moment the processor is ready to receive or write the data. The control signals previously referred to which are delivered by the processor to the pipeline memory and part of the address are delivered over another latch 122 to column decode 104. Latch 122 operates to synchronize the operation of column decode 104 with the pipeline latch 28.
In one application, memory 12, Figs. 6A and 6B, includes four memory banks 150, 152, 154 and 156, and processor 14 is a microprocessor which includes a data fetch unit data address generators 158 and 160, cache memory 162, and an instruction fetch unit, program sequencer 164. Microprocessor 14 also includes conventional flags 166, timers 168 and interrupts 170. There is also a register file 172, floating point fixed multiplier and a fixed point accumulator 174, barrel shifter 176,
and a floating point and fixed point arithmetic logic unit 178. The other processor, processor 16, is actually an I/O processor which includes I/O processor register 180, serial port 182, nibble port 184, and data memory address control 186. The external port 190 includes a pair of MUXes 192, 194 and an external port control 196. In this application data address generator 158 and 160 and programmer sequencer 164 each are considered a processor, although in this case they are actually subprocessors of microprocessor 14, so that there are four processors in the eyes of memory 12: data address generators 158 and 160, program sequencer 164, and I/O processor 16. In this application there are two bus pairs: the program address and the program memory data buses 200 and 202 are one pair, and the data memory address and data memory data buses 206 and 208 are the other pair.
In accordance with this invention, during phase one, data address generator 160 or program sequencer 164 can access the program memory bus pair 200 and 202. Also during phase one the data address generator 158 can access the data memory address pair 206, 208. During phase two, only I/O processor 16 accesses the memory banks of memory 12 via program memory address pair 200 and 202. If both the program sequencer 164 and data address generator 160 need to address the program memory address bus 200, 202, then one must access it in phase one of a first clock period and the other obtain its access during phase one of the next clock period.
An even more advantageous application of the invention to
the system of Figs. 6A and 6B can be seen with respect to Fig. 7, where the competition between program sequencer 164 and data address generator 160 for the program memory address pair 200 and 202 can be avoided by having data address generator 158 and 160 deliver their addresses to the two different buses 202 and 206 during the first phase, and then have the program sequencer 164 and I/O processor 16 deliver their address requests during the second phase to buses 200 and 206. This is achievable because program sequencer 164 is only loosely coupled to the data address generators 158 and 160. This underlies all applications of this invention as the phase distribution of the requests depends upon the ability of the various processors to work synchronously but out of phase with one another in order to obtain the multiple access to the pipeline memory without interrupting one another or requiring the memory to run at double or multiple speeds. With this implementation, as shown in Fig. 7, the bus pairs 200, 202 and 206, 208 are more efficiently used, data address generator 160 and program sequencer 164 are no longer in competition, and cache memory 162 can be eliminated. While thus far the illustrative examples have all been with respect to a dual access pipeline memory system, this is not a necessary limitation of the invention as the invention applies to any multi-access pipeline memory system as shown in phantom in Fig. 7 where an additional pair of processors such as I/O processor 220 and microprocessor 224 are shown utilizing program memory address bus 200 and data memory address bus 206 during a third phase, ø3, where of course the system clock signal has been divided into three phases or
4/24628
13
more, depending upon the number of processors to be accommodated and the number of buses available in conjunction with the pipeline memory.
Although specific features of this invention are shown in some drawings and not others, this is for convenience only as some feature may be combined with any or all of the other features in accordance with the invention.
Other embodiments will occur to those skilled in the art an are within the following claims: What is claimed is:
Claims
1. A dual phase, dual access pipeline memory system comprising: first and second processors; a pipeline memory including latch means; bus means for interconnecting said processors and pipeline memory; a clock circuit responsive to a system clock signal for providing a first clock signal in phase with said system clock signal for operating said first processor, a second clock signal out of phase with said system clock signal for operating said second processor out of phase with said first processor, and a third clock signal at twice the system clock signal rate for clocking said pipeline memory through said latch means to allow an address to be supplied to said pipeline memory by said first processor during one phase while said first processor is accessing data from the previous cycle in the same phase and, conversely, to allow an address to be supplied to said pipeline memory by said second processor during the other phase while the said second processor is accessing data from the previous cycle in the same phase.
2. The dual phase, dual access pipeline memory system of claim 1 in which said processors are an I/O processor and a microprocessor.
