WO2004008328A1 - Acces a une memoire depuis differents domaines d'horloge - Google Patents

Acces a une memoire depuis differents domaines d'horloge Download PDF

Info

Publication number
WO2004008328A1
WO2004008328A1 PCT/EP2003/006642 EP0306642W WO2004008328A1 WO 2004008328 A1 WO2004008328 A1 WO 2004008328A1 EP 0306642 W EP0306642 W EP 0306642W WO 2004008328 A1 WO2004008328 A1 WO 2004008328A1
Authority
WO
WIPO (PCT)
Prior art keywords
clock signal
clock
memory
processing unit
data lines
Prior art date
Application number
PCT/EP2003/006642
Other languages
English (en)
Inventor
Andreas Ekvall
Original Assignee
Telefonaktiebolaget Lm Ericsson (Publ)
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from EP02388045A external-priority patent/EP1380960B1/fr
Application filed by Telefonaktiebolaget Lm Ericsson (Publ) filed Critical Telefonaktiebolaget Lm Ericsson (Publ)
Priority to AU2003242750A priority Critical patent/AU2003242750A1/en
Publication of WO2004008328A1 publication Critical patent/WO2004008328A1/fr

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
    • G06F13/14—Handling requests for interconnection or transfer
    • G06F13/16—Handling requests for interconnection or transfer for access to memory bus
    • G06F13/1668—Details of memory controller
    • G06F13/1689—Synchronisation and timing concerns

