Electrical computers and digital data processing systems: input/ – Intrasystem connection – Bus expansion or extension
Reexamination Certificate
2000-04-27
2003-07-15
Dharia, Rupal (Department: 2189)
Electrical computers and digital data processing systems: input/
Intrasystem connection
Bus expansion or extension
C710S305000
Reexamination Certificate
active
06594719
ABSTRACT:
FIELD OF THE INVENTION
The present invention is generally related to data processing systems, and more particularly to computer systems having at least one host processor and connectable to a plurality of peripherals devices including notebook computers, storage devices, displays, keyboards, mouse's and so forth.
BACKGROUND OF THE INVENTION
Computer systems today are powerful, but are rendered limited in adapting to changing computing environments. The PCI bus is pervasive in the industry, but as a parallel data bus is not easily bridged to other PCI based devices. Full bridges are known, such as used in traditional laptop computer/docking stations. However, separating the laptop computer from the docking station a significant distance has not been possible. Moreover, the processing power of computer systems has been resident within the traditional computer used by the user because the microprocessor had to be directly connected to and resident on the PCI motherboard. Thus, upgrading processing power usually meant significant costs and/or replacing the computer system.
PCI
The PCI bus is primarily a wide multiplexed address and data bus that provides support for everything from a single data word for every address to very long bursts of data words for a single address, with the implication being that burst data is intended for sequential addresses. Clearly the highest performance of the PCI bus comes from the bursts of data, however most PCI devices require reasonable performance for even the smallest single data word operations. Many PCI devices utilize only the single data mode for their transfers. In addition, starting with the implementation of the PCI 2.1 version of the specification, there has been at least pseudo isochronous behavior demanded from the bus placing limits on an individual device's utilization of the bus, thus virtually guaranteeing every device gets a dedicated segment of time on a very regular interval and within a relatively short time period. The fundamental reason behind such operation of the PCI bus is to enable such things as real time audio and video data streams to be mixed with other operations on the bus without introducing major conflicts or interruption of data output. Imagine spoken words being broken into small unconnected pieces and you get the picture. Prior to PCI 2.1 these artifacts could and did occur because devices could get on the bus and hold it for indefinite periods of time. Before modification of the spec for version 2.1, there really was no way to guarantee performance of devices on the bus, or to guarantee time slot intervals when devices would get on the bus. Purists may argue that PCI is still theoretically not an isochronous bus, but as in most things in PC engineering, it is close enough.
Traditional High Speed Serial
Typical high speed serial bus operation on the other hand allows the possibility of all sizes of data transfers across the bus like PCI, but it certainly favors the very long bursts of data unlike PCI. The typical operation of a serial bus includes an extensive header of information for every data transaction on the bus much like Ethernet, which requires on the order of 68 bytes of header of information for every data transaction regardless of length. In other words, every data transaction on Ethernet would have to include 68 bytes of data along with the header information just to approach 50% utilization of the bus. As it turns out Ethernet also requires some guaranteed dead time between operations to “mostly” prevent collisions from other Ethernet devices on the widely disperse bus, and that dead time further reduces the average performance.
The typical protocol for a serial bus is much the same as Ethernet with often much longer header information. Virtually all existing serial bus protocol implementations are very general and every block of data comes with everything needed to completely identify it. FiberChannel (FC) has such a robust protocol that virtually all other serial protocols can be transmitted across FC completely embedded within the FC protocol, sort of like including the complete family history along with object size, physical location within the room, room measurements, room number, street address, city, zip code, country, planet, galaxy, universe, . . . etc. and of course all the same information about the destination location as well, even if all you want to do is move the object to the other side of the same room. Small transfers across all of these protocols, while possible, are extremely expensive from a bandwidth point of view. Of course the possibility of isochronous operation on the more general serial bus is not very reasonable.
Recreating High Speed Serial for PCI
In creating the proprietary Split-Bridge™ technology, Mobility electronics of Phoenix, Ariz., the present applicant, actually had to go back to the drawing board and design a far simpler serial protocol to allow a marriage to the PCI bus, because none of the existing implementations could coexist without substantial loss of performance. For a detailed discussion of Applicant's proprietary Split-Bridge™ technology, cross reference is made to Applicant's co-pending commonly assigned patent applications identified as Ser. No. 09/130,057 and 09/130,058 both filed Jun. 6, 1998, the teachings of each incorporated herein by reference. The Split-Bridge™ technology approach is essentially custom fit for PCI and very extensible to all the other peripheral bus protocols under discussion like PCIx, and LDT™ of AMD corporation. Split-Bridge™ technology fundamentals are a natural for extending anything that exists within a computer. It basically uses a single-byte of overhead for 32 bits of data and address—actually less when you consider that byte enables, which are not really “overhead”, are included as well.
Armed with the far simpler protocol, all of the attributes of the PCI bus are preserved and made transparent across a high speed serial link at much higher effective bandwidth than any existing serial protocol. The net result is the liberation of a widely used general purpose bus, and the new found ability to separate what were previously considered fundamental inseparable parts of a computer into separate locations. When the most technical reviewers grasp the magnitude of the invention, then the wheels start to turn and the discussions that follow open up a new wealth of opportunities. It now becomes reasonable to explore some of the old fundamentals, like peer-to-peer communication between computers that has been part of the basic PCI specification from the beginning, but never really feasible because of the physical limits of the bus prior to Split-Bridge™ technology. The simplified single-byte overhead also enables very efficient high speed communication between two computers and could easily be extended beyond PCI.
The proprietary Split-Bridge™ technology is clearly not “just another high speed link” and distinguishing features that make it different represent novel approaches to solving some long troublesome system architecture issues.
First of all is the splitting of a PCI bridge into two separate and distinct pieces. Conceptually, a PCI bridge was never intended to be resident in two separate modules or chips and no mechanism existed to allow the sharing of setup information across two separate and distinct devices. A PCI bridge requires a number of programmable registers that supply information to both ports of a typical device. For the purpose of the following discussion, the two ports are defined into a north and south segment of the complete bridge.
The north segment is typically the configuration port of choice and the south side merely takes the information from the registers on the north side and operates accordingly. The problem exists when the north and south portions are physically and spatially separated and none of the register information is available to the south side because all the registers are in the north chip. A typical system solution conceived by the applicant prior to the invent
Ahern Frank W.
Doss Jeff
Mollo Charles
Dharia Rupal
Jackson & Walker, LLP
Klinger Robert C.
Mobility Electronics Inc.
LandOfFree
Extended cardbus/pc card controller with split-bridge... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Extended cardbus/pc card controller with split-bridge..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Extended cardbus/pc card controller with split-bridge... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3041552