US 5,790,831 AGrant
VL-bus/PCI-bus Bridge
Issue Date:1998-08-04
•33 Claims
•15 Drawing Sheets
Abstract
A PCI-bus is added to a VESA local bus (VL-bus) computer system using a VL-bus/PCI-bus bridge. The VL-bus/PCI-bus bridge claims a VL-bus cycle by asserting LDEV# to the VL-bus/system-bus bridge. If no other VL-bus device claims the cycle as well, then the VL-bus/PCI-bus bridge translates the cycle onto the PCI-bus and awaits a response from a PCI device. If no PCI device claims a cycle by the PCI-bus device claiming deadline, then the VL-bus/PCI-bus bridge asserts BOFF# to the host and suppresses its assertion of LDEV# when the host repeats the cycle on the VL-bus. The VL-bus/system-bus bridge therefore can translate the repetition of the cycle onto the system bus. When asserting BOFF# to the host, the VL-bus/PCI-bus bridge also asserts the VL-bus device ready signal LRDY# after assertion of BOFF# and releases LRDY# before releasing BOFF#. The VL-bus controller does not receive BOFF# necessarily, but responds to LRDY# by asserting RDYRTN# onto the VL-bus, thereby signifying to all other VL-bus devices that the VL-bus cycle has ended and permitting them to restart their state machines in anticipation of a new VL-bus cycle. The host ignores RDYRTN# while, and only while, BOFF# is asserted.
Metadata
Assignee
- Opti Inc.
Inventors
- Fong Lu (David) Lin
- Cherng-Yeuan (Henry) Tsay
- David H. Doan
Application Information
Application Number:US 7159013
Filing Date:1996-09-18
Priority Date:1994-11-01
Art Unit:234
Classifications
IPC:
G06F 1314G06F 1342
Field of Search:
395500;309;311
Patent Drawings (15 sheets)
Description
BACKGROUND
1. Field of the Invention
The invention relates to IBM PC AT-compatible computer architectures, and more particularly, to enhancements thereof for communicating with I/O peripheral devices.
2. Description of Related Art
The IBM PC/AT.RTM. computer architecture has become an industry standard architecture for personal computers and is typically built around a host CPU such as an 80386, 80486 or Pentium.RTM. microprocessor manufactured by Intel Corporation, or similar microprocessors manufactured by others. The host CPU is coupled to a host bus, capable of performing memory accesses and data transfers at high rates of speed (i.e., on the order of 25-100 MHz with today's technology). The host bus includes 32 or, in the case of computers built around the Pentium, 64 data lines, a plurality of address lines, and various control lines. The typical IBM PC AT-compatible platform also includes DRAM main memory, and in many cases a timer, a real-time clock, and a cache memory.
The typical IBM PC AT-compatible computer also includes an I/O bus, also know as a system bus or AT-bus, which is separate and distinct from the host bus. The system bus usually conforms to one of two industry-established standards known as ISA (Industry Standard Architecture) and EISA (Extended ISA). The system bus is coupled to the host bus via a host-bus/system-bus bridge, and includes, depending on the system bus standard adhered to, 16 or 32 data lines, a plurality of address lines, as well as control lines. The I/O address space is logically distinct from the memory address space and if the CPU desires to access an I/O address, it does so by executing a special I/O instruction. Such an I/O instruction generates memory access signals on the host bus, but also activates an M/IO# signal on the host bus to indicate that this is an access to the I/O address space. The host-bus/system-bus bridge recognizes the I/O signals thereby generated by the CPU, performs the desired operation over the system bus, and if appropriate, returns results to the CPU over the host bus.
In practice, some I/O addresses may reside physically on the host bus and some memory addresses may reside physically on the system bus. More specifically, the devices which respond to accesses to certain I/O space addresses may be connected to the control lines (and usually the address and data lines as well) of the host bus, while the devices which respond to accesses to certain memory space addresses may be connected to the control lines (and usually the address and data lines as well) of the system bus. The host-bus/system-bus bridge is responsible for recognizing that a memory or I/O address access must be emulated by an access to the other bus, and is responsible for doing such emulation. For example, a ROM (or EPROM) BIOS may reside physically on the system bus, but actually form part of the local memory address space. During system boot, when the CPU sends out a non-I/O address which is physically within the ROM BIOS, the host-bus/system-bus bridge recognizes such, enables a buffer (considered herein as being part of the host-bus/system-bus bridge) which couples the address onto the system bus, and activates the chip select for the ROM. The bridge then assembles a data word of the size expected by the host CPU, from the data returned by the ROM, and couples the word onto the host bus for receipt by the CPU. In many systems, at some point during the ROM-based boot-up procedure, the ROM BIOS is copied into equivalent locations in the DRAM main memory, which does reside on the host bus, and thereafter accessed directly.
In the standard architecture, the logical main memory address space is divided into a low memory range (0h-9FFFFh), a reserved memory range (A0000h-FFFFFh) and an extended memory range (100000h to the top of memory). In a typical system the system ROM BIOS is located logically at memory space addresses F0000h-FFFFFh, and resides physically on the system bus. Addresses C0000h-EFFFFh contain ROM BIOS portions for specific add-on cards and reside physically on their respective cards on the system bus. Addresses A0000h-BFFFFh contain the video buffer, which is a part of a video controller residing on the system bus. Duplicate memory space is typically provided in DRAM on the host bus for addresses C0000h-FFFFFh, and the user of the system can select during a setup procedure, which portions of the ROM BIOS are to be "shadowed" by being copied into the duplicate DRAM space during boot-up.
In addition to the above elements, a keyboard controller typically also resides on the system bus, as does the timer and real-time clock. A typical IBM PC AT-compatible system may also include a DMA controller which permits peripheral devices on the system bus to read or write directly to or from main memory, as well as an interrupt controller for transmitting interrupts from various add-on cards to the CPU. The add-on cards are cards which may be plugged into slot connectors coupled to the system bus to increase the capabilities of the system. Add-on cards are also sometimes referred to as expansion cards or accessory cards.
General information on the various forms of IBM PC AT-compatible computers can be found in IBM, "Technical Reference, Personal Computer AT" (1985), in Sanchez, "IBM Microcomputers: A Programmer's Handbook" (McGraw-Hill: 1990), in MicroDesign Resources, "PC Chip Sets" (1992), and in Solari, "AT Bus Design" (San Diego: Annabooks, 1990). See also the various data books and data sheets published by Intel Corporation concerning the structure and use of the iAPX-86 family of microprocessors, including Intel Corp., "Pentiumn.TM. Processor", Preliminary Data Sheet (1993); Intel Corp., "Pentiumn.TM. Processor User's Manual" (1994); "i486 Microprocessor Hardware Reference Manual", published by Intel Corporation, copyright date 1990, "386 SX Microprocessor", data sheet, published by Intel Corporation (1990), and "386 DX Microprocessor", data sheet, published by Intel Corporation (1990). All the above references are incorporated herein by reference.
The various signals on the host bus include the input/output signals of whichever microprocessor the system is built around. Such signals are therefore well known in the industry and can be determined by reference to the above-incorporated publications. The various signals on the system bus also are well known in the industry. The Solari book incorporated above describes the lines in detail. For present purposes, only the following signals are important:
Note that some of the signals described in this specification are asserted high, whereas others are asserted low. As used herein, signals which are asserted low are given a `#` or `B` suffix in their names, whereas those asserted high (or for which an assertion polarity has no meaning) lack a `#` or `B` suffix. Also, two signal names mentioned herein that are identical except that one includes the `#` or `B` suffix while the other omits it, are intended to represent logical compliments of the same signal. It will be understood that one can be generated by inverting the other, or both can be generated by separate logic in response to common predecessor signals.
Recently, efforts have been made to reduce the size and improve the manufacturability of PC AT-compatible computers. Specifically, efforts have been made to minimize the number of integrated circuit chips required to build such a computer. Several manufacturers have developed "PC AT chipsets" (also known as "core logic chipsets" or "I/O bus interface circuitry"), which integrate a large amount of the host-bus/system-bus bridge circuitry and other circuitry onto only a few chips. An example of such a chipset is the 386WB PC/AT chipset manufactured by OPTi, Inc., Santa Clara, Calif. These chipsets implement the host-bus/system bus bridge, the timer, real-time clock (RTC), DMA controller, as well as some additional functionality.
In the original IBM PC AT computers manufactured by IBM Corp., the system bus operated with a data rate of 8 MHz (BCLK=8 MHz). This was an appropriate data rate at that time since it was approximately equivalent to the highest data rates which the CPUs of that era could operate with on the host bus. Numerous third party vendors have since developed peripheral devices and controller cards which are intended to be plugged into an AT (ISA) slot on the system bus, and which rely upon the 8 MHz maximum data rate. The AT standard also requires a wait state (i.e. 125 nS) for 16-bit data transfers, and four wait states (500 nS) for 8-bit data transfers. A zero wait state data transfer is also available, but only if the peripheral device signals, by activating the SRDY# control line on the system bus, that it can handle such fast data transfers.
In the years since the IBM PC AT was originally introduced, technology has improved dramatically to the point where host buses on high-end PC AT-compatible computers can operate on the order of 100 MHz. Despite these advances, however, such computers are still manufactured with a system bus operating at around 8 MHz because of the need to maintain compatibility with previously designed peripheral devices. These devices were designed in reliance upon the 8 MHz data rate and AT wait state protocol, and many such devices are not capable of operating faster. Even modern designs for AT bus peripherals often rely on the 8 MHz maximum data rate, even though very little additional effort or cost would be involved to design them to operate faster.
In addition to the large disparity between data transfer rates on the system bus as compared to the host bus in modern PC AT-compatible computers, the host-bus/system-bus bridge circuitry needs to delay its handling of requests and responses from one bus to the other merely because the clocks are not synchronized. The circuitry therefore must hold a request or response until the appropriate clock edge on the destination bus appears. This can add on the order of 30-200 nS to each system bus cycle. Accordingly, it can be seen that any access to a peripheral device on the system bus imposes a substantial penalty on the performance of PC AT-compatible computers. This penalty will only become worse as the disparity between the host bus and system bus data rates continues to increase.
The penalty applies for most types of peripheral devices, but in the past it has been most noticeable for video display controllers. Video display controllers have a command port which responds to accesses in the I/O address space, as well as a video memory port which responds to accesses in the memory address space. Manufacturers have traditionally placed both ports on the system bus, however, thereby imposing the speed limitations of the system bus on the video memory port as well as the command port. U.S. patent application Ser. No. 07/851,444, filed Mar. 16, 1992, now U.S. Pat. No. 5,309,568, (Attorney Docket No. OPTI3030WSW), owned by the assignee of the present application and incorporated herein by reference in its entirety, describes certain attempts to permit accesses to the video memory port to take place over the host bus instead of the system bus. In addition, some graphics chip vendors have tried incorporating features into their chips for connection directly to a host bus. For example, see S3, Inc., "86C911 GUI Accelerator", Databook (April 1992), incorporated herein by reference.