Definitions

  • the invention relates to a method of controlling access to a common memory module from a number of different processing units clocked by different clock signals, wherein the number of different processing units share common address lines and data lines connected to the memory module.
  • the method comprises the steps of controlling the access to the common address and data lines by control signals generated by a memory control circuit so that only one of said processing units is allowed to drive said common address and data lines at a given time, and generating said control signals in dependence of the first clock signal when a processing unit clocked by the first clock signal is allowed to drive said common address and data lines, and in de- pendence of the second clock signal when a processing unit clocked by the second clock signal is allowed to drive said common address and data lines.
  • the invention further relates to a memory control circuit for controlling access to a common memory module from a number of different processing units clocked by different clock signals.
  • a proc- essing unit can be any digital circuit capable of exchanging data with a memory.
  • central processing units, controllers and input/output de ⁇ vices may be mentioned.
  • Control circuits of the type mentioned here are often referenced to as memory arbiters or memory arbitration circuits, and as an example they are often used in digital ASIC (Application Specific Inte- grated Circuit) designs.
  • CONFIRMATION COPV data lines connecting the processing unit to the memory with a number of control lines, which may comprise e.g. clock signals, select signals and a signal indicating whether the processing unit is reading or writing data from/to the memory.
  • control lines may comprise e.g. clock signals, select signals and a signal indicating whether the processing unit is reading or writing data from/to the memory.
  • the processing unit When, for instance, the processing unit is going to write data into the memory, it drives the address lines with the relevant memory location, the data lines with the data to be written, and the control lines appropriately.
  • reading the processing unit drives the address lines, and the memory responds by driving either the same data lines or a separate set of reading data lines with the read data.
  • Processing units needing access to the memory send a request signal to the arbiter, and based on the arbitration scheme used the arbiter grants access to the lines for a clock cycle to one of the processing units which is then allowed to drive the address, data and control lines during this clock cycle.
  • US 5 179 667 discloses a memory control apparatus using two different clock rates and intended to solve these synchronization problems.
  • a CPU is running at a fast speed (i.e. at a high clock rate), while an input/output device is running at a slow speed (i.e. at a low clock rate).
  • the CPU's clock rate governs, and when an input/output device is accessing memory the input/output device's clock rate governs.
  • this apparatus is able to control the access to the memory from the two clock domains, the memory access will be delayed because wait states will have to be inserted each time the control apparatus switches between requests from the different clock domains.
  • the object is achieved in that the method further comprises the steps of sampling the second clock signal by the first clock signal, generating from results obtained by the sampling a sampled version of the second clock signal having edges defined by edges of the first clock signal, and synchronizing said control signals with the sampled version of the second clock signal when a processing unit clocked by the second clock signal is allowed to drive said common address and data lines.
  • the speed of the memory access can be improved further, because the address lines and the data lines for data to be read from the memory module can be driven by a processing unit in every clock cycle. Otherwise a further clock cycle would have been occupied for each read request, because the requesting unit and the memory cannot drive the same data lines in the same clock cycle. This is particularly relevant when clocked memory modules are used.
  • the method further comprises the step of storing in a register data to be read by a processing unit clocked by the second clock signal for at least one clock period of the sampled version of the second clock signal, the data lines for read data will only be occupied for one clock cycle of the first clock signal, even when a processing unit clocked by the second (slower) clock signal reads data from the memory.
  • the method further comprises the step of generating said control signals allowing a processing unit to drive said common address and data lines such that control signals generated in dependence of the first clock signal as well as control signals generated in dependence of the second clock signal are active for a time period equal to one clock cycle of the first clock signal, the address lines may be driven by a processing unit clocked by the first clock signal again even before a memory access governed by the second clock signal is completed.
  • a request may be granted immediately in the clock cycle in which it occurs, thus improving the memory access further.
  • the method further comprises the step of generating said control sig- nals allowing a processing unit to drive said common address and data lines in a number of different state machines comprising at least one state machine for each of said different clock rates, each state machine depends on only one clock signal, which simplifies the design.
