US20020004882A1 - Semiconductor circuit and method of controlling the same - Google Patents
Semiconductor circuit and method of controlling the same Download PDFInfo
- Publication number
- US20020004882A1 US20020004882A1 US09/242,049 US24204999A US2002004882A1 US 20020004882 A1 US20020004882 A1 US 20020004882A1 US 24204999 A US24204999 A US 24204999A US 2002004882 A1 US2002004882 A1 US 2002004882A1
- Authority
- US
- United States
- Prior art keywords
- data
- dram
- row
- rows
- refreshing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 239000004065 semiconductor Substances 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims description 70
- 230000014759 maintenance of location Effects 0.000 claims description 49
- 230000002265 prevention Effects 0.000 abstract description 8
- 230000006866 deterioration Effects 0.000 abstract description 6
- 230000007423 decrease Effects 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 17
- 238000005457 optimization Methods 0.000 description 16
- 230000003247 decreasing effect Effects 0.000 description 9
- 238000012545 processing Methods 0.000 description 6
- 238000012546 transfer Methods 0.000 description 6
- 238000013519 translation Methods 0.000 description 6
- 238000001514 detection method Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000004904 shortening Methods 0.000 description 3
- 230000002123 temporal effect Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 239000003990 capacitor Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000002035 prolonged effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C11/00—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor
- G11C11/21—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements
- G11C11/34—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices
- G11C11/40—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors
- G11C11/401—Digital stores characterised by the use of particular electric or magnetic storage elements; Storage elements therefor using electric elements using semiconductor devices using transistors forming cells needing refreshing or charge regeneration, i.e. dynamic cells
- G11C11/406—Management or control of the refreshing or charge-regeneration cycles
-
- G—PHYSICS
- G11—INFORMATION STORAGE
- G11C—STATIC STORES
- G11C7/00—Arrangements for writing information into, or reading information out from, a digital store
- G11C7/20—Memory cell initialisation circuits, e.g. when powering up or down, memory clear, latent image memory
Definitions
- the invention relates to a semiconductor circuit and a control method thereof and more particularly to a semiconductor circuit and a control method thereof wherein the interval between refreshing operations of a DRAM is prolonged.
- Each memory cell of DRAM is small, and made of a capacitor.
- Their advantage over the implementing of SRAM is that the chip area can be greatly reduced.
- SRAM Static RAM
- the memory cells of DRAM in general are arranged in a matrix array. The data stored in the memory cells are read out all at once for each row, detected by a sense amplifier, and the values of the data read out are written into the memory cells from which the data were read out. Such a sequence of operations is called “refresh”.
- the DRAM does not accept any access to itself
- the refreshing operation is executed by a DRAM controller outside the DRAM, but in a sleeping mode such as when backed up by battery, the refreshing operation is performed by a refresh controller inside the DRAM LSI. In the sleeping mode, any read/write access from outside the DRAM is not accepted.
- Method 1 since the refreshing cycle time in the normal mode is adjusted to the memory cell having the shortest data retention time in the DRAM, there still remains the unresolved problem of power consumption in the normal mode. Further, Method 2 does not cope with variations in the data retention time.
- the merged DRAM/logic LSI has a connection at a wide memory band width to achieve highly improved processing performance of the logic portion.
- LSI there arises a problem of conflict between refresh and access to the DRAM made by the logic circuit, thereby prolonging the time required for DRAM access and, hence, the processing performance of the logic circuit is deteriorated. This is a serious problem, comparable with the problem of achieving reduction in power consumption.
- the first object of the present invention in a semiconductor circuit or, more particularly, in an LSI on which a DRAM and a logic circuit are merged, is to decrease the number of refreshing operations by refreshing only the rows storing necessary data, and thereby achieve both reduction in power consumption and prevention of deterioration in the performance of the logic circuit caused by an increase in the access time due to the confliction between refresh and access to the DRAM.
- a second object of the invention is to determine the rows storing data thereon according to the degree of importance of the data to, thereby, ensure saving important data without excessively shortening the refresh cycle time.
- a third object of the invention is to have refreshing operations carried out at suitable cycle time in conformity with the temperatures to thereby achieve both reduction in power consumption and prevention of deterioration in the processing performance of the logic circuit.
- the invention in a control method of a merged DRAM/logic LSI, is characterized by comprising the steps of disposing data, which are combined so that the number of the rows storing data thereon may be reduced, on each of the rows of the DRAM and refreshing each of the rows having data stored thereon.
- Another method to achieve the first object in a control method of a merged DRAM/logic LSI, is characterized by comprising the steps of disposing arbitrary data of which periods from being written in to being read out are overlapping or close to each other are disposed on the same row of the DRAM and refreshing the row only during the time period from the writing in of the data at the first to the reading out thereof at the end.
- a method to achieve the first and the second objects, in a control method of a merged DRAM/logic LSI comprises the steps of obtaining the memory capacity required by an application using DRAM and storing data in DRAM in order from a row having a longer data retention time, with reference made to a table storing previously obtained data retention time of each row of DRAM, and setting a refreshing cycle adapted to the row whose data retention time is the shortest of the rows storing data thereon.
- data may be disposed, when it is stored in the DRAM, on specific rows in accordance with the importance of the data.
- a method to achieve the third object, in each of the above methods, is characterized by comprising the step of detecting the temperature of the semiconductor circuit and setting the refreshing cycle time of the DRAM according on the temperature.
- a semiconductor circuit to achieve the first object is characterized by comprising means for disposing data, which are combined so that the number of the rows storing the data thereon may be reduced, on each of the rows of the DRAM and means for refreshing each of the rows having the data stored thereon.
- Another semiconductor circuit which has achieved the first object is characterized by comprising means for disposing arbitrary data of which periods from being written in to being read out are overlapping or close to each other on the same row of DRAM, and means for refreshing the row only during the time period from the writing in of the data to the reading out thereof at the end.
- a semiconductor circuit which has achieved the first and the second objects is characterized by comprising means for obtaining the memory capacity required by an application using DRAM and storing data in the DRAM in an order starting from a row having longer data retention time, with reference made to a table storing the previously obtained data retention time of each row of DRAM, and means for setting a refreshing cycle time adapted to the row whose data retention time is the shortest of the rows storing data thereon.
- a semiconductor circuit which has achieved the third object, in each of the above semiconductor circuits, is characterized by comprising means for detecting the temperature of the semiconductor circuit and setting the refreshing cycle of the DRAM according to the temperature.
- FIG. 1 is a block diagram showing the general configuration of an information processing apparatus to which a semiconductor circuit of the invention is applied.
- FIG. 2 is a block diagram showing an internal configuration of the main memory unit 4 shown in FIG. 1.
- FIG. 3 is a block diagram showing an internal configuration of the DRAM 7 .
- FIG. 4( a ) is a diagram showing the relationship between stored positions of data and rows before optimization and FIG. 4( b ) is a diagram showing the relationship between stored positions of data and rows after optimization.
- FIG. 5 is a functional diagram showing the principle of memory allocation made by a compiler in a first embodiment.
- FIG. 6 is a functional diagram showing the principle of memory allocation made by an operating system in the first embodiment.
- FIG. 7 is an explanatory drawing of a second embodiment, of which (a) is a graph showing lifetimes of data in a DRAM, (b) is an explanatory drawing of a case where data are disposed without taking the lifetimes of the data into consideration, (c) is a graph showing lifetimes of data disposed on each row, (d) is an explanatory drawing of a case where data whose lifetimes are similar are disposed on the same row and stored in the DRAM, and (e) is a graph showing the lifetimes of data on each of the rows after the disposition of the data A-H is optimized.
- FIG. 8 is a functional diagram showing the principle of memory allocation made by a compiler in the second embodiment.
- FIG. 9 is a functional diagram showing the principle of memory allocation made by an operating system in the second embodiment.
- FIG. 10 is a block diagram of an embodiment where a data retention time storing table is provided in a third embodiment.
- FIG. 11 is a functional diagram showing the principle of memory allocation made by a compiler in the third embodiment.
- FIG. 12 is a functional diagram showing the principle of memory allocation made by an operating system in the third embodiment.
- FIG. 13 is a functional diagram showing the principle of memory allocation made by a compiler in a fourth embodiment.
- FIG. 14 is a block diagram of a fifth embodiment provided with a means for detecting temperature and a means for controlling the refreshing cycle time.
