PatentDe  


Dokumentenidentifikation EP0542318 26.10.2000
EP-Veröffentlichungsnummer 0542318
Titel Digitales Verarbeitungsgerät und Steuerverfahren dafür
Anmelder NEC Corp., Tokio/Tokyo, JP
Erfinder Goto, Junichi, Minato-ku, Tokyo, JP
Vertreter Glawe, Delfs, Moll & Partner, Patentanwälte, 80538 München
DE-Aktenzeichen 69231463
Vertragsstaaten DE, FR, GB
Sprache des Dokument EN
EP-Anmeldetag 16.11.1992
EP-Aktenzeichen 921195731
EP-Offenlegungsdatum 19.05.1993
EP date of grant 20.09.2000
Veröffentlichungstag im Patentblatt 26.10.2000
IPC-Hauptklasse G06F 9/34
IPC-Nebenklasse G06T 1/60   

Beschreibung[en]
Background of the Invention Field of the invention

The present invention relates to a digital processor and a method for controlling the same, and more specifically to a digital signal processor having a high performance, and a method for controlling the same.

Description of related art

In conventional digital processors, generation of address is based on an adding operation, and therefore, both of generation for a memory address and an arithmetic and logical operation of data have been performed in common by an arithmetic logic unit (ALU). However, this combined use has become a hindrance in elevation of performance. In particular, in a digital signal processor (DSP) in which processings including many repeated processings, such as a sum-of-products operation indispensable for signal processings, are executed, it is important that the address generation is made independently of the arithmetic and logical operation of data, so that tile processing is effectively performed.

Recently, Kaneko et al proposed a digital signal processor in "1987 IEEE International Solid-State Circuit Conference", "Digest of Technical Papers", February 1987, pp 158-159, published by IEEE. The proposed digital signal processor includes an address generation circuit composed of an addition circuit having one input connected to an output of an address register, whose input is connected to an output of the addition circuit itself. The other input of the addition circuit is connected to receive through a selector a fixed value "+1", an output of a first displacement register, or an output of a second displacement register. A selection signal for the selector is supplied from a controller within the processor at each operation clock period.

This type of address generation is important in an image processing in which each one pixel is stored at one individual address. Here, consider that a rectangular region is accessed by assuming that the memory has M(=16) pixels in a horizontal direction and N(=20) pixels in a vertical direction and the rectangular region has "m"(=3) pixels in a horizontal direction and "n"(=4) pixels in a vertical direction. If the rectangular region is accessed from an upper left corner, a first horizontal access from the upper left corner to an upper right corner of the rectangular region can be performed by adding "+1" to a current address stored in the address register. Therefore, the selector selects "+1". When the access is moved from the upper right corner to a left end of a second line of of the rectangular region, a required address change is obtained by adding {M-(m-1)} (=14) to the current address. Therefore, this displacement value of "14" is previously registered in the first displacement register, and the selector selects this first displacement register. When the access reaches to a lower right corner, the address is moved to return to the upper left corner of the next rectangular region. For this purpose, a required address change is obtained by subtracting {(n-1)M+(m-1)} (=50) from the current address. Therefore, this displacement value of "-50" is previously registered in the second displacement register, and the selector selects this second displacement register. Thus, a triple loop processing is performed.

For the loop processing, the number of loops and the end discrimination are described in a program, and the controller decodes the program at each step so as to supply a necessary selection signal to the selector.

In the above mentioned digital processor and the controlling method therefor, if the number of the memories to be accessed simultaneously becomes large, a corresponding number of selection signals must be simultaneously supplied. For elevation of the processing capacity of the processor, it can be sufficiently considered that a simultaneous access to a number of memories is required. However, a controller of many processors is configured so that a so-called horizontal microinstruction is read out at each operation clock period, and then decoded to generate various control signals including the selection signal, so that the various control signals are supplied to different parts of the processor. Therefore, in order to simultaneously supply a number of selection signals, it is necessary to elongate each one microinstruction, which means that it is required to increase the capacity of a microprogram memory storing the microinstructions. This will result in a decreased reading speed and in an increased cost.

In addition, in order to improve the operation clock period, it may become necessary to cause various circuits of the processor including the ALU to operate in a pipelined mode. In this case, because of the number of pipelined steps, the loop processing for generating an address for the memory supplying the data to the ALU and the loop processing for generating an address for the memory storing the result of the operation of the ALU must be executed with a time difference therebetween. However, the two loop processings (or more than two loop processings in some cases) having a time difference from each other cannot be controlled in the conventional manner in which one microinstruction is read out for each one clock period and decoded to perform a necessary processing.

WO-87/0486 discloses a digital processor according to the precharacterizing part of claim 1.

1987 IEEE Int. Conf. on Solid-State Circuits, February 1987, Coral Gables, FL, USA, pages 158 - 159. Kaneko K. et al.: "A 50ns DSP with Parallel Processor Architecture" discloses a digital processor and a method for controlling such a digital processor as shown by Fig. 5A and 5B and discussed above.

Summary of the Invention

Accordingly, it is an object of the present invention to provide a digital processor and a control method therefor, which have overcome the above mentioned defect of the conventional ones.