  • the invention also relates to a memory control circuit for controlling access to a common memory module from a number of different processing units arranged to be clocked by at least two different clock signals, a first one of said different clock signals having a first clock rate and a second clock signal having a second clock rate which is lower than the rate of the first clock signal, where the number of different processing units share common address lines and data lines connected to the memory module, the memory control circuit further being arranged to control the access to the common address and data lines by control signals generated by the memory control circuit so that only one of said processing units is allowed to drive said common address and data lines at a given time, and generate said control signals in dependence of the first clock signal when a processing unit clocked by the first clock signal is allowed to drive said common address and data lines, and in dependence of the second clock signal when a processing unit clocked by the second clock signal is allowed to drive said common ad- dress and data lines.
  • the memory control circuit is further arranged to sample the second clock signal by the first clock signal; generate from results obtained by the sampling a sampled version of the second clock signal having edges defined by edges of the first clock signal; and synchronize said control signals with the sampled version of the second clock signal when a processing unit clocked by the second clock signal is allowed to drive said common address and data lines, a faster memory access is provided even when the control is switched between requests from the different clock domains.
  • the speed of the memory access can be improved further, because the address lines and the data lines for data to be read from the memory module can be driven by a processing unit in every clock cycle. Otherwise a further clock cycle would have been occupied for each read request, because the requesting unit and the memory cannot drive the same data lines in the same clock cycle. This is particularly relevant when clocked memory modules are used.
  • the memory control circuit further comprises a register for storing data to be read by a processing unit clocked by the second clock signal for at least one clock period of the sampled version of the second clock signal, the data lines for read data will only be occupied for one clock cycle of the first clock signal, even when a processing unit clocked by the second (slower) clock signal reads data from the memory.
  • the memory control circuit is further arranged to generate said control signals allowing a processing unit to drive said common address and data lines such that control signals generated in dependence of the first clock sig- nal as well as control signals generated in dependence of the second clock signal are active for a time period equal to one clock cycle of the first clock signal
  • the address lines may be driven by a processing unit clocked by the first clock signal again even before a memory access governed by the second clock signal is completed.
  • a request may be granted immediately in the clock cycle in which it occurs, thus improving the memory access further.
  • each state machine depends on only one clock signal, which simplifies the design.
  • figure 1 shows a block diagram of a known memory control system
  • figure 2 shows an example of a timing diagram for the system of figure 1 ,
  • figure 3 shows a block diagram of a memory control system handling requests from two different clock domains
  • figure 4 shows an example of a timing diagram for the system of figure 3
  • figure 5 shows a block diagram of a memory control circuit according to the invention
  • figure 6 shows an example of a timing diagram for a system with the memory control circuit of figure 5
  • figure 7 shows a block diagram of a memory control system in which data outputs to the slower clock domain are registered
  • figure 8 shows an example of a timing diagram for the system of figure 7,
  • figure 9 shows an example of a timing diagram for a memory control system in which all bus grant signals are clocked on the faster clock signal
  • figure 10 shows a state diagram for a state machine handling the faster clock domain in a system corresponding to the timing diagram of figure 9
  • figure 11 shows a state diagram for a state machine handling the slower clock domain in a system corresponding to the timing diagram of figure 9, and
  • figure 12 shows a state diagram for an acknowledge signal to the slower clock domain in a system corresponding to the timing diagram of figure 9.
  • Figure 1 shows a known memory control system 1 in which two processing units 2, 3 (Unit A and Unit B) can both access a memory 4, which is here a RAM (Random Access Memory) of the clocked type, which means that read data will be available the clock cycle after an address is clocked in.
  • a common address bus with a number of address lines allows the processing units 2, 3 to supply addresses to the memory 4, while two separate data buses are used for reading and writing data, respectively.
  • the separate buses for read- ing and writing data are normally used with clocked RAMs, as in this example, while a common data bus for reading and writing data is often sufficient for other types of memory.
  • a Wr/Rd signal indicates to the memory whether a processing unit is reading or writing data.
  • the access to the memory is controlled by a memory control circuit 5, which in the following is also called a memory arbiter and works according to an arbitration scheme, such as random, First In-First Out, round-robin or priority based arbitration.
  • a single clock signal is used by both processing units 2, 3 as well as the memory 4 and the memory arbiter 5.
  • the processing units 2, 3 request access to the memory by means of request signals (A Req and B Req) sent to the memory arbiter, which in turn can allow one of the processing units to drive the address and data buses for one clock cycle by means of control signals in the form of bus grant and acknowl- edge signals (A BusGr, B BusGr, A Ackn and B Ackn).
  • a bus grant signal to one of the units means that this unit may drive the common buses/lines for address, data and the Wr/Rd signal, while an acknowledge signal tells the requesting unit that read data are available on the read data bus.
  • a number of control lines may also be connected from the memory arbiter 5 to the memory 4. These lines may include the clock signal, select signals, output enable signals or similar control signals depending on the type of memory used, and for some types of memory they may be avoided.