However, these solutions all suffer from the problem that they are non-standard. That is, if a vendor of I/O interface chipsets provides for a host bus capability, there is no assurance that it will interface directly with products made by more than one peripheral device controller vendor. A layer of buffers and glue logic therefore may be required to enable such peripheral device controllers to take advantage of the host bus feature, and the glue logic may be different for each different peripheral controller. On the other hand, if a maker of peripheral device controllers, such as a maker of a VGA (Video Graphics Adapter) controller, provides for a host bus capability in the peripheral controller, there is no guarantee that it will interface correctly with the host-bus/system-bus bridge chipsets made by more than one chipset manufacturer. Again, different buffers and glue logic may be required for each vendor of chipsets.
In two different efforts for ameliorating the above problem, instead of creating a private standard, two different organizations have defined different bus protocols and attempted to promulgate them as standards for the entire personal computer industry. One such standard, referred to herein as the VESA (Video Electronics Standards Association) or VL-Bus standard, is defined in VESA, "VESA VL-Bus Local Bus Standard", Revision 1.0 (1992), incorporated herein by reference. Significant aspects of the VL-Bus specifications are described in Appendix A hereto. Further revisions of the VESA standard are in preparation, one recent version being VESA, "VESA VL-Bus Proposal, Version 2.0 p, Revision 0.8 p (May 17, 1993), also incorporated herein by reference. The other such standard, referred to herein as the PCI standard, is defined in Intel Corp., "Peripheral Component Interconnect (PCI), revision 1.0 Specification" (Jun. 22, 1992) and in PCI Special Interest Group, "PCI Local Bus Specification", Revision 2.0 (Apr. 30, 1993), both incorporated herein by reference. Significant aspects of the PCI 2.0 Bus specifications are described in Appendix B hereto. Each standard has advantages and disadvantages over the other, and depending on the application, one standard or the other may be more beneficial to have in a system.
For example, one advantage of the VL-bus is that it is relatively simple to include in a personal computer system, especially those built around an Intel 486 microprocessor. This is because the VL-bus signal lines are similar to signal lines of the 486 CPU, except for a few additional signal lines included on the VL-bus. Thus the only additional expense required to add a VL-bus to a pre-existing 486-based computer design, is a very small VL-bus controller to handle the additional signal lines. Such a controller has already been included in chipsets. An example of such a chipset includes the OPTi 82C802G and either the 82C601 or 82C602, all incorporated herein by reference. The 82C802G is described in OPTi, Inc., "OPTi PC/AT Single Chip 82C802G Data Book", Version 1.2a (Dec. 1, 1993), and significant aspects are also set forth in Appendix C hereto. The 82C601 and 82C602 are described in OPTi, Inc., "PC/AT Data Buffer Chips, Preliminary, 82C601/82C602 Data Book", Version 1.0e (Oct. 13, 1993), and significant aspects are also set forth in Appendix D hereto. Both data books are incorporated herein by reference in their entirety.
While a minimum VL-bus system requires no additional circuitry, the insertion of a simple host-bus/VL-bus bridge provides buffering for additional VL-bus devices. For Pentium.RTM.--based systems, the host bus of which has a 64-bit wide data path, the bridge could also include circuitry to break up a 64-bit host-bus originated access, into two 32-bit VL-bus accesses. Such circuitry is still relatively simple. (Extension to the 32-bit VL-bus standard have been proposed in order to accommodate 64-bit access in a single VL-bus cycle, but in general, as the data path of host CPUs continues to expand, it can be expected that at least some future system designs will continue to employ a bridge which breaks up a wider-data-path host bus access into two, four, or some other number of narrower-data-path cycles on the VL-bus.)
A primary advantage of the PCI-bus, on the other hand, is its processor independence. The PCI-bus was intended to provide very high-speed accesses using a standard bus protocol, and to interface those accesses with any host processor bus using an appropriate host-bus/PCI-bus bridge. The host-bus/PCI-bus bridge is significantly more expensive than the circuitry required to implement a VL-bus, but the independence it provides from ever-faster and ever-more varied host processor buses provides a stable target for designers of peripheral products. A peripheral device designed for the PCI-bus would not need to be redesigned for each new version of an Intel microprocessor, or indeed, for each new microprocessor that might in the future power a personal computer system.
To date, neither the VL-bus standard nor the PCI-bus standard has achieved dominance in the marketplace for personal computer systems or in the marketplace for peripheral devices. Thus peripheral device manufacturers designing cards intended to bypass the slow system bus, still usually must design one version of the card for the PCI-bus and one version for the VL-bus. Similarly, computer system integrators and chipset manufacturers often find themselves having to double their product offerings since each market segment for VL-bus systems can have an equivalent but separate market segment for PCI-bus systems.
It is possible to overcome these problems by designing a computer system which incorporates both VL-bus expansion slots and PCI-bus expansion slots, in addition to the standard ISA-- or EISA-bus expansion slots. The motherboard circuitry to implement this would be expected to include programmable registers which would indicate whether a particular valid cycle definition on the host bus is to be handled by a device on the host bus (such as main memory), a device on the VL-bus (which may be the same as the host bus in 486 systems), by a device on the PCI-bus, or by a device on the system bus. Such motherboard circuitry would be expensive, however, and may require an entirely new chipset design.
In Acer Laboratories, Inc., "M1435 PCI-VL.sub.-- Bus Bridge, Preliminary Datasheet" (Sep. 20, 1993), incorporated by reference herein, there is described a VL-bus/PCI-bus bridge chip which, together with the Acer M1429kG/M1429 VESA chip, permits both a VL-bus and a PCI-bus to be included in a single system. According to the datasheet, when the M1435 chip detects a valid VL-bus cycle, it first determines whether the cycle is intended for system memory or for another VL-bus device. The chip is believed to perform a positive decode of the address to determine whether the cycle is intended for system memory, and it observes the LDEV signal to determine whether the cycle has been claimed by another VL-bus device. If neither is the case, then the M1435 translates the cycle to the PCI-bus. If no PCI agent claims the translated PCI cycle, then the M1435 asserts an ISA REQJ signal to the M1429 chip, thereby informing the M1429 to start an ISA cycle. See also Acer Laboratories, Inc., "M1429G/M1431/M1435 Data Sheet" (October 1993), incorporated herein by reference.
The Acer technique for accommodating both the VL-bus and PCI-bus in a single system is limited in that it operates only with a host-bus/system-bus interface chipset which observes and understands the ISA REQJ signal asserted by the M1435 bridge. Other inter-chip signals may also be required between the M1435 and M1429. Since most interface chipsets do not understand these signals, such chipsets would have to be modified by the chipset manufacturer before they could be used with the M1435 bridge. It would be desirable, therefore, to provide a VL-bus/PCI-bus bridge which does not require modification of any existing VL-bus/system-bus chipset. Such a bridge could be used in conjunction with the chipset of any manufacturer which supports the VL-bus.
SUMMARY OF THE INVENTION
The invention takes advantage of a feature of Intel 486 and Pentium microprocessors, known as the back-off feature. These microprocessors sense a BOFF# signal, which is not considered part of the VL-bus. When the microprocessor detects assertion of BOFF#, while a host bus cycle is in progress, it aborts the cycle and restarts it after it detects BOFF# negated. According to the invention, roughly described, a VL-bus/PCI-bus bridge claims a VL-bus cycle by asserting LDEV# to the VL-bus/system-bus bridge. If no other VL-bus device claims the cycle as well, and the cycle is not to system memory, then the VL-bus/PCI-bus bridge translates the cycle onto the PCI-bus and awaits a response from a PCI device. If no PCI device claims a cycle by the PCI-bus device claiming deadline, then the bridge asserts BOFF# to the host. The bridge then awaits the host's repetition of the cycle which was subject to the BOFF#, and suppresses its assertion of LDEV# so that the VL-bus/system-bus bridge can translate the cycle onto the system bus.
In another aspect of the invention, a VL-bus device can take advantage of the back-off feature of the 486 and Pentium CPU's by asserting the VL-bus device ready signal LRDY# after assertion of BOFF#, and negating LRDY# before or at the same time as negating BOFF#. The VL-bus controller does not receive BOFF# necessarily, but responds LRDY# by asserting RDYRTN# onto the VL-bus, thereby signifying to all other VL-bus devices that the VL-bus cycle has ended and permitting them to restart their state machines in anticipation of a new VL-bus cycle.
BRIEF DESCRIPTION OF THE DRAWINGS
The invention will be described with respect to particular embodiments thereof, and reference will be made to the drawings, in which:
FIGS. 1 and 2 are block diagrams of a computer system incorporating the invention;
FIGS. 3, 4 and 5 are timing diagrams of signals in the computer system of FIG. 2;
FIGS. 6-10 are schematic diagrams of circuitry in the VL-bus/PCI-bus bridge of FIG. 2;
FIGS. B1-B5 are timing diagrams for certain signals on the PCI-bus of FIG. 2.
DETAILED DESCRIPTION
FIG. 1 is a block diagram illustrating pertinent features of a computer system incorporating the invention. The system includes a VL-bus host 102, which may, for example, be an Intel 80386 or 80486 microprocessor, or it may include an Intel Pentium microprocessor in combination with a conventional host-bus/VL-bus bridge. The VL-bus host 102 is connected to a VL-bus 104, comprising VL-bus control lines 106, VL-bus data lines 108, and VL-bus address lines 110. The system can also include a math coprocessor 112 connected to the VL-bus 104; if the VL-bus host includes a bridge, then the math coprocessor 112 can instead be coupled to the host bus (not shown). The system of FIG. 1 also includes a core logic chipset comprising an 82C802G chip 118 and an 82C602 chip 120, both manufactured by OPTi, Inc., Santa Clara, Calif., as well as two sets of buffers 114 and 116. The chipset (referred to collectively herein as 122) is coupled to the VL-bus 104 and provides control and address signals to main memory DRAM 124, cache SRAM 126, as well as other components in the system. The cache SRAM 126 also receives address signals from the VL-bus address lines 110 via buffer 128, and both the cache SRAM 126 and the main memory DRAM 124 are coupled bi-directionally to the VL-bus data lines 108. The system also includes an ISA bus 130, which includes address lines 132, a 16-bit SD data bus 134, and an 8-bit XD data bus 136. The XD data bus 136, together with accompanying control and address signals (collectively the X-bus), are considered herein to form part of the ISA bus 130.