- FIG. 1 is a block diagram showing a general configuration of an information processing apparatus to which the semiconductor circuit of the invention is applied.
- a processor 1 a ROM 3 , a main memory unit 4 , a temperature detection means 5 , an I/O (input/output interface) 6 , and the like are connected by a bus 2 .
- the main memory unit 4 includes a DRAM 7 for storing data, a DRAM controller 8 for controlling write and read operations on the DRAM 7 , and a data retention time storing means 9 for storing data retention times of each of the rows in the DRAM 7 .
- a logic portion 33 is constituted of non-memory circuits such as logic circuits, including the processor 1 , the I/O 6 , and the like.
- FIG. 2 is a block diagram showing an internal configuration of the main memory unit 4 shown in FIG. 1.
- the DRAM controller 8 within the main memory unit 4 includes a row address generating means 10 for generating the row address of the row of the DRAM 7 , a timer 11 for deciding the timing of generation of the row address, and a row flag storing portion 12 for setting the row as the object of refreshing in the DRAM 7 .
- the DRAM 7 is connected to the bus 2 through the interface (i/f) 13 .
- One DRAM controller 8 is implemented within the chip of a merged DRAM/logic LSI and performs control of the refreshing operation and the like at all times irrespective of the operation mode of the DRAM.
- the row address generating means 10 outputs a refresh address to thereby update the address of the row.
- the row flag 12 includes the flags corresponding to each of the rows in the DRAM.
- the DRAM controller 8 refers to the value of the flag corresponding to the row indicated by the row address generating means 10 every time a predetermined time period is counted out by the timer 11 , and carries out a refreshing operation only when the value of the flag is set so that a refreshing operation should be made and, otherwise, it makes no operation.
- the predetermined time period the time obtained by dividing the refreshing cycle time by the number of the rows, for example, is used.
- the DRAM controller 8 also has a function to coordinate the contention made between access to the DRAM through the bus 2 and a refreshing operation.
- the processor 1 and the DRAM 7 within the main memory unit 4 may have a path to connect with each other not by way of the bus 2 .
- the DRAM controller 8 also performs the coordination of the contention between access to the DRAM through this path and a refreshing operation.
- FIG. 3 is a block diagram showing an internal configuration of the DRAM 7 .
- a control signal line 14 to which a control signal for controlling write and read is supplied, an address line 15 to which an address signal of (A+B) bits is supplied, and a data line 16 to which data is supplied are connected to an input/output interface 17 .
- An address of (A+B) bits is separated by the input/output interface 17 into a row address of A bits and a column address of B bits, of which the row address of A bits is supplied to a row decoder 18 and the column address of B bits is supplied to a column decoder 19 .
- a memory cell 24 is disposed in the vicinity of the intersection of the word line 21 and a bit line 23 .
- the output of the row decoder 18 is supplied to the word line 21 selected from a memory cell array 20 and the data in the memory cells on the selected row is output over 2 B pieces of the bit lines 23 . Then, a sense amplifier 22 amplifies the data. Further, the data on the bit lines selected by the column decoder 19 are output to outside the main memory unit 4 .
- the size of the column address is in agreement with the number of the bit lines and the minimum amount of data accessible at one time is one bit.
- the amount of data accessible at one time may be either a predetermined value or a value that is to be determined each time of access by a signal for specifying the size included in the control signal 14 .
- the size of the space which the column address has may be smaller than the number of bit lines.
- the minimum amount of data accessible at one time is the value obtained by dividing the number of bit lines by the size of the column address space.
- FIG. 4( a ) shows the relationship between the stored position of data and the row when optimization of disposition of data has not been made.
- the position indicated by the circle shows the position in which data is written and an oval indicates an aggregate of individual data.
- FIG. 4( a ) shows an example where the positions to which the data are allocated are randomly determined. In this example, data are written on each of five rows, which are arranged in succession.
- the disposition of the data is determined, as shown in FIG. 4( b ), so that the number of rows storing data thereon decreases while a plurality of data are disposed so as not to overlap each other on the same row.
- the method to determine the combination of the data allocated to each row with the aim of reducing the number of the rows is solved as a combinatorial optimization problem.
- rows 21 a and 21 d to which data are allocated are indicated by thick lines and rows 21 b, 21 c, and 21 e to which no data are allocated are indicated by thin lines.
- Each field of the row flag 12 corresponds to each row of the DRAM.
- the DRAM controller 8 determines whether refreshing is to be done or not for each row in accordance with the flag value set in the row flag 12 , and performs the refreshing control for the row which is set to be refreshed by the flag.
- Each flag of the row flag 12 can be realized by one bit (on/off).
- the memory allocation optimization process is applied to the intermediate format 42 by the memory allocation means 41 as one of the functions of the compiler.
- the process for the combinatorial optimization of data is performed to minimize the number of rows to which data are actually allocated, with the number of rows and the number of memory cells on each row, as the structure of the memory, used as constraints to thereby determine relative addresses of individual data.
- the intermediate format 43 is output with the operation for setting flags, corresponding to rows on which the data are to be stored, inserted therein.
- the intermediate format 43 is converted into a sequence of instructions in the last step of the compiling process.
- a method for actually setting the flag there is a method which uses an instruction to automatically set/reset the flag for an accessed row, or a method which establishes the value by an immediate instruction or a data transfer instruction.
- disposition of pages is determined by the address translation means 44 , when it maps the pages represented by the virtual address 45 onto the physical address 46 , such that the number of the pages, i.e., the number of the rows to which the data are allocated, is minimized.
- the setting of the row flag 12 is realized by incorporating an instruction that sets the flag as described above, an immediate instruction, or a data transfer instruction, into the address translation means 43 such that the flag is set for the row on which data are actually mapped.
- refreshing operations are carried out only for the rows on which the data are stored, whereby unnecessary refreshing operations can be decreased and both reduction in power consumption and prevention of increase in memory access time due to contention between refresh and memory access can be achieved.
- the position in which the compiler or the operating system operates may be either inside or outside the semiconductor circuit, provided that it is able to know the structure of the memory.
- FIG. 7( a ) is a graph showing lifetimes of data A-H to be stored in a DRAM.
- FIG. 7( c ) is a drawing in which the lifetimes of the data in FIG. 7( a ) are rearranged for each row. Since, for example, row R 1 has data B, C, and F stored thereon, it becomes necessary to constantly refresh row R 1 as long as any of data B, C, and F are live.
- data A and C are disposed on row R 0 , data B and D on row R 1 , data F and G on row R 2 , and data E on row R 3 .
- FIG. 7( e ) is a graph showing the lifetimes of the data of each row after the positions, in which data A-H are written, have been altered.
- Each field of the row flag 12 is corresponding to each row of the DRAM, and the DRAM controller 8 controls the refreshing operation for the relevant row only during the period set by the flag to indicate that the refreshing operation should be made.
- Each flag of the row flag 12 is realized by one bit (on/off).
- a memory allocation optimization process is applied to the intermediate format 48 by the memory allocation means 47 as one of the functions of the compiler.
- the memory allocation means 47 obtains the lifetimes of the individual data on the basis of a flow graph as the intermediate format 49 .
- the process for combinational optimization of data is performed to minimize the sum total for all of the rows of the lifetimes of each of the rows where at least a set of data is still live, with the number of rows and the number of memory cells as the structure of the memory used as the constraints, to thereby determine relative address of individual data.
- the intermediate format 49 is output with the operation to set the flag corresponding to the row on which the data is to be stored at the start of the lifetime of the data and the operation for resetting the flag of the row on which the data is stored at the end of the lifetime of the data inserted therein.
- the intermediate format 49 is converted into a sequence of instructions in the last step of the compiling process.
- a method for actually setting the flag there is a method which uses an instruction to automatically set/reset the flag for an accessed row or a method which establishes the value by an immediate instruction or a data transfer instruction.
- disposition of pages is determined by the address translation means 50 , when it maps the pages represented by the virtual address 51 onto the physical address 52 so that the sum total, for all the rows, of the lifetimes for each of the rows where at least one page is live is minimized, with the number of the rows and the number of the memory cells, as the structure of the memory used as the constraints.
- the lifetime of each page is given, when the page includes at least one set of data, by the time period during which the data is live, and it can be obtained by the compiler by analyzing the lifetime of each individual datum when it allocates the data to each page.