  • Figure 2 shows a timing diagram illustrating an example of how the memory arbiter controls the access to the memory from the two processing units.
  • the clock signal Clk is shown at the top of the diagram, while the other signals correspond to the signals shown in figure 1. It is noted that all signal shifts are initiated by a positive going edge of the clock signal Clk, because all circuits are clocked by this clock signal.
  • the selected arbitration scheme en- sures that the granted unit should change every clock cycle, when more than one unit is requesting access simultaneously. If two new requests arrive at the same time, it is decided according to the arbitration scheme which unit should be granted the first clock cycle. In this case it is assumed that Unit B is given a higher priority than Unit A.
  • Unit B At the time ti requests to read data from the memory are received from Unit A as well as Unit B. Since Unit B has the higher priority it is granted the first clock cycle by activating the signal B BusGr. This signal is generated combi- natorially so that it is available directly in the same clock period. As a re- sponse to this signal, Unit B drives the address bus with the address of the relevant memory location, and also this happens combinatorially in the same clock period. In the next clock period Unit A is allowed to drive the address bus, and at the same time the data requested by Unit B are now available on the RdData bus, which is acknowledged to Unit B by activating the signal B Ackn. Since both units are requesting a further access, Unit B is again granted the next clock period followed by Unit A.
  • FIG. 1 both processing units belonged to a single clock domain, i.e. they were clocked by the same clock signal that was also used for the memory arbiter and the memory.
  • Figures 3 and 4 illustrate some of the problems that arise when processing units belonging to different clock domains request access to a common memory.
  • the processing unit 2 (Unit A) is a processing unit similar to Unit A in figure 1 , and it belongs to a clock domain clocked by the clock signal Clk1 corresponding to the signal Clk shown in figure 2.
  • further processing units e.g. Unit B, Unit C and Unit D, may belong to the same clock domain.
  • the processing unit 13 (Unit 1 ) belongs to a different clock domain clocked by the clock signal Clk2. Although not shown, it is assumed that further processing units, e.g. Unit 2 and Unit 3, may belong to the same clock domain. As seen in figure 4, Clk2 has a lower clock rate than Clk As examples, Unit A, Unit B, etc. may be CPUs, e.g. running on a clock frequency in the range 10-40 MHz, while Unit 1 , Unit 2, etc., may be input/output devices such as Ethernet interfaces, disk controllers and printer interfaces, which are normally running on considerably lower clock frequencies. Further, each of the two clock signals will often be free running, i.e. they are not synchronized to each other.
  • the memory arbiter 15 is necessarily more complex than the memory arbiter 5 of figure 1 , because it must be able to handle both clock domains. It should be governed by Clk1 when Unit A is accessing memory, and by Clk2 when Unit 1 is accessing memory, and both clock signals are therefore connected to the memory arbiter 15 and through the memory control bus also to the memory 4.
  • Unit A requests a number of read operations, and since no other units are accessing the memory at this time, access is granted immediately for a number of consecutive clock cycles. However, at time t 3 , which is in the middle of a Clk1 clock cycle, Unit 1 requests a write operation to the memory. Although it is assumed in this example that the units belonging to the Clk2 clock domain always have a higher priority than the units belonging to the Clk1 clock domain, the read operation already in progress by Unit A cannot be interrupted. This operation is completed at time , and the memory arbiter 15 now switches to be governed by Clk2.
  • FIGs 5 and 6 show how this situation can be improved according to the invention.
  • the memory arbiter 25 shown in figure 5 is similar to the memory arbiter 15 in figure 3, except for the important fact that it includes a sample/hold circuit 26 allowing the clock signal Clk2 to be sampled by the clock signal Clk1.
  • the result is a sampled clock signal Clk2s having its edges aligned with the edges of Clk1.
  • the sampled clock signal Clk2s may have some jitter, so that the duration of the clock cycles may vary, but this does not create any problems. Actually, jitter would also be allowed in the original clock signal Clk2.
  • Clk2 is sampled on the positive edges of Clk1 , but of course it could just as well be sampled on the negative edges.
  • the memory arbiter 25 is governed by Clk1 when Unit A is accessing memory, but when Unit 1 is accessing memory the memory arbiter 25 is now governed by Clk2s instead of CIK2.
  • the important difference is that the two governing clock signals Clk1 and Clk2s are now "synchronized" to each other so that a switch from one clock domain to the other will always take place on coinciding edges of both clock signals.
  • the memory arbiter circuit 25 can be implemented as two state machines, one running on Clk1 and one running on Clk2s.
  • Unit A requests a number of read operations at time t 2 , and again access is granted immediately, because no other units are accessing the memory at this time.
  • the request from Unit 1 for a write operation is also again received at time t 3 .
  • the read operation already in progress is completed at time t as before, and the memory arbiter 25 now switches to be governed by Clk2s.
  • this switch occurs at a positive edge of Clk2s, and thus there is no need to wait for a subsequent positive clock edge before the memory arbiter can perform the actions governed by the clock signal Clk2s.
  • bus access is granted to Unit 1 at time t 4 instead of waiting until ts as in figure 4, and Unit 1 can perform the write operation in the Clk2s clock cycle following t 4 .
