Data processing system with synchronous dynamic memory in integrated circuit technology

BACKGROUND OF THE INVENTION AND PRIOR ART

0001The present application relates to a data processing system employing a synchronous dynamic memory for storing data and instructions, wherein system components of preferably universal design are used which are manufactured in integrated circuit technology.

0002In conventional data processing systems the main storage apparatus for storing data and instructions is designed for the preferred use of standardized cheap general purpose dynamic memory components of conventional design. An architecture of a main storage apparatus of this kind in a work station using a plurality of dynamic memory components can be found, for example, in a paper by L. Johnson et al., entitled "System Level ASIC Design for Hewlett-Packard's Low Cost PA-RISC Workstations", Proceedings 1991 Internat. Conf. on Computer Design (ICCD), pp. 132-133.

0003Specifications of such a general purpose dynamic memory can be found in the Hitachi IC Memory Handbook 2, "DRAM, DRAM Module" (1991.9), pp. 389-393. In general, the conventional general purpose dynamic memory does not have a clock input which serves as an input signal to a chip. Instead, during read/write, an internal operation clock must be generated in the chip from other control input signals. Further, a mode register for prescribing the operation mode of the dynamic memory is, in general, not provided therein and, as a consequence, the operation mode of the conventional dynamic memory is fundamentally single. Moreover, the dynamic memory is, in general, designed in form of a single internal memory bank.

0004On the other hand, in another article in NIKKEI ELECTRONICS, 1992.5.11 (No. 553), pp. 143-147, a dynamic memory is described which is accessible with a 2 or 4 times higher speed than before. This synchronous dynamic memory has a plurality of memory banks and a built-in register for setting the operation mode of these memory banks, such as: delay from /RAS or /CAS transition, sequential access of words (wrap length) and the order of addresses of input/output data pieces.

0005Another conventional dynamic memory is disclosed in the US Patent 4,394,753 to Siemens. It contains a plurality of memory areas which can be accessed in parallel. Only one address is input so that corresponding cells in each memory are simultaneously accessed.

0006In data processing systems, wherein the main storage apparatus employs general purpose dynamic memory units or components without clock input as described above, it is impossible to input a clock signal directly to the respective general purpose dynamic memory chips and cause each chip to be operated in synchronism with the clock signal.

0007Accordingly, control signals for the general purpose dynamic memory units or components must be generated by special circuitry externally of the chip with a timing which meets an AC characteristic of the chip, on the basis of the processor system's clock.

0008On the other hand, inside the general purpose dynamic chip or component an internal operation clock must be generated from the control signal to ensure control of the internal operation. Consequently, in data processing systems using general purpose dynamic memory chips or components the overhead covering the system clock up to the internal operation clock is substantially increased, making it difficult to design a main storage apparatus capable of operating at high speed in synchronism with the system clock.

0009Also, in the above data processing systems wherein the main storage apparatus is designed and constructed of general purpose dynamic memory chips or components, this memory is of single mode not having incorporated a mode register for prescribing the operation mode of the dynamic memory, and has incorporated a single memory bank only. If the main storage apparatus needs to be set up so as to comply with an operation mode of a plurality of memory banks, it is difficult under the aspects of performance and costs to design and construct an optimized main storage apparatus operating efficiently with high speed.

0010If a synchronous dynamic memory is designed for having a plurality of memory banks and a built-in register for setting the operation mode, then the problem arises that both, a conventional processor and the synchronous dynamic memory lack concrete circuitry means to realize control of access to the plurality of memory banks and control of setting an operation mode to the built-in register. If such concrete circuitry means would be incorporated either in the conventional processor or in the synchronous dynamic memory, those units loose compatibility with general applicability which is most desirable in data processing system design.

0011It is, therefore, an object of the present invention to overcome the above problems and providing a data processing system having processing means and a main storage apparatus designed in integrated circuit technology of general applicability and optimized from the standpoint of performance and costs.

