US 5,249,277 AGrant
Optimized Performance Memory Method and System
Issue Date:1993-09-28
•19 Claims
•19 Drawing Sheets
Abstract
A system and method for optimizing the performance of a computer memory system substitutes faster memory for a certain portion of the main memory. The substituted memory section automatically assumes a predesignated portion of the main memory and allows the main memory to become sequentially reorganized. Using this arrangement, for example, all the DOS applications that reside in the first 640 Kbytes of memory can be resident in the faster memory and the original main memory can then be used for other applications.
Metadata
Assignee
- Compuadd Corporation
Inventors
- James H. Leftwich
- Gregory D. Roberts
- James M. O'Quinn
Application Information
Application Number:US 5644313
Filing Date:1990-08-08
Priority Date:1990-08-08
Art Unit:232
Classifications
IPC:
G06F 1202
Field of Search:
395365364425;400230.01;230.03200 MS File;900 MS File
Patent Drawings (19 sheets)
Description
Technical Field of the Invention
This invention relates to computer memory systems and more particularly to such a system which automatically adapts to the presence of additional fast performance memory.
Background of the Invention
The IBM PC family is based on the 8088, 80286 and 80386 Intel processors. A popular operating system for these PC's is the disk operating system ("DOS"). This operating system and all the applications that run under it are stored in the first 640 kilobytes of random access memory ("RAM"). (DOS 4.01 and earlier).
In order to speed up performance of DOS and DOS applications, many computer designers have installed memory caches on their systems. These systems typically use a small high speed memory as the "cache". This technique increases memory access speed when the program is loaded into the cache. This solution partially fails, however, when the size of DOS and the DOS applications exceeds the size of the cache memory. The system is then forced to wait to address data in the slower speed memories. Another disadvantage of cache memory systems are DRAM refresh delays. These delays are created by the need to periodically refresh dynamic RAM's. Because the caches are small and cannot contain entire DOS and DOS application data, they must be linked with the DRAM's. The DRAM's often contain spillover data from the cache. The cache memory then is idle while the DRAM's are refreshed.
It is one objective of the present invention to optimize performance of the memory containing the DOS and DOS applications.
This is a second objective of this inventor to have a zero wait state during memory reads and writes.
It is a still further objective of this inventor to avoid overhead involved in maintaining the slower dynamic random access memory ("DRAM") memory that contains the remaining system memory.
Summary of the Invention
To achieve performance better than cache based systems, the first 640 kilobytes of memory are mapped to a high speed static random access memory ("SRAM") array. These addresses contain DOS and DOS applications. The most frequently accessed addresses are now within a high speed SRAM array. The array can achieve zero wait state reads and writes because of its high operating speed and static memory characteristics. There is no refresh overhead associated with the SRAM array. Because the array is large enough to contain all of DOS and of DOS applications, it may operate independently of the DRAM refresh cycle.
Therefore, it is a technical advantage of this invention that, when installed, a signal from the invention is passed to the mother board telling it to sequentially map the first 640 kilobytes of memory to the card. The remaining memory is mapped sequentially to the cache and DRAM. When the card is removed, the absence of the signal tells the mother board to map all memory to the DRAM and cache system. It is a further technical advantage of this invention to provide high speed memory cells for the most frequently addressed data.
Brief Description of the Drawings
The foregoing objects, features and technical advantages, as well as others, of the invention will be more apparent from the following description of the drawings in which:
FIG. 1 shows a schematic representation of the host computer and of the fast performance memory;
FIG. 2 illustrates the location of system memory with and without the SRAM card installed; and
FIGS. 3-7, 8a, 8b, 9a, 9b, 10a, 10b, 11a, 11b, 12, 13, 14a, 14b show a specific embodiment of the invention operating in conjunction with a computer system.
FIG. 15 illustrates a flowchart of the steps performed by a preferred embodiment of the invention.