  • Unit 1 needed three complete clock cycles of Clk2 or Clk2s to perform a write and a read operation, because the read data must be available to Unit 1 for a full clock cycle after the address has been clocked in.
  • RdData data bus
  • the registers 37 may be clocked on either a positive or a negative edge of Clk1. Here they are shown clocked on a negative edge.
  • Unit 1 drives the address and write data buses for the two full Clk2s clock cycles and it is granted thereto by the bus grant signal (1 BusGr). However, this is not needed. A Clk1 clock cycle would be sufficient for the memory 4 to catch the information.
  • Unit 1 drives the buses combinatorially based on the bus grant signal, it will thus be sufficient to provide the bus grant signal (1 BusGr) for one Clk1 clock cycle, and Unit 1 will then only drive the buses for that Clk1 clock cycle.
  • the bus grant and acknowledge signals of the Clk2 domain may be clocked by the Clk1 clock signal. This means that both state machines are clocked on Clk1 , while the conditions for switching from one state machine to another is still governed by Clk2s. In this way the address and data buses may be made available for units clocked by Clk1 even during the time where Unit 1 is performing a read or write operation.
  • Unit A no longer has to wait for Unit 1 to complete its write and read operations before the remaining read operation can be performed. It also means that the read data will actually be available to the registers 37 earlier than before, i.e. already at time t ⁇ 2 , and to ensure the correct function of the slower Unit 1 the registers 37 may therefore be clocked on the negative edge of Clk1 , as can be seen in the figure.
  • the memory arbiter corresponding to the timing diagram of figure 9 can be implemented as two separate state machines, each of them handling one clock domain.
  • a connection between the two state machines allows the state machine for the faster clock to know when to wait for the slower clock.
  • a request from the faster clock domain will be served immediately, if possible, to avoid wait states, while an access from the slower clock only needs to be served before the next slower clock cycle.
  • the sampling of the slower clock signal by the faster clock signal ensures that the arbitration will be clock skew tolerant.
  • the slower units i.e. those belonging to the Clk2 clock domain, will have a bus grant signal clocked on Clk1 , but they will still manage a read or write operation in one Clk2 clock cycle. In this way the state machine handling the Clk2 clock domain can tell the state machine handling the Clk1 clock domain when it gives bus grant so that there will be no bus collisions, and this is achieved without introducing any extra guard clock cycles.
  • Figure 10 shows a state diagram for the state machine handling the Clk1 clock domain. It is seen that four processing units, i.e. Unit A, Unit B, Unit C and Unit D, belong to this clock domain in the example shown. For clarity reasons not all signals and conditions are shown. As an example, the condition for the transition from the IdleState to UnitA ReqState is shown, while the corresponding conditions for the other units are similar. It is also seen that the Clk2 clock domain has a higher priority so that a request from this domain will transfer control to the other state machine by a transition to the Clk2 ReqState. A state diagram for the state machine handling the Clk2 clock domain is shown in figure 11. Here it is seen that among the three units, i.e.
  • Unit 1 Unit 1 , Unit 2 and Unit 3, belonging to this domain in the example, Unit 1 has the highest priority and Unit 3 the lowest.
  • the state where read data are made available to the requesting Clk2 unit by means of the register enable signal is also shown.
  • the acknowledge signal to the units in the Clk2 clock domain can be clocked as shown in figure 12, which corresponds to the timing diagram of figure 9.
  • the acknowledge signals are clocked so that they are stable and free of glitches for both clock domains.
  • the multiplexing of the address, write data and write/read signals can be performed by OR-gates.
  • the units are supposed to drive zero on all these signals when not receiving a bus grant signal.
  • the OR multiplexing can be used, which saves gates compared to normal multiplexers.
  • the multiplexing can also be distributed by performing the OR functions in multiple steps. This will minimize the routing need.
  • the drive outputs of the units may be tri-stated as usual.
  • This design will arbitrate between all requesting units from both clock do- mains without any delays and with minimum throughput delay.
  • a write from the fast clock domain will be performed in one clock cycle, and a read will be performed in two clock cycles.
  • a request from the slower clock domain may be completed in one slow clock cycle for both read and write operations.
  • the arbitration is fair, which means that the granted unit switches every clock cy- cle if more than one unit is requesting at the same time.
  • the initial priority could be fixed since it is only the first access that some unit may get before the others. After that all the other units get their turn in e.g. a Round-Robin fashion. The initial priority could also be programmable if this would be advantageous to the design.
  • the slower clock signal Clk2 is sampled on the negative (falling) edges of the faster clock signal Clk1. This requires that in each Clk2 clock cycle the Clk2 signal must be high as well as low for at least one full Clk1 clock cycle in order for the sampling to work correctly. Therefore, the Clk2 clock cycle needs to be at least twice as long as the Clk1 clock cycle (i.e. half the clock rate).
  • Memories of the mentioned DRAM type, or other memory types including ROMs and flash memories, may of course be used instead of the clocked RAM in the examples described above.
  • ROMs and flash memories may of course be used instead of the clocked RAM in the examples described above.
  • a common data bus for reading and writing data will often be used instead of the two separate buses described above.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Bus Control (AREA)