The core logic chipset 122 is responsible for many functions in the system, including recognizing VL-bus cycles intended for devices which reside on the ISA bus, and translating such cycles onto the ISA bus 130. Address signals are transferred between the VL-bus address leads 110 and the ISA-bus address leads 132 via buffers 116 and the 82C602 data buffer 120, and data is transmitted between the VL-bus data lines 108 and the ISA-bus data lines 134 and 136 via the buffers 114 and the 82C602 data buffer 120. A BIOS ROM and flash EPROM 138 reside on the ISA bus 130, as does a keyboard controller 140 and other ISA-bus peripherals 142. Residing on the VL-bus 104 is a VL-bus device 144, as well as a VL-bus/PCI-bus bridge 146 (also referred to herein as a PCI bridge 146) described in more detail below. The PCI-bus bridge is also described in OPTi, Inc., "82C822 PCIB VESA Local Bus to PCI Bridge Interface," Data Sheet (April, 1994), incorporated herein in its entirety, and is described in Appendix E hereto. The VL-bus/PCI-bus bridge 146 is also connected to a PCI-bus 148, comprising control lines 150 and a multiplexed AD address/data bus 152. A PCI-bus device 154 resides on the PCI-bus 148. Although VL-bus 144 is shown residing exclusively on the VL-bus 104, it will be understood that in accordance with the VL-bus specification incorporated above, such device may also be connected to the ISA bus 130. The signals on the VL-bus 104 and the PCI-bus 148 conform to those set forth above and in the appendices with respect to the respective standards.
FIG. 2 is a block diagram showing some of the components of the system of FIG. 1, and providing additional detail regarding the control signals. Only signals pertinent to an understanding of the invention are illustrated, and it will be understood that other signals exist as well. Referring to FIG. 2, it can be seen that the VL-bus host 102 has its CPU address (CA) lines connected to the VL-bus VA lines 202 and its CD data lines connected to the VL-bus data lines 204. It also has its ADS#, R/W#, M/IO# and D/C# outputs connected to the corresponding VL-bus signal lines 206, 207, 208 and 210. The CLK output of the VL-bus host 102 drives the VL-bus LCLK lines 212, and the VL-bus host 102 BLAST# and BRDY# inputs are connected to receive the signals on the corresponding lines 214 and 216 of the VL-bus. As in conventional VL-bus system implementations, the VL-bus host 102 is not connected to the VL-bus LRDY# signal line 218, but rather has its RDY# input connected the VL-bus RDYRTN# signal line 220. Additionally, although the BOFF# input of the VL-bus host 102 is not used in many conventional VL-bus implementations, in the system of FIGS. 1 and 2, it is connected to a corresponding BOFFB output of the VL-bus/PCI-bus bridge 146 for reasons hereinafter described. The BOFFB output of the VL-bus/PCI bridge 146 is also coupled to a WBACK# input of VL-bus device 144. In Pentium systems, the coupling of BOFFB to WBACK# takes place via an AND gate 223, the other input of which is connected to the HITM# output of the Pentium host.
The VL-bus address lines 202 are also connected to the core logic chipset 122, as are the VL-bus data lines 204, the D/C# line 206, the W/R# line 207, the M/IO# line 208, and the ADS# line 210. The core logic chipset 122 also has its BRDY# and BLAST# lines connected to the respective lines 216 and 214 of the VL-bus 104, has its LRDYI# input connected to the LRDY# line 218 of VL-bus 104, and has its RDY# line connected to the VL-bus RDYRTN# line 220. It also has its HCLK input connected to receive the signal on LCLK line 212. The LDEV# input of core logic chipset 122 is connected as hereinafter described.
As previously mentioned, the VL-bus/PCI-bus bridge 146 has a BOFFB output which is connected to the BOFF# input of the VL-bus host 102. If the host 102 separates the VL-bus 104 from the host bus (not shown) connected to the host CPU, and if such bridge does not translate the BOFF# signal between the two buses, then BOFFB is connected directly to the BOFF# input of the host CPU itself. The VL-bus/PCI-bus bridge 146 also includes an LADSIB signal line connected to the ADS# signal line of the VL-bus 104, an MIOB signal line connected to the M/IO# signal line of VL-bus 104, a DCIB signal line connected to the D/C# signal line of VL-bus 104, VD signal lines connected to the data lines of VL-bus 104, VA signal lines connected to the address lines 202 of VL-bus 104, and a clock input connected to the LCLK line 212 of VL-bus 104. The VL-bus/PCI-bus bridge 146 also includes a BLASTIB signal line connected to the VL-bus BLAST# line, a BRDYIB signal line connected to the BRDY# signal line, an RDYIB input connected to the VL-bus RDYRTN# line 220, and an LRDYOB output connected the VL-bus LRDY# signal line 218.
In a conventional VL-bus system implementation, the VL-bus signal lines LDEV<.times.># are typically ANDed together before being provided to the VL-bus/ISA-bus core logic chipset 122 LDEV# input. In the present embodiment, however, the combined signal (called LDEVI# is intercepted by the VL-bus/PCI-bus bridge 146 before being provided (by the name LDEVO#) to the LDEV# input of the VL-bus/ISA-bus core logic chipset 122. Accordingly, in FIG. 2, the VL-bus signal lines LDEVO#, LDEV1# and LDEV2# are connected to respective inputs of a 3-input AND gate 222, the output of which forms the LDEVI# signal. This signal is provided to an LDEVIB input of the VL-bus/PCI-bus bridge 146, which produces an LDEVOB output on a signal line LDEVO#. The LDEVO# is connected to the LDEV# input of VL-bus/ISA-bus core logic chipset 122.
On the PCI-bus side of VL-bus/PCI-bus bridge 146, the bridge is connected to the 32-bit multiplexed address/data bus 152. The bridge 146 also has a FRAMEB signal line connected to the FRAME# control signal line of the PCI-bus 148, an IRDYB signal line connected to the IRDY# signal line of the PCI-bus 148, a TRDYB signal line connected to the TRDY# signal line of PCI-bus 148, and a DEVSELIB signal line connected to the DEVSEL# signal line of the PCI-bus 148.
FIG. 2 also illustrates an ISA-bus device 142 connected to the ISA-bus 130 which is connected to the VL-bus/ISA-bus core logic chipset 122, a PCI-bus device 154 which is connected to all of the signal lines of PCI-bus 148, and a VL-bus device 144 which is connected to appropriate signal lines of the VL-bus 104. Pertinent to the present description, the VL-bus device 144 is able to listen to the VL-bus cycle definition signals (VA, BE#, R/W#, D/C#, and M/IO#) . If it recognizes a valid cycle to which it can respond, the VL-bus device 144 can assert LDEV# onto the VL-bus LEDV0# signal line, and can terminate its VL-bus cycle by asserting LRDY# on the VL-bus LRDY# signal line 218. The PCI-bus device 154 and the VL-bus device 144 can constitute any desired expansion devices, but typically they will include a video display driver, possibly SCSI adapters, disk controllers, LAN adapters, and so on.
The operation of the system of FIGS. 1 and 2 will be described with respect to certain timing diagrams beginning with FIG. 3. FIG. 3 simplified is a timing diagram illustrating the timing of signals on the VL-bus 104 for a VL-bus read cycle initiated by the VL-bus host 102 and claimed by a conventional VL-line target device. Possible variations, and the timing of various signals not shown in FIG. 3, can be gleaned from the above-incorporated VL-bus specifications. The timing is the same for a VL-bus master-initiated cycle, except that it is preceded by a bus arbitration procedure.
Referring to FIG. 3, the LCLK signal is indicated at line 302. Six LCLK cycles are shown, labeled T1, T21, . . . , T25. Cycle T1 corresponds to the T1 cycle of an Intel 80486 microprocessor, and cycles T21 . . . T25 correspond to five consecutive T2 cycles of such a microprocessor (see the Intel databooks incorporated above).
In LCLK cycle T1, as indicated on line 304, the host 102 places the cycle definition signals (ADR31:2!, M/IO#, W/R#, D/C# and BE3:0!#) on the appropriate signal lines of the VL-bus 104. It then strobes the ADS# signal to assert the validity of the cycle definition signals and to start the VL-bus cycle as shown in line 306. If a VL-bus target device, such as 144 in FIG. 2, recognizes the cycle definition and desires to claim the VL-bus cycle, it must do so within 20 nanoseconds after sampling ADS# low at the rising edge which begins T21. To claim the cycle, VL-bus target device 144 asserts its LDEV# output onto the VL-bus LDEV<.times.># signal to which it is connected. This causes AND gate 222 (FIG. 2) to assert LDEVI# low, which signal is passed through the VL-bus/PCI-bus bridge 146 to the LDEVO# signal line and onto the LDEV# input of VL-bus/ISA-bus core logic chipset 122. The core logic chipset 122 samples LDEV# on the first rising edge of LCLK following the 20 nanosecond period to determine whether any VL-bus device has claimed the cycle. Such rising edge is referred to herein as the VL-bus device bus-claiming deadline, and is relative to the LCLK rising edge at which ADS# is sampled asserted. For systems with LCLK.ltoreq.33.3 MHz, the VL-bus device claiming deadline is typically the first rising edge of LCLK following the rising edge on which ADS# is sampled asserted, and for systems in which LCLK.ltoreq.40 MHz, the VL-bus device claiming deadline is typically the second rising edge of LCLK following the rising edge in which ADS# is sampled asserted.
Accordingly, referring again to FIG. 3, the VL-bus device 144, through the AND gate 222, asserts LDEVI# within T21 for 33.3 MHz systems (arrow 308), or within T21 or T22 for 40 MHz systems and above (arrow 310), as indicated on line 312. In either case, VL-bus/PCI-bus bridge 146 asserts LDEVO# to the VL-bus/ISA-bus core logic chipset 122 immediately thereafter as indicated on line 314 (arrows 316 and 318, respectively). (Note that the VL-bus/PCI-bus bridge 146 is at the same time decoding the address and cycle definition signals as described elsewhere herein, and may assert LDEVO# as a result of the decode as well. The bridge 146 therefore may actually assert LDEVO# before receipt of LDEVI# as indicated by arrow 315.)
The VL-bus host 102 unasserts (negates) ADS# 306 after being sampled asserted, i.e. within LCLK cycle T21. The VL-bus device 144 therefore detects ADS# negated at the rising edge which begins T22, and within T22, begins driving its read data onto the VD lines 204 of VL-bus 104 as indicated on line 320.
The cycle definition signals remain active until RDYRTN# is sampled asserted on a rising edge of LCLK, i.e. in LCLK cycle T24 or T25 as indicated by arrows 330 and 331.
After the end of the second T2 state (i.e., in LCLK cycle T23), the VL-bus device 144 begins asserting LRDY# as indicated in line 322 for one LCLK cycle. This signal is detected at the LRDYI# input of the VL-bus/ISA-bus core logic chipset 122, which responds by asserting RDYRTN# on the VL-bus 104 (line 324 and arrow 326 in FIG. 3). For 40 MHz systems and faster, the core logic chipset 122 may wait until the first rising edge of LCLK following detection of LRDY# asserted to assert RDYRTN# (arrow 328), thereby resynchronizing the ready signals with LCLK. For 33 MHz systems, LRDYI# can be connected directly to RDYRTN#. The VL-bus host detects RDYRTN# asserted on its RDY# input, and terminates the VL-bus cycle.