Detailed Description of the Invention
FIG. 1 shows system 10 with address bus 17 and data bus 16. When SRAM 11 is not installed, processor 101 only uses cache memory 12 and DRAM 13 to store information. Individual memory cells in cache 12 and DRAM 13 are identified by processor 101 through address bus 17. The contents of individual cells in cache 12 and DRAM 13 are passed to and from processor 101 through data bus 16. Buffers 14 and 15 isolate the circuitry of DRAM 13 from processor 101 and from any other device connected to buses 16 and 17. Buffers 14 and 15 also latch highspeed signals on buses 14 and 15 for use by DRAM 13.
When SRAM card 11 is installed in system 10, processor 101 reads and writes data to SRAM 11, cache 12 and DRAM 13. SRAM card 11 is connected to buses 16 and 17. Neither SRAM 11 nor cache 12 needs to be refreshed by processor 101 and buffers 14 and 15 operate as if SRAM 11 was not present.
FIG. 2 depicts how system memory is allocated when SRAM card 11 is installed and when it is absent. When SRAM card 11 is not installed, processor 101 sequentially writes all data to the cache memory 12 and to DRAM 13. As shown, the address space is all within memory 12 and 13, sequentially.
When the SRAM card is installed, the first 640 kilobytes of system memory is written to the SRAM card and thus the first 640 kilobytes of memory 11 and 12 are shifted upward in sequence as shown. The memory cells of memory 11 then contain DOS and DOS application software. The original memory is then allocated to the cache memory and to the DRAM. In either case, provision for 384 kilobytes of video Read Only Memory ("ROM") is permanently located at the addresses immediately following the first 640 kilobytes of system memory.
FIGS. 3 through 13 show one embodiment of a circuit that performs the functions discussed above. This embodiment relies on a number of programmable logic array circuits (PALS), programming for which is shown in Appendix A.
As shown on FIG. 3, the chip enables for the SRAMS on the SRAMCARD come out of DOSDEC pal. This pal decodes cycles to the SRAMCARD when it is installed, and asserts the proper chip enable. Only one chip enable is asserted at a time. The chip enables correspond to rows of SRAMS on the SRAMCARD. The exact definition is given below.
MEMDEC pal 302 decodes memory space for the motherboard. It also generates a signal which indicates to offsheet logic that the current cycle is to local memory.
Local memory is memory resources which are on the motherboard and have special interfaces. The SRAMCARD qualifies as local memory.
The following Signal Description section describes individual signals on the page.
FIG. 4 shows logic which completes the control for the SRAMCARD. DOSGATE pal 403 controls the gating of addresses and chip enables to the card. DOSWRITE pal 401 controls the write strobes to the card. DOSREAD 402 pal controls the output enables to the card. The write strobes and the output enables connect directly to the SRAMS on the SRAMCARD.
FIG. 5 shows the connector for the SRAMCARD on the mother board. All the control signals discussed above go to this connector. Address and data from the 386 and the DMA controller also go to this connector.
FIG. 6 shows the connector for the SRAMCARD on the card.
FIG. 7 shows the address latches for the SRAM on the SRAMCARD.
FIGS. 8-11 show the SRAM array and FIG. 12 shows the termination resistors for the control signals.
FIG. 13 shows bypass capacitors to reduce switching noise.
FIG. 14 contains logic which decodes the cycles to the 32-Bit memory card. The three pals 1401, 402 and 403 use switches from the 32-Bit card, and the signal BDOSRAMEN- to determine where to locate the DRAMS in memory. If the SRAMCARD is installed, BDOSRAMEN- is active and the pals begin decoding the 32-BIT DRAM card at 1 MBYTE. If the SRAMCARD is not installed, BDOSRAMEN- is inactive and the pals begin decoding the 32-BIT DRAM card at the bottom of memory space (address 0).
ROW3DEC pal decodes cycles to ROW 3 on the DRAMCARD. ROW12DEC pal decodes cycles to ROWs 1 and 2 on the DRAMCARD. ROWODEC pal decodes cycles to ROW 0 on the DRAMCARD.
