Electrical computers and digital processing systems: processing – Processing architecture – Microprocessor or multichip or multimodule processor having...
Reexamination Certificate
1998-10-23
2003-12-02
Ellis, Richard L. (Department: 2183)
Electrical computers and digital processing systems: processing
Processing architecture
Microprocessor or multichip or multimodule processor having...
C711S165000
Reexamination Certificate
active
06658552
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The invention relates to digital data processing circuits. In particular, the invention relates to the performance of data manipulation functions performed on strings of data elements.
2. Description of the Related Art
Conventional microprocessing circuits include several common building blocks. Essentially all such systems include a main memory storage area for storing data and instructions, and an execution unit for operating on the data in accordance with the instructions. After the function specified by a given instruction is performed, processed data is returned to the main memory storage area.
Increases in processor performance have been obtained by enhancements to this fundamental scheme. The processor may include two or more separate execution units which can process multiple instructions in parallel. The Intel Pentium and Pentium Pro are two examples of this type of processor. In some cases, different execution units are dedicated to different functions. The Intel Pentium Pro, for example, includes separate execution units for floating point and fixed point arithmetic operations. Another performance enhancement in almost universal use is the provision of data and instruction caches which provide local storage of recently used data and instructions. This speeds the fetching and storing of data and instructions by reducing the number of memory accesses required from a typically much slower main memory storage area.
Still, some types of operations are performed inefficiently by these processor architectures. One inefficiently performed class of instructions is data string manipulation instructions. In these instructions, operations involving a sequence of data elements are performed. For instance, a block of data may be moved from one series of memory addresses to another series of memory addresses. Alternatively, the elements of a block of data may be compared to a test data element or a string of test data elements. In the Intel Pentium Pro, assembly language instructions are provided to perform these functions on a specified string of data. Although the total length of the processed string can be very large, data is moved and/or analyzed as short string portions of at most 32 bits long due to the bus width and 32 bit execution unit. Performing a string move on the Pentium Pro thus involves a sequential process of reading and writing pieces of the data string to and from main memory (or to and from the cache for those portions of the string which are present there). String scans for matching data are similarly performed. Short pieces of the string are read from memory and compared to the desired test string. Thus, the Pentium Pro architecture includes useful string instructions as part of its instruction set, but is incapable of performing string operations on large strings as quickly as memory technology might allow.
Another commercially available device which includes string manipulation features is the TMS320C80 digital signal processor from Texas Instruments. Because this device is adapted for use in video and multimedia applications, features for improving the speed of movement of large blocks of data such as a set of image pixels are included. In the TMS320C80, the programmer may write string movement parameters to a memory location. These string movement parameters can then be transferred to the memory controller portion of the device, and the string movement is performed by the memory controller without further involvement of the execution unit. This feature helps speed up the movement of data blocks, but setting up the transfer parameters requires preliminary write operations, which is inconvenient for the programmer, and results in the need to use several instructions to initiate a block move. Furthermore, although the TMS320C80 includes a data cache, these memory move operations are not performed utilizing cached data, and no mechanism to ensure cache coherency with the main memory where the data move occurs is provided.
Other implementations of memory systems which can perform data manipulation have been described. In U.S. Pat. No. 5,590,370, a system is disclosed which includes “active memory elements” that incorporate processing logic for performing searches and other data manipulations outside of the host processing circuit. U.S. Pat. No. 4,731,737 also discusses memory elements which can receive data manipulation commands from an external host processor. However, neither of these systems provide for cache coherency, and they further do not describe the use of assembly language instruction sets which provide simple and efficient programming of data string manipulations. Thus, there is a continuing need for improvements in processor architectures to allow the processing of data strings quickly and efficiently.
SUMMARY OF THE INVENTION
A digital processing system optimized for string manipulations comprises an instruction fetch unit coupled to an external memory, a first execution unit coupled to receive, decode, and perform assembly language arithmetic and logic instructions received from external memory via the instruction fetch unit, and a second execution unit coupled to receive, decode, and perform assembly language string manipulation instructions received from external memory via the instruction fetch unit. Instructions may be analyzed to detect data string operations for routing to the appropriate execution unit.
In systems with data caching, data may be reassigned from a first memory location to a second memory location by writing a value to an entry in a cache tag memory without changing the content of an entry in a cache data memory associated with the entry in the cache tag memory. In some embodiments, data move operations include reading a cache line containing at least a portion of data from a data cache; shifting the cache line a selected amount, and storing the cache line in the data cache.
Compare operations in systems with data caching are also optimized. Accordingly, a cache memory system may comprise a data memory configured to hold cache lines comprising a plurality of bytes of data and a plurality of comparators, wherein each comparator has a first input coupled to the data memory such that each comparator receives one of the plurality of bytes of data via its associated first input, and wherein each comparator has a second input coupled to a second data source, whereby the cache line may be compared to data received from the second data source. The second data source may comprise external string manipulation circuitry.
REFERENCES:
patent: 3593304 (1971-07-01), Gardner et al.
patent: 3787817 (1974-01-01), Goldberg
patent: 3820079 (1974-06-01), Bergh et al.
patent: 4068305 (1978-01-01), Cutler
patent: 4096566 (1978-06-01), Borie et al.
patent: 4631660 (1986-12-01), Woffinden et al.
patent: 4731737 (1988-03-01), Witt et al.
patent: 4760523 (1988-07-01), Yu et al.
patent: 4876643 (1989-10-01), McNeill et al.
patent: 4896133 (1990-01-01), Methvin et al.
patent: 4907189 (1990-03-01), Branson et al.
patent: 5131085 (1992-07-01), Eikill et al.
patent: 5134711 (1992-07-01), Asthana et al.
patent: 5163142 (1992-11-01), Mageau
patent: 5197135 (1993-03-01), Eickemeyer et al.
patent: 5226165 (1993-07-01), Martin
patent: 5353424 (1994-10-01), Partovi et al.
patent: 5355460 (1994-10-01), Eickemeyer et al.
patent: 5418529 (1995-05-01), De Luigi et al.
patent: 5442802 (1995-08-01), Brent et al.
patent: 5519873 (1996-05-01), Butter et al.
patent: 5590370 (1996-12-01), Asthana et al.
patent: 5675765 (1997-10-01), Malamy et al.
patent: 5752071 (1998-05-01), Tubbs et al.
patent: 5754878 (1998-05-01), Asghar et al.
patent: 5887137 (1999-03-01), Kadoya
TMS320C80 (MVP) Transfer Controller User's Guide, Chapters 1-4, pp. TC:1-1 through 4-54, Copyright© 1995, Texas Instruments Incorporated.
Ellis Richard L.
Knobbe Martens Olson & Bear LLP
Micro)n Technology, Inc.
LandOfFree
Processing system with separate general purpose execution... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Processing system with separate general purpose execution..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Processing system with separate general purpose execution... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3151401