- the setting of the flag when the page is disposed on the physical memory and the resetting of the flag when the physical memory is released can be realized by incorporating a flag setting instruction, an immediate instruction, a data transfer instruction, or the like, as described in the first embodiment, into the address translation means 50 .
- refreshing operations are carried out for the rows storing data only for the time period during which the data is live, whereby unnecessary refreshing operations can be decreased, power consumption reduced, and increase in the memory access time due to contention between refresh and memory access prevented.
- the compiler or the operating system may work either at inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- the data retention time storing table 9 a storing the data retention times tr 0 -tr 7 on each of a plurality of rows R 0 -R 7 of the DRAM 7 .
- the storing table 9 a corresponds to the data retention time storing means 9 shown in FIG. 1 and can be realized by a device which can be written onto such as a PROM, an EPROM, a FLASH memory, or an FPGA. On the table 9 a, there is stored, for each row, the data retention time of the memory cell which has the shortest data retention time among all of the memory cells included in the row.
- the measurement of the data retention time is carried out when the DRAM is tested.
- the results of the measurement are stored on the storing table 9 a.
- the information stored in table 9 a, as the data retention time may be the time itself, or an encoded value which is divided by some value.
- Each field of the row flag 12 corresponds to each row of the DRAM.
- the DRAM controller 8 determines whether refreshing is to be made or not for each row in accordance with the flag value set in the row flag 12 and performs the control of refreshing of the row which is set by the flag to be refreshed.
- Each flag of the row flag 12 can be realized by one bit (on/off).
- a process of memory allocation optimization is applied to the intermediate format 54 by the memory allocation means 53 as one of the functions of the compiler.
- the values stored on the storing table 9 a are referred to and the physical address corresponding to the row having the longer data retention time is preferentially taken as the object of data allocation, with the number of rows and the number of memory cells as the structure of the memory used as the constraints. Then, referring to the data retention time of the row which was taken lastly as the object of allocation, the refresh cycle time to be set in the timer 8 is determined.
- the intermediate format 55 including the operation to set the flag for the row on which actual data is to be stored and the operation to set the refreshing cycle time, is output.
- the intermediate format 55 is converted into a sequence of instructions in the last step of the compiling process.
- a method for actually setting the flag there are two methods. One is to use an instruction to automatically set/reset the flag for an accessed row, the other is to establish the value by an immediate instruction or a data transfer instruction.
- the row having the longer data retention time is preferentially taken as the object of allocation of the page, i.e., the data, with the number of rows and the number of memory cells on each row as the structure of the memory taken as the constraints. Then, reference is made to the data retention time of the row lastly taken as the object of allocation and the refreshing cycle time is set in the timer 8 and thus the method can be realized.
- the setting of the flag and the refreshing cycle time is realized by incorporating an instruction, such as a flag setting instruction, an immediate instruction, or a data transfer instruction, as described in the first embodiment, into the address translation means 56 .
- the position in which the compiler or the operating system operates may be either inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- the rows to store data thereon are determined according to the degree of importance of the data. Namely, those data which can be corrected for error or those data which allow for some error present therein are allocated to rows that have shorter holding times. On the other hand, those data which cannot be corrected for error or those data which allow for no numerical error present therein are allocated to such rows that have longer data retention times.
- the frequency of the refreshing operations can be decreased.
- the determination of the disposition of the data according to their degree of importance can be achieved, as shown in FIG. 13, by a compiler corresponding to the intermediate format 60 having a function indicative of the degree of importance of the data.
- the degree of importance of the data can be described in a program by for example the provision of types showing the degrees of importance.
- the memory allocation means 59 of the compiler shown in FIG. 13, making reference to the data retention time of each row stored in the data retention time storing means 9 , disposes data on the intermediate format 60 in accordance with the degree of importance of the data, determines the physical address of individual data, and outputs the intermediate format 61 .
- the intermediate format 61 is converted into a sequence of instructions in the last step of the compiling process.
- the position in which the compiler operates may be either inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- the semiconductor circuit formed of the DRAM 7 and the logic portion 33 merged on a chip is provided therein with the temperature detection means 5 for detecting the temperature of the semiconductor circuit.
- the refreshing cycle time is suitably set.
- the temperature detection means 5 can be realized by a method in which a leakage current of a memory cell having the same temperature characteristic as the memory cell for actually storing data is monitored, a method in which a ring oscillator is used, or the like (1994. Ito, “VLSI Memory”, Baifukan).
- the invention can be utilized in the field of the semiconductor circuits employing DRAM.
Landscapes
- Engineering & Computer Science (AREA)
- Microelectronics & Electronic Packaging (AREA)
- Computer Hardware Design (AREA)
- Dram (AREA)
- Memory System (AREA)
Abstract
Description
- The invention relates to a semiconductor circuit and a control method thereof and more particularly to a semiconductor circuit and a control method thereof wherein the interval between refreshing operations of a DRAM is prolonged.
- With the progress in the integration of semiconductors, it has become possible to integrate a processor, a memory, or other circuit onto one chip. Further, with the progress in the technology of processes enabling a logic circuit such as a processor and DRAM (Dynamic RAM) to be merged on one chip, it has become possible to implement DRAM on a memory.
- Each memory cell of DRAM is small, and made of a capacitor. Their advantage over the implementing of SRAM (Static RAM) is that the chip area can be greatly reduced. On the other hand, they bring such a disadvantage in that the electric charge stored thereon as data discharges with the passage of time and, hence, the data is lost. Therefore, it is necessary to make an operation to keep data from being lost. The memory cells of DRAM in general are arranged in a matrix array. The data stored in the memory cells are read out all at once for each row, detected by a sense amplifier, and the values of the data read out are written into the memory cells from which the data were read out. Such a sequence of operations is called “refresh”.
- During the refreshing operation, the DRAM does not accept any access to itself In the normal mode which allows read/write from outside the DRAM, the refreshing operation is executed by a DRAM controller outside the DRAM, but in a sleeping mode such as when backed up by battery, the refreshing operation is performed by a refresh controller inside the DRAM LSI. In the sleeping mode, any read/write access from outside the DRAM is not accepted.
- Here, some problems arise when the DRAM and logic circuit are merged on the same LSI.
- Merged DRAM/logic LSIs, however, have some disadvantages such that the DRAM portion of merged DRAM/logic LSIs might suffer from shorter data retention time. Heat and noise dissipated by the logic portion on the same chip could be harmful to the data retention time of the one-tip DRAM. When the ambient temperature rises, for example, from 25° C. to 70° C., the leakage current of the charge stored on the memory cell of the DRAM is increased by 30 fold and, hence, it becomes necessary to carry out the refreshing operation 30 times more frequently (Ito, “VLSI Memory Design”, Baifukan). In a system using conventional DRAM, refreshing operations are made at short intervals assuming the worst operating environmental conditions. Therefore, under normal temperature conditions, refreshing operations are being made at a great frequency.
- There is great variation of the data retention time among the memory cells of DRAM and few memory cells have short data retention times (Iwata et al. “Circuit Techniques for Super Low Retention Current DRAM”, Technical Report of Institute of Electronics, Information, and Communication Engineers, ICD 95-50). However, refreshing operations are being made at the same cycle time for all of the rows. This means that refreshing operations are being made at a greater frequency than is needed for many of the rows, often including no memory cells that have a small actual capacity in terms of data retention time.
- Further, all of the rows are refreshed whether the data held in DRAM are necessary for the logic circuit or not. In reality, only necessary data need be held in memory and, hence, unnecessary data need not be refreshed.
- Such excessive refreshing invites wasteful power consumption.
- Under these circumstances, means for decreasing the frequency of refreshing operations are being investigated. As a technique to decrease power consumption in the data retention mode of a conventional DRAM, there is a method1 in which self-refreshing is conducted at a refreshing cycle time corresponding to temperature (Japanese Patent Laid-Open No. 6-215561). Further, as a technique to decrease power consumption in the normal mode of a conventional DRAM, there is a
method 2 in which a control for the power supply and the decision as to whether a refreshing operation should be carried out is executed according to a flag set up in each memory area (Japanese Patent Laid-Open No. 5-324140, U.S. Pat. No. 5469559). - However, in Method1 above, since the refreshing cycle time in the normal mode is adjusted to the memory cell having the shortest data retention time in the DRAM, there still remains the unresolved problem of power consumption in the normal mode. Further,
Method 2 does not cope with variations in the data retention time. - What is more important is that the merged DRAM/logic LSI has a connection at a wide memory band width to achieve highly improved processing performance of the logic portion. In such an LSI, there arises a problem of conflict between refresh and access to the DRAM made by the logic circuit, thereby prolonging the time required for DRAM access and, hence, the processing performance of the logic circuit is deteriorated. This is a serious problem, comparable with the problem of achieving reduction in power consumption.