3. The dual phase, dual access pipeline memory system of claim 1 in which said processors are both microprocessors.
4. The dual, phase dual access pipeline memory system of claim 1 in which said processors include subprocessors in the same microprocessor.
5. The dual phase, dual access pipeline memory system of claim 4 in which said subprocessors include an instruction fetch unit and a data fetch unit.
6. The dual phase, dual access pipeline memory system of claim l in which said pipeline memory includes a plurality of memory banks, said bus means includes a plurality of data address bus pairs, and there are further included third and fourth processors.
7. The dual phase, dual access pipeline memory system of claim 6 in which said first and third processors include data fetch units which access different memory banks from each other in the same phase with each other, and said second and fourth processors include an instruction fetch unit and an I/O processor which access different memory banks from each other in the same phase with each other but out of phase with said first and third processors.
8. A multi-phase, multi-access pipeline memory system, comprising: a number, n, of processors; a pipeline memory including latch means; bus means for interconnecting said processors and pipeline memory; a clock circuit, responsive to a system clock signal for dividing said system clock signal into n phases for providing multiple clock signals corresponding to the n phases of said system clock signal for operating each processor to allow data and address to be transferred only during its assigned phase thereby enabling said memory and each processor to operate at the system clock rate while allowing n accesses to said memory during each system clock signal period, one access for each processor.
9. The multi-phase, multi-access pipeline memory system of claim 8 in which said processors include subprocessors in the same microprocessor.
10. The multi-phase, multi-access pipeline memory system of claim 9 in which said subprocessors include an instruction fetch unit and a data fetch unit.
11. The multi-phase, multi-access pipeline memory system of claim 8 in which said pipeline memory includes a plurality of memory banks, and said bus means includes a plurality of data address bus pairs.
12. The multi-phase, multi-access pipeline memory system of claim 8 in which said latch means includes a plurality of latches to enable pipelining of n accesses of said pipeline memory during each system clock signal period.
Priority Applications (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE69432014T DE69432014T2 (en) | 1993-04-22 | 1994-04-22 | MULTI-PHASE MULTIPLE ACCESS PIPELINE STORAGE SYSTEM |
EP94914890A EP0695444B1 (en) | 1993-04-22 | 1994-04-22 | Multi-phase multi-access pipeline memory system |
JP6517429A JPH08509082A (en) | 1993-04-22 | 1994-04-22 | Multi-phase, multi-access pipeline memory system |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US08/052,073 US5471607A (en) | 1993-04-22 | 1993-04-22 | Multi-phase multi-access pipeline memory system |
US08/052,073 | 1994-04-22 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1994024628A1 true WO1994024628A1 (en) | 1994-10-27 |
Family
ID=21975282
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/US1994/004455 WO1994024628A1 (en) | 1993-04-22 | 1994-04-22 | Multi-phase multi-access pipeline memory system |
Country Status (5)
Country | Link |
---|---|
US (1) | US5471607A (en) |
EP (1) | EP0695444B1 (en) |
JP (1) | JPH08509082A (en) |
DE (1) | DE69432014T2 (en) |
WO (1) | WO1994024628A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0649100A2 (en) * | 1993-10-15 | 1995-04-19 | Hitachi, Ltd. | Data processing system and image processing system |
WO1999019875A2 (en) * | 1997-10-10 | 1999-04-22 | Rambus Incorporated | Apparatus and method for pipelined memory operations |
Families Citing this family (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5761466A (en) * | 1994-05-09 | 1998-06-02 | Lsi Logic Corporation | Soft programmable single-cycle/pipelined micro-programmed control system |