Although this description describes the invention with reference to the above specified embodiments, the claims and not this description limit the scope of the invention. Various modifications of the disclosed embodiment, as well as alternative embodiments of the invention, will become apparent to persons skilled in the art upon reference to the above description. Therefore, the appended claims will cover such modifications that fall within the true scope of the invention. ##SPC1##
Claims
What is claimed is:
1. A device for optimizing memory performance comprising: a first memory storage unit operable for storing information at defined memory locations therein, said defined locations of said first memory storage unit being sequentially ordered; a second memory storage unit also operable for storing information at defined memory locations therein, said defined locations of said second memory storage unit being sequentially ordered, said second memory storage unit having a faster operating speed than said first memory storage unit to optimize a specific operating system, said second memory storage unit generating a signal indicating its presence and availability; and means responsive to said signal for sequentially ordering said second memory storage unit prior to the sequential ordering of said first memory storage unit such that an operating system and operating system applications are stored within said second memory storage unit, wherein said second memory storage unit is sequentially ordered beginning with address number one.
2. The system of claim 1 wherein said first memory storage unit includes dynamic random access memories; and wherein said second memory storage unit includes at least 640K of high speed static random access memories.
3. The device of claim 1, wherein the sequential ordering of said first memory storage unit begins at a one-millionth address in response to a presence of said second memory storage unit.
4. The device of claim 1, wherein said operating system is a disk operating system.
5. The device of claim 1, further comprising: a cache memory having sequentially ordered addresses after sequentially ordered addresses of said second memory storage unit.
6. The device of claim 1, further comprising: a video read only memory having addresses sequentially ordered immediately after a last address of said second memory storage unit.
7. A method of optimizing performance of a memory system comprising the steps of: ascertaining a presence of a fixed memory and a separate other memory, wherein said separate other memory has a faster operating speed than said fixed memory; addressing said fixed memory in sequential address order beginning with address number one when said separate other memory is not present; addressing said ascertained separate other memory sequentially beginning with address number one such that an operating system is placed in faster separate other memory instead of slower fixed memory for optimization of said operating system; and addressing said fixed memory sequentially beginning with an address greater than the last address of said ascertained separate other memory.
8. The method set forth in claim 7, wherein said fixed memory is dynamic random access memory corresponding to operating system extended memory; and wherein said ascertained separate other memory is high speed static random access memory corresponding to operating system real memory where an operating system and its applications reside.
9. The method of claim 8, wherein said operating system is a disk operating system.
10. The method of claim 8, further comprising the step of: addressing a video read only memory in sequential address order immediately following said separate other memory and before said fixed memory.
11. The method of claim 10, further comprising the step of: addressing a cache memory in sequential order after said video read only memory.
12. The method of claim 7, wherein said fixed memory is addressed beginning with the one-millionth address.
13. A high speed memory system for use with a digital computer comprising: sequentially addressable dynamic random access memory; 640K of high speed sequentially addressable static random access memory for storing an operating system and its applications; means for determining when said high speed memory is installed in said computer; means for addressing said high speed memory, when present, beginning with said computer's first memory address; means for addressing said dynamic random access memory, when said high speed memory is present, beginning with said computer's one-millionth address; and means for addressing said computer's dynamic random access memory, when said high speed memory is not present, beginning with said computer's first memory address.
14. The memory system of claim 13, wherein said operating system is a disk operating system.
15. The memory system of claim 13, wherein said determining means includes generating a signal in response to a presence of said high speed memory.
16. The memory system of claim 15, wherein said signal activates said addressing means such that address number one of the computer is within said high speed memory.
17. The memory system of claim 16, further comprising: a cache memory having addresses sequentially ordered after sequentially ordered addresses of said high speed memory.
18. The memory system of claim 17, further comprising: a video read only memory having addresses sequentially ordered immediately following a last address of said high speed memory.
19. The memory system of claim 18, wherein said video read only memory includes 384K of address space such that said dynamic random access memory is sequentially addressed beginning with said computer's one-millionth address.