- The first object of the present invention, in a semiconductor circuit or, more particularly, in an LSI on which a DRAM and a logic circuit are merged, is to decrease the number of refreshing operations by refreshing only the rows storing necessary data, and thereby achieve both reduction in power consumption and prevention of deterioration in the performance of the logic circuit caused by an increase in the access time due to the confliction between refresh and access to the DRAM.
- A second object of the invention is to determine the rows storing data thereon according to the degree of importance of the data to, thereby, ensure saving important data without excessively shortening the refresh cycle time.
- A third object of the invention is to have refreshing operations carried out at suitable cycle time in conformity with the temperatures to thereby achieve both reduction in power consumption and prevention of deterioration in the processing performance of the logic circuit.
- In order to achieve the first object, the invention, in a control method of a merged DRAM/logic LSI, is characterized by comprising the steps of disposing data, which are combined so that the number of the rows storing data thereon may be reduced, on each of the rows of the DRAM and refreshing each of the rows having data stored thereon.
- Another method to achieve the first object, in a control method of a merged DRAM/logic LSI, is characterized by comprising the steps of disposing arbitrary data of which periods from being written in to being read out are overlapping or close to each other are disposed on the same row of the DRAM and refreshing the row only during the time period from the writing in of the data at the first to the reading out thereof at the end.
- A method to achieve the first and the second objects, in a control method of a merged DRAM/logic LSI, comprises the steps of obtaining the memory capacity required by an application using DRAM and storing data in DRAM in order from a row having a longer data retention time, with reference made to a table storing previously obtained data retention time of each row of DRAM, and setting a refreshing cycle adapted to the row whose data retention time is the shortest of the rows storing data thereon.
- In this method, data may be disposed, when it is stored in the DRAM, on specific rows in accordance with the importance of the data.
- A method to achieve the third object, in each of the above methods, is characterized by comprising the step of detecting the temperature of the semiconductor circuit and setting the refreshing cycle time of the DRAM according on the temperature.
- A semiconductor circuit to achieve the first object is characterized by comprising means for disposing data, which are combined so that the number of the rows storing the data thereon may be reduced, on each of the rows of the DRAM and means for refreshing each of the rows having the data stored thereon.
- Another semiconductor circuit which has achieved the first object is characterized by comprising means for disposing arbitrary data of which periods from being written in to being read out are overlapping or close to each other on the same row of DRAM, and means for refreshing the row only during the time period from the writing in of the data to the reading out thereof at the end.
- A semiconductor circuit which has achieved the first and the second objects is characterized by comprising means for obtaining the memory capacity required by an application using DRAM and storing data in the DRAM in an order starting from a row having longer data retention time, with reference made to a table storing the previously obtained data retention time of each row of DRAM, and means for setting a refreshing cycle time adapted to the row whose data retention time is the shortest of the rows storing data thereon.
- In this semiconductor circuit, there may be provided means for disposing data, when it is stored in the DRAM, on specific rows in accordance with the degree of importance of the data.
- A semiconductor circuit which has achieved the third object, in each of the above semiconductor circuits, is characterized by comprising means for detecting the temperature of the semiconductor circuit and setting the refreshing cycle of the DRAM according to the temperature.
- According to this invention, the following meritorious effects can be obtained:
- (1) Since only the rows storing necessary data are refreshed, the number of refreshing operations can be decreased and, hence, reduction in power consumption and prevention of deterioration in the processing performance of the logic circuit caused by contention between refresh and DRAM access can be attained.
- (2) Since the rows on which data are written are determined according to the degree of importance of the stored data, the need for excessively shortening the refreshing cycle can be eliminated and, hence, reduction in power consumption and prevention of deterioration in the processing performance of the logic circuit caused by contention between refresh and DRAM access can be attained.
- (3) Since the refreshing cycle is set according to temperature, the need for excessively shortening the refreshing cycle can be eliminated and, hence, reduction in power consumption and prevention of deterioration in the processing performance of the logic circuit caused by contention between refresh and DRAM access can be attained.
- FIG. 1 is a block diagram showing the general configuration of an information processing apparatus to which a semiconductor circuit of the invention is applied.
- FIG. 2 is a block diagram showing an internal configuration of the
main memory unit 4 shown in FIG. 1. - FIG. 3 is a block diagram showing an internal configuration of the
DRAM 7. - FIG. 4(a) is a diagram showing the relationship between stored positions of data and rows before optimization and FIG. 4(b) is a diagram showing the relationship between stored positions of data and rows after optimization.
- FIG. 5 is a functional diagram showing the principle of memory allocation made by a compiler in a first embodiment.
- FIG. 6 is a functional diagram showing the principle of memory allocation made by an operating system in the first embodiment.
- FIG. 7 is an explanatory drawing of a second embodiment, of which (a) is a graph showing lifetimes of data in a DRAM, (b) is an explanatory drawing of a case where data are disposed without taking the lifetimes of the data into consideration, (c) is a graph showing lifetimes of data disposed on each row, (d) is an explanatory drawing of a case where data whose lifetimes are similar are disposed on the same row and stored in the DRAM, and (e) is a graph showing the lifetimes of data on each of the rows after the disposition of the data A-H is optimized.
- FIG. 8 is a functional diagram showing the principle of memory allocation made by a compiler in the second embodiment.
- FIG. 9 is a functional diagram showing the principle of memory allocation made by an operating system in the second embodiment.
- FIG. 10 is a block diagram of an embodiment where a data retention time storing table is provided in a third embodiment.
- FIG. 11 is a functional diagram showing the principle of memory allocation made by a compiler in the third embodiment.
- FIG. 12 is a functional diagram showing the principle of memory allocation made by an operating system in the third embodiment.
- FIG. 13 is a functional diagram showing the principle of memory allocation made by a compiler in a fourth embodiment.
- FIG. 14 is a block diagram of a fifth embodiment provided with a means for detecting temperature and a means for controlling the refreshing cycle time.
- FIG. 1 is a block diagram showing a general configuration of an information processing apparatus to which the semiconductor circuit of the invention is applied. A processor1, a
ROM 3, amain memory unit 4, a temperature detection means 5, an I/O (input/output interface) 6, and the like are connected by abus 2. Themain memory unit 4 includes aDRAM 7 for storing data, aDRAM controller 8 for controlling write and read operations on theDRAM 7, and a data retention time storing means 9 for storing data retention times of each of the rows in theDRAM 7. Here, alogic portion 33 is constituted of non-memory circuits such as logic circuits, including the processor 1, the I/O 6, and the like. - FIG. 2 is a block diagram showing an internal configuration of the
main memory unit 4 shown in FIG. 1. TheDRAM controller 8 within themain memory unit 4 includes a row address generating means 10 for generating the row address of the row of theDRAM 7, atimer 11 for deciding the timing of generation of the row address, and a rowflag storing portion 12 for setting the row as the object of refreshing in theDRAM 7. TheDRAM 7 is connected to thebus 2 through the interface (i/f) 13. - One
DRAM controller 8 is implemented within the chip of a merged DRAM/logic LSI and performs control of the refreshing operation and the like at all times irrespective of the operation mode of the DRAM. - In a conventional system in which a logic circuit and a DRAM are connected with each other on a print circuit board, there are circuits for separately performing refreshing both outside and inside the DRAM, and they may operate in different cycles depending on the modes (normal mode/sleep mode) of the DRAM. This is the point in which it is different from the DRAM controller in the merged DRAM/logic LSI.