The rising edge of LRDY# in T24 causes the core logic chipset 122 to unassert RDYRTN# either immediately thereafter (arrow 334), or for faster systems, in response to the first LCLK rising edge following detection of LRDY# unasserted (arrow 336). The VL-bus device 144 drives the VL-bus LRDY# signal line to a high logic level for one-half LCLK cycle following deassertion of LRDY#, and then floats LRDY#.
The VL-bus device 144 continues to drive data onto the VD lines 204 until the first LCLK rising edge at which it detects RDYRTN# asserted (arrow 338 or 340). It also cause AND gate 222 to de-assert LDEVI# at the same time (arrows 342, 344), and the VL-bus/PCI-bus bridge 146 de-asserts LDEVO# to the core logic chipset 122 at the same time (arrows 346, 348).
Thus it can be seen that the VL-bus/PCI-bus bridge 146 does not interfere with VL-bus read accesses to VL-bus target devices such as 144. The same is true with VL-bus write accesses to VL-bus target devices, and to VL-bus read and write accesses regardless of whether the originator is the VL-bus host 102, the VL-bus/ISA-bus core logic chipset 122 (e.g. acting for an ISA-bus master), or a VL-bus master device. Note that in the timing diagram of FIG. 3, if no VL-bus device had asserted its corresponding LDEV<.times.># signal by the VL-bus device claiming deadline, then the chipset 122 translates the cycle onto the ISA-bus 130 and returns any read data onto the VD lines 204 of the VL-bus 104.
FIG. 4 is a timing diagram illustrating the operation of the system of FIGS. 1 and 2 when the target device of a VL-bus access resides on the PCI-bus 148. The timing diagram of FIG. 4, like that of FIG. 3, has been simplified for clarity of description. In addition, a system with LCLK .ltoreq.33.3 MHz is assumed, and it is assumed that the PCI-bus clock is the same as the VL-bus clock, LCLK, but is treated asynchronously.
Referring to FIG. 4, as with FIG. 3, an LCLK signal is indicated on line 402. One T1 cycle is shown, followed by nine T2 cycles labeled T21 . . . , T29. The VL-bus host 102 drives a cycle definition onto the VL-bus 104 prior to T1 as indicated on line 404, and asserts ADS# as indicated on line 406. It is assumed that no VL-bus device claims the cycle, so LDEVI# remains high throughout the VL-bus cycle as indicated on line 408. Nevertheless, the VL-bus/PCI-bus bridge 146 asserts LDEVO# to VL-bus/ISA-bus core logic chipset 122 in response to detection of ADS# asserted at the beginning of the first T2 cycle. The core logic chipset 122 thus considers the VL-bus cycle as having been claimed by a VL-bus device, and withholds any translation of the cycle onto the ISA-bus 130. The VL-bus/PCI-bus bridge 146 also drives the address from VA lines 202 onto the AD lines 152 of PCI-bus 148 (line 412 in FIG. 4), and drives an appropriate command onto the C/BE# lines of PCI-bus 148 (line 414 in FIG. 4). After a delay for synchronizing to the PCI-bus clock signal, in T23, VL-bus/PCI-bus bridge 146 asserts FRAME# on the PCI-bus 148 (line 416 in FIG. 4). It also asserts the PCI-bus IRDY# signal in T24 (line 418). It is assumed that only one data phase will occur on the PCI-bus in response to the present example VL-bus cycle, so the bridge 146 de-asserts FRAME# in T24.
Prior to the PCI-bus device claiming deadline, the target PCI-bus device 154 (FIG. 2) claims the PCI-bus cycle by asserting DEVSEL#. The PCI-bus device 154 is assumed to be a fast device, so as indicated on line 420 (FIG. 4), it asserts DEVSEL# during T24 in response to detection of FRAME# asserted at the rising edge which begins T24. The bridge 146 releases the address and command from the PCI-bus AD and C/BE# signal lines, respectively, in response to the assertion of DEVSEL#, and asserts BE# onto the C/BE# signal lines. In T25, the PCI-bus device 154 drives the read data onto the PCI-bus AD lines 152 and asserts TRDY# for one clock cycle as indicated on line 422. The bridge 146 latches the data from AD lines 152 onto the VL-bus VD data lines 204 as indicated by line 430 and arrow 424 (although initially invalid data may be driven from AD to VD as early as T22). At the beginning of T26, the VL-bus/PCI-bus bridge 146 samples TRDY# asserted, and in response thereto, releases C/BE# and de-asserts IRDY# (arrows 426 and 428, respectively). The PCI-bus device 154 also releases the AD lines 152 and negates DEVSEL# in T26.
After another synchronization delay, in T28, the VL-bus/PCI-bus bridge 146 asserts LRDY# as indicated on line 432 for one LCLK cycle. The VL-bus/ISA-bus core logic chipset 122 responds by asserting RDYRTN# on the VL-bus 104 (line 434 and arrow 436 in FIG. 4); it de-asserts RDYRTN# in response to the de-assertion of LRDY# (arrow 438). The VL-bus/PCI-bus bridge 146 samples RDYRTN# asserted at the beginning of T29, and in response thereto, negates LDEVO# to the core logic chipset 122 (arrow 440).
Accordingly, it can be seen that the VL-bus/PCI-bus bridge translates VL-bus cycles onto the PCI-bus 148 for response by PCI-bus devices. This is accomplished without the need for any special signals between the VL-bus/PCI-bus bridge 146 and either the VL-bus host 102 or the VL-bus/ISA-bus core logic chipset 122. It will be understood that the same is true of VL-bus data write cycles, and VL-bus cycles originated by VL-bus device masters and by the core logic chipset 122 for an ISA-bus master. Essentially the VL-bus/PCI-bus bridge appears to each of the VL-bus devices as a VL-bus target and appears to each PCI-bus device as the host-bus/PCI-bus bridge. It will also be understood that numerous types of PCI-bus transactions are defined in the PCI-bus specification incorporated above, and while only one such transaction type is illustrated in FIG. 4, different situations may result in different PCI-bus transaction types as appropriate or desirable.
FIG. 5 is a timing diagram illustrating the operation of the system of FIGS. 1 and 2 for a VL-bus read access to a device which resides on the ISA-bus 130 (or to an address to which no device responds). As with FIGS. 3 and 4, the timing diagram of FIG. 5 is simplified for clarity of illustration. Possible variations, and the timing of various signals not shown in FIG. 5, can be gleaned from the above-incorporated references. As with FIGS. 3 and 4, the timing shown in FIG. 5 is the same for a VL-bus master-initiated cycle, except that it is preceded by a bus arbitration procedure.
Referring to FIG. 5, the LCLK signal is illustrated as line 502. Certain signals, such as the VL-bus signals VD and LDEVI#, and PCI-bus signals AD, C/BE#, TRDY#, are omitted from FIG. 5 for clarity. Signal line 504 illustrates the cycle definition signals on the VL-bus 104, and line 506 illustrates the ADS# signal. The VL-bus host 102 drives the cycle definition lines and asserts ADS# during the T1 cycle in order to define a valid VL-bus cycle. At the beginning of T21, the VL-bus/PCI-bus bridge 146 samples ADS# asserted, and asserts LDEVO# in response thereto (line 508 and arrow 540). It is assumed that no other VL-bus device intends to claim the VL-bus cycle, so LDEVI# (not shown in FIG. 5) remains high for the entire process of FIG. 5.
As with the VL-bus cycle illustrated in FIG. 4, after a synchronization delay, the VL-bus/PCI-bus bridge 146 asserts FRAME# during T23 on the PCI-bus 148 as shown at line 510 in FIG. 5. Again, since only one PCI-bus data phase is assumed to be desired, the bridge 146 negates FRAME# during T24. The bridge 146 also asserts IRDY# on the PCI-bus 148 during T24 as indicated at line 512 in FIG. 5.
According to the PCI-bus specification, a PCI-bus controller can be programmed to expect a PCI-bus device claiming signal asserted prior to a PCI-bus claiming deadline which is programmable to be 1, 2 or 3 PCI-clock rising edges after the rising edge at which FRAME# is sampled asserted. This is true of the bridge 146, and it is assumed that the bridge 146 has been programmed for the medium PCI-bus device claiming deadline of two PCI-bus clock cycles. Such programming was accomplished during system boot-up, in which the boot-up code surveyed the PCI-bus devices included in the system and determined that none required longer than two PCI-bus clock cycles to claim a PCI-bus cycle; this information was written into the register address offset 52h(1:0) of the PCI-bus bridge 146. Accordingly, since no PCI-bus device claims the cycle by asserting DEVSEL# by the LCLK rising edge which begins T26, (line 514 in FIG. 5), the VL-bus/PCI-bus bridge 146 issues a "master abort". This negates RDY# (arrow 516).
At this point, since the VL-bus/PCI-bus bridge 146 has asserted LDEVO# to the VL-bus/ISA-bus core logic chipset 122, the chipset 122 will not know that it needs to try finding the destination device on the ISA-bus 130. One solution would be to provide additional signals between the VL-bus/PCI-bus bridge 146 and the core logic chipset 122 in order to so inform the chipset 122, but this would require a modification of standard VL-bus/ISA-bus core logic chipsets such as 122. Accordingly, in the system of FIGS. 1 and 2, the VL-bus/PCI-bus bridge 146, in response to its failure to detect DEVSEL# asserted at the LCLK rising edge which begins T26, asserts BOFF# in T27 to the VL-bus host 102 (line 518 and arrow 520 in FIG. 5). As indicated on line 522 and arrow 524, after assertion of BOFF#, the VL-bus/PCI-bus bridge 146 asserts LRDY# in LCLK cycle T28 to cause the termination of the VL-bus cycle. LRDY# is delayed by one LCLK cycle (programmable to 0, 1 or 2 LCLK cycles) after BOFF# for reasons which will become apparent. As indicated by arrow 528, the VL-bus controller in 122 asserts RDYRTN# in response to detection of LRDY# asserted, and negates RDYRTN# in T29 in response to the negation of LRDY# in T29 (arrow 530). After the PCI-bus bridge 146 detects RDYRTN# asserted at the beginning of LCLK period T29, it negates LDEV# in LCLK period T29 as indicted by arrow 534. It also negates BOFF# in LCLK period T210 as indicated by arrow 532. The delay from RDYRTN# to negation of BOFF#, like the delay from BOFF# to LRDY#, is programmable in PCI-bus bridge 146 in order to be assured that the host CPU has itself received the BOFF# signal despite any time delays in a host-bus/VL-bus bridge in VL-bus host 102.
At the first LCLK rising edge in which BOFF# is sampled negated, the VL-bus host 102 repeats the VL-bus cycle which was aborted by the PCI-bus bridge 146. It does this by starting a new T1 cycle in which it once again drives the cycle definition signals on the VL-bus 104 as indicated on line 504 in FIG. 5, and asserts ADS# as indicated on line 506 in FIG. 5 (arrows 536 and 538, respectively). The VL-bus host 102 negates ADS# in the following LCLK cycle and withdraws the cycle definition signals in due course. Unlike the initial assertion of the VL-bus cycle, however, the VL-bus/PCI-bus bridge 146 does not assert LDEVO# in response to the repetition of the cycle, nor does it assert any signals on the PCI-bus 148. Instead, the VL-bus/ISA-bus core logic chipset 122 performs its usual duties of translating the cycle to the ISA-bus 130 and returning any read data back to the VL-bus VD lines 204. If no ISA-bus device responds to the access, then as in a conventional system, the data which the chipset 122 returns to the VL-bus VD lines 204 will be undefined. The chipset 122 then assert RDYRTN# for one LCLK cycle to end the VL-bus cycle.