SUMMARY OF THE INVENTION

0012Thus, the present application is concerned with an improved data processing system comprising a processing means and a main storage apparatus wherein system components are manufactured in integrated circuit technology and wherein the component chips are designed for general applicability thereby avoiding special concrete circuitry to solve special design requirements, in particular, in connection with the provision of a plurality of memory banks and setting the operation mode in a build-in register, at the same time achieving an operation in an enhanced manner aiming at permitting a higher data throughput in the data processing system.

0013The above and other objects are solved in an advantageous manner basically by applying the features laid down in claim 1. Further enhancements, features and aspects of the invention are defined in the associated subclaims.

0014In a preferred embodiment of the present invention, the memory (501) is a synchronous dynamic IC type memory which operates in synchronism with a clock signal (CLK) applied to its input terminal (552). The circuitry for controlling the access to a plurality of memory banks (502, 503) and the setting of an operation mode to the built-in mode register (CMR) is embodied in the main storage controller (MC) which is coupled to the processor (MPU) and the main storage apparatus (MS).

0015This design ensures the use of general purpose conventional processor components (CHIP) of high generality and also of general purpose conventional memory components (MS) of high generality.

0016Further, in an embodiment of the present invention the processor (MPU) and the main storage controller (MC) are respectively formed of separate chips, while in a still further embodiment the processor (MPU) and the main storage controller (MC) are respectively formed of independent cores inside the same chip. The use of the conventional general purpose processor core and the conventional general purpose memory chip can be ensured when a core of the main storage controller (MC) is incorporated into the same processor chip.

0017Other objects and features of the present invention will become apparent from the different embodiments to be described hereinafter.

BRIEF DESCRIPTION OF THE DRAWINGS

0018Fig 1 is a block diagram showing an architecture of a data processing system according to a first embodiment of the present invention.

0019Fig. 2 is a block diagram showing an internal architecture of a microprocessor unit (MPU).

0020Fig. 3 is a diagram showing area assignment in a processor bus space.

0021Fig. 4 is an illustrative diagram of an MS area and an MC register area.

0022Figs. 5A and 5B are diagrams showing an internal architecture of a synchronous dynamic memory and a field organization of a command register included in the synchronous dynamic memory.

0023Fig. 6 is a diagram showing an architecture of a main storage apparatus (MS).

0024Fig. 7 is a diagram showing an internal architecture of a main storage controller.

0025Figs. 8A and 8B are diagrams showing examples of bit assignment of row, column and bank addresses.

0026Fig. 9 is a time chart of mode setting and refresh cycle.

0027Fig. 10 is a time chart of two read block transfer cycles.

0028Fig. 11 is a time chart of a read block transfer cycle/write block transfer cycle.

0029Fig. 12 is a diagram showing an architecture of a processor system according to another embodiment of the present invention.

DESCRIPTION OF PREFERRED EMBODIMENTS

Overall Architecture of Processor System

0030Fig. 1 is a diagram showing a construction of a processor system.

0031Reference numeral 101 designates a microprocessor unit (hereinafter abbreviated as MPU) constructed of a single chip.

0032Reference numeral 102 designates a main storage apparatus (hereinafter abbreviated as MS) which includes a plurality of synchronous dynamic memory chips.

0033Reference numeral 104 designates a controller for MS 102 which is constructed of a single chip.

0034Reference numeral 103 designates a clock generator (hereinafter abbreviated as CG) of the processor system. The CG 103 supplies clock signals 150, 151 and 152 to the MPU 101, the MS 102 and the MC 104. These clock signals are synchronous with each other. In the present embodiment, 150, 151 and 152 are clock signals which are in synchronism with each other at the same frequency. However, the relation between 150 and 151 and the relation between 150 and 152 may be allowed to be l:N (N being integer) or N:l. Denoted by 150, 151 and 152 are signals which are synchronous with each other. Therefore, the individual components of the processor system operate in synchronism with a single system clock.

