KR960006282B1 - Dynamic memory device - Google Patents

Abstract

No content.

Description

Dynamic Memory Device

1 is a block diagram showing a memory circuit according to an embodiment of the present invention.

2 is a block diagram showing a memory circuit of a conventional example.

3 is a block diagram showing the structure of a conventional graphic display system.

4 is a diagram for explaining that both processing apparatuses perform figure drawing.

FIG. 5 is a diagram for explaining that one processing apparatus performs figure drawing, and the other processing apparatus performs character drawing.

6 is a block diagram showing a configuration of a graphic display system by a plurality of processing apparatuses of this embodiment.

7 is a diagram for explaining arithmetic functions of this embodiment.

Fig. 8 is a block diagram showing the structure of a frame buffer memory of the prior art.

9 is a block diagram showing the configuration of the memory circuit of the embodiment;

FIG. 10 shows the write mask circuit of FIG. 9. FIG.

FIG. 11 is a diagram for explaining a member frame buffer with the memory circuit of FIG. 9. FIG.

Fig. 12 is a block diagram showing an example of the configuration of a graphic display system for explaining the set of operation codes in this embodiment.

13 is a timing chart showing memory access timing of the operation of this embodiment.

14 is a timing chart showing generation of a select signal and an operation code set signal of this embodiment from memory access timing.

FIG. 15 is a timing chart showing memory write timing when a select signal is added to FIG.

The present invention relates to a dynamic memory device.

In recent years, with the improvement of display resolution, graphic display devices have begun to require a large capacity of memory for displaying information, that is, a frame buffer. When displaying graphic data, a large-capacity frame buffer is required, and therefore, many memory accesses are performed, and a need for high-speed read / write is required. Conventionally, as a method of coping with such a problem, there is, for example, decentralization of processing.

As an example of decentralization of this process, there is a method of executing a part of the process with a frame buffer. 2 shows an example of the configuration of the frame buffer memory circuit by this method. In Fig. 2, reference numeral 1 denotes an operator, reference numeral 2 denotes a memory, reference numeral 3 denotes a functional designation register of the operator, and reference numeral 6 denotes a write mask register. When data is written to the frame buffer, the data width is in units of bits and is often different from the number of bits in the word structure of the memory. Therefore, the write processing to the frame buffer must be performed in bit units and bit units. In the example of FIG. 2, a bit-wise operation is executed by the operator 1 and the extension function designation register 3, and the bit-by-bit write is performed by writing only the bits for which the write by the mask register 6 is valid. Run In addition, in this frame buffer, since the memory read / modify / write is executed in a cycle of writing data D from the data processing apparatus, it is necessary to perform the same operation in the normal memory. The lead drive of the data DO of the memory 2 is also unnecessary, and the speed can be increased.

3 shows another example of the decentralization of the process. 3 shows an example of the configuration of a graphic display device in which two data processing apparatuses 10 and 10 'are connected to a frame buffer memory 9' 'via a common bus 11. In the example of FIG. 3, the system divides the area of the frame buffer memory 9 "into two of a and b, and writes a to the data processing device 10 and b to the data processing device 10 '. An example of this system drawing is shown in Fig. 4. The frame buffer memory 9 " is displayed on the CRT, and the two divided areas are divided up and down as shown in Figs. In the case of drawing a circle on the memory 9 '', for example, arcs αα'α '' are parallel to the data processing apparatus 10 and arcs ββ'β '' are parallel to the data processing apparatus 10 '. The drawing process of a circle can be roughly divided into two operations: calculation of a circle and writing of a coordinate point into a frame buffer. If the calculation process takes longer than the writing process, the calculation process is performed by two processing apparatuses (10). ) And (10 '), which increases the writing time, but when the writing process becomes longer, the two processing apparatuses write a frame buffer memo. Since the access contention with respect to the 9 "is reduced, the effect of processing with two becomes smaller. In recent years, with advances in the LSI technology, the calculation processing time of the data processing apparatus is shortened, and the write processing time is relatively long. Therefore, there is a need to use a frame buffer memory 9 'that can reduce the number of memory accesses as shown in FIG.