- Every time a predetermined time period is counted out by the
timer 11 being a component of theDRAM controller 8, the row address generating means 10 outputs a refresh address to thereby update the address of the row. Therow flag 12 includes the flags corresponding to each of the rows in the DRAM. TheDRAM controller 8 refers to the value of the flag corresponding to the row indicated by the row address generating means 10 every time a predetermined time period is counted out by thetimer 11, and carries out a refreshing operation only when the value of the flag is set so that a refreshing operation should be made and, otherwise, it makes no operation. As the predetermined time period, the time obtained by dividing the refreshing cycle time by the number of the rows, for example, is used. - The
DRAM controller 8 also has a function to coordinate the contention made between access to the DRAM through thebus 2 and a refreshing operation. Incidentally, the processor 1 and theDRAM 7 within themain memory unit 4 may have a path to connect with each other not by way of thebus 2. In this case, theDRAM controller 8 also performs the coordination of the contention between access to the DRAM through this path and a refreshing operation. - FIG. 3 is a block diagram showing an internal configuration of the
DRAM 7. Acontrol signal line 14 to which a control signal for controlling write and read is supplied, anaddress line 15 to which an address signal of (A+B) bits is supplied, and adata line 16 to which data is supplied are connected to an input/output interface 17. An address of (A+B) bits is separated by the input/output interface 17 into a row address of A bits and a column address of B bits, of which the row address of A bits is supplied to arow decoder 18 and the column address of B bits is supplied to acolumn decoder 19. Amemory cell 24 is disposed in the vicinity of the intersection of theword line 21 and abit line 23. The output of therow decoder 18 is supplied to theword line 21 selected from amemory cell array 20 and the data in the memory cells on the selected row is output over 2B pieces of the bit lines 23. Then, asense amplifier 22 amplifies the data. Further, the data on the bit lines selected by thecolumn decoder 19 are output to outside themain memory unit 4. - In the description above, the size of the column address is in agreement with the number of the bit lines and the minimum amount of data accessible at one time is one bit. The amount of data accessible at one time may be either a predetermined value or a value that is to be determined each time of access by a signal for specifying the size included in the
control signal 14. Further, the size of the space which the column address has may be smaller than the number of bit lines. In this case, the minimum amount of data accessible at one time is the value obtained by dividing the number of bit lines by the size of the column address space. - Some control methods used in the above described semiconductor circuit will be described below.
- First Embodiment
- The basic concept of the control method in the present invention will be described below as compared with the control method in the prior art.
- FIG. 4(a) shows the relationship between the stored position of data and the row when optimization of disposition of data has not been made. Referring to FIG. 4(a), the position indicated by the circle shows the position in which data is written and an oval indicates an aggregate of individual data.
- FIG. 4(a) shows an example where the positions to which the data are allocated are randomly determined. In this example, data are written on each of five rows, which are arranged in succession.
- Here in this embodiment, attention is paid to the disposition of data on the DRAM. With respect to the disposition of the data on the memory cells on each row, when the state of overlapping of the data is examined in the direction of the
bit line 23, there are no overlaps of data onrows - Then, the disposition of the data is determined, as shown in FIG. 4(b), so that the number of rows storing data thereon decreases while a plurality of data are disposed so as not to overlap each other on the same row. The method to determine the combination of the data allocated to each row with the aim of reducing the number of the rows is solved as a combinatorial optimization problem. In FIG. 4(b),
rows 21 a and 21 d to which data are allocated are indicated by thick lines androws 21 b, 21 c, and 21 e to which no data are allocated are indicated by thin lines. - Thus, after having data disposed as described above, setting for the control of execution of the refreshing operation is made for each row.
- Each field of the
row flag 12 corresponds to each row of the DRAM. TheDRAM controller 8 determines whether refreshing is to be done or not for each row in accordance with the flag value set in therow flag 12, and performs the refreshing control for the row which is set to be refreshed by the flag. Each flag of therow flag 12 can be realized by one bit (on/off). - As a method for determining data disposition, there is one performed by a compiler and another performed by an operating system.
- In the former method performed by the compiler as shown in FIG. 5, the memory allocation optimization process is applied to the
intermediate format 42 by the memory allocation means 41 as one of the functions of the compiler. The process for the combinatorial optimization of data is performed to minimize the number of rows to which data are actually allocated, with the number of rows and the number of memory cells on each row, as the structure of the memory, used as constraints to thereby determine relative addresses of individual data. Then, theintermediate format 43 is output with the operation for setting flags, corresponding to rows on which the data are to be stored, inserted therein. Theintermediate format 43 is converted into a sequence of instructions in the last step of the compiling process. As a method for actually setting the flag, there is a method which uses an instruction to automatically set/reset the flag for an accessed row, or a method which establishes the value by an immediate instruction or a data transfer instruction. - In the latter method, performed by the operating system as shown in FIG. 6, disposition of pages is determined by the address translation means44, when it maps the pages represented by the
virtual address 45 onto thephysical address 46, such that the number of the pages, i.e., the number of the rows to which the data are allocated, is minimized. The setting of therow flag 12 is realized by incorporating an instruction that sets the flag as described above, an immediate instruction, or a data transfer instruction, into the address translation means 43 such that the flag is set for the row on which data are actually mapped. - After the data have been disposed as described above, refreshing operations are carried out only for the rows on which the data are stored, whereby unnecessary refreshing operations can be decreased and both reduction in power consumption and prevention of increase in memory access time due to contention between refresh and memory access can be achieved.
- The position in which the compiler or the operating system operates may be either inside or outside the semiconductor circuit, provided that it is able to know the structure of the memory.
- Second Embodiment
- Now, a control method in which the disposition of data is determined taking note of the lifetime of the data, i.e., the period of time from first writing in of a variable to the reading out thereof at the end.
- FIG. 7(a) is a graph showing lifetimes of data A-H to be stored in a DRAM.
- If the data are mapped without considering their lifetimes data whose lifetimes are different will be allocated to the same row as shown in FIG. 7(b).
- FIG. 7(c) is a drawing in which the lifetimes of the data in FIG. 7(a) are rearranged for each row. Since, for example, row R1 has data B, C, and F stored thereon, it becomes necessary to constantly refresh row R1 as long as any of data B, C, and F are live.
- Therefore, in the present embodiment, as shown in FIG. 7(d), attention is paid to the lifetimes of data A-H, and the data are stored so that those having lifetimes close to or overlapping each other are disposed on the same row to be stored in the DRAM. The method for determining the combination of data to be allocated to each row with the aim of minimizing the sum total of the time periods, during which the live data are kept stored on each of the rows, is solved as a combination optimization problem.
- In the example shown in FIG. 7(d), data A and C are disposed on row R0, data B and D on row R1, data F and G on row R2, and data E on row R3.
- FIG. 7(e) is a graph showing the lifetimes of the data of each row after the positions, in which data A-H are written, have been altered.
- Thus, after allocating the data whose lifetimes are overlapping or close to each other to the same row, the setting for controlling the execution of the refreshing operation is made for each row.
- Each field of the
row flag 12 is corresponding to each row of the DRAM, and theDRAM controller 8 controls the refreshing operation for the relevant row only during the period set by the flag to indicate that the refreshing operation should be made. Each flag of therow flag 12 is realized by one bit (on/off). - As a method for determining the data disposition, there is one performed by a compiler and another performed by an operating system.
- In the former method, performed by the compiler as shown in FIG. 8, a memory allocation optimization process is applied to the
intermediate format 48 by the memory allocation means 47 as one of the functions of the compiler. The memory allocation means 47 obtains the lifetimes of the individual data on the basis of a flow graph as theintermediate format 49. The process for combinational optimization of data is performed to minimize the sum total for all of the rows of the lifetimes of each of the rows where at least a set of data is still live, with the number of rows and the number of memory cells as the structure of the memory used as the constraints, to thereby determine relative address of individual data. Then theintermediate format 49 is output with the operation to set the flag corresponding to the row on which the data is to be stored at the start of the lifetime of the data and the operation for resetting the flag of the row on which the data is stored at the end of the lifetime of the data inserted therein. Theintermediate format 49 is converted into a sequence of instructions in the last step of the compiling process. As a method for actually setting the flag, there is a method which uses an instruction to automatically set/reset the flag for an accessed row or a method which establishes the value by an immediate instruction or a data transfer instruction. - In the latter method, performed by the operating system as shown in FIG. 9, disposition of pages is determined by the address translation means50, when it maps the pages represented by the
virtual address 51 onto thephysical address 52 so that the sum total, for all the rows, of the lifetimes for each of the rows where at least one page is live is minimized, with the number of the rows and the number of the memory cells, as the structure of the memory used as the constraints. The lifetime of each page is given, when the page includes at least one set of data, by the time period during which the data is live, and it can be obtained by the compiler by analyzing the lifetime of each individual datum when it allocates the data to each page. The setting of the flag when the page is disposed on the physical memory and the resetting of the flag when the physical memory is released can be realized by incorporating a flag setting instruction, an immediate instruction, a data transfer instruction, or the like, as described in the first embodiment, into the address translation means 50. - After the data have been disposed as described above, refreshing operations are carried out for the rows storing data only for the time period during which the data is live, whereby unnecessary refreshing operations can be decreased, power consumption reduced, and increase in the memory access time due to contention between refresh and memory access prevented.