Another object of the present invention is to provide a digital processor and a control method therefor, which can perform a simultaneous access to a number of memories with high efficiency.

The above and other objects of the present invention are achieved by a digital processor according to claim 1, and by a method for controlling such a digital processor according to claim 8. The further dependent claims are related to different advantageous aspects of the present invention.

The digital processor comprises a memory group composed of "m" memories, each memory "i" (i=1 to "m") of the "m" memories having Pi address ports "ij" (j=1 to Pi), an arithmetic and logic unit group composed of "n" arithmetic and logic units, a first arbitration circuit receiving and arbitrating outputs of the memory group for outputting the arbitrated outputs of of the memory group to the arithmetic and logic unit group, a second arbitration circuit receiving and arbitrating outputs of the arithmetic and logic unit group for outputting the arbitrated outputs of the arithmetic and logic unit group to the memory group, a plurality of address generation circuits each for supplying a generated address to the memory group, a clock supply circuit for supplying respective clock signals to at least the address generation circuits, and a control circuit supplying respective control signals to the memory group, the arithmetic and logic unit, the first arbitration circuit and the second arbitration circuit in synchronism with the clock signals generated by the clock supply circuit, the control circuit operating to previously set address generation rules to the address generation circuits so that the address generation circuits set with the address generation rules operate to generate an address in accordance with the set address generation rule and in synchronism with the clock supplied from the clock signal supply circuit, independently of the control circuit, and the set address generation rules are maintained effective until they are reset.

In one preferred embodiment, each one address generation circuit "ij" is connected to each one of the address ports "ij" in one-to-one relation. In addition, the address generation circuits "ij" are configured to supply a read enable signal or a write enable signal to the memory "i".

In another preferred embodiment, the digital processor further includes an address arbitration circuit receiving the addresses generated by the address generation circuits each of which generates the address in accordance with the set address generation rule and in synchronism with the clock supplied from the clock signal supply circuit. The address arbitration circuit supplies the received addresses to a portion or all of the address ports "ij" in accordance with an address arbitration rule which is set by the control circuit and which is maintained effective until it is reset. In addition, wherein the address generation circuits "ij' are configured to supply a read enable signal or a write enable signal to the address arbitration circuit, and the address arbitration circuit supplies the received read enable signal or the received write enable signal to the portion or all of the address ports in accordance with the preset address arbitration rule.

Preferably, a portion or all of the arithmetic and logic units is configured to have one or more pipeline stages.

According to another aspect of the present invention, there is provided a method for controlling a digital processor which comprises a memory group composed of "m" memories, each memory "i" (i=1 to "m") of the "m" memories having Pi address ports "ij" (j=1 to Pi), an arithmetic and logic unit group composed of "n" arithmetic and logic units, a first arbitration circuit receiving and arbitrating outputs of the memory group for outputting the arbitrated outputs of the memory group to the arithmetic and logic unit group, a second arbitration circuit receiving and arbitrating outputs of the arithmetic and logic unit group for outputting the arbitrated outputs of the arithmetic and logic unit group to the memory group, a plurality of address generation circuits each for supplying a generated address to the memory group, a clock supply circuit for supplying respective clock signals to at least the address generation circuits, and a control circuit supplying respective control signals to the memory group, the arithmetic and logic unit, the first arbitration circuit and the second arbitration circuit in synchronism with the clock signals generated by the clock supply circuit, the control circuit operating to previously set address generation rules to the address generation circuits so that the address generation circuits set with the address generation rules operate to generate an address in accordance with the set address generation rule and in synchronism with the clock supplied from the clock signal supply circuit, independently of the control circuit, and the set address generation rules are maintained effective until they are reset, the method including the steps of selecting one or more of the address generation circuits, setting an address generation rule to each of the selected address generation circuits from the control circuit, supplying address generation start signals to the selected address generation circuits from the control circuit simultaneously or with a time difference so that each of the selected address generation circuits starts its address generation and continues to supply an address to a corresponding memory in synchronism with the clock and independently of the control circuit until the address generation indicated by the address generation rule is completed, each of the selected address generation circuits generating an address generation end signal when an address generation is ended, and a selected one of address generation end signals generated by the selected address generation circuits being supplied to the control circuit.

In one preferred embodiment, each one address generation circuit "ij" is connected to each one of the address ports "ij" in one-to-one relation. In addition, the address generation circuits "ij" are configured to supply a read enable signal or a write enable signal to the memory "i".

In another preferred embodiment, the digital processor further includes an address arbitration circuit receiving the addresses generated by the address generation circuits each of which generates the address in accordance with the set address generation rule and in synchronism with the clock supplied from the clock signal supply circuit. The address arbitration circuit supplies the received addresses to a portion or all of the address ports "ij" in accordance with the address arbitration rule which is set by the control circuit and which is maintained effective effective until it is reset. The address arbitration rule is set to the address arbitration circuit from the control circuit before the address generation rules are set, and each of the selected address generation circuits continues to supply an address to the address arbitration circuit in synchronism with the clock signal and independently of the control circuit until the address generation indicated by the address generation rule is completed, the address arbitration circuit continuing to supply the received addresses to the portion or all of the address ports in accordance with the address arbitration rule which is set by the control circuit and which is maintained effective effective until it is reset. In addition, the address generation circuits "ij" are configured to supply a read enable signal or a write enable signal to the address arbitration circuit, and the address arbitration circuit supplies the received read enable signal or the received write enable signal to the portion or all of the address ports in accordance with the preset address arbitration rule.