JPH0816530A (en) * | 1994-07-04 | 1996-01-19 | Kurieiteibu Design:Kk | Coprocessor system and external memory device with auxiliary operation function |
JP3013714B2 (en) * | 1994-09-28 | 2000-02-28 | 日本電気株式会社 | Semiconductor storage device |
JP2970434B2 (en) * | 1994-10-31 | 1999-11-02 | 日本電気株式会社 | Synchronous semiconductor memory device and sense control method |
US5896543A (en) * | 1996-01-25 | 1999-04-20 | Analog Devices, Inc. | Digital signal processor architecture |
US5954811A (en) * | 1996-01-25 | 1999-09-21 | Analog Devices, Inc. | Digital signal processor architecture |
GB2310738B (en) * | 1996-02-29 | 2000-02-16 | Advanced Risc Mach Ltd | Dynamic logic pipeline control |
JP2959482B2 (en) * | 1996-08-19 | 1999-10-06 | 日本電気株式会社 | Large-scale integrated circuits |
US5790838A (en) * | 1996-08-20 | 1998-08-04 | International Business Machines Corporation | Pipelined memory interface and method for using the same |
WO1999019805A1 (en) | 1997-10-10 | 1999-04-22 | Rambus Incorporated | Method and apparatus for two step memory write operations |
EP1327991A3 (en) * | 1997-10-10 | 2005-05-11 | Rambus Inc. | Apparatus and method for pipelined memory operations |
AU9798798A (en) * | 1997-10-10 | 1999-05-03 | Rambus Incorporated | Power control system for synchronous memory device |
US6401167B1 (en) | 1997-10-10 | 2002-06-04 | Rambus Incorporated | High performance cost optimized memory |
US6002882A (en) * | 1997-11-03 | 1999-12-14 | Analog Devices, Inc. | Bidirectional communication port for digital signal processor |
US6061779A (en) * | 1998-01-16 | 2000-05-09 | Analog Devices, Inc. | Digital signal processor having data alignment buffer for performing unaligned data accesses |
EP1031988A1 (en) * | 1999-02-26 | 2000-08-30 | Texas Instruments Incorporated | Method and apparatus for accessing a memory core |
US6629223B2 (en) | 1998-10-06 | 2003-09-30 | Texas Instruments Incorporated | Method and apparatus for accessing a memory core multiple times in a single clock cycle |
US7126874B2 (en) * | 2004-08-31 | 2006-10-24 | Micron Technology, Inc. | Memory system and method for strobing data, command and address signals |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3896418A (en) * | 1971-08-31 | 1975-07-22 | Texas Instruments Inc | Synchronous multi-processor system utilizing a single external memory unit |
US4050058A (en) * | 1973-12-26 | 1977-09-20 | Xerox Corporation | Microprocessor with parallel operation |
US4459688A (en) * | 1979-11-26 | 1984-07-10 | Hitachi, Ltd. | Access request selecting circuit |
US4547845A (en) * | 1982-04-21 | 1985-10-15 | The United States Of America As Represented By The Secretary Of The Navy | Split-BUS multiprocessor system |
US4621362A (en) * | 1984-06-04 | 1986-11-04 | International Business Machines Corp. | Routing architecture for a multi-ring local area network |
US4817006A (en) * | 1986-03-28 | 1989-03-28 | Thomas Engineering, Inc. | Pharmaceutical tablet press control mechanism |
US4866604A (en) * | 1981-10-01 | 1989-09-12 | Stratus Computer, Inc. | Digital data processing apparatus with pipelined memory cycles |
US4999769A (en) * | 1987-08-20 | 1991-03-12 | International Business Machines Corporation | System with plural clocks for bidirectional information exchange between DMA controller and I/O devices via DMA bus |
US5010476A (en) * | 1986-06-20 | 1991-04-23 | International Business Machines Corporation | Time multiplexed system for tightly coupling pipelined processors to separate shared instruction and data storage units |
US5029289A (en) * | 1987-12-21 | 1991-07-02 | Ncr Corporation | Character display system |
Family Cites Families (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4392200A (en) * | 1980-01-28 | 1983-07-05 | Digital Equipment Corporation | Cached multiprocessor system with pipeline timing |
FR2491703B1 (en) * | 1980-10-03 | 1988-04-29 | Thomson Csf | COMPRESSION DEVICE AND TEMPORAL DATA DECOMPRESSION DEVICE AND TRANSMISSION SYSTEM COMPRISING AT LEAST ONE OF SUCH DEVICES |
JPS6068461A (en) * | 1983-09-21 | 1985-04-19 | Nec Corp | Memory multiplex access device |
