Systems and methods for buffering memory transactions

Details Systems and methods for buffering memory transactions Systems and methods for buffering memory transactions

2001-11-20

2004-06-08

Sparks, Donald (Department: 2187)

Electrical computers and digital processing systems: memory

Storage accessing and control

Hierarchical memories

C711S140000, C711S169000, C711S146000

Reexamination Certificate

active

06748497

ABSTRACT:

BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates in general to data processing systems, and in particular, to data processing systems and methods for mitigation latencies in data processing systems including.
2. Description of the Related Art
Modern signal processing systems, such as those found, for example, in commercial and consumer audio and multimedia products, are moving, with improvements in Very Large Scale Integration (VLSI) fabrication processes, to “system on a chip” (SoC) implementations. Such implementations may include one or more processors which may perform signal processing and control functions, on-chip memory, and signal amplification whereby an amplified signal may be delivered directly to the user's listening device, a speaker or a headphone set, for example.
As the sources of digital audio, video and multimedia data have become more sophisticated, the tasks required of the play back systems have correspondingly become more complex. For example, the source stream may be delivered in a compressed format in accordance with one or more standardized compression formats, such as those promulgated by the Motion Picture Experts Group (MPEG). Additionally, the compressed digital audio data may be embedded in a multiplexed bitstream that includes additional data, for example, conditional access information which may be used to limit the access to the underlying content to users who have subscribed thereto. Consequently, the digital signal processing demands placed upon the SoC may be significant. Thus, such an SoC may incorporate a DSP engine to perform the computationally intensive signal processing required to extract and recover the uncompressed digital data. Instructions and data for the DSP engine may be stored in memory which may be on chip, off chip, or a combination of both. Typically, the speed of the DSP exceeds that of the memory devices, and in modem DSP systems the memory latency can be long enough to stall the DSP engine while the memory transaction (read/write) completes. Buffers inserted between the memory system and the DSP may be used to reduce latency penalties associated with memory reads by speculatively prefetching and storing instructions or data. However, systems using such buffer mechanisms have, heretofore remained vulnerable to memory latencies with respect to writes to memory.
Consequently, there is a need in the art for systems and methods to shield a DSP(or similar high-performance processor) from memory latencies. In particular, there is a need for such systems and methods adapted for both read and write transactions.
SUMMARY OF THE INVENTION
According to the principles of the present invention, a buffer apparatus is disclosed that includes a read buffer unit configured for storing at least one data value read from a memory device, and a write buffer unit configured for storing at least one data value for writing to the memory device. The read buffer unit is operable for updating with the at least one data value for writing to the memory device in response to a write to the write buffer unit.
The inventive concept addresses a problem modem signal processing systems, such as those found, for example, in commercial and consumer audio and multimedia products, particularly, with improvements in Very Large Scale Integration (VLSI) fabrication processes, “system on a chip” (SoC) implementations. As the sources of digital audio, video and multimedia data have become more sophisticated, the tasks required of the play back systems have correspondingly become more complex. Consequently, the digital signal processing demands placed upon the SoC may be significant, and such an SoC may incorporate a DSP engine to perform the computationally intensive signal processing required to extract and recover the uncompressed digital data. Instructions and data for the DSP engine may be stored in memory which may be on chip, off chip, or a combination of both. Typically, the speed of the DSP exceeds that of the memory devices, and in modern DSP systems the memory latency can be long enough to stall the DSP engine while the memory transaction (read/write) completes. The read and write buffers units of the present invention may mitigate against memory latencies while maintaining coherency between the data therein.


REFERENCES:
patent: 5588128 (1996-12-01), Hicok et al.
patent: 6187483 (2001-02-01), Capodieci et al.
patent: 6216200 (2001-04-01), Yeager
patent: 6260095 (2001-07-01), Goodrum
patent: 6282626 (2001-08-01), Platko et al.
patent: 6341335 (2002-01-01), Kanai et al.
patent: 6370073 (2002-04-01), Leung
patent: 6415353 (2002-07-01), Leung
patent: 6449685 (2002-09-01), Leung
Farkas, et al., “How Useful are Non-blocking Loads, Stream Buffers and Speculative Execution in Multiple Issue Processors?”, © 1995 IEEE, pp. 78-89.*
Lepak et al., “Silent Stores and Store Value Locality”, © IEEE 2001, pp. 1174-1190.*
Lepak et al., “On the Value Locality of Store Instructions”, © 2000 ACM, pp. 182-191.*
Jouppi, Norman P., Digital Equipment Corporation Western Research Lab, “Improving Direct-Mapped Cache Performance by the Addition of a Small Fully Associative Cache and Prefetch Buffers,” 1990 IEEE, pp. 364-373.
Palacharla, Subbrano, Computer Sciences Department, University of Wisconsin-Madison and R.E. Kessler, Cray Research, Inc., “Evaluating Stream Buffers as a Secondary Cache Replacement,” 1994 IEEE, pp. 24-33.

Affiliated with

Also associated with

Rating

Systems and methods for buffering memory transactions does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Systems and methods for buffering memory transactions, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Systems and methods for buffering memory transactions will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3307197