With the above mentioned arrangement, if one address generation circuit is provided in a one-to-one relation for each of address ports of each of all the memories, it becomes possible to simultaneously access a number of memories. Each of the address generation circuits is previously supplied from the control circuit with the address generation rule typified by the number of loops, etc., which are explained in the "Description of related art". In response to an address generation start signal supplied from the control circuit, each of the address generation circuits starts its address generation. Therefore, by simultaneously supplying the address generation start signals to a number of address generation circuits, it is possible to simultaneously access a number of memories. During execution of the address generation, each address generation circuit performs the counting of the loops and the discrimination of the loop end, which were described as a program in the prior art and which were decoded and executed by the control circuit.

In addition, In the case that the ALU performs a pipelined operation, the loop processing for generating an address for the memory supplying the data to the ALU and the loop processing for generating an address for the memory storing the result of the operation of the ALU must be executed with a time difference therebetween, because of the number of the pipelined steps. Also in this case, it is possible that after the address generation start signal is supplied to the address generation circuit for the data supplying memory, the address generation start signal is supplied to the address generation circuit for the operation result storing memory with a delay corresponding to the above mentioned time difference. If each of the address generation circuits performs the address generation of the number corresponding to the number of access designated by the control circuit, the address generation circuit supplies an address generation end signal to the control circuit. From the start of the address generation to the generation of the address generation end signal, the address generation circuit continues to generate the address independently of the control circuit.

The above and other objects, features and advantages of the present invention will be apparent from, the following description of preferred embodiments of the invention with reference to the accompanying drawings.

Brief Description of the Drawings

  • Figure 1 is a block diagram of a first embodiment of the digital processor in accordance with the present invention;
  • Figure 2 is a flow chart illustrating an operation of the first embodiment shown in Figure 1;
  • Figure 3 is a block diagram of a second embodiment of the digital processor in accordance with the present invention;
  • Figure 4 is a flow chart illustrating an operation of the first embodiment shown in Figure 3; and
  • Figures 5A and 5B illustrate an example of a rectangular memory and a rectangular region within the rectangular memory, and steps for performing the triple loop operation, respectively.

Description of the Preferred embodiments

Referring to Figure 1, there is shown a block diagram of a first embodiment of the digital processor in accordance with the present invention.

The shown digital processor comprises a memory group 1 including a number of memories 11, 12, &peseta; &peseta; &peseta; , 1m, which are labelled "MEMORY 1", MEMORY 2", &peseta; &peseta; &peseta;, "MEMORY m". The memory 11 includes P1 address ports of "Address Port 211" to "Address Port 21P1", and the memory 12 includes P2 address ports of "Address Port 221" to "Address Port 22P2". Similarly, the memory 1m includes Pm address ports of "Address Port 2ml "to "Address Port 2mPm". Each address port 2ij (i=1 to m, and j=1 to Pi) includes an address generation circuit "1ij". For example, the address port 211 includes an address generation circuit 111, and the address port 2mPm includes an address generation circuit 1mPm. Each of the address generating circuit 1ij is configured to receive a parameter signal through a signal line 5ij from a control circuit 5 and to output an address generation termination signal through a signal line 6ij to the control circuit 5. The control circuit 5 receives a clock signal from a clock signal supply circuit 6, which also supply a clock signal to each of all the address generation circuits.

The shown digital processor also includes an ALU group 2 composed of ALUs 21, 22, &peseta;&peseta;&peseta; , 2n, which are labelled "ALU 1", ALU 2", &peseta;&peseta;&peseta;, "ALU n". As shown, outputs of the memory group 1 are supplied through a first arbiter 3 to the ALU group 2, and outputs of the ALU group 2 are supplied through a second arbiter 4 to the memory group 1. Therefore, the first arbiter 3 receives and arbitrates outputs of the memory group 1 for outputting the arbitrated outputs of of the memory group 1 to the ALU group 3. Similarly, the second arbiter 4 receives and arbitrates outputs of the ALU group 2 for outputting the arbitrated outputs of of the ALU group 2 to the memory group 2. In addition, the clock supply circuit 6 supplies respective clock signals to not only the address generation circuits but also the other circuits as shown. The control circuit 5 also supplies respective control signals to the memory group, the arithmetic and logic unit, the first arbitration circuit and the second arbitration circuit in synchronism with the clock signals generated by the clock supply circuit 6.

Now, operation of the first embodiment will be described with reference to the flow chart of Figure 2.

Steps A and B shown in the flow chart of Figure 2 correspond to an operation for sequentially setting an individual address generation rule to each of one or more address generation circuits desired to generate an address.