0035Reference numeral 153 designates a processor bus through which the MPU 101 and the MC 104 are coupled together and which consists of an address, data and control signals. Of them, a data bus 154 is also coupled to the MS 102. Through this data bus 154, data from the MS 102 is transmitted directly to the MPU 101.

0036Reference numeral 156 designates addresses and control signal which are supplied from the MC 104 to the synchronous dynamic memory MS 102.

0037The MC 104 is also coupled to an I/O bus 157. Coupled to this I/O bus 157 are an I/O device 106 and a read only memory (hereinafter abbreviated as ROM) 105 in which initial program loading, operation system boot and a system initializing program are stored.

Internal Architecture of MPU and Processor Bus

0038Fig. 2 shows an internal architecture of the MPU 101 and breakdown of the processor bus 153. An instruction processing unit 201 is a unit which decodes an instruction and performs, on the basis of decoded information, such processings as an operation, fetch of data (operand), and store of data and branch. Denoted by 202 is an instruction cache for storing instructions temporarily and supplying the instructions at a high speed in accordance with a request from the instruction processing unit 201. Denoted by 203 is a data cache for storing data temporarily and supplying the data at a high speed in accordance with a request from the instruction processing unit 201. The block length of cache is 16 bytes in both of the instruction cache 202 and the data cache 203. Namely, since the processor bus 153 has a data width of 4 bytes, 16 bytes of a block timed to occurrence of a cache miss is divided by four and transfer from the MS 102 to each division of cache is carried out. Denoted by 204 is a bus control unit for controlling the processor bus. The bus control unit 204 responds to a request from the instruction cache 202, data cache 203 or instruction processing unit 201 to start the processor bus 153 in order that a necessary instruction and necessary data are fetched from the outside or transferred to the outside.

0039Breakdown of the processor bus 153 is as follows. <ul id="ul0001" list-style="none" compact="compact"><li>PD0-PD31 (154): Data bus of 4-byte width. Input/output signal. The data bus 154 is coupled directly to the MS 102. PD0 is the most significant bit and PD31 is the least significant bit.</li><li>PA0-PA31 (250): Address bus of 32-bit width, permitting 4-gigabyte addressing. Output signal. PA0 is the most significant bit and PA31 is the least significant bit.</li><li>PBS (251): Bus start signal. Output signal.</li><li>PR/W (252): Read/write request signal. During H, read and during L, write. Output signal.</li><li>PBL (253): Block transfer request. Output signal.</li><li>PDC (254): Transfer ending. Input signal.</li></ul>

Area Assignment in Processor Bus Space

0040In the present system, a 4-gigabyte space addressable through PA0-PA31 (250) is divided into four areas as shown in Fig. 3 in accordance with upper two bits of addresses. <ul id="ul0002" list-style="none" compact="compact"><li>MS area (301): Area to which the MS 102 is assigned.</li><li>MC register area (302): Area to which an internal register of the MC 104 is assigned.</li><li>I/O register area (303): Area to which an internal register of the I/O device 106 is assigned.</li><li>ROM area (304): Area to which the ROM 105 is assigned.</li></ul>

Internal Assignment in MS Area and MC Register Area

0041Fig. 4 shows internal assignment in the MS area 301 and the MS register area 302. An area between H'00000000 and H'003FFFFF is a sub-area for bank 0. This bank corresponds to one of banks inside the synchronous dynamic memory. An area between H'0040000 and H'007FFFFF is a sub-area for bank 1. This bank corresponds to the other bank inside the synchronous dynamic memory. Assigned to an address H'40000000 of the MC register area 302 is a MODE register of 8-bit length. When the MPU 101 writes a suitable value in this MODE register, the value is set in a mode register inside the synchronous dynamic memory and an operation mode of thereof is determined.