DE3334797A1 (en) * | 1983-09-26 | 1985-01-03 | Siemens AG, 1000 Berlin und 8000 München | MULTIPROCESSOR COMPUTER, ESPECIALLY MULTIPROCESSOR CENTRAL CONTROL UNIT OF A TELEPHONE SWITCHING SYSTEM |
GB8401807D0 (en) * | 1984-01-24 | 1984-02-29 | Int Computers Ltd | Pipelined data processing apparatus |
US4685088A (en) * | 1985-04-15 | 1987-08-04 | International Business Machines Corporation | High performance memory system utilizing pipelining techniques |
JPS63155340A (en) * | 1986-12-19 | 1988-06-28 | Fujitsu Ltd | Reading system for storage device |
JPS6421786A (en) * | 1987-07-15 | 1989-01-25 | Nec Corp | Semiconductor memory |
JPS6443894A (en) * | 1987-08-10 | 1989-02-16 | Nec Corp | Semiconductor memory |
US5206833A (en) * | 1988-09-12 | 1993-04-27 | Acer Incorporated | Pipelined dual port RAM |
US5220201A (en) * | 1990-06-26 | 1993-06-15 | Canon Kabushiki Kaisha | Phase-locked signal generator |
JP2765245B2 (en) * | 1991-02-07 | 1998-06-11 | 日本電気株式会社 | Serial clock generation circuit |
JPH04262450A (en) * | 1991-02-15 | 1992-09-17 | Omron Corp | Processor |
JP2776098B2 (en) * | 1991-11-27 | 1998-07-16 | 松下電器産業株式会社 | Clock recovery circuit and time axis error correction device |
-
1993
- 1993-04-22 US US08/052,073 patent/US5471607A/en not_active Expired - Lifetime
-
1994
- 1994-04-22 DE DE69432014T patent/DE69432014T2/en not_active Expired - Lifetime
- 1994-04-22 WO PCT/US1994/004455 patent/WO1994024628A1/en active IP Right Grant
- 1994-04-22 JP JP6517429A patent/JPH08509082A/en active Pending
- 1994-04-22 EP EP94914890A patent/EP0695444B1/en not_active Expired - Lifetime
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US3896418A (en) * | 1971-08-31 | 1975-07-22 | Texas Instruments Inc | Synchronous multi-processor system utilizing a single external memory unit |
US4050058A (en) * | 1973-12-26 | 1977-09-20 | Xerox Corporation | Microprocessor with parallel operation |
US4459688A (en) * | 1979-11-26 | 1984-07-10 | Hitachi, Ltd. | Access request selecting circuit |
US4866604A (en) * | 1981-10-01 | 1989-09-12 | Stratus Computer, Inc. | Digital data processing apparatus with pipelined memory cycles |
US4547845A (en) * | 1982-04-21 | 1985-10-15 | The United States Of America As Represented By The Secretary Of The Navy | Split-BUS multiprocessor system |
US4621362A (en) * | 1984-06-04 | 1986-11-04 | International Business Machines Corp. | Routing architecture for a multi-ring local area network |
US4817006A (en) * | 1986-03-28 | 1989-03-28 | Thomas Engineering, Inc. | Pharmaceutical tablet press control mechanism |
US5010476A (en) * | 1986-06-20 | 1991-04-23 | International Business Machines Corporation | Time multiplexed system for tightly coupling pipelined processors to separate shared instruction and data storage units |
US4999769A (en) * | 1987-08-20 | 1991-03-12 | International Business Machines Corporation | System with plural clocks for bidirectional information exchange between DMA controller and I/O devices via DMA bus |
US5029289A (en) * | 1987-12-21 | 1991-07-02 | Ncr Corporation | Character display system |
Non-Patent Citations (1)
Title |
---|
See also references of EP0695444A4 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6466221B1 (en) | 1993-10-15 | 2002-10-15 | Hitachi, Ltd. | Data processing system and image processing system |
US7254737B2 (en) | 1993-10-15 | 2007-08-07 | Renesas Technology Corp. | Data processing system and image processing system |
EP0649100A2 (en) * | 1993-10-15 | 1995-04-19 | Hitachi, Ltd. | Data processing system and image processing system |
US8332683B2 (en) | 1993-10-15 | 2012-12-11 | Renesas Electronics Corporation | Data processing system and image processing system |
US6550014B2 (en) | 1993-10-15 | 2003-04-15 | Hitachi, Ltd. | Data processing system and image processing system |
US5999197A (en) * | 1993-10-15 | 1999-12-07 | Hitachi, Ltd. | Synchronized data processing system and image processing system |
EP1003107A1 (en) * | 1993-10-15 | 2000-05-24 | Hitachi, Ltd. | Data processing system and image processing system |
US6097404A (en) * | 1993-10-15 | 2000-08-01 | Hitachi, Ltd. | Data processing system and image processing system |