Here, referring to Figure 5A and 5B, the triple loop as mentioned hereinbefore will be explained. Figure 5A illustrates a rectangular memory having M(=16) pixels in a horizontal direction and N(=20) pixels in a vertical direction. Within this memory, a rectangular region is set, which has "m"(=3) pixels in a horizontal direction and "n"(=4) pixels in a vertical direction. If the rectangular region is accessed from an upper left corner, a first horizontal access designated by the arrow 1 from the upper left corner to an upper right corner of the rectangular region can be performed by adding "+1" to a current address stored in an address register. When the access is moved from the upper right corner to a left end of a second line of of the rectangular region as shown by an arrow 2 , a required address change is obtained by adding {M-(m-1)} (=14) to the current address. Therefore, this displacement value of "14" is previously registered. When the access reaches to a lower right corner, the access returns to the upper left corner as shown by an arrow 3 . For this purpose, a required address change is obtained by subtracting {(n- 1)M+(m-1)} (=50) from the current address. Therefore, this displacement value of "-50" is previously registered. Thus, a triple loop processing is performed as shown in Figure 5B.

Turning to Figures 1 and 2, in the case of performing the access in the triple loop as mentioned hereinbefore and just explained with reference to Figures 5A and 5B, the address generation rule means parameters such as an initial address, three loop numbers and two displacement values (in this case, one displacement value is fixed to "+1"). In this connection, one way can be considered, in which the parameters are stored in a bit pattern in the microinstruction, and read out by the control circuit so that these parameters are supplied to each of the address generation circuits. Here, assuming that the address generation circuit(s) desired to be set are designated by "1ij", the parameters are supplied from the control circuit 5 through an signal line 51 and the signal line 5ij. Therefore, each of the signal lines 51 and 5ij is preferred to be a parallel bit signal bus of for example 16 bits or 32 bits. If the signal lines 51 and 5ij are coupled by a selector or a bus, the parameter can be supplied to a destination address generator by switching over the selector or the bus. This operation is repeated in a required number, namely, the number corresponding to the number of the address generation circuits desired to be set with the address generation rule.

The succeeding steps C, D and E are to supply an address generation start signal to the address generation circuits set with the address generation rule. This address generation start signal can be expressed in one bit. For example, the address generation start signal is normally of "0", and when it becomes "1", it means the address generation starting. As mentioned above, since the address generation start signal can be expressed in one bit, it is easy that a number of address generation start signals are stored in the microinstruction, and therefore, it is possible to simultaneously bring a plurality of address generation circuits into an address generation starting condition. In Figure 1, each of the signal lines 51 and 5ij includes a signal line for this one bit signal.

In the pipelined operation, in the case that the generation of an address for a memory for supplying data to the pipeline-operating ALU and the generation of an address for a memory for storing the result of the operation of the ALU are performed independently, the address generation start signal is first supplied to the address generation circuit for the former, and then, after a delay of clock periods of the number corresponding to the number of pipelined stages, the address generation start signal is supplied to the address generation circuit for the latter.

Thereafter, the address generation is executed by each address generation circuit independently of the control circuit 5, until the triple loop processing defined by the address generation rule is completed. When the address generation ends, each of the address generation circuits generates a one-bit address generation end signal, which is "0" during the address generation but becomes "1" after the address generation ends, for example.

On the other hand, after the address generation start signal has been supplied, the control circuit 5 is brought into a wait condition, by the fact that the loop is executed by a conditional jump instruction which is flagged with the address generation end signal. The address generation end signal to be used as the flag is one selected from a plurality of address generation end signals. For example, the address generation end signal to be used as the flag is the address generation end signal generated by the address generation circuit for the memory storing the result of the operation of the ALU. In Figure 1, the address generation end signal is supplied through each signal line 6ij and also through the selector and the signal line 52 to the control circuit 5. This selection processing of the selector is shown in the step F of Figure 2. If it is constructed that the selector is effective until the select signal is reset (it is also possible to store the select signal in a register), the processing in the step F can be performed before the address generation starting. In any case, the precessing of the step G is performed by a conditional jump flagged with one of the address generation end signals.

In the first embodiment, each of the address generation circuits "1ij" can be configured to supply a read enable signal or a write enable signal to the associated memory "i".

Referring to Figure 3, there is shown a block diagram of a second embodiment of the digital processor in accordance with the present invention. In Figure 3, elements similar to those shown in Figure 1 are given the same Reference Numerals, and explanation thereof will be omitted for simplification of the description.

The second embodiment is configured so that the number of address generation circuits is made smaller than the total number of all address ports of memories for the purpose of reducing the amount of hardware. The reason for this is that there is less necessity that all the memories are completely simultaneously operated.

Now, a portion of the second embodiment differing from the first embodiment is that tile second embodiment will be explained. Only one feature of the second embodiment differing from the first embodiment is that the second embodiment is required to include means for determining which of the address generation circuits is selected and to which of the address ports an address generated by the selected address generation circuit should be supplied. This means is an address arbiter 110 shown in Figure 3, which can be realized in the form of a selector or a bus structure. From which of the address generation circuits to which of the address ports an address should be supplied, namely, an address arbitration rule can be set by supplying a control signal from the control circuit to the address arbiter 110. If this control signal is stored in a register, the resetting can be made unnecessary until the address generation circuit terminates its address generation.