Therefore, when the frame buffer memory 9 'of FIG. 2 is applied to the frame buffer of the system shown in FIG. 3, when both processing apparatuses perform the figure drawing and share the same processing as in FIG. Modulation of memory is the same, so there is no problem. However, as shown in FIG. 5, when one processing apparatus performs figure drawing a 'and the other processing apparatus performs character drawing b', a problem arises. In general, when the types of drawing are different, the calculation of the memory modulation is also different. When two processing apparatuses alternately access the frame buffer memory, the set of modulation operations and the read / modify / write operations are performed. It must be performed as a unit of drawing processing, and the set of operations is the same as that of the memory access from the processing apparatus, so that two memory accesses are made and the speed cannot be increased. It is also possible to consider the method of reducing the number of sets of operations by moving the access rights of the two buffers to the framebuffer instead of alternately, by shifting the access right to the other when one processor and the other access the same time. Processing to transfer the right of access is required, and it takes extra time compared with performing the drawing process with the same modeling function. That is, in the past, the same processing as in FIG. 4 has been shared among a plurality of data processing apparatuses, but recently, as represented by a multi-window system or the like, a plurality of data processing apparatuses perform different processing as shown in FIG. There is a need to divide and execute the memory circuit, and there is no memory circuit in consideration of this.

And as an example of the frame buffer using this kind of read-modify-write operation, for example, Nikkei Electronics '84. 8. "Design of a frame buffer for a 1280x1024 pixel graphics display in 64 KRAM with nibble mode" (p. 227 to 245) of No. 27.

An object of the present invention is to provide a dynamic memory device capable of setting a predetermined mode without providing a dedicated terminal in the dynamic memory device.

Now, in general, when any one resource is shared by a plurality of processing apparatuses, it is necessary to perform exclusive use control of the resource. In addition, when a plurality of processing apparatuses speeds up by sharing a single process, the processing and resource utilization must be cooperatively performed. Since such exclusive control and cooperative control are generally realized by a program of a processing apparatus, some processing time is required.

Here, the common resources are classified into two types of peripheral devices and storage devices, but the peripheral devices take the form of usage which is temporarily occupied when the processing device starts to use them, and the storage devices have access rights by priority control at the time of access. It takes the form of ignoring occupancy and uses a form that can be used at any time asynchronously. In this way, the usage pattern is different because the peripheral device has an operation mode internally when the operation starts, and the moving operation mode transitions. Therefore, it is difficult to stop the process on the way. This is because data reading or writing ends at the timing at which the processing apparatus has accessed, and the internal operation mode does not continue after the access ends. If a storage device that performs read / modify / write operations such as the above-described memory circuit is applied to a resource having such a classification, this memory device has a modulating function as an internal state, but the internal operation mode is accessed. As a peripheral device having a property of not continuing afterwards, it occupies a position as a peripheral device which performs a higher speed operation than the processing device. Therefore, in the case of exclusive control or cooperative control of a memory device that performs high-speed operation by a program of a low-speed processing device, the overhead for switching each other becomes large, so that it is essential to cope with the inside of the memory circuit. The memory circuit for performing such read / modify / write does not require an internal operation mode instructed from the outside, and the internal state can be switched in response to the processing apparatus only by the internal operation of the memory circuit.

Therefore, the dynamic memory device of the present invention is the dynamic memory device 9, 9 ', 9 "controlled by the lower address strobe signal RAS, the column address strobe signal CAS and the write enable signal WE. Prior to the read or write operation of the dynamic memory devices 9, 9 ', and 9 ", the control circuit 3 responsive to the low level of the write enable signal WE at the time of lowering of the lower address strobe signal RAS 3; 4), and the operation of the dynamic memory devices 9, 9 ', 9 "is set to a predetermined mode according to the output of the control circuits 3, 4 by the response. do.

Next, an embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a configuration of a frame buffer memory circuit according to an embodiment of the present invention. (1) denotes an arithmetic unit (LU) for realizing the read / modify / write operation modulating function, (2) a memory for storing data, and (3) and (4) arithmetic function for specifying the arithmetic function of the calculator. Designated register, (5) is an arithmetic function selector for selecting the arithmetic function, (6) and (7) is a write mask register for holding the write mask data, (8) is a write mask selector for selecting the write mask data, (D ) Denotes write data from the common bus, and (C) denotes select signals for the arithmetic function selector 5 and the write mask selector 8.

FIG. 6 is a block diagram showing a configuration example in which the frame buffer memory circuit 9 of the embodiment shown in FIG. 1 is applied to a multiprocessor system. Reference numerals 10 and 10 'denote data processing apparatuses, 11 denotes a common bus, and 12 denote address decoding circuits.