The timing relationship between the PCI-bus bridge's assertion of LRDY# and BOFF# (lines 522 and 518 in FIG. 5) solves a problem which could otherwise preclude the use of LDEVO# and BOFF# to add PCI-bus 148 to a conventional system without modifying the core logic chipset 122. In particular, if the PCI-bus bridge 146 were to assert BOFF# without asserting LRDY#, then the VL-bus controller in VL-bus/PCI-bus core logic chipset 122 would not know one VL-bus cycle has ended and a new VL-bus cycle will soon begin, since the chipset 122 does not necessarily receive BOFF#. Other VL-bus devices would have the same problem, since they too do not necessarily receive BOFF#. VL-bus devices need only observe the RDYRTN# line (and BRDY# with BLAST#) to determine that a VL-bus cycle has ended, and RDYRTN# will not occur unless LRDY# is asserted first. On the other hand, the PCI-bus bridge 146 cannot assert LRDY# without asserting BOFF#, since LRDY# causes the VL-bus controller in core logic chipset 122 to assert RDYRTN#, which is connected to the VL-bus host 102 RDY# input. Thus the VL-bus host 102 would consider the VL-bus cycle terminated successfully, and would have no reason to repeat it. Thus the PCI-bus bridge 146 needs to assert both BOFF# and LRDY#.
If the bridge 146 were to assert LRDY# before asserting BOFF#, then once again, the VL-bus controller in chipset 122 would assert RDYRTN# and the VL-bus host 102 would receive RDY# possibly before it receives BOFF#. In this case the VL-bus host 102 would consider the original VL-bus cycle terminated successfully, begin a subsequent (non-repeat) VL-bus cycle, and apply the back-off to the subsequent cycle instead of the original cycle. On the other hand, if the PCI-bus bridge 146 asserts LRDY# too late, such that the core logic chipset 122 does not assert RDYRTN# until after BOFF# has been negated at the input of the host, then the VL-bus host 102 might repeat the back-offed VL-bus cycle before the VL-bus devices, including that in the VL-bus/ISA-bus core logic chipset 122, can reset their VL-bus state machines and prepare to detect it.
According to an aspect of the invention, therefore, the PCI-bus bridge 146 does not assert LRDY# until after it has asserted BOFF#. The 80486 and Pentium-compatible processors ignore their RDY# input while BOFF# is being asserted. Advantageously, the PCI-bus bridge 146 delays assertion of LRDY# by a number of LCLK cycles in order to ensure that the microprocessor itself has had an opportunity to sample BOFF# asserted before (or no later than the LCLK rising edge that) it detects an active RDY# from the RDYRTN# line. This delay, which is programmable in the PCI-bus bridge 146, accommodates any timing delays which might occur between the BOFFB output of PCI-bus bridge 146 and the BOFF# input of the microprocessor itself. Additionally, the PCI-bus bridge 146 does not negate BOFF# until after it detects RDYRTN# asserted. As previously mentioned, advantageously the PCI-bus bridge 146 can be programmed to delay negation of BOFF# by one or two additional LCLK cycles in order to ensure that BOFF# is still asserted at the host microprocessor BOFF# input when the microprocessor's RDY# input is asserted.
The timing diagram of FIG. 5 also describes a VL-bus master-initiated access to an ISA-bus device. VL-bus devices are not required to receive BOFF#, but they do receive a WBACK# signal. In a conventional system, WBACK# is typically connected to the HITM# output of a Pentium-based VL-bus host. The Pentium contains an on-chip write-back cache memory, and it asserts HITM# whenever it has an internal cache miss and a write-back to system memory is required. When a bus master detects HITM# asserted (e.g. via a WBACK# input), the bus master is required to immediately abort its cycle and float all address data and control signals that it drives as master. When the device later samples WBACK# inactive, the device may restart the bus cycle with a new ADS#.
Conveniently, this required operation of a VL-bus master in response to WBACK# is the same as the operation of the VL-bus host in response to BOFF#. Thus VL-bus masters can be accommodated in a system incorporating the invention, by coupling the BOFFB output of the PCI-bus bridge 146 to the WBACK# inputs of each such device. In systems having a write-back cache, the BOFFB signal can be ANDed with HITM# such that either HITM# or BOFFB will force the VL-bus master to perform a backoff function.
As previously mentioned, some VL-bus implementations are built around a host microprocessor whose data path is twice as wide as the VD lines 204 on VL-bus 104. Such is the case for systems built around the Intel Pentium microprocessor, for example. In order to accommodate these arrangements, a conventional VL-bus system includes a host-bus/VL-bus bridge (not shown) within the VL-bus host 102 of FIG. 2. Such a bridge will assert two accesses on the VL-bus in response to each access on the host bus if the data path on the host bus is twice as wide as that on the VL-bus 104. Described more generally, since it is also possible that a future system could be built around a microprocessor whose data path is four times as wide as the VD lines 204, or more, such bridge will assert a "plurality" of accesses on the VL-bus 104 in response to each access on the host bus (not shown).
In order to accommodate a situation where the host-bus data path is twice as wide as the VL-bus data path, the VL-bus/PCI-bus bridge 146 uses its BE4# input pin to determine whether two VL-bus "sub-cycles" are being issued for each host-bus cycle. In systems whose CPUs have only a 32-bit wide data bus, such as 486 systems, the BE4# input pin is strapped high. In Pentium systems, the BE4# input of the bridge 146 is connected to the BE4# address lead of VL-bus 104. If BE4# is high in a particular cycle, then only one sub-cycle will occur. If BE4# is low in the particular cycle, then two sub-cycles will occur. In the latter case the VL-bus/PCI-bus bridge 146 will assert LDEVO# and subsequently LRDY# for both of the sub-cycles, but will assert BOFF# only for the last such sub-cycle. On repetition, the VL-bus/PCI-bus bridge 146 blocks assertion of LDEVO# for both sub-cycles of the host bus cycle repetition. In this manner, the host-bus/VL-bus bridge (not shown) also need not be modified in any way to accommodate the addition of a VL-bus/PCI-bus bridge such as 146.
In the computer system illustrated in FIGS. 1 and 2, any ISA-bus device, PCI-bus device or VL-bus device, in addition to the VL-bus host 102, can be a master for any particular access. The target for such accesses can be main memory (not shown in FIG. 2), or any ISA-bus device, PCI-bus device, or VL-bus device. The VL-bus/PCI-bus bridge therefore accommodates a wide variety of different kinds of accesses, only some of which have been described in detail with respect to FIGS. 3, 4 and 5. Such descriptions are sufficient for an understanding of the invention, but for completeness, an overview of the remainder of the cycle permutations is included in Appendix E.
FIGS. 6 through 10 are schematic diagrams of pertinent portions of the VL-bus/PCI-bus bridge 146 which control various signals used for implementing the invention. While the descriptions above and in the appendices are sufficient to enable implementation of the invention, descriptions at the schematic level for some aspects are provided for those interested in more details about an example implementation.
FIG. 6 is a schematic diagram for circuitry which generates the LDEVOB output of the PCI-bus bridge 146. It comprises a 3-input NOR gate 602 which generates such a signal, the three inputs of which receive a QN output of a D-flip-flop 604, a DRDEVLO signal, and an LDEVI signal. The LDEVI signal is the inverse of the LDEVIB input of PCI-bus bridge 146. The DRDEVLO signal is generated by a 5-input NOR gate 606, the five inputs of which receive signals labeled: SAMPLEB, H2PENB, BLOCK, ISADEV and HMEMCYC. The SAMPLEB signal is generated by gates 608 in response to WS, T21B and T22B signals according to the formula:
WS is a programmable signal which is asserted high to indicate that the LCLK signal is operating at a high enough frequency such that the local device claiming deadline is to be delayed by one wait state. T21B is a signal generated by means not shown in the PCI-bus bridge 146, which is asserted low at the beginning of T21 (see FIG. 3, for example) and returns high at the beginning of T22. The T22B signal is asserted low by means not shown in the PCI-bus bridge 146 at the beginning of T22, and returns high at the beginning of T23.
The H2PENB signal is generated by logic 610 in response to the following signals from the VL-bus 104 (FIG. 2): WR, MIO, and DC. The formula to produce H2PENB is:
The BLOCK signal is generated by circuitry described below with respect to FIG. 10, and the ISADEV signal is generated by circuitry (not shown) which, based on a positive decode of the cycle definition signals on VL-bus 104 asserts ISADEV when the intended recipient of the bus cycle is known to reside on the ISA-bus 130 (FIG. 2). The HMEMCYC signal is generated by NOR gate 612, the two inputs of which receive an HMEMB signal and an ENLDEVO signal, respectively. The HMEMB signal corresponds to the LMEM# input of the PCI-bus bridge 146 (see Appendix E) or, depending on a register bit selection, a positive decode of the VL-bus cycle definition signals. If the PCI-bus bridge 146 is programmed to perform such a positive decode, then circuitry (not shown) in PCI-bus bridge 146 asserts HMEMB low when the destination of the cycle is known to be host memory on the host memory bus or VL-bus 104. ENLDEVO is a register bit which is programmable to enable (when asserted high) or disable (when low) the use of HMEMB in the determination of whether to assert LDEVOB.
The flip-flop 604 has a clock input connected to receive CLK, which is the LCLK signal on VL-bus 104. The flip-flop 604 also has a D input connected to receive the output of logic 614, which generates the D input according to the formula:
In this formula, Q is the Q output of flip-flop 604, RST is a chip-wide reset signal, and DRDEVLO is a signal generated by NOR gate 606 discussed above. RDYB is generated by logic circuitry 616 according to the inverse of the formula:
where BRDYI and BLASTI are the VL-bus BRDY# and BLAST# signals, respectively; RDYI is connected to the VL-bus RDYRTN# signal line; and "Other" represents other ready signals not important for an understanding of the present invention. BOFFRDY is given by the formula:
where FBOLRDY and BOFFP are both generated by circuitry described below with respect to FIGS. 9 and 7, respectively, and BOFFEN is a register bit which enables the entire back-off process when set.
In operation, the circuitry of FIG. 6 asserts LDEVOB low in response to a logic high value on any of the three inputs of NOR gate 602. Thus due to the LDEVI input, it can be seen that whenever any VL-bus device asserts one of the LDEV<.times.># signal lines in FIG. 2, the LDEVI# signal provided to the PCI-bus bridge 146 will pass therethrough to the LDEV# input of the VL-bus/ISA-bus core logic chipset 122. The other circuitry in FIG. 6 also will activate LDEVOB in most cases, so LDEVOB will remain asserted until the VL-bus device unasserts LDEV<.times.># or until the flip-flop 604 QN output goes low, whichever occurs later.