The flow chart of Figure 4 illustrates an operation of the second embodiment. The steps "a" to "g" of Figure 4 correspond to the steps "A" to "G" of Figure 2, and therefore, explanation thereof will be omitted. In addition, the step "h" of Figure 2 illustrates the setting of the address arbitration rule to the address arbiter 110.

In the second embodiment, each of the address generation circuits can be configured to supply a read enable signal or a write enable signal to the address arbitration circuit, and the address arbitration circuit can supply the received read enable signal or the received write enable signal to the portion or all of the address ports in accordance with the preset address arbitration rule.

As will be apparent from the above, the digital processor in accordance with the present invention having a plurality of memories and a plurality of ALUs is characterized in that each of address ports of each of the memories is associated with an address generation circuit capable of executing a loop processing required for address generation. With this arrangement, it is possible to access a plurality of memories, and therefore, the processing of excellent efficiency can be realized.

In addition, it is not necessarily required that each of address ports of each of the memories is associated with one address generation circuit. It is possible to distribute, by use of a selector or the like, addresses generated by a limited number of address generation circuits to memories which need an address. In this case, the amount of hardware can be reduced.

The invention has thus been shown and described with reference to the specific embodiments. However, it should be noted that the present invention is in no way limited to the details of the illustrated structures but changes and modifications may be made within the scope of the appended claims.


Anspruch[de]
  1. Digitaler Prozessor mit einer Speichergruppe, die besteht aus "m"-Speichern, wobei jeder Speicher "i" (i = 1 bis "m") der "m"-Speicher Pi Adressenports "ij" (j = 1 bis Pi) aufweist, eine arithmetische und logische Einheitengruppe, die aus "n"-arithmetischen und logischen Einheiten besteht, eine erste Arbiterschaltung, die die Ausgangssignale der Speichergruppe empfängt und beurteilt zum Ausgeben des beurteilten Ausgangssignals der Speichergruppe an die arithmetische und logische Einheitengruppe, und eine zweite Arbiterschaltung, die die Ausgänge der arithmetischen und logischen Einheitengruppe empfängt und beurteilt, um die beurteilten Ausgänge der arithmetischen und logischen Einheitengruppe an die Speichergruppe zu übergeben,

    gekennzeichnet durch

    eine Vielzahl von Adressenerzeugungsschaltungen, die jeweils eine erzeugte Adresse an die Speichergruppe zuführt, eine Taktzuführschaltung zum Zuführen entsprechender Taktsignale an mindestens die Adressenerzeugungsschaltungen und eine Steuerschaltung, die jeweilige Steuersignale an die Speichergruppe zuführt, wobei die arithmetische und logische Einheitengruppe, die erste Arbiterschaltung und die zweite Arbiterschaltung synchron mit den Taktsignalen von der Taktzuführschaltung erzeugt werden, wobei die Steuerschaltung so arbeitet, daß sie zuerst Adressenerzeugungsregeln für die Adressenerzeugungsschaltung setzt, so daß die Adressenerzeugungsschaltungen, die mit den Adressenerzeugungsregeln gesetzt sind, so arbeiten, daß sie eine Adresse erzeugen in Übereinstimmung mit der gesetzten Adressenerzeugungsregel und in Synchronität mit dem Takt, der von der Takterzeugungsschaltung zugeführt wurde, unabhängig von der Steuerschaltung, und wobei die gesetzten Adresserzeugungsregeln effektiv gehalten werden, bis sie zurückgesetzt werden.
  2. Digitaler Prozessor nach Anspruch 1, wobei jede Adressenerzeugungsschaltung "ij" in 1:1-Verhältnis mit jedem der Adressenports "ij" verbunden ist.
  3. Digitaler Prozessor nach Anspruch 1, welche ferner aufweist eine Adressenarbiterschaltung, die die Adressen empfängt, die von den Adressenerzeugungsschaltungen erzeugt wurden, welche jeweils die Adresse erzeugt in Übereinstimmung mit den gesetzten Adresserzeugungsregeln und in Synchronität mit dem Takt, der von der Taktsignalerzeugungsschaltung zugeführt wird, wobei die Adress-Arbiterschaltung die empfangenen Adressen an einen Teil oder alle der Adressenports "ij" zuführt in Übereinstimmung mit der Adressenarbiterregel, die durch die Steuerschaltung gesetzt wurde und effektiv gehalten wird, bis sie rückgesetzt wird.
  4. Digitaler Prozessor nach Anspruch 1, wobei die Adressenerzeugungsschaltungen "ij" configuriert sind, um ein Leseaktiviersignal oder ein Lesedeaktiviersignal an den Speicher "i" zuzuführen.
  5. Digitaler Prozessor nach Anspruch 3, wobei die Adressenerzeugungsschaltungen "ij" so configuriert sind, daß sie ein Leseaktiviersignal oder ein Lesedeaktiviersignal an die Adressenarbiterschaltung liefern und die Adressenarbiterschaltung das empfangene Leseaktiviersignal oder das empfangene Lesedeaktiviersignal an einen Teil oder alle Adressports zuführt in Übereinstimmung mit der voreingestellten Adressenarbiterregel.
  6. Digitaler Prozessor nach Anspruch 1, wobei ein Teil oder alle der arithmetischen und logischen Einheiten mit einer oder mehreren Pipelinestufen konfiguriert sind.
  7. Digitaler Prozessor nach Anspruch 2, wobei ein Teil oder alle der arithmetischen und logischen Einheiten mit einer oder mehreren Pipelinestufen konfiguriert sind.
  8. Verfahren zum Steuern eines digitalen Prozessors gemäß einem der Ansprüche 1 bis 7, welcher eine Speichergruppe aufweist, die aus "m"-Speichern zusammengesetzt ist, wobei jeder Speicher "i' (i = 1 bis "m") der "m"-Speicher Pi Adressenports "ij" (j = 1 bis Pi) aufweist, eine arithmetische und logische Einheitengruppe, die aus "n"-arithmetischen und logischen Einheiten besteht, eine erste Arbiterschaltung, die die Ausgangssignale der Speichergruppe empfängt und beurteilt zum Ausgeben des beurteilten Ausgangssignals der Speichergruppe an die arithmetische und logische Einheitengruppe, und eine zweite Arbiterschaltung, die die Ausgänge der arithmetischen und logischen Einheitengruppe empfängt und beurteilt, um die beurteilten Ausgänge der arithmetischen und logischen Einheitengruppe an die Speichergruppe zu übergeben, eine Vielzahl von Adressenerzeugungsschaltungen, die jeweils eine erzeugte Adresse an die Speichergruppe zuführt, eine Taktzuführschaltung zum Zuführen entsprechender Taktsignale an mindestens die Adressenerzeugungsschaltungen und eine Steuerschaltung, die jeweilige Steuersignale an die Speichergruppe zuführt, wobei die arithmetische und logische Einheitengruppe, die erste Arbiterschaltung und die zweite Arbiterschaltung synchron mit den Taktsignalen von der Taktzuführschaltung erzeugt werden, wobei die Steuerschaltung so arbeitet, daß sie zuerst Adressenerzeugungsregeln für die Adressenerzeugungsschaltung setzt, so daß die Adressenerzeugungsschaltungen, die mit den Adressenerzeugungsregeln gesetzt sind, so arbeiten, daß sie eine Adresse erzeugen in Übereinstimmung mit der gesetzten Adressenerzeugungsregeln und in Synchronität mit dem Takt, der von der Takterzeugungsschaltung zugeführt wurde, unabhängig von der Steuerschaltung, und wobei die gesetzten Adresserzeugungsregeln effektiv gehalten werden, bis sie zurückgesetzt werden, wobei das Verfahren die Schritte aufweist:

    Auswählen eines oder mehrerer der Adressenerzeugungsschaltungen, Setzen einer Adressenerzeugungsregel für jede der ausgewählten Adressenerzeugungsschaltungen von der Steuerschaltung, Zuführen von Adressenerzeugungs-Startsignalen an die ausgewählten Adresserzeugungsschaltungen simultan oder mit einer Zeitdifferenz, so daß jede der ausgewählten Adressenerzeugungsschaltungen ihre Adressenerzeugung startet und fortführt, um Adressen an einen entsprechenden Speicher zuzuführen in Synchronität mit dem Takt und unabhängig von der Steuerschaltung bis die Adressenerzeugung, die durch die Adressenerzeugungsregel angedeutet ist, vollendet ist, wobei jeder der ausgewählten Adresserzeugungsschaltungen ein Adresserzeugungs-Endsignal erzeugt, wenn die Adressenerzeugung beendet ist, und wobei ein ausgewähltes der Adressenerzeugungs-Endsignale, die durch die ausgewählten Adressenerzeugungsschaltungen erzeugt wurden, zur Steuerschaltung zugeführt wird.
Anspruch[en]
  1. A digital processor comprising a memory group composed of "m" memories, each memory "i" (i=1 to "m") of said "m" memories having Pi address ports "ij" (j=1 to Pi), an arithmetic and logic unit group composed of "n" arithmetic and logic units, a first arbitration circuit receiving and arbitrating outputs of said memory group for outputting the arbitrated outputs of said memory group to said arithmetic and logic unit group, and a second arbitration circuit receiving and arbitrating outputs of said arithmetic and logic unit group for outputting the arbitrated outputs of said arithmetic and logic unit group to said memory group,