- The method of optimization in the temporal domain as described in the second embodiment above and the method of optimization in the spatial domain as described in the first embodiment may be practiced at the same time.
- The compiler or the operating system may work either at inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- Third Embodiment
- A control method for determining the disposition of data, with attention paid to the data retention time of each of the rows of a DRAM, will be described.
- In this example, there is provided, as schematically shown in FIG. 10, the data retention time storing table9 a storing the data retention times tr0-tr7 on each of a plurality of rows R0-R7 of the
DRAM 7. - The storing table9 a corresponds to the data retention time storing means 9 shown in FIG. 1 and can be realized by a device which can be written onto such as a PROM, an EPROM, a FLASH memory, or an FPGA. On the table 9 a, there is stored, for each row, the data retention time of the memory cell which has the shortest data retention time among all of the memory cells included in the row.
- The measurement of the data retention time is carried out when the DRAM is tested. The results of the measurement are stored on the storing table9 a. The information stored in table 9 a, as the data retention time may be the time itself, or an encoded value which is divided by some value.
- When the times are divided by rank, an advantage is obtained in that the size of the table can be decreased.
- When data are stored in a
main memory unit 4, the data retention times of each of the rows stored on the storing table 9 a are referred to and the data are stored in order from the row having the longer data retention time. Then, the refreshing cycle time adapted to the shortest data retention time of the row on which data has actually been stored is set in thetimer 11. - Upon setting of the
timer 11 as described above, setting for controlling the execution of refreshing is made for each row. - Each field of the
row flag 12 corresponds to each row of the DRAM. TheDRAM controller 8 determines whether refreshing is to be made or not for each row in accordance with the flag value set in therow flag 12 and performs the control of refreshing of the row which is set by the flag to be refreshed. Each flag of therow flag 12 can be realized by one bit (on/off). - As a method for determining the data disposition, there is one performed by a compiler and another performed by an operating system.
- In the former method performed by the compiler as shown in FIG. 11, a process of memory allocation optimization is applied to the
intermediate format 54 by the memory allocation means 53 as one of the functions of the compiler. When data are allocated to the memory, the values stored on the storing table 9 a are referred to and the physical address corresponding to the row having the longer data retention time is preferentially taken as the object of data allocation, with the number of rows and the number of memory cells as the structure of the memory used as the constraints. Then, referring to the data retention time of the row which was taken lastly as the object of allocation, the refresh cycle time to be set in thetimer 8 is determined. Finally, the intermediate format 55, including the operation to set the flag for the row on which actual data is to be stored and the operation to set the refreshing cycle time, is output. The intermediate format 55 is converted into a sequence of instructions in the last step of the compiling process. As a method for actually setting the flag, there are two methods. One is to use an instruction to automatically set/reset the flag for an accessed row, the other is to establish the value by an immediate instruction or a data transfer instruction. - In the latter method, performed by the operating system as shown in FIG. 12, when a page represented by the
virtual address 57 is mapped onto thephysical address 58 by the address translation means 56, the row having the longer data retention time is preferentially taken as the object of allocation of the page, i.e., the data, with the number of rows and the number of memory cells on each row as the structure of the memory taken as the constraints. Then, reference is made to the data retention time of the row lastly taken as the object of allocation and the refreshing cycle time is set in thetimer 8 and thus the method can be realized. The setting of the flag and the refreshing cycle time is realized by incorporating an instruction, such as a flag setting instruction, an immediate instruction, or a data transfer instruction, as described in the first embodiment, into the address translation means 56. - Thus, by using the rows preferentially from those having longer data retention time, the number of the refreshing operations can be decreased, power consumption reduced, and increase in the memory access time due to contention between refresh and memory access prevented.
- It is also possible to use a method for determining the disposition of data by taking note of the data retention time of each row as described in the third embodiment in parallel with the method for optimization in the spatial domain as described in the first embodiment and the method for optimization in the temporal domain as described in the second embodiment.
- The position in which the compiler or the operating system operates may be either inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- Fourth Embodiment
- A method by which disposition of data, when the data are stored in a DRAM, is determined in accordance with the degree of importance of the data will be described.
- In the data to be stored in the DRAM, there are, for example, those that can be corrected for error, such as communication packets, and those that cannot be corrected for error.
- There are also those data that allow for no error produced therein such as those for numerical calculation and control purposes, whereas there are those that cause no serious inconvenience even if some error is present therein such as picture/voice data.
- Therefore, the rows to store data thereon are determined according to the degree of importance of the data. Namely, those data which can be corrected for error or those data which allow for some error present therein are allocated to rows that have shorter holding times. On the other hand, those data which cannot be corrected for error or those data which allow for no numerical error present therein are allocated to such rows that have longer data retention times. Thus, by disposing data according to the degree of importance of the data thereby suitably prolonging the cycle time of refreshing, the frequency of the refreshing operations can be decreased.
- The determination of the disposition of the data according to their degree of importance can be achieved, as shown in FIG. 13, by a compiler corresponding to the
intermediate format 60 having a function indicative of the degree of importance of the data. The degree of importance of the data can be described in a program by for example the provision of types showing the degrees of importance. The memory allocation means 59 of the compiler shown in FIG. 13, making reference to the data retention time of each row stored in the data retention time storing means 9, disposes data on theintermediate format 60 in accordance with the degree of importance of the data, determines the physical address of individual data, and outputs theintermediate format 61. Theintermediate format 61 is converted into a sequence of instructions in the last step of the compiling process. - The position in which the compiler operates may be either inside or outside the semiconductor circuit provided that it is able to know the structure of the memory.
- Fifth Embodiment
- It is a characteristic of ordinary DRAM that the data retention time of each row is not stationary, but variable with the temperature. Namely, as described above, the lower the temperature, the longer the data retention time. The interval between the refreshing operations of a conventional DRAM is generally set to be very short assuming the worst operating condition, namely a high-temperature condition. However, since the actual data retention time at normal temperatures is great, the refreshing cycle time adapted to the high-temperature condition becomes very short as compared with the actual data retention time, which means that excessive refreshing is then being practiced.
- Thus, not only is electric power overly is consumed, but average memory access time increases, because of the more frequent contention occurring between refresh and memory access and, thereby, the performance of the
logic portion 33 merged with theDRAM 7 deteriorates. - Therefore, in the embodiment shown in FIG. 14, the semiconductor circuit formed of the
DRAM 7 and thelogic portion 33 merged on a chip is provided therein with the temperature detection means 5 for detecting the temperature of the semiconductor circuit. Thus, by controlling thetimer register 8 a of theDRAM controller 8 according to the temperature, the refreshing cycle time is suitably set. - The temperature detection means5 can be realized by a method in which a leakage current of a memory cell having the same temperature characteristic as the memory cell for actually storing data is monitored, a method in which a ring oscillator is used, or the like (1994. Ito, “VLSI Memory”, Baifukan).
- Thus, by setting the refreshing cycle time upon detection of the temperature, thereby decreasing the frequency of the refreshing operations, execution of unnecessary refreshing in the DRAM/logic hybrid LSI can be decreased and, thereby, both reduction in power consumption and prevention of increase in the access time due to contention between refresh and memory access can be achieved.
- It is also possible to use the method as described in the fifth embodiment, in which the refreshing cycle time is set upon detection of the temperature, in parallel with the method of optimization in the spatial domain as described in the first embodiment, the method of optimization in the temporal domain as described in the second embodiment, and the method for determining the disposition of data in accordance with the data retention time as described in the third and fourth embodiments.
- The invention can be utilized in the field of the semiconductor circuits employing DRAM.