The remainder of the circuitry of FIG. 6 operates by asserting DRDEVLO before the VL-bus device claiming deadline, to thereby assert LDEVOB, and by maintaining LDEVOB asserted with the flip-flop 604 until an appropriate ready signal has been received by the PCI-bus bridge 146. More specifically, NOR gate 606 will assert DRDEVLO only when all five of its inputs are low. The SAMPLEB signal is a temporal qualifier on the assertion of DRDEVLO, since SAMPLEB is asserted low only during T22 (for systems with LCLK operating at frequencies greater than or equal to 40 MHz) or during T21 and T22 (for systems operating at LCLK frequencies less than or equal to 33 MHz). The inverse of SAMPLEB is given by:
Thus if NOR gate 606 asserts DRDEVLO at all, it unasserts DRDEVLO at the beginning of T23.
The H2PENB input of NOR gate 606 is asserted low unless the cycle definition signals on VL-bus 104 indicate that the present cycle is an interrupt acknowledge cycle, a halt/special cycle, or a reserved cycle, all as defined at page 7 of the above-incorporated "i486 MICROPROCESSOR" Databook. These are VL-bus cycles which are known in advance not to be destined for any device residing on the PCI-bus bridge 148.
The BLOCK signal input of NOR gate 606 will block assertion of DRDEVLO only if the current VL-bus cycle definition is a repeat of a prior VL-bus cycle for which the PCI-bus bridge 146 asserted BOFF# to the VL-bus host 102. In Pentium-based systems, BLOCK remains asserted for the repetition of both sub-cycles in a repeat of a BOFF#'d host cycle. The BLOCK signal therefore prevents the PCI-bus bridge 146 from claiming the repetition of a VL-bus cycle which the PCI-bus bridge 146 has already attempted on the PCI-bus 148. The circuitry which generates BLOCK is described below with respect to FIG. 10.
The ISADEV signal of NOR gate 606 is asserted high on a positive decode indicating that the target of the current VL-bus cycle is already known to reside on the ISA-bus 130. The ISADEV signal thus prevents the PCI-bus bridge 146 from claiming the cycle and trying it on the PCI-bus first. Note that another embodiment of the invention need not prevent this attempt, but rather can rely on the BOFF# procedures described herein for the cycle eventually to reach the ISA-bus 130.
The HMEMCYC signal is asserted high on a positive decode of the VL-bus cycle, performed either inside or outside the PCI-bus bridge chip 146, to indicate that the access is to host memory. The positive decode, if performed internally to the PCI-bus bridge 146, is performed identically, with a parallel set of configuration registers, to that in the VL-bus/ISA-bus core logic chipset 122. The PCI-bus bridge 146 further includes a register bit ENLDEVO which can be used to permit NOR gate 606 to assert DRDEVLO even if the positive decode indicates that the VL-bus cycle destination is in fact host memory; an attempt by the PCI-bus bridge 146 to claim such a cycle by asserting LDEVO# would not have any effect in most systems because most VL-bus/ISA-bus core logic chipsets ignore their LDEV# inputs when their own internal decode indicates that the VL-bus cycle is to host memory.
Thus it can be seen that the circuitry of FIG. 6 will assert LDEVOB in either the T21 or T22 states of the VL-bus host 102, whichever is appropriate, unless the current VL-bus cycle is a repeat of a prior VL-bus cycle which was subject to BOFF#, and unless it is known in advance that the target of the VL-bus cycle is not on the PCI-bus 148. The latter condition, which may be omitted in whole or in part in a different embodiment, occurs in the present embodiment when it is known in advance that the cycle is a special kind of cycle defined only for targets which do not reside on the PCI-bus, if at all (H2PENB), when the target is known in advance to reside on the ISA-bus 130 (ISADEV), and when the VL-bus cycle is known in advance to be destined for host memory.
When DRDEVLO is asserted, its inverse is written into D-flip-flop 604 on the next LCLK rising edge in order to maintain a logic 1 input of NOR gate 602 from the QN output of flip-flop 604. The logic 614 repeatedly rewrites this value into flip-flop 604 until either a system-wide reset occurs, or RDYB is asserted low by circuitry 616. Circuitry 616 asserts RDYB low upon receipt of RDYRTN# asserted by the VL-bus controller 122. It asserts RDYB low also when it detects both BRDY# and BLAST# asserted on VL-bus 104. Thus the circuitry of FIG. 6 maintains LDEVOB asserted until the rising edge of LCLK which follows completion of the VL-bus cycle which the circuitry claimed by asserting LDEVOB. Note that as described in more detail below, if the PCI-bus bridge 146 did translate the cycle onto the PCI-bus bridge 148, then the RDYRTN# signal will be asserted by the VL-bus controller 122 only after having detected LRDY# asserted by the PCI-bus bridge 146. If the circuitry of FIG. 6 had asserted LDEVOB only because another VL-bus device had caused the assertion of LDEVI, then the cycle will have been terminated only in response to ready signals from the VL-bus device which had claimed the cycle.
The logic circuitry 616 asserts RDYB also in response to a pulse on FBOLRDY or BOFFP, but only if the PCI-bus bridge 146 has been programmed to defeat the back-off feature (BOFFEN=0). FBOLRDY and BOFFP each carry a high-going pulse for one LCLK cycle in response to a master-abort on the PCI-bus 148, although the FBOLRDY pulse occurs only if the current VL-bus cycle is the first sub-cycle of a Pentium cycle, and the BOFFP pulse occurs only if the current VL-bus cycle is anything other than the first sub-cycle of a Pentium cycle. These pulses occur only after the PCI-bus bridge 146 has translated a cycle onto the PCI-bus 148, and failed to detect a PCI-bus device claiming signal prior to the PCI-bus device claiming deadline. Accordingly, if the back-off feature of PCI-bus bridge 146 has been disabled, the logic circuitry 616 causes the release of LDEVO# if the PCI-bus bridge 146 attempted, but failed, to find a target device on the PCI-bus 148. In this case, with BOFFEN=0, it is then up to external circuitry (not shown) to handle the VL-bus cycle.
FIG. 7 is a schematic diagram of circuitry in the PCI-bus bridge 146 which generates the BOFFP signal used in the circuitry of FIG. 6. FIG. 7 also shows circuitry which generates an LBOLRDY signal, and this will be described subsequently. Referring to FIG. 7, the BOFFP signal is the output of a NOR gate 702, one input of which is connected to the Q output of a D-flip-flop 704, the D input of which is connected to the Q output of a flip-flop 706 via an inverter 708. The second input of NOR gate 702 is connected to the Q output of flip-flop 706. Both flip-flops 704 and 706 are clocked by the CLK signal which is connected to the VL-bus LCLK line.
The D input to flip-flop 706 connected to the output of a 4-input NAND gate 710, the inputs of which are connected to signals labeled RSTB, TNH, LASTDW and SMSABORT. The RSTB signal is the system-wide reset signal as previously described. TNH indicates the state of the state machine in PCI-bus bridge 146 which controls PCI-bus cycles. While the details of this state machine are unimportant for an understanding of the present invention, it will be useful to understand that the state machine is in state TNH (TNH=1) while the PCI-bus bridge is controlling a PCI-bus cycle, and is in a TH state (TH=1) while a VL-bus cycle is in process but the PCI-bus bridge 146 will not translate it onto the PCI-bus. The LASTDW signal is generated by circuitry described below with respect to FIG. 8, and basically is asserted during all VL-bus cycles in non-Pentium-based systems, and only during the last data word sub-cycle in Pentium-based systems. Thus LASTDW indicates that the present VL-bus cycle is the last (or only) sub-cycle of a 32-bit or 64-bit access. The SMSABORT signal is generated by circuitry described below with respect to FIG. 9. Essentially it is a PCI-bus master-abort signal, synchronized to the VL-bus LCLK if necessary.
The flip-flop 704 and 706, together with the inverter 708 and NOR gate 702, produce a high-going pulse on the BOFFP in response to a high-to-low transition on the output of NAND gate 710. The output of NAND gate 710 will go low only when all of its inputs are high. Thus it can be seen that the circuitry in FIG. 7 will generate a high-going BOFFP pulse, for one LCLK cycle, in response to a master-abort which occurs while the PCI-bus bridge 146 is translating the last VL-bus cycle of a 32-bit or 64-bit access, onto the PCI-bus 148. As previously mentioned, the circuitry does not generate a pulse on BOFFP if the system is Pentium-based, and the current VL-bus cycle is only the first sub-cycle of a Pentium access. (Note that while reset signals are incorporated throughout the schematic diagrams set forth herein, they are asserted only in unusual situations not relative to an understanding of the present invention. Thus, while their effect can be determined from the schematics, it is unnecessary to describe such effects as well. Hereinafter, reset signals will always be assumed to be unasserted.)
The LASTDW signal is generated by circuitry 802 in FIG. 8 according to the formula:
LHA2 is the 2.sup.2 address bit on the VL-bus 104, and therefore indicates whether the current VL-bus cycle is to an even- or odd-numbered data word address. In non-Pentium-based systems, VL-bus cycles can be addressed to any 32-bit data word. But in Pentium-based systems, a 64-bit access always appears on a 32-bit VL-bus as an access to an even-numbered data word followed by an access to an odd-numbered data word. Thus for Pentium-based systems only, LHA2 indicates which sub-cycle is currently taking place on the VL-bus in response to a 64-bit access.
LBE4 is an input pin which, when asserted, indicates that the system is a Pentium-based system. Thus the circuitry 802 asserts LASTDW whenever the system is not a Pentium-based system (LBE4=1), and if it is a Pentium-based system, asserts LASTDW only during the second sub-cycle of a 64-bit access.
The SMSABORT signal used in the circuitry of FIG. 6 is generated by circuitry 902 in FIG. 9. Circuitry 902 includes an inverting multiplexer 904, the output of which produces SMSABORT. The multiplexer 904 has a `0` input, a `1` input and a select input S, the last of which is connected to receive an ASYNC signal from a register bit. The `0` input of multiplexer 904 is connected to the output of a NAND gate 906, the three inputs of which are connected to receive TNH, MSABORT and an ATSPACE signal. The TNH signal has been described previously, and the MSABORT signal is generated by the PCI-bus bridge 146 in response to the failure of any PCI-bus device to claim a cycle on the PCI-bus 148 prior to the PCI-bus device claiming deadline. The ATSPACE signal is produced by circuitry 908 according to the formula:
where LMIO is a latched version of the VL-bus M/IO# signal. ATSPACE is asserted for VL-bus I/O cycles to an address within the 64 k I/O memory space, and for VL-bus memory cycles to an address within the 16-megabyte memory address space which can be supported on a standard ISA bus. (In an EISA implementation, the 16-megabyte limitation would be omitted since the EISA-bus supports a full 4 GB of memory address space.) Thus ATSPACE is high for accesses which could possibly be directed to devices residing on the ISA-bus 130; more intuitively, if ATSPACE is low, then the current VL-bus cycle cannot be directed to a device on the ISA-bus 130.