    characterized by

    a plurality of address generation circuits each for supplying a generated address to said memory group, a clock supply circuit for supplying respective clock signals to at least said address generation circuits, and a control circuit supplying respective control signals to said memory group, said arithmetic and logic unit group, said first arbitration circuit and said second arbitration circuit in synchronism with the clock signals generated by said clock supply circuit, said control circuit operating to previously set address generation rules to said address generation circuits so that the address generation circuits set with the address generation rules operate to generate an address in accordance with the set address generation rule and in synchronism with the clock supplied from said clock signal supply circuit, independently of said control circuit, and the set address generation rules are maintained effective until they are reset.
  2. A digital processor claimed in Claim 1 wherein each one address generation circuit "ij" is connected to each one of the address ports "ij" in one-to-one relation.
  3. A digital processor claimed in Claim 1 further including an address arbitration circuit receiving the addresses generated by said address generation circuits each of which generates the address in accordance with the set address generation rule and in synchronism with the clock supplied from said clock signal supply circuit, said address arbitration circuit supplying said received addresses to a portion or all of said address ports "ij" in accordance with an address arbitration rule which set by said control circuit and which is maintained effective until it is reset.
  4. A digital processor claimed in Claim 1 wherein said address generation circuits "ij" are configured to supply a read enable signal or a write enable signal to the memory "i".
  5. A digital processor claimed in Claim 3 wherein said address generation circuits "ij" are configured to supply a read enable signal or a write enable signal to said address arbitration circuit, and said address arbitration circuit supplies the received read enable signal or the received write enable signal to the portion or all of said address ports in accordance with the preset address arbitration rule.
  6. A digital processor claimed in Claim 1 wherein a portion or all of said arithmetic and logic units is configured to have one or more pipeline stages.
  7. A digital processor claimed in Claim 2 wherein a portion or all of said arithmetic and logic units is configured to have one or more pipeline stages.
  8. A method for controlling a digital processor according to any of claims 1-7, which comprises a memory group composed of "m" memories, each memory "i" (i=1 to "m") of said "in" memories having Pi address ports "ij" (j=1 to Pi), an arithmetic and logic unit group composed of "n" arithmetic and logic units, a first arbitration circuit receiving and arbitrating outputs of said memory group for outputting the arbitarted outputs of said memory group to said arithmetic and logic unit group, a second arbitration circuit receiving and arbitrating outputs of said arithmetic and logic unit group for outputting the arbitrated outputs of said arithmetic and logic unit group to said memory group, a plurality of address generation circuits each for supplying a generated address to said memory group, a clock supply circuit for supplying respective clock signals to at least said address generation circuits, and a control circuit supplying respective control signals to said memory group, said arithmetic and logic unit group, said first arbitration circuit and said second arbitration circuit in synchronism with the clock signals generated by said clock supply circuit, said control circuit operating to previously set address generation rules to said address generation circuits so that the address generation circuits set with the address generation rules operate to generate an address in accordance with the set address genertion rule and in synchronism with the clock supplied from said clock signal supply circuit, independently of said control circuit, and the set address generation rules are maintained effective until they are reset, the method including the steps of selecting one or more of said address generation circuits, setting an address generation rule to each of the selected address generation circuits from said control circuit, supplying address generation start signals to said selected address generation circuits from said control circuit simultaneously or with a time difference so that each of said selected address generation circuits starts its address generation and continues to supply an address to a corresponding memory in synchronism with said clock and independently of said control circuit until the address generation indicated by the address generation rule is completed, each of said selected address generation circuits generating an address generation end signal when an address generation is ended, and a selected one of address generation end signals generated by said selected address generation circuits being supplied to said control circuit.
Anspruch[fr]
  1. Processeur numérique comprenant un groupe de mémoires composé de "m" mémoires, chaque mémoire "i" (i=1 à "m") desdites "m" mémoires ayant Pi ports d'adresse "ij" (j=1 à Pi), un groupe d'unités arithmétiques et logiques composé de "n" unités arithmétiques et logiques, un premier circuit d'arbitrage recevant et arbitrant des sorties dudit groupe de mémoires pour sortir les sorties arbitrées dudit groupe de mémoires vers ledit groupe d'unités arithmétiques et logiques, et un deuxième circuit d'arbitrage recevant et arbitrant des sorties dudit groupe d'unités arithmétiques et logiques pour sortir les sorties arbitrées dudit groupe d'unités arithmétiques et logiques vers ledit groupe de mémoires,