Claims (10)
Applications Claiming Priority (3)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP9-155125 | 1997-06-12 | ||
JP15512597 | 1997-06-12 | ||
PCT/JP1998/002543 WO1998057332A1 (en) | 1997-06-12 | 1998-06-08 | Semiconductor circuit and method of controlling the same |
Publications (2)
Publication Number | Publication Date |
---|---|
US20020004882A1 true US20020004882A1 (en) | 2002-01-10 |
US6446159B2 US6446159B2 (en) | 2002-09-03 |
Family
ID=15599113
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US09/242,049 Expired - Lifetime US6446159B2 (en) | 1997-06-12 | 1998-06-08 | Semiconductor circuit and method of controlling the same |
Country Status (7)
Country | Link |
---|---|
US (1) | US6446159B2 (en) |
EP (1) | EP0917152B1 (en) |
JP (1) | JP4246812B2 (en) |
KR (1) | KR100419462B1 (en) |
CN (1) | CN100336134C (en) |
DE (1) | DE69835794T2 (en) |
WO (1) | WO1998057332A1 (en) |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20040268031A1 (en) * | 2001-09-20 | 2004-12-30 | Lawrence Richard H | System and method for managing data in memory for reducing power consumption |
US6938118B1 (en) * | 1998-10-28 | 2005-08-30 | Imsys Technologies Ab | Controlling access to a primary memory |
US20050243626A1 (en) * | 2004-04-29 | 2005-11-03 | M-Systems Flash Disk Pioneers Ltd. | Refreshing data stored in a flash memory |
US20060117162A1 (en) * | 2004-12-01 | 2006-06-01 | Sauber William F | System and method for information handling system memory page mapping optimization |
US20070012602A1 (en) * | 2002-05-16 | 2007-01-18 | United Parcel Service Of America, Inc. | Systems and Methods for Package Sortation and Delivery Using Radio Frequency Identification Technology |
DE10304673B4 (en) * | 2002-02-05 | 2010-03-11 | Qimonda Ag | Refresh circuit for dynamic memory |
US20120272036A1 (en) * | 2011-04-22 | 2012-10-25 | Naveen Muralimanohar | Adaptive memory system |
US20160027523A1 (en) * | 2013-07-29 | 2016-01-28 | Western Digital Technologies, Inc. | Data retention monitoring using temperature history in solid state drives |
US11977098B2 (en) | 2009-03-25 | 2024-05-07 | Aehr Test Systems | System for testing an integrated circuit of a device and its method of use |
Families Citing this family (40)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DK0757336T3 (en) | 1995-08-04 | 2001-03-19 | Belle Gate Invest B V | Data Exchange System comprising portable data processing units |
US6385645B1 (en) | 1995-08-04 | 2002-05-07 | Belle Gate Investments B.V. | Data exchange system comprising portable data processing units |
CN1322430A (en) | 1998-09-29 | 2001-11-14 | 太阳微系统公司 | Superposition of data over voice |
ATE475139T1 (en) | 1999-06-10 | 2010-08-15 | Belle Gate Invest B V | DEVICE FOR STORING DIFFERENT VERSIONS OF DATA SETS IN SEPARATE DATA AREAS AND METHOD FOR UPDATE A DATA SET IN A MEMORY |
WO2001040910A1 (en) * | 1999-12-06 | 2001-06-07 | De Jong, Eduard, Karel | Computer arrangement using non-refreshed dram |
KR100759257B1 (en) | 1999-12-06 | 2007-09-17 | 선 마이크로시스템즈 인코포레이티드 | Computer device using non refresh DRAM |
WO2001043080A1 (en) | 1999-12-07 | 2001-06-14 | Sun Microsystems Inc. | Secure photo carrying identification device, as well as means and method for authenticating such an identification device |
US7080261B1 (en) | 1999-12-07 | 2006-07-18 | Sun Microsystems, Inc. | Computer-readable medium with microprocessor to control reading and computer arranged to communicate with such a medium |
JP2004505354A (en) | 2000-07-20 | 2004-02-19 | ベル ゲイト インヴェストメント ベスローテン フェンノートシャップ | METHOD AND SYSTEM FOR PROTECTING DATA TRANSFER AND COMMUNICATION OF DEVICE |
JP2002056671A (en) | 2000-08-14 | 2002-02-22 | Hitachi Ltd | Data holding method of dynamic RAM and semiconductor integrated circuit device |
WO2002032231A1 (en) * | 2000-10-19 | 2002-04-25 | Edens, Luppo | Protein hydrolysates |
CN1231919C (en) * | 2001-02-16 | 2005-12-14 | 皇家菲利浦电子有限公司 | Method for updating dynamic memory |
JP4212257B2 (en) | 2001-04-26 | 2009-01-21 | 株式会社東芝 | Semiconductor integrated circuit |
JP4700223B2 (en) * | 2001-05-18 | 2011-06-15 | 株式会社バッファロー | DRAM device and DRAM device refresh method |
KR100413761B1 (en) * | 2001-05-31 | 2003-12-31 | 삼성전자주식회사 | Semiconductor memory device capable of controlling refresh cycle by variation of temperature and process and method thereof |
KR100406555B1 (en) * | 2001-06-29 | 2003-11-20 | 주식회사 하이닉스반도체 | Semiconductor memory device and method for testing therefor |
US6990342B2 (en) * | 2002-08-29 | 2006-01-24 | Motorola, Inx. | Method and apparatus for cell reselection within a communications system |
US6920523B2 (en) * | 2002-10-07 | 2005-07-19 | Infineon Technologies Ag | Bank address mapping according to bank retention time in dynamic random access memories |
CN100487816C (en) | 2003-04-23 | 2009-05-13 | 富士通微电子株式会社 | Semiconductor memory and refreshing method thereof |
US6876593B2 (en) * | 2003-07-01 | 2005-04-05 | Intel Corporation | Method and apparatus for partial refreshing of DRAMS |
US7345940B2 (en) * | 2003-11-18 | 2008-03-18 | Infineon Technologies Ag | Method and circuit configuration for refreshing data in a semiconductor memory |
CN100409365C (en) * | 2003-12-05 | 2008-08-06 | 晶豪科技股份有限公司 | Data refreshing method for pseudo static random access memory |
US7145823B2 (en) | 2004-06-30 | 2006-12-05 | Intel Corporation | Method and apparatus to implement a temperature control mechanism on a memory device |
JP4786155B2 (en) * | 2004-08-18 | 2011-10-05 | ルネサスエレクトロニクス株式会社 | Semiconductor device and semiconductor device refresh processing method |
US7342841B2 (en) | 2004-12-21 | 2008-03-11 | Intel Corporation | Method, apparatus, and system for active refresh management |
KR100803352B1 (en) | 2006-06-12 | 2008-02-14 | 주식회사 하이닉스반도체 | Refresh control device and method of semiconductor memory |
US20080016272A1 (en) * | 2006-07-03 | 2008-01-17 | Stmicroelectronics Sa | Method of refreshing dynamic random access memory, in particular in standby mode and in active operating mode, and corresponding dynamic random access memory device, for example incorporated into a cellular mobile telephone |
KR100832029B1 (en) * | 2006-09-28 | 2008-05-26 | 주식회사 하이닉스반도체 | Temperature information output device and semiconductor device having same |
KR100832006B1 (en) | 2006-09-29 | 2008-05-26 | 주식회사 하이닉스반도체 | A semiconductor memory device having a cell refresh cycle generation device and a driving method thereof |
JP5439788B2 (en) * | 2008-10-03 | 2014-03-12 | 富士通株式会社 | Computer device and memory usage setting program |
US8984217B2 (en) * | 2010-08-24 | 2015-03-17 | Qualcomm Incorporated | System and method of reducing power usage of a content addressable memory |
JP5737753B2 (en) * | 2011-03-30 | 2015-06-17 | 日本電気株式会社 | Network device, network device management method, management program, and packet network system |
JP2014030110A (en) * | 2012-07-31 | 2014-02-13 | Toshiba Corp | Reconfigurable integrated circuit device and method of writing to the same |
CN104143355B (en) * | 2013-05-09 | 2018-01-23 | 华为技术有限公司 | A kind of method and apparatus of refreshed dram |
US9336855B2 (en) * | 2013-05-14 | 2016-05-10 | Qualcomm Incorporated | Methods and systems for smart refresh of dynamic random access memory |
CN104239220B (en) * | 2013-06-13 | 2017-11-24 | 华为技术有限公司 | Memory refress treating method and apparatus |
CN103810126B (en) * | 2014-01-27 | 2017-06-13 | 上海新储集成电路有限公司 | Mixing DRAM memory and the method for reducing power consumption when the DRAM memory refreshes |
US10447433B2 (en) | 2014-11-21 | 2019-10-15 | Telefonaktiebolaget Lm Ericsson (Publ) | Signal processing apparatus and method |