Abstract

Système permettant de contrôler l'accès à un module commun de mémoire depuis des unités de traitement rythmées par au moins deux signaux d'horloge différents. Ces différentes unités de traitement partagent des lignes d'adresse et des lignes de données communes. L'accès aux lignes communes est contrôlé par des signaux de contrôle générés par un circuit de contrôle de mémoire (25), de sorte que seule une unité est autorisée à commander les lignes communes à un moment donné. Les signaux de contrôle sont générés en fonction du premier signal d'horloge quand une unité rythmée par ce signal d'horloge commande lesdites lignes communes et en fonction du deuxième signal d'horloge quand une unité rythmée par ce signal d'horloge commande la ligne. Le deuxième signal d'horloge est échantillonné par le premier signal d'horloge et les signaux de contrôle sont synchronisés avec une version échantillonnée du deuxième signal d'horloge quand une unité de traitement rythmée par le deuxième signal d'horloge commande les lignes communes.
PCT/EP2003/006642 2002-07-12 2003-06-24 Acces a une memoire depuis differents domaines d'horloge WO2004008328A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
AU2003242750A AU2003242750A1 (en) 2002-07-12 2003-06-24 Memory access from different clock domains

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
EP02388045.3 2002-07-12
EP02388045A EP1380960B1 (fr) 2002-07-12 2002-07-12 Accès à mémoire des domaines d'horloge différents
US39677702P 2002-07-18 2002-07-18
US60/396,777 2002-07-18

Publications (1)

Publication Number Publication Date
WO2004008328A1 true WO2004008328A1 (fr) 2004-01-22

Family

ID=30116931

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2003/006642 WO2004008328A1 (fr) 2002-07-12 2003-06-24 Acces a une memoire depuis differents domaines d'horloge

Country Status (2)

Country Link
AU (1) AU2003242750A1 (fr)
WO (1) WO2004008328A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7707592B2 (en) 2003-10-10 2010-04-27 Telefonaktiebolaget L M Ericsson (Publ) Mobile terminal application subsystem and access subsystem architecture method and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2053537A (en) * 1979-07-10 1981-02-04 Lucas Industries Ltd Digital Computing Apparatus
US5671393A (en) * 1993-10-01 1997-09-23 Toyota Jidosha Kabushiki Kaisha Shared memory system and arbitration method and system
US5761533A (en) * 1992-01-02 1998-06-02 International Business Machines Corporation Computer system with varied data transfer speeds between system components and memory

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
GB2053537A (en) * 1979-07-10 1981-02-04 Lucas Industries Ltd Digital Computing Apparatus
US5761533A (en) * 1992-01-02 1998-06-02 International Business Machines Corporation Computer system with varied data transfer speeds between system components and memory
US5671393A (en) * 1993-10-01 1997-09-23 Toyota Jidosha Kabushiki Kaisha Shared memory system and arbitration method and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7707592B2 (en) 2003-10-10 2010-04-27 Telefonaktiebolaget L M Ericsson (Publ) Mobile terminal application subsystem and access subsystem architecture method and system

Also Published As

Publication number Publication date
AU2003242750A1 (en) 2004-02-02

Similar Documents

Publication Publication Date Title
US4050097A (en) Synchronization technique for data transfers over an asynchronous common bus network coupling data processing apparatus
US8078781B2 (en) Device having priority upgrade mechanism capabilities and a method for updating priorities
EP0476990A2 (fr) Arbitrage dynamique de bus
US20010042178A1 (en) Data path architecture and arbitration scheme for providing access to a shared system resource
JP2510799B2 (ja) デ―タ処理装置およびメモリコントロ―ラ
EP1812868A1 (fr) Procede et dispositif pour operation atomique semaphore
US6973078B2 (en) Method and apparatus for implementing low latency crossbar switches with integrated storage signals
US7007121B1 (en) Method and apparatus for synchronized buses
Yun et al. A high-performance asynchronous SCSI controller
US5627968A (en) Data transfer apparatus which allows data to be transferred between data devices without accessing a shared memory
US8799699B2 (en) Data processing system
US20100169525A1 (en) Pipelined device and a method for executing transactions in a pipelined device
US6189076B1 (en) Shared synchronous memory with a switching circuit controlled by an arbiter and method for glitch free switching of a clock signal
EP1380960B1 (fr) Accès à mémoire des domaines d'horloge différents
EP0886218B1 (fr) Schéma à multiplexage temporel de résolution d'interblocage dans l'arbitrage distribué
US5377334A (en) Fast asynchronous resource master-slave combination
WO2004008328A1 (fr) Acces a une memoire depuis differents domaines d'horloge
US5446847A (en) Programmable system bus priority network
JPH10307787A (ja) バッファメモリ装置
US6910091B1 (en) Arbitration method of a bus bridge
JPH09223103A (ja) 情報処理システム
JP2502030B2 (ja) 同期式デ―タ処理システム用の同期化装置
EP0284094B1 (fr) Dispositif de décision de priorité en cascade
WO2003014948A1 (fr) Architecture systeme d'un module de commutation a debit binaire eleve entre des unites fonctionnelles dans un systeme sur puce
JP2001167049A (ja) バス調停装置

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP

WWW Wipo information: withdrawn in national office

Country of ref document: JP