The output of NAND gate 906, in addition to being connected to the `0`input of multiplexer 904, is also connected to one input of a 2-input NAND gate 910, the other input of which is connected to the output of a 3-input NAND gate 912. The output of NAND gate 910 is cross-coupled to one of the inputs of NAND gate 912. The other two inputs of NAND gate 912 are connected to receive RSTB and a ready signal RDYB previously described with respect to FIG. 6. The output of NAND gate 912 is connected to the D input of a flip-flop 914 which is also clocked by LCLK. The output of flip-flop 914 is connected to the `1` input of multiplexer 904.
The circuitry 902 operates in either a synchronous mode or an asynchronous mode, depending on the programming of register bit ASYNC. Basically, if the clock signal on the VL-bus and the PCI-bus are the same, then ASYNC=0 and multiplexer 904 selects the output of NAND gate 906 directly for driving the SMSABORT signal. If the two clocks are asynchronous, either because they operate at different speeds or because clock skew considerations suggest that re-synchronization be performed, the VL-bus host 102 will have programmed ASYNC=1. In this situation, multiplexer 904 selects the LCLK-synchronized output of flip-flop 914 to drive SMSABORT. Essentially cross-coupled NAND gates 910 and 912 latch the signal output of NAND gate 906, and flipflop 914 synchronizes the result with the VL-bus LCLK signal. The latch is cleared when circuitry 616 (FIG. 6) asserts RDYB, indicating either completion of the VL-bus cycle or, if the back-off feature of PCI-bus bridge 146 is disabled, an appropriate time to clear SMSABORT.
Thus circuitry 902 asserts SMSABORT whenever the output of NAND gate 906 goes low. This occurs only when all three inputs of NAND gate 906 are high, thereby requiring (1) that a master abort have occurred on the PCI-bus 148 (MSABORT=1); (2) that the VL-bus cycle not be to a memory or I/O address which clearly cannot reside on the ISA- bus 130 (ATSPACE=1); and (3) that the PCI-bus bridge 146 presently be translating a cycle onto the PCI-bus 148 (TNH=1) (i.e. SMSABORT will not be asserted if the bridge 146 is not acting as the PCI-bus master).
The FBOLRDY signal, the inverse of which is used in FIG. 6, is generated by circuitry 916 in FIG. 9. Circuitry 916 includes a pulse generator comprising D flip-flops 918 and 920 separated by inverter 922, and 4-input NOR gate 924. The QN outputs of flip-flops 918 and 920 are connected to respective inputs of NOR gate 924, the other two inputs of which are connected to receive LASTDW (previously described) and the inverse of TNH (previously described). It can be seen that if LASTDW=0 and TNH=1, then circuitry 916 will produce a high-going pulse for one LCLK period on FBOLRDY, in response to a high-to-low transition of SMSABORT. The high-going pulse on FBOLRDY will be prevented if TNH=0, indicating that any PCI-bus cycle which resulted in a master-abort was not generated by PCI-bus bridge 146. The high-going pulse on FBOLRDY will be prevented also if LASTDW=1, which occurs for all VL-bus cycles except for the first sub-cycle of a 64-bit Pentium access. Thus considering the circumstances under which circuitry 902 generates SMSABORT, circuitry 916 will produce a high-going pulse on FBOLRDY only in response to a master-abort on the PCI-bus 148, occurring during a PCI-bus cycle originated by PCI-bus bridge 146 (TNH=1) in response a VL-bus cycle which constitutes the first subcycle of a Pentium-based 64-bit access.
Circuitry 804 of FIG. 8 generates the BOFFB signal for driving the BOFF# input of VL-bus host 102. Referring to FIG. 8, BOFFB is produced by a 2-input NAND gate 806, one input of which is connected to receive the register programmable BOFFEN signal previously described. The other input of NAND gate 806 is connected to the Q output of a D-flip-flop 808 clocked by LCLK. The D input of flip-flop 808 is driven by circuitry 810 according to the formula:
All signals in this formula have been previously described except Q and CPURDY. Q is the output of flip-flop 808, and CPURDY is the output of circuitry 812 (FIG. 8), comprising sequential D-flip-flops 814 and 816 and 4:1 multiplexer 818. The input to the circuitry 812 is the RDYI signal, which is connected to receive RDYRTN# from the VL-bus controller in 122.
The two select inputs of multiplexer 818 are connected to receive two register bits LRDYDL0 and LRDYDL1. In a system with no bridge between the host CPU and the VL-bus 104, the LRDYDL0 and LRDYDL1 signals (collectively, the LRDYDL signals) would be programmed such that multiplexer 818 passes RDYRTN# directly through to CPURDY. In systems which include a host-bus/VL-bus bridge, the LRDYDL signals can be programmed to have multiplexer 818 pass to CPURDY either the Q output of flip-flop 814 or the Q output of flip-flop 816. These represent versions of RDYRTN# delayed by one and two LCLK cycles, respectively. Thus by proper programming of the LRDYDL register bits, the circuitry 812 asserts CPURDY only after it is certain that the RDYRTN# signal on VL-bus 104 has reached and been detected by the host CPU itself on its RDY# input.
In the operation of circuitry 804, referring to the above formula for the D input of flip-flop 808, the NAND gate 806 will never assert BOFFB unless the back-off feature of PCI-bus bridge 146 is enabled (BOFFEN=1). If it is enabled, then the circuitry will assert BOFFB when the first term of the above formula is true. It will hold BOFFB asserted until both terms of the above-formula are false--usually the second term will be true for the longest period of time. The first term will be true in response to a master-abort on the PCI-bus 148 (SMSABORT=1) which was controlled by the PCI-bus bridge (TNH=1) in response to either the only sub-cycle of a 32-bit access or the last sub-cycle of a Pentium 64-bit access (LASTDW=1). CPURDY must also be negated, but this is always the case at the time of SMSABORT. The circuitry 804 will hold D asserted due to the second term of the above formula, until CPURDY is asserted to indicate that RDYRTN# has already reached the host CPU.
As previously described and as described in more detail with reference to schematics below, the assertion a by PCI-bus bridge 146 of LRDY# to terminate the VL-bus cycle in the view VL-bus devices, must reach the host CPU neither before nor after BOFF# in order to ensure that the CPU will ignore it. The VL-bus controller 122 generates the signal (RDYRTN#) which drives the RDY# input of the CPU for one LCLK cycle, in response the PCI-bus bridge's assertion of LRDY#. Thus by maintaining BOFFB asserted until 0, 1 or 2 LCLK cycles (as appropriate) after the PCI-bus bridge 146 detects RDYRTN# asserted, circuitry 804 ensures that the BOFF# input of the host CPU will still be active when the host CPU receives an active RDY#.
The LRDYOB signal output of the PCI-bus bridge 146, which drives the LRDY# line of VL-bus 104, is generated by circuitry 926 in FIG. 9, according to the inverse formula:
The circuitry 926 will assert LRDYOB in response to any one of several readies. In particular, it will assert LRDYOB when a PCI-bus-bridge-originated PCI-bus cycle terminates normally (H2PLRDYB=0), as well as when any of several other readies are received (collectively called "Others"). If the back-off feature is enabled (BOFFEN=1), then circuitry 926 asserts LRDYOB also in response to a pulse on either FBOLRDY or LBOLRDY. FBOLRDY, as previously explained, carries a pulse in response to SMSABORT only if the current VL-bus cycle, which the PCI-bus bridge 146 attempted on the PCI-bus 148 prior to master-abort, was the first sub-cycle of a Pentium 64-bit access. Like BOFFP, described above, LBOLRDY carries a pulse in response to SMSABORT only if the current VL-bus cycle, which the PCI-bus bridge 146 attempted on the PCI-bus 148 prior to master-abort, was the only sub-cycle of a non-Pentium 32-bit access or the last sub-cycle of a Pentium 64-bit access. LBOLRDY is generated from BOFFP by circuitry 712 in FIG. 7, which merely delays the BOFFP pulse by 0, 1 or 2 LCLK cycles depending the programming of the LRDYDL bits. By delaying LBOLRDY in this manner, the PCI-bus bridge 146 ensures that for VL-bus cycles in which BOFF# is asserted to the host CPU, the RDYRTN# signal will not reach the CPU's RDY# input any earlier than BOFF#.
As can be seen, for non-Pentium systems, PCI-bus bridge 146 will generate both BOFF# and LRDY# when its attempt to find a target device on the PCI-bus 148 fails. In a Pentium-based system, the PCI-bus bridge 146 will assert LRDY# in both such sub-cycles, but will assert BOFF# only in the last sub-cycle of the Pentium 64-bit access.
FIG. 10 illustrates circuitry which produces the BLOCK signal used in FIG. 6 to block DRDEVLO when the VL-bus host 102 repeats a cycle which has been subject to BOFF#. As shown in FIG. 10, BLOCK is produced by 2-input NAND gate 1002, one input of which is connected to receive a BLOCK2B signal, and the other of which is connected to the output of an OR gate 1004. One input of OR gate 1004 is connected to the output of logic 1006, which asserts its output low according to the formula:
SAMECYC is a signal generated by circuitry (not shown) in the PCI-bus bridge 146 which indicates that the current VL-bus cycle definition signals match those previously latched for the VL-bus which is subject to the BOFF#. For Pentium-based systems, SAMECYC is asserted on both sub-cycles of a Pentium 64-bit access.
The other input of OR gate 1004 is connected to the QN output of a D-flip-flop 1008, the D input of which is connected the output of circuitry 1010 which drives D according to the formula:
UNBLOCK is the QN output of a D flip-flop 1012, the D input of which is connected to the output of a 3-input NOR gate 1014. The three inputs of NOR gate 1014 are connected to receive RSTB, TH and SAMECYC.
The BLOCK2B signal is the Q output of a D flip-flop 1016, the D input of which is connected to the output of logic 1018 implementing the inverse of the formula:
In operation, NAND gate 1002, in combination with OR gate 1004, asserts BLOCK whenever BLOCK2B goes low and whenever both the output of circuitry 1006 and the QN output of flip-flop 1008 are low. Essentially, the QN output of flip-flop 1008 goes low when BOFFB is asserted, and remains low until flip-flop 1012 asserts UNBLOCK. Flip-flop 1012 asserts UNBLOCK when the cycle repetition appears. Circuitry 1006 temporally qualifies BLOCK by allowing the QN output of flip-flop 1008 to pass through OR gate 1004 only during T21 and T22 of the cycle repetition.