US10691572B2 (en) | 2017-08-30 | 2020-06-23 | Nvidia Corporation | Liveness as a factor to evaluate memory vulnerability to soft errors |
EP4365738A1 (en) * | 2022-11-04 | 2024-05-08 | Aptiv Technologies AG | Method to dimension a buffer memory |
Family Cites Families (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4172282A (en) * | 1976-10-29 | 1979-10-23 | International Business Machines Corporation | Processor controlled memory refresh |
JPS60157353A (en) * | 1984-01-26 | 1985-08-17 | Citizen Watch Co Ltd | Communication system for inquiry of printer information |
US4920489A (en) * | 1987-08-14 | 1990-04-24 | Cardiodata Inc. | Apparatus and method for solid state storage of episodic signals |
JPH01290193A (en) * | 1988-05-17 | 1989-11-22 | Pfu Ltd | DRAM refresh control method |
DE69127518T2 (en) * | 1990-06-19 | 1998-04-02 | Dell Usa Lp | Digital computer that has a system for the successive refresh of an expandable dynamic RAM memory circuit |
US5276843A (en) * | 1991-04-12 | 1994-01-04 | Micron Technology, Inc. | Dynamic RAM array for emulating a static RAM array |
JPH05242671A (en) * | 1992-02-27 | 1993-09-21 | Toshiba Corp | Device for refreshing dram |
JP2546161B2 (en) * | 1992-07-22 | 1996-10-23 | 日本電気株式会社 | Dynamic memory device |
JP2546138B2 (en) * | 1993-05-17 | 1996-10-23 | ティアック株式会社 | Disk drive |
US6085283A (en) * | 1993-11-19 | 2000-07-04 | Kabushiki Kaisha Toshiba | Data selecting memory device and selected data transfer device |
JPH09306164A (en) * | 1996-05-13 | 1997-11-28 | Internatl Business Mach Corp <Ibm> | Memory refresh system |
US5907857A (en) * | 1997-04-07 | 1999-05-25 | Opti, Inc. | Refresh-ahead and burst refresh preemption technique for managing DRAM in computer system |
US6094704A (en) * | 1997-06-17 | 2000-07-25 | Micron Technology, Inc. | Memory device with pipelined address path |
-
1998
- 1998-06-05 JP JP15803398A patent/JP4246812B2/en not_active Expired - Fee Related
- 1998-06-08 EP EP98923184A patent/EP0917152B1/en not_active Expired - Lifetime
- 1998-06-08 KR KR10-1999-7001225A patent/KR100419462B1/en not_active Expired - Fee Related
- 1998-06-08 US US09/242,049 patent/US6446159B2/en not_active Expired - Lifetime
- 1998-06-08 CN CNB988009846A patent/CN100336134C/en not_active Expired - Fee Related
- 1998-06-08 DE DE69835794T patent/DE69835794T2/en not_active Expired - Lifetime
- 1998-06-08 WO PCT/JP1998/002543 patent/WO1998057332A1/en active IP Right Grant
Cited By (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US6938118B1 (en) * | 1998-10-28 | 2005-08-30 | Imsys Technologies Ab | Controlling access to a primary memory |
US20040268031A1 (en) * | 2001-09-20 | 2004-12-30 | Lawrence Richard H | System and method for managing data in memory for reducing power consumption |
US7380048B2 (en) * | 2001-09-20 | 2008-05-27 | Intel Corporation | System and method for managing data in memory for reducing power consumption |
DE10304673B4 (en) * | 2002-02-05 | 2010-03-11 | Qimonda Ag | Refresh circuit for dynamic memory |
US20070012602A1 (en) * | 2002-05-16 | 2007-01-18 | United Parcel Service Of America, Inc. | Systems and Methods for Package Sortation and Delivery Using Radio Frequency Identification Technology |
US20050243626A1 (en) * | 2004-04-29 | 2005-11-03 | M-Systems Flash Disk Pioneers Ltd. | Refreshing data stored in a flash memory |
US7325090B2 (en) * | 2004-04-29 | 2008-01-29 | Sandisk Il Ltd. | Refreshing data stored in a flash memory |
US20060117162A1 (en) * | 2004-12-01 | 2006-06-01 | Sauber William F | System and method for information handling system memory page mapping optimization |
US8250296B2 (en) * | 2004-12-01 | 2012-08-21 | Dell Products L.P. | System and method for information handling system memory page mapping optimization |
US11977098B2 (en) | 2009-03-25 | 2024-05-07 | Aehr Test Systems | System for testing an integrated circuit of a device and its method of use |
US20120272036A1 (en) * | 2011-04-22 | 2012-10-25 | Naveen Muralimanohar | Adaptive memory system |
US20160027523A1 (en) * | 2013-07-29 | 2016-01-28 | Western Digital Technologies, Inc. | Data retention monitoring using temperature history in solid state drives |
US9595347B2 (en) * | 2013-07-29 | 2017-03-14 | Western Digital Technologies, Inc. | Data retention monitoring using temperature history in solid state drives |
Also Published As
Publication number | Publication date |
---|---|
DE69835794D1 (en) | 2006-10-19 |
US6446159B2 (en) | 2002-09-03 |
JPH1166845A (en) | 1999-03-09 |
JP4246812B2 (en) | 2009-04-02 |
EP0917152A4 (en) | 2000-09-06 |
EP0917152A1 (en) | 1999-05-19 |
CN100336134C (en) | 2007-09-05 |
EP0917152B1 (en) | 2006-09-06 |
KR100419462B1 (en) | 2004-02-19 |
WO1998057332A1 (en) | 1998-12-17 |
CN1234901A (en) | 1999-11-10 |
KR20000068150A (en) | 2000-11-25 |
DE69835794T2 (en) | 2007-09-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US6446159B2 (en) | Semiconductor circuit and method of controlling the same | |
US11631450B2 (en) | Partial refresh technique to save memory refresh power | |
JP5063041B2 (en) | Dynamic semiconductor memory with improved refresh mechanism | |
US6167484A (en) | Method and apparatus for leveraging history bits to optimize memory refresh performance | |
KR20100022061A (en) | Methods, circuits, and systems to select memory regions | |
JP5000844B2 (en) | Memory controller and refresh cycle control method for controlling refresh cycle of memory | |
WO2008144609A1 (en) | Methods, circuits, and systems to select memory regions | |
JP2012518242A (en) | Dynamic random access memory (DRAM) refresh | |
JP2010534897A (en) | System and method for reducing dynamic RAM power consumption through the use of valid data indicators | |
KR100515072B1 (en) | Semiconductor memory device for saving power consumption in refresh operation | |
KR0142795B1 (en) | Dram refresh circuit | |
JP2000021162A (en) | Volatile memory and embedded dynamic random-access memory | |
US6650586B1 (en) | Circuit and system for DRAM refresh with scoreboard methodology | |
US20230037415A1 (en) | Memory system testing, and related methods, devices, and systems | |
JP2001093278A (en) | Semiconductor memory | |
US6741516B2 (en) | Semiconductor memory | |
US20240329846A1 (en) | Memory sprinting | |
JP2868464B2 (en) | Semiconductor memory device and refresh control method thereof | |
JPH04195890A (en) | Memory system, memory element and refresh method | |
JPH0660645A (en) | Power saving storage device | |
JPH07254272A (en) | Semiconductor device | |
JP2002269981A (en) | Semiconductor memory | |
JPS6231091A (en) | System for controlling refresh of dynamic memory | |
JPH09251775A (en) | Memory control device and memory device | |
JPS61142590A (en) | Semiconductor memory device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: MATSUSHITA ELECTRIC INDUSTRIAL CO., LTD., JAPAN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:KAI, KOJI;OHSAWA, TAKU;MURAKAMI, KAZUAKI;REEL/FRAME:012816/0582 Effective date: 19990205 |
|
STCF | Information on status: patent grant |
Free format text: PATENTED CASE |
|
FPAY | Fee payment |
Year of fee payment: 4 |
|
FEPP | Fee payment procedure |
Free format text: PAYOR NUMBER ASSIGNED (ORIGINAL EVENT CODE: ASPN); ENTITY STATUS OF PATENT OWNER: LARGE ENTITY |
|
FPAY | Fee payment |
Year of fee payment: 8 |
|
FPAY | Fee payment |
Year of fee payment: 12 |
|
AS | Assignment |
Owner name: SOCIONEXT INC., JAPAN Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:PANASONIC CORPORATION;REEL/FRAME:035294/0942 Effective date: 20150302 |