Internal Architecture of Synchronous DRAM

0042Fig. 5A shows an internal architecture of a synchronous dynamic memory 501 in a single chip for formation of the MS 102. The MS 102 is comprised of four of the above chips. The memory of this chip has two memory banks which are a bank 0 (502) and a bank 1 (503). Each memory bank is of 1,048,576 words×8 bits. Therefore, the whole chip has a capacity of 16M bits (=8M bytes). Denoted by RFADR 504 is an address counter adapted to prepare a row address for refresh. Denoted by CMR 505 is a mode register for determining an operation mode of the chip 501. Denoted by 506 is an internal control circuit for the chip 501. This circuit responds to control signals from the outside of the chip and a value set in the CMR 505 to prepare an internal operational signal in synchronism with a clock signal inputted externally of the chip.

Interface Signals of Synchronous DRAM

0043Interface signals of the synchronous dynamic memory are as follows. <ul id="ul0003" list-style="none" compact="compact"><li>A0-A10 (550): Address signal. Input. A row address and a column address are inputted. Used as a row address are 11 bits of A0-A10. Used as a column address are 9 bits of A0-A8. During inputting of a column address, A10 is used for bank designation. During setting of the CMR 505, mode information is inputted through A0-A7.</li><li>I/O0-I/O7 (551): Data signal. Input/output. Interface for data signal during read/write.</li><li>CLK (552): Clock signal. Input. In synchronism with a rising edge of this signal, a value on an input signal to the chip is fetched internally thereof. Or, in synchronism with a rising edge of this signal, an output is transmitted externally of the chip.</li><li>/WE (553): Write enable signal. Input. Asserted (Low level, hereinafter referred to as L) when requesting data write.</li><li>/CAS (554): Column address strobe signal. Input. Asserted (L) when supplying a column address.</li><li>/RAS0, /RAS1 (555): Row address strobe signal. Input. Asserted (L) when supplying a row address. This signal corresponds to the respective banks and constitutes an operation start signal of each bank.</li><li>/DQM (556): Data mask signal. Input. During read, this signal behaves as an enable signal for the output I/O0-I/O7 (551). Unless this signal is asserted (L) during read, the output 551 remains at a high impedance state. During write, this signal behaves as a write enable signal. During write, with this signal asserted (L), data is written actually.</li></ul>

Field Organization of Mode Register

0044Fig. 5B shows a field organization of the CMR 505 and the contents thereof. An RL field, a CL field and a WL field are respectively associated with addresses defined by bits A0-A2, A3-A4 and A5-A7 and during mode setting, each of the fields fetches values on corresponding address bits. The RL field indicates an /RAS delay. For example, if 100 is set here, data is read out during read operation 4-clock after the /RAS has been asserted. The CL field indicates a /CAS delay. For example, 10 is set here, data is read out during read operation 2-clock after the /CAS has been asserted. The WL field indicates a wrap length. This chip has the function to sequentially read, in synchronism with the clock, data pieces on a row designated by the same row address, beginning with a site designated by a column address. At that time, the column address is wrapped around at a length designated by the WL field. For example, if 000 is designated by the WL field, the wrap length becomes 4 and wraparound of 0-1-2-3, 1-2-3-0, 2-3-0-1 and 3-0-1-2 proceeds.

Architecture of Main Storage

0045Fig. 6 shows an architecture of the MS 102 using four (601, 602, 603 and 604) synchronous dynamic memories 501. 8-bit data signals of individual chips are coupled to respective byte positions of the data bus 154. The clock signal 151 connects to the CLK 552 of each chip, and A0-A10 (651), /WE, /CAS (652), /RAS0, /RAS1 (653) and /DQM (654) connect to corresponding input signals which are common to the respective chips. Denoted by 651, 652, 653 and 654 are output signals from the MC 104.

Internal Architecture of Main Storage Controller and Bit Assignment to Row, Column and Bank

0046Fig. 7 shows an internal architecture of the MC 104. The internal architecture is comprised of a request control unit 701, an internal register control unit 702, an MS control unit 704 and an I/O control unit 709. The request control unit 701 analyzes upper address two bits of a bus cycle issued from the MPU 101 onto the processor bus 153 to decide which of the MS area 301, MS register area 302, I/O register area 303 and ROM area 304 the bus cycle is destined for and then transfers control to a corresponding control unit.