Next, an operation example of the present embodiment will be described. In Figs. 1 and 6, the data bus for reading memory data, the address decoding circuit of the memory block, the read / modify / write control circuit, etc., which are unnecessary for the description of the present invention, are omitted for ease of viewing. In the present embodiment, the physical capacity of the memory circuit 9 is 1M bytes, but the memory circuit 9 assigns addresses from address 800000H to 9FFFFFH, and address space of 2M bytes from 800000H to 9FFFFFH. The memory circuit 9 occupies a double address. In the double occupancy method of the memory circuit 9, the 800000H address and the 900000H address are the same byte data, and similarly the 8FFFFFH address and the 9FFFFFH address are configured to be the same byte data. The data read from the address of ×××× H and the data from the reading of the 9 ××××× H benches are the same as long as ××××× is the same. The reason why the memory circuit 9 occupies double from the address 800000H and the address 900000H is to distinguish the accesses of the data processing apparatus 10 and 10 '. That is, the data processing apparatus 10 regards 1M bytes from address 800000H, and the data processing apparatus 10 'makes 1M bytes from address 900000H. The means for distinguishing between these accesses is the address decoder 12.

The address decoder 12 outputs " 0 " when the upper one digit of the address is 8H (even), and outputs " 1 " when 9H (odd). The calculation function of the calculator 1 is 16 kinds of logical operations shown in FIG.

In order to designate these 16 kinds of operations, the operation code data FC is 4 bits of data, and the operation function designation registers 3, 4 and the operation function selector 5 also have a 4-bit configuration. Since the memory 2 is composed of 16 bits as a word, the write mask registers 6, 7 and the write mask selector 8 are also 16 bits.

Next, an operation example in the case where the data processing apparatus 10 writes to the frame buffer memory 9 in the structure of FIG. 6 will be described. The data processing apparatus 10 sets the function code F0 in the arithmetic function specifying register 3 and the mask data M0 in the write mask register 6 in advance. When the data processing apparatus 10 performs write access, for example, at address 800000H, the timing of the memory access is executed in the order of read / modify / write as shown in FIG. The data processing apparatus 10 outputs the address 800000H on the address bus. The address decoder 12 outputs "0", and the arithmetic function selector 5 selects the arithmetic function designation register 3 to operate. F0 is output to the calculator 1 as the code data FC.

At this time, the write mask selector 8 selects the write mask register 6 and outputs M0 as the WE to the memory 2. In FIG. 15, data of address 800000H is read out in the read period, and the write data D from the data processing apparatus 10 and the modulated data are calculated by the calculator 1 in accordance with F0 in the period. Write is performed in accordance with the data of M0 in the period of. In the write mask data, " 0 " is forbidden to write and " 1 " is writable, so FFH is designated as M0 for normal writing. When the data processing apparatus 10 'accesses the frame buffer 9, the function code F1 is assigned to the arithmetic function specification register 4 beforehand and the mask data Ml is applied to the write mask register 7 in advance. Set it. In order to process the same data as the 800000H address accessed by the data processing apparatus 10, the data processing apparatus 10 'write accesses to the 900000H address. The timing chart of the write access of the data processing apparatus 10 'is as shown in FIG. 15, wherein the output signal C of the address decoder 12 is 1 during access, the function code of the modulator is F1, It is different that the light mask is M1.

By changing the addresses accessed by the data processing apparatuses 10 and 10 'in this manner, the calculation and mask data can be made different, so even when different drawing is performed between the processing apparatuses as shown in FIG. It is not necessary to perform a set of calculation functions every time.

Next, the structure of the frame buffer memory 9 and the method of setting arithmetic functions in the present embodiment will be described.

8 is a configuration example of a general frame buffer. Conventionally, a memory is composed of a plurality of memory ICs (Integrated Circuits), and an arithmetic operator 1, arithmetic function designation register 3, and a write mask register 6 are added as external parts of the memory. The reason why the memory is composed of a plurality of memory ICs is that the memory capacity is large and cannot be realized with one IC. At this time, the memory is divided into 1, 2, and 4 bits in units of the data bit direction (here, 16 bits). For example, at least 16 memory ICs are required for one-bit division.

In the case where the entire frame buffer shown in FIG. 8 is IC, it is necessary to divide the IC into a plurality of ICs due to the capacity.

Next, a description will be given of a method of setting arithmetic functions and writing masks corresponding to division in this embodiment.

In the set method, it is not important that there are a plurality of arithmetic function designation registers 3, 4, and write mask registers 6, 7, so that they are described as one by one.

