Combined cache with main memory and a control method thereof

Details Combined cache with main memory and a control method thereof Combined cache with main memory and a control method thereof

1998-11-17

2001-12-25

Kim, Matthew (Department: 2186)

Electrical computers and digital processing systems: memory

Storage accessing and control

Control technique

C711S209000, C711S219000

Reexamination Certificate

active

06334173

ABSTRACT:

BACKGROUND OF THE INVENTION
The present invention relates to a cache device, more particularly to a combined cache with main memory and a control method thereof, in which an internal cache of a microprocessor is used for both caching function and main memory function.
A large capacity of cache is provided within a high performance microprocessor in order to improve the performance of the processor. As is well known, cache is a storage medium for storing a data to be frequently used, in which the data is previously read out from another storage medium having relatively lower access speed (i.e. main memory). Since the access to the cache is faster than that to the main memory, a required data can be quickly transferred to the microprocessor, thereby improving the performance of the system.
Most of computer systems use as a main memory a DRAM (Dynamic Random Access Memory) which is cheap and can be easily used for implementing large quantity of main memory. Alternatively, as a cache is used a SRAM (Static Random Access Memory), which is faster in access speed though the cost is high. Due to the development of personal computers, 16 kilobytes of internal cache is embedded into Pentium processor, while 512 kilobytes of SRAM cache are further embedded into Pentium Pro processor other than 16 kilobytes of internal cache.
‘Cache hit’ represents the case that the data required can be fetched from a cache without access to a main memory because the data is stored in the cache, while ‘cache miss’ represents the case that the data required should be fetched from the main memory because the data is not stored in the cache. In the case of cache miss, if there is any vacant space in the cache, some parts located nearby the currently accessed data are read out from the main memory to be stored into the cache, considering their locality. Alternatively, if there is no vacant space in the cache, a replacement algorithm can be used for selecting a needless data block from the cache, and the selected block of the cache is then used for storing the data block related to the currently accessed data from the main memory.
As cache memory, there is included a fully associative mapped cache, a directly mapped cache, an N-way cache and so on. The fully associative mapped cache uses the entire address as tag so that the entire address of data is stored into a tag register of cache when the data is loaded into the cache. The directly mapped cache is proposed for improving the hardware complexity of the fully associative mapped cache, in which the entire address of data can be divided into 2 fields: a tag field and an offset field, and used for the control of the cache. Also, the N-way cache that increases data storage by n-times is proposed for increasing the cache hit rate.
FIG. 1
is a diagram for schematically illustrating the address bus of the directly mapped cache and the N-way cache, in which the address bus can be divided into a tag field and an offset field, for the control of the directly mapped cache and the N-way cache.
Here, during data loading, the directly mapped cache stores the tag field of address into the tag register of the cache. During data reading, the directly mapped cache responds to the offset field to determine if the data can be read out from which data register within the cache.
The N-way cache stores the tag field of the address into the tag register of the cache during data loading, and determines which data can be read out from a certain way in response to the offset field during data reading.
For the convenience of explanation, it is assumed that the address is 16 bits, the offset field 2 bits, the tag field 14 bits, the data bus 8 bits and there are 4 ways.
FIG. 2
is an internal block diagram of a conventional 4-way cache used only as cache function, which includes 4 ways (way_
0
, way_
1
, way_
2
and way_
3
), an OR gate
10
and a multiplexer
11
. Each of 4 ways receives corresponding tag field and corresponding offset field of the address bus, respectively. The OR gate
10
performs a logic OR operation on output signals applied from four ways, so as to generate a cache hit signal when any one of four ways allows a way hit to occur. The multiplexer
11
selectively transfers one of data applied from four ways, in response to way hit signals applied from four ways.
FIG. 3
is an internal block diagram of one of 4 ways shown in FIG.
2
. The way includes a tag register
30
, a comparator
31
, a tag hit latch
32
, four 8-bit data registers
33
, a multiplexer
34
, four 1-bit valid bit registers
35
, a multiplexer
36
, an AND gate
37
. The tag register
30
stores 14-bit tag field applied through the address bus when the data is loaded into the cache. The comparator
31
compares the tag value stored in the tag register
30
with the tag value of the address inputted through the address bus for reading out from the cache. The tag hit latch
32
stores the compared result signal from the comparator
31
as a tag hit signal. Each of four 8-bit data registers
33
stores the data loaded from the main memory into the cache. The multiplexer
34
selectively outputs one of the data stored in four 8-bit data registers
33
,in response to the 2-bit offset field of the address inputted through the address bus. Each of four 1-bit valid bit registers
35
stores corresponding valid bit indicating whether the data stored in the corresponding one of four data registers
33
is valid or not. The multiplexer
36
selectively outputs one of the content of four 1-bit valid bit registers
35
, in response to the 2-bit offset field of the address inputted through the address bus. The AND gate performs an logic AND operation on the tag hit signal from the tag hit latch
32
and the valid bit from the multiplexer
36
, so as to produce the result as a way hit signal.
The operation of the 4-way cache as configured above will be explained.
First, the comparator in the respective ways compares the value of tag field of the address bus applied for reading out from the cache with the content of the tag register
30
. When the compared result indicates that the value of the tag field is the same as the value stored in the tag register
30
, ‘tag hit signal’ is generated so that the tag hit latch
32
is set to “1”, and the multiplexer
34
selectively produces one among four data registers
33
to the data bus, in response to the 2-bit offset field of the inputted address bus. Also, the multiplexer
36
selects and produces the corresponding valid bit to the selected data register as an offset field. Here, when the tag hit signal stored in the tag hit latch
32
is “1” and the selected valid bit from the multiplexer
36
is “1”, the way hit signal becomes “1”.
Next, a cache hit signal is produced, based on respective way hit signals generated as above, and the data outputted from the way generating the way hit signal is finally transferred to the data bus for the microprocessor.
The capacity of cache becomes larger, and the memory must be required in order to store the data for the control of microprocessor and micro-controller. Considering this point, when a microprocessor embedded with large amount of cache is used as a CPU (Central Process Unit) of a computer system having a large amount of memory; the internal cache memory is used for cache-unique purpose. On the other hand, when the microprocessor embedded with cache is used for system control that requires relatively small amount of memory but high speed processing, the internal cache is used as a main memory, thereby increasing the processing speed of the system and maximizing the utility of the cache.
SUMMARY OF THE INVENTION
The present invention is proposed, considering the above-mentioned requirements. The object of the present invention is to provide a combined cache with main memory and a control method thereof, which can be configured by only adding a minimized control circuit to one of various structures of cache that is used for main memory function.
Further, the other object of the present invention is to provide a combined

Affiliated with

Also associated with

Rating

Combined cache with main memory and a control method thereof does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Combined cache with main memory and a control method thereof, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Combined cache with main memory and a control method thereof will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-2575509