0047Provided in the internal register control unit 702 are control registers included in the MC 104. One of them is a MODE register 703 for determining an operation mode of the synchronous dynamic memory. The internal register control unit 702 watches an address signal on the address bus PA0-PA31 (250) to detect that an address from the processor 101 accesses the mode register 505 of the synchronous dynamic memory 501,and responsive to a result of this detection, it transfers setting information (information from the data bus PD0-PD31 (154)) during this accessing to the mode register 505 of the synchronous dynamic memory 501. More particularly, when a value from the MPU 101 is written in this MODE 703, the internal register control unit 702 sends an indication to the MS control unit 704 and sends information written in the MODE 703 to the A0-A7 through a selector 706 to execute a write cycle to the CMR 505 of the synchronous dynamic memory 501.

0048The MS control unit 704 controls an address signal A0-A10 (651) of a synchronous dynamic memory 501 constituting the MS 102, and a DRAM access control unit 707 generates control signals /WE, /CAS (652), /RAS0, /RAS1 (653) and /DQM (654).

0049Denoted by MADR0 (705a) and MADR1 (705b) are registers for holding access addresses of bus cycles issued from the MPU 101 to the MS area. The two registers are constructed in the form of a FIFO (first in first out). An address of a preceding bus cycle is latched in the MADR1 (705b) and an address of a succeeding bus cycle is latched in the MADR0 (705a). As holding of the address of the preceding bus cycle becomes unneeded, the contents of the MADR0 (705a) is shifted to the MADR1 (705b). The contents of 705b is divided into a row address field, a column address field and a bank field.

0050The bit position of each field is shown in Fig. 8A. The 9-th bit represents the bank field CA10, the 10-th to 20-th bits represent the row address field RA0-RA10, and the 21st to 29-th bits represent the column address field CA0-CA8.

0051When the MS control unit 704 transmits a row address, the RA0-RA10 is transferred to the A0-A10 (651) by means of the selector 706.

0052When the MS control unit 704 transmits a column address, the CA0-CA8 is transferred to the A0-A8 (651) by means of the selector 706 and at the same time, the bank field CA10 is transferred to the A10 (651).

0053Denoted by CMP 714 is a comparator for comparing bank fields in the MADR0 (705a) and MADR1 (705b). When a comparison results in coincidence, accesses are destined for the same bank and therefore two cycles of one synchronous dynamic memory cannot be operated in parallel. But when a comparison results in non-coincidence, indicating that accesses are destined for different banks and therefore parallel operations of two cycles are permitted, the DRAM control 707 generates a control signal (/RAS0, /RAS1) which enables the parallel operations. This improves the throughput of the MS 102.

0054Denoted by RFTIME 708 is a refresh timer. This timer issues a refresh request to the DRAM control 707 at constant time intervals in order to cause it to execute a refresh cycle of the synchronous dynamic memory 501.

0055The I/O control unit 709 generates an I/O control signal 758 for controlling a bus cycle on the input/output bus 157.

0056Apart from the present embodiment, bit assignment to a row address field, a column address field and a bank field can be effected as shown in Fig. 8B.

0057During initial operation of this processor system, an initial operation program is read out of the ROM 105 and executed. In this program, mode setting of the synchronous dynamic memory 501 is first carried out.

During Initial Operation of Processor System

0058A time chart in this phase is shown in Fig. 9. The MPU 101 issues onto the processor bus 153 an address MA of the MODE register 703 included in the MC 104 and a write bus cycle of a mode setting value MD (clocks 2-4). In response thereto, the MS control unit 704 of the MC 104 asserts /RAS0, /RAS1, /CAS and /WE for the MS 102 and passes a set value to the A0-A7, thereby issuing a mode setting cycle. Through this, the mode setting of all of the synchronous dynamic memories 501 can be accomplished (clock 5). Indicated at a clock 10 is a refresh cycle. This is executed by asserting /RAS0, /RAS1 and /CAS.