       caractérisé par une pluralité de circuits de génération d'adresse, chacun étant destiné à fournir une adresse générée audit groupe de mémoires, un circuit de fourniture d'horloge pour fournir des signaux d'horloge respectifs au moins auxdits circuits de génération d'adresse, et un circuit de commande fournissant des signaux de commande respectifs audit groupe de mémoires, audit groupe d'unités arithmétiques et logiques, audit premier circuit d'arbitrage et audit deuxième circuit d'arbitrage en synchronisme avec les signaux d'horloge générés par ledit circuit de fourniture d'horloge, ledit circuit de commande fonctionnant de manière à instaurer auparavant des règles de génération d'adresse dans lesdits circuits de génération d'adresse de telle sorte que les circuits de génération d'adresse instaurés avec les règles de génération d'adresse fonctionnent de manière à générer une adresse conformément à la règle de génération d'adresse instaurée et en synchronisme avec l'horloge fournie depuis ledit circuit de fourniture de signal d'horloge, indépendamment dudit circuit de commande, et les règles de génération d'adresse instaurées sont maintenues effectives jusqu'à ce qu'elles soient restaurées.
  2. Processeur numérique selon la revendication 1, dans lequel chacun des circuits de génération d'adresse "ij" est connecté à chacun des ports d'adresse "ij" en une relation biunivoque.
  3. Processeur numérique selon la revendication 1, incluant en outre un circuit d'arbitrage d'adresse recevant les adresses générées par lesdits circuits de génération d'adresse dont chacun génère l'adresse conformément à la règle de génération d'adresse instaurée et en synchronisme avec l'horloge fournie depuis ledit circuit de fourniture de signal d'horloge, ledit circuit d'arbitrage d'adresse fournissant lesdites adresses reçues à une partie ou à la totalité desdits ports d'adresse "ij" conformément à une règle d'arbitrage d'adresse qui est instaurée par ledit circuit de commande et qui est maintenue effective jusqu'à ce qu'elle soit restaurée.
  4. Processeur numérique selon la revendication 1, dans lequel lesdits circuits de génération d'adresse "ij" sont configurés de manière à fournir un signal de validation de lecture ou un signal de validation d'écriture à la mémoire "i".
  5. Processeur numérique selon la revendication 3, dans lequel lesdits circuits de génération d'adresse "ij" sont configurés de manière à fournir un signal de validation de lecture ou un signal de validation d'écriture audit circuit d'arbitrage d'adresse, et ledit circuit d'arbitrage d'adresse fournit le signal de validation de lecture reçu ou le signal de validation d'écriture reçu à la partie ou à la totalité desdits ports d'adresse conformément à la règle d'arbitrage d'adresse préinstaurée.
  6. Processeur numérique selon la revendication 1, dans lequel une partie ou la totalité desdites unités arithmétiques et logiques est configurée de manière à posséder un ou plusieurs étages en pipeline.
  7. Processeur numérique selon la revendication 2, dans lequel une partie ou la totalité desdites unités arithmétiques et logiques est configurée de manière à posséder un ou plusieurs étages en pipeline.
  8. Méthode pour commander un processeur numérique selon l'une quelconque des revendications 1 à 7 qui comprend un groupe de mémoires composé de "m" mémoires, chaque mémoire "i" (i=1 à "m") desdites "m" mémoires ayant Pi ports d'adresse "ij" (j=1 à Pi), un groupe d'unités arithmétiques et logiques composé de "n" unités arithmétiques et logiques, un premier circuit d'arbitrage recevant et arbitrant des sorties dudit groupe de mémoires pour sortir les sorties arbitrées dudit groupe de mémoires vers ledit groupe d'unités arithmétiques et logiques, un deuxième circuit d'arbitrage recevant et arbitrant des sorties dudit groupe d'unités arithmétiques et logiques pour sortir les sorties arbitrées dudit groupe d'unités arithmétiques et logiques vers ledit groupe de mémoires, une pluralité de circuits de génération d'adresse, chacun étant destiné à fournir une adresse générée audit groupe de mémoires, un circuit de fourniture d'horloge pour fournir des signaux d'horloge respectifs au moins auxdits circuits de génération d'adresse, et un circuit de commande fournissant des signaux de commande respectifs audit groupe de mémoires, audit groupe d'unités arithmétiques et logiques, audit premier circuit d'arbitrage et audit deuxième circuit d'arbitrage en synchronisme avec les signaux d'horloge générés par ledit circuit de fourniture d'horloge, ledit circuit de commande fonctionnant de manière à instaurer auparavant des règles de génération d'adresse dans lesdits circuits de génération d'adresse de telle sorte que les circuits de génération d'adresse instaurés avec les règles de génération d'adresse fonctionnent de manière à générer une adresse conformément à la règle de génération d'adresse instaurée et en synchronisme avec l'horloge fournie depuis ledit circuit de fourniture de signal d'horloge, indépendamment dudit circuit de commande, et les règles de génération d'adresse instaurées sont maintenues effectives jusqu'à ce qu'elles soient restaurées, la méthode incluant les étapes de sélection d'un ou de plusieurs desdits circuits de génération d'adresse, d'instauration d'une règle de génération d'adresse dans chacun des circuits de génération d'adresse sélectionnés depuis ledit circuit de commande, de fourniture de signaux de début de génération d'adresse auxdits circuits de génération d'adresse sélectionnés depuis ledit circuit de commande simultanément ou avec une différence de temps de telle sorte que chacun desdits circuits de génération d'adresse sélectionnés commence sa génération d'adresse et continue à fournir une adresse à une mémoire correspondante en synchronisme avec ledit signal d'horloge et indépendamment dudit circuit de commande jusqu'à ce que la génération d'adresse indiquée par la règle de génération d'adresse soit achevée, chacun desdits circuits de génération d'adresse sélectionnés générant un signal de fin de génération d'adresse lorsqu'une génération d'adresse est finie, et un signal sélectionné des signaux de fin de génération d'adresse générés par lesdits circuits de génération d'adresse sélectionnés étant fourni audit circuit de commande.






IPC
A Täglicher Lebensbedarf
B Arbeitsverfahren; Transportieren
C Chemie; Hüttenwesen
D Textilien; Papier
E Bauwesen; Erdbohren; Bergbau
F Maschinenbau; Beleuchtung; Heizung; Waffen; Sprengen
G Physik
H Elektrotechnik

Anmelder
Datum

Patentrecherche

Patent Zeichnungen (PDF)

Copyright © 2008 Patent-De Alle Rechte vorbehalten. eMail: info@patent-de.com