In the graphic display of the present situation, since arithmetic operations are mainly performed in bit units, it is possible to divide arithmetic units into bit units of operation data. In addition, even when arithmetic operation is used, a circuit for handling the digit raising signal is added, and in principle, it is possible to divide an operator into units of bits. Since the write mask register 6 is a circuit for performing write control in units of bits, it is obvious that the write mask register 6 can be divided into units of bits. However, the operation function specifying register 3 is determined by the type of operation function of the operator 1; A bit length indicating a number is stored, and since the bit length is not related to the bit length of the operation data (here 16), it cannot be divided into bit units of the operation data. Therefore, the arithmetic function specification register 3 needs to have the same thing for every unit divided by 1 bit unit. As such, it is wasteful to have the same function for each divided unit, but the degree of integration of the IC increases year by year, so that the ratio of the number of elements used as peripheral circuits to the number of memory elements in the case of integration is a little less than 1%. It doesn't matter. In the case of integration, it is not a problem to have the arithmetic function designation register 3 for each bit division unit as described above, but it is not a problem to divide the frame buffer shown in FIG. have. The reason for this is that data input of the arithmetic function specification register 3 is performed with the data signals D _{15 to} D ₀ . When dividing by bit of data, simply dividing by 1 bit unit, the input data of the arithmetic function specification register 3 becomes 1 bit, and 4 bits of FIG. 7 cannot be designated. Conversely, if the number of bits necessary for the input of the arithmetic function designation register 3 is input as input data, a signal valid only for the designation of the arithmetic function must be used as a package pin at the time of IC, resulting in a large package.

In the frame memory according to the present embodiment, when arithmetic function designation is performed on the data bus, the number of arithmetic functions depends on the bit division of the data. Thus, by using an address signal that does not depend on the bit division and does not depend on the data bus, Specifies the operation function.

Ninth turn an example of a memory for a frame buffer that specifies the portion of the address signal to the calculation function (Dj) is a one-bit signal of the data signal 16 bits of the data processing units for rendering graphics, (A ₂₃ ~A ₁ ) Is an address signal of the data processing apparatus, (WE) is a write control signal of the data processing apparatus, (FS) is a data set control signal for the arithmetic function specification register 3 and the write mask register 6, and (DOj) is The read data of the memory device 2, DIj, is the calculation result data of the calculator 1, and Wj is a write control signal for the memory device 2.

10 shows an example of the construction of a write mask register. Reference numeral 61 denotes a write mask data storage register, and 62 denotes a gate for suppressing the write control signal WE.

11 is a structural example of a frame buffer using the memory circuit of FIG. In FIG. 11, a 4-bit configuration is shown to clarify the connection relationship.

12 is an example in which the memory circuit of the embodiment is applied to a graphic display system for explaining a set of operational codes. Denoted at 10 is a data processing apparatus, and at 13 is a decode circuit for generating the set signal FS.

Next, an operation example of this memory circuit will be described. In the embodiment, the calculation functions of the calculator 1 are 16 types shown in FIG. 7 as described above. When the data processing apparatus 10 writes F0FFH at, for example, A00014H, the decode circuit 13 outputs the set signal FS, and the address signals A _{4 to} A _{1 to the} arithmetic function designation register 3. That is, 0101B (B is bit data) is set. As a result, the calculator 1 selects logicalization as the calculation function, as shown in the calculation function table in FIG. In the write mask register 6, one bit of the 16-bit data of the data 0F00H from the data processing apparatus 10 is set in the write mask data storage register 61. One bit to be set is the same position as the bit position of the memory element. As a result, F0FFH is set as the write mask data.

Next, the case where the data processing apparatus 10 writes F3FFH at address 800000H will be described. It is assumed that 0512H is stored at address 800000H. The memory access timing of the data processing apparatus 10 is shown in FIG. The write access to the memory circuit 9 of the data processing apparatus 10 is in the read / modify / write operation as shown in FIG. 0512H is read out to the DO bus and F3FFH is input to the D bus at the timing of the lead modulated light read. At the timing of the next modulation, the calculator 1 calculates data of the D bus and the DO bus, and outputs the calculation result to the DI bus. In this case, since the value of the D bus is F3FFH and the DO bus is 0512H, the data on the DI bus is F7FFH. This is because the arithmetic operator 1 selects logicalization as an arithmetic function by the operation mentioned above. Finally, the data of the DI bus F7FFH is written at the timing of the read / modify / write write. However, in the above-described set operation, the write mask data has F0FFH set, and as shown in FIG. 10, the mask data is " 0 ""bits of the gate 62 is turned on, and" 1 "bits, so that a gate 62 off, D _11, only 4 bits of the ₈ ~D perform the actual write operation, and the remaining 12 bits in The write operation does not occur. As a result, the data at address 800000H becomes 0712H.