Parallel Operations of Two Accesses in Two Different Memory Banks

0059Fig. 10 shows a case of two read block transfer cycles. In this case, /RAS delay is 4 clocks, /CAS delay is 1 clock and the wrap length is 4. At clocks 2 and 6, read block transfer cycle (with PBL asserted) requests are issued from the MPU 101. This issuance is done in the event that, for example, the internal cache of the MPU 101 misses. The preceding block transfer cycle is for the bank 0 and therefore, /RAS0 is asserted for the MS 102 at clock 3 to start the bank 0. Concurrently therewith, a row address Ar is passed through the A0-A10. At clock 6, /CAS is asserted and at the same time, a column address Ac is passed. In order to pass read data to data bus PD0-PD31, /DQM is started to be asserted at clock 7. One block read data of 4 words, that is, A, A+1, A+2 and A+3 are sequentially read in synchronism with clocks 8, 9, 10 and 11. During read-out of this one block, start of a succeeding bus cycle (access to the bank 1) is initiated (/RAS1 is asserted at clock 8) and data for this, that is, B, B+1, B+2 and B+3 are sequentially read during 4 clocks which begin with clock 13. By asserting PDC, the MPU 101 can be informed of arrival of read data.

0060Fig. 11 shows a case where after a read block transfer cycle of data A, A+1, A+2 and A+3, a write block transfer cycle of data B, B+1, B+2 and B+3 is issued. In this case, /RAS delay is 4 clocks, /CAS delay is 1 clock and the wrap length is 4. At clock 6, a write block transfer cycle (<maths id="math0001" num=""><math display="inline"><mrow><mtext>PR/WL=L</mtext></mrow></math><img file="EP0809252A2_D0001.tif" /></maths>) request is issued from the MPU 101. This issuance is done in the event that, for example, the internal cache of the MPU 101 misses. The preceding block transfer cycle is for the bank 0 and therefore, /RAS0 is asserted for the MS 102 at clock 3 to start the bank 0. Concurrently therewith, a row address Ar is passed through the A0-A10. At clock 6, /CAS is asserted and at the same time, a column address Ac is passed. In order to pass read data to data bus PD0-PD31, /DQM is started to be asserted at clock 7. The read data is sequentially read in synchronism with clocks 8, 9, 10 and 11. During read-out of this data, start of a succeeding bus cycle (access to the bank 1) is initiated (/RAS1 is asserted at clock 8) and when PDC is asserted at clock 12, the MPU 101 sequentially delivers data onto onto the data bus PD0-PD31 during 4 clocks which begin with clock 13.

0061Since the parallel operations of the two banks can be permitted as shown in Figs. 10 and 11, the main storage apparatus of high throughput can be constructed.

Other Embodiments

0062The present invention has been described by way of example but the invention is in no way limited to the foregoing specified embodiments and may obviously be modified in various ways within the scope of the fundamental technical idea of the present invention. For example, the following embodiment can be adopted in accordance with the present invention.

0063Fig. 12 is a diagram showing an architecture of a processor system according to another embodiment of the invention and this embodiment differs from the embodiment of Fig. 1 in that a processor (MPU) and a main storage controller (104) are respectively formed of independent cores inside the same chip. Accordingly, by adding the core of the main storage controller (MC) into the same chip, the use of the conventional processor core of highly generality and the conventional memory chip of high generality can be ensured.

0064As has been described, according to the typical embodiment form of the present invention, means to realize controlling of access to a plurality of banks of the memory (MS) and controlling of setting of an operation mode to the built-in register is arranged in the main storage controller (MC) coupled to the processor (MPU) and the main storage apparatus (MS) and therefore the use of the conventional processor of high generality and the conventional memory of thigh generality can be ensured.