US6288728B1 (en) | 1993-10-15 | 2001-09-11 | Hitachi, Ltd. | Data processing system and image processing system |
US7711976B2 (en) | 1993-10-15 | 2010-05-04 | Renesas Technology Corp. | Data processing system and image processing system |
US5713011A (en) * | 1993-10-15 | 1998-01-27 | Hitachi, Ltd. | Synchronized data processing system and image processing system |
EP0649100A3 (en) * | 1993-10-15 | 1997-02-26 | Hitachi Ltd | Data processing system and image processing system. |
US6789210B2 (en) | 1993-10-15 | 2004-09-07 | Renesas Technology Corp. | Data processing system having memory including mode register |
US6785833B2 (en) | 1993-10-15 | 2004-08-31 | Renesas Technology Corp. | Data processing system and image processing system |
WO1999019875A2 (en) * | 1997-10-10 | 1999-04-22 | Rambus Incorporated | Apparatus and method for pipelined memory operations |
US6963956B2 (en) | 1997-10-10 | 2005-11-08 | Rambus Inc. | Apparatus and method for pipelined memory operations |
WO1999019875A3 (en) * | 1997-10-10 | 1999-09-02 | Rambus Inc | Apparatus and method for pipelined memory operations |
US7330951B2 (en) | 1997-10-10 | 2008-02-12 | Rambus Inc. | Apparatus and method for pipelined memory operations |
US7353357B2 (en) | 1997-10-10 | 2008-04-01 | Rambus Inc. | Apparatus and method for pipelined memory operations |
US6356975B1 (en) | 1997-10-10 | 2002-03-12 | Rambus Incorporated | Apparatus and method for pipelined memory operations |
US6718431B2 (en) | 1997-10-10 | 2004-04-06 | Rambus Inc. | Apparatus and method for pipelined memory operations |
Also Published As
Publication number | Publication date |
---|---|
US5471607A (en) | 1995-11-28 |
EP0695444B1 (en) | 2003-01-15 |
DE69432014D1 (en) | 2003-02-20 |
EP0695444A4 (en) | 2001-09-12 |
DE69432014T2 (en) | 2003-11-20 |
EP0695444A1 (en) | 1996-02-07 |
JPH08509082A (en) | 1996-09-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US5471607A (en) | Multi-phase multi-access pipeline memory system | |
US5787488A (en) | Multi-phase multi-access pipeline memory system | |
US6061779A (en) | Digital signal processor having data alignment buffer for performing unaligned data accesses | |
US5408626A (en) | One clock address pipelining in segmentation unit | |
EP0053457B1 (en) | Data processing apparatus | |
JP2569323B2 (en) | Coprocessors-Architecture | |
EP0345325B1 (en) | A memory system | |
US4197579A (en) | Multi-processor for simultaneously executing a plurality of programs in a time-interlaced manner | |
US5381360A (en) | Modulo arithmetic addressing circuit | |
US4229801A (en) | Floating point processor having concurrent exponent/mantissa operation | |
US3753236A (en) | Microprogrammable peripheral controller | |
JP3531167B2 (en) | System and method for assigning tags to instructions to control instruction execution | |
US5960455A (en) | Scalable cross bar type storage controller | |
KR880001170B1 (en) | Microprocessor | |
GB1600633A (en) | Data processing systems | |
WO1994003860A1 (en) | Massively parallel computer including auxiliary vector processor | |
JPH05224929A (en) | Computer, pipeline processing method and instruction processing method | |
JP2008181551A (en) | Vector tailgating for computer provided with vector register | |
JP3510729B2 (en) | Instruction execution method and instruction execution device | |
GB2153558A (en) | Pipelined data processing | |
US4089052A (en) | Data processing system | |
US4974157A (en) | Data processing system | |
US5363490A (en) | Apparatus for and method of conditionally aborting an instruction within a pipelined architecture | |
US4740892A (en) | Microcomputer having peripheral functions | |
EP0057312B1 (en) | Subroutine control circuitry |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): JP |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): AT BE CH DE DK ES FR GB GR IE IT LU MC NL PT SE |
|
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 1994914890 Country of ref document: EP |
|
WWP | Wipo information: published in national office |
Ref document number: 1994914890 Country of ref document: EP |
|
WWG | Wipo information: grant in national office |
Ref document number: 1994914890 Country of ref document: EP |