According to this embodiment, the following effects are obtained. Corresponding to each of the data processing apparatuses 10 and 10 ', the arithmetic function designation registers 3 and 4 and the write mask registers 6 and 7 are provided. Even when (10 ') is asynchronously independent and write access is performed to the frame buffer memory 9, the designation of the modulator of the read / modify / write operation and the mask light are made for each data processing apparatus. There is no need for cooperative control between the data processing apparatuses, and drawing processing can be executed without mutual interference except for access delay due to contention for the frame buffer memory 9.

The present embodiment is a frame buffer memory of a graphic display. In the data processing of the data processing apparatuses 10 and 10 ', the coordinate calculation of the dots to be drawn is mainly the case, and when calculation processing such as coordinate calculation takes time, By dividing the processing by two data processing apparatuses, the computation processing time can be shortened, and the writing time is shortened. In addition, when the frame buffer writing process takes time, the number of accesses can be reduced by using lead modality light, so that the drawing time can be shortened and a high speed graphic display system can be realized.

In the present embodiment, part of the address signal is used as a control signal, so that a memory circuit for performing read / modify / write, which can specify arithmetic functions, can be realized regardless of the data division method. In other words, when IC is used, the structure of the memory element can be determined without depending on the functions of read, modulate, and write.

In the present embodiment, two data processing apparatuses are used, but of course, three or more data processing apparatuses can be realized in the same manner.

It is also apparent that the system can be applied to a system for performing parallel drawing by initiating a plurality of tasks with one data processing apparatus and assigning different addresses to the tasks.

The memory circuit of the embodiment differs from the conventional memory IC in the set signal FS for setting the arithmetic function and the write mask data, and the signal C for selecting the arithmetic function and the write mask. In this case, when IC is formed, two pins are larger than a normal memory. Although it may implement | achieve by increasing two pins, in order to make a package small, you may substitute for the said signal using the timing of memory access. An example of access timing when this method is used is shown in FIG. Timings that do not appear in the timing sequence of a normal dynamic RAM as shown in FIG. 14 are used to distinguish the processing apparatus (when the WE signal is set to L level due to the falling of RAS), or operation codes and writing By using the mask data set (with CAS and WE at L level due to the falling of RAS), the FS or C signals can be generated and the number of pins of the IC package can be reduced.

Incidentally, although the data width is 16 bits and the unit of division is 1 bit in this embodiment, it is obvious that any value may be a value other than the values described in this embodiment.

In the present embodiment, the calculation function and the writing mask are simultaneously specified, but may be specified separately.

It is also apparent that the data width of the function specification of the calculator may be other than 4 bits.

In addition, the present embodiment may be applied to a memory having a built-in shift register and having a serial output.

As is apparent from the above description, according to the present invention, the present invention includes a control circuit for setting a predetermined mode in response to a combination of a low address strobe signal and a write enable signal. As a result, it is possible to set the predetermined mode in a specific combination application mode of the same control signal as that of the normal dynamic memory device, so that the same number of pinouts and pins as the dynamic memory device having no predetermined mode can be realized. In addition, it is possible to commonize a package having a large proportion in the price of the dynamic memory device, thereby providing a low-cost dynamic memory device.

Claims (3)

The lower memory strobe signal RAS, the column address strobe signal CAS, and the write enable signal WE are controlled by the dynamic memory device 9, 9 ', 9' '. Prior to the read or write operation for " 9 ", the control circuits 3 and 4 responding that the write enable signal WE is at a low level at the time of the lower address strobe signal RAS falling. And the operation of the dynamic memory device (9, 9 ', 9' ') is set to a predetermined mode in accordance with the output of the control circuit (3, 4) by the response. 2. The dynamic memory device according to claim 1, wherein a write operation of the dynamic memory device (9, 9 ', 9 ") is executed in accordance with the predetermined mode. The operation of the dynamic memory device according to claim 1 or 2, wherein a mode setting signal supplied from an address terminal is stored in a register controlled by the output of the control circuits 3 and 4, wherein the operation of the dynamic memory device is performed in the predetermined mode. Dynamic memory device, characterized in that is set to.