Since flip-flop 1008 is cleared when the cycle repetition appears, its QN output is no longer low during the second sub-cycle of a Pentium-based 64-bit access. In this situation, BLOCK2B is low. Referring to the above formula for the D input of flip-flop 1016, it can be seen from the first term that a logic 0 will be written into the flip-flop on the first sub-cycle (LHA2=0) of the cycle repetition (SAMECYC=1) of a Pentium-based (LBE4=1) host memory cycle (TH=1) in a system where the back-off feature of PCI-bus bridge 146 is enabled (BOFFEN=1). Thereafter, since BLOCK2=1, the second term of the formula will control the writing of a logic 1 into the flip-flop 1016. Specifically, the circuitry 1018 will maintain BLOCK2B=0 until the end of the second (LHA2=1) host memory (TH=1) VL-bus cycle of the Pentium 64-bit access.
Thus the circuitry of FIG. 10 will correctly block assertion by the PCI-bus bridge 146 of LDEVO# for the only sub-cycle of the repetition of a BOFF#'d non-Pentium-originated access, and for both sub-cycles of a repetition of a BOFF#'d Pentium 64-bit access.
As used herein, a given signal or event is "responsive" to a predecessor signal or event if the predecessor signal or event influenced the given signal or event. If there is an intervening processing element or time period, the given event or signal can still be "responsive" to the predecessor signal or event. If the intervening processing element combines more than one signal or event, the signal output of the processing element is considered "responsive" to each of the signal or event inputs. If the given signal or event is the same as the predecessor signal or event, this is merely a degenerate case in which the given signal or event is still considered to be "responsive" to the predecessor signal or event.
The foregoing description of preferred embodiments of the present invention has been provided for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations will be apparent to practitioners skilled in this art. For example, adaptations will be apparent for using the invention with an EISA bus instead of an ISA bus. The embodiments described herein were chosen and described in order to best explain the principles of the invention and its practical application, thereby enabling others skilled in the art to understand the invention for various embodiments and with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents.
APPENDIX A VL-BUS SIGNALS
The following signals are defined in the VL-bus standard 1.0.
APPENDIX B PCI-BUS 2.0
The PCI Local Bus is a high performance, 32-bit or 64-bit bus with multiplexed address and data lines. It is intended for use as an interconnect mechanism between highly integrated peripheral controller components, peripheral add-in boards, and processor/memory systems.
Configuration registers are specified for PCI components and add-in cards. A system with embedded auto configuration software can automatically configure PCI add-in cards at power on.
Typical PCI Local Bus implementations will support up to three add-in board connectors, although expansion capability is not required. The PCI add-in board connector is a Micro Channels (MC)-style connector. The same PCI expansion board can be used in ISA-, EISA-, and MC-based systems.
Relevant PCI-bus signals are defined below. The second column indicates a signal type as is defined in as follows:
Interrupt Pins (Optional)
Interrupts on PCI are optional and defined as "level sensitive", asserted low (negative true), using open drain output drivers. The assertion and deassertion of INTx# is asynchronous to CLK. PCI defines one interrupt line for a single function device and up to four interrupt lines for a multi-function device or connector. For a single function device, only INTA# may be used while the other three interrupt lines have no meaning.
Any function on a multi-function device can be connected to any of the INTx# lines. The file Interrupt Pin register defines which INTx# line the function uses to request an interrupt. If a device implements a single INTx# line, it is called INTAKE if it implements two lines, then they are called INTA# and INTB#, and so forth. For a multi-function device, all functions may use the same INTx# line or each may have its own (up to a maximum of four functions) or any combination thereof. A single function can never generate an interrupt request on more than one INTx# line.
The system vendor is free to combine the various INTx# signals from PCI connector(s) in any way to connect them to the interrupt controller. They may be wire-ORed or electronically switched under program control, or any combination-thereof. This means the device driver may not make any assumptions about interrupt sharing. All PCI device drivers must be able to share an interrupt (chaining) with any other logical device, including devices in the same multi-function package.
Cache Support Pins (Optional)
A cacheable PCI memory should implement both cache support pins as inputs, to allow it to work with either write-through or write-back caches. If cacheable memory is located on PCI, a bridge connecting a write-back cache to PCI must implement both pins as outputs; a bridge connecting a write-through cache may only implement one pin.
64-Bit Bus Extension Pins (Optional)
The 64-bit extension pins are collectively optional. That is, if the 64 bit extension is used, all the pins in this section are required.
Bus Commands
Bus Commands indicate to the target the type of transaction the master is requesting. Bus Commands are encoded on the C/BE3::0!# lines during the address phase, and include the following:
The Special Cycle command provides a simple message broadcast mechanism on PCI. It is designed to be used as an alternative to physical signals when sideband communication is necessary.
The I/O Read command is used to read data from an agent mapped in I/O address space. AD31::00! provide a byte address. All 32 bits must be decoded. The Byte Enables indicate the size of the transfer and must be consistent with the byte address.
The I/O Write command is used to write data to an agent mapped in I/O address space. All 32 bits must be decoded The Byte Enables indicate the size of the transfer and must be consistent with the byte address.
The Memory Read command is used to read data from an agent mapped in the memory address space. The target is free to do an anticipatory read for this command only if it can guarantee that such a read will have no side effects. Furthermore, the target must ensure the coherency (which includes ordering) of any data retained in temporary buffers after this PCI transaction is completed. Such buffers must be invalidated before any synchronization events (e.g. updating an I/O status register or memory flag) are passed through this access path.
The Memory Write command is used to write data to an agent mapped in the memory address space. When the target returns "ready", it has assumed responsibility for the coherency (which includes ordering) of the subject data. This can be done either by implementing this command in a fully synchronous manner, or by insuring any software transparent posting buffer will be flushed before synchronization events (e.g., updating an I/O status register or memory flag) are passed through this access path. This implies that the master is free to create a synchronization event immediately after using this command.
The Configuration Read command is used to read the configuration space of each agent. An agent is selected when its IDSEL signal is asserted and AD1::0! are 00. During the address phase of a configuration cycle, AD7::2! address one of the 64 DWORD registers (where byte enables address the byte(s) within each DWORD) configuration space of each device and AD31::11! are logical don't cares. AD10::08! indicate which device of a multi-function agent is being addressed.
The Configuration write command is used to transfer data to the configuration space of each agent. An agent is selected when its IDSEL signal is asserted and AD1::0! are 00. During the address phase of a configuration cycle, the AD7::2! lines address the 64 DWORD (where byte enables address the byte(s) within each DWORD) configuration space of each device and AD31::11! are logical don't cares. AD10::08! indicate which device of a multi-function agent is being addressed.
The Memory Read Multiple command is semantically identical to the Memory Read command except that it additionally indicates that the master may intend to fetch more than one cache line before disconnecting. The memory controller should continue pipelining memory requests as long as FRAME# is asserted. This command is intended to be used with bulk sequential data transfers where the memory system (and the requesting master) might gain some performance advantage by sequentially reading ahead an additional cache line when a software transparent buffer is available for temporary storage.
The Dual Address Cycle (DAC) command is used to transfer a 64-bit address to devices that support 64-bit addressing. Targets that support only 32-bit addresses must treat this command as reserved and not respond to the current transaction in any way.
The Memory Read Line command is semantically identical to the Memory Read command except that it additionally indicates that the master intends to complete more than two 32-bit PCI data phases. This command is intended to be used with bulk sequential data transfers where the memory system (and the requesting master) might gain some performance advantage by reading up to a cache line boundary in response to the request rather than a single memory cycle. As with the Memory Read command, pre-fetched buffers must be invalidated before any synchronization events are passed through this access path.
The Memory Write and Invalidate command is semantically identical to the Memory Write command except that it additionally guarantees a minimum transfer of one complete cache line; i.e. the master intends to write all bytes within the addressed cache line in a single PCI transaction. The master may allow the transaction to cross a cache line boundary only if it intends to transfer the entire next line also. This command requires implementation of a configuration register in the master indicating the cache line size. It allows a memory performance optimization by invalidating a "dirty" line in a write-back cache without requiring the actual write-back cycle, thus shortening access time.
PCI Protocol Fundamentals
The basic bus transfer mechanism on PCI is a burst. A burst is composed of an address phase and one or more data phases. PCI supports bursts in both memory and I/O address spaces. The host bridge (that resides between the host processor and PCI) may merge (or assemble) memory write accesses into a single transaction when no side effects exist. A device indicates no side effects (allow prefetching of read data and merging of write data in any order) by setting the prefetch bit in the base address register. A bridge may distinguish where merging is allowed and where it is not, by an address range which could be provided by configuration software during initialization. Merging of data into that buffer must stop (and the buffer flushed) when a subsequent write occurs that is not prefetchable or a read (to any range). Write transactions following either of these two events may be merged with subsequent writes, but not to previously merged data, if in the prefetchable range.
Since I/O accesses from the processor cannot be combined, they will normally only have a single data phase. However, it will not be precluded. There is no implied addressing on I/O bursts. When I/O bursts are done, the target and master must understand the implied addressing. PCI devices that do not deal with multiple I/O data phases must disconnect the access after the first data phase. To ensure that I/O devices will operate correctly, bridges may never merge or combine sequential I/O accesses into a single PCI access or burst. All I/O accesses must appear on PCI exactly as the host generated them. (If a target of an I/O access is selected by its address but the byte enables indicate a transfer larger than the device supports, the target terminates with target-abort.)
All signals are sampled on the rising edge of the clock, except RST#, INTA#, INTB#, INTC#, and INTD#. Each signal has a setup and hold aperture with respect to the rising clock edge, in which transitions are not allowed. Outside this aperture, signal values or transitions have no significance. This aperture occurs only on "qualified" clock edges for AD31::0!, AD63::32!, PAR2, PAR64, and IDSEL signals and on every clock edge for LOCK#, IRDY#, TRDY#, FRAME# DEVSEL#, STOP#, REQ#, GNT#, REQ64#, ACK64#, SBO#, SDONE, SERR# (on falling edge only), and PERR#. C/BE3::0!#, C/BE7::4!# (as bus commands) are qualified on the clock edge that FRAME# is first asserted. C/BE3::0!#, C/BE7::4!# (as byte enables) are qualified on each rising clock edge following the completion of an address phase or data phase. RST#, IRQA#, IRQB#, IRQC#, and IRQD# are not qualified or synchronous.
Basic Transfer Control
The fundamentals of all PCI data transfers are controlled with three signals. Refer to FIG. A1.
The interface is IDLE when both FRAMES and IRDY# are deasserted. The first clock edge on which FRAME# is asserted is the address phase, and the address and bus command code are transferred on that clock edge. The next clock edge begins the first of one or more data phases, during which data is transferred between master and target on each clock edge for which both IRDY# and TRDY# are asserted. Wait cycles may be inserted in a data phase by either the master or the target with IRDY# and TRDY# signals respectively.
The source of the data is required to assert its xRDY# signal unconditionally when data is valid (IRDY# on a write transaction, TRDY# on a read transaction). The receiving agent may assert its xRDY# as it chooses.
Once a master has asserted IRDY# it cannot change IRDY# or FRAMES# until the current data phase completes regardless of the state of TRDY#. Once a target has asserted TRDY# or STOP# it cannot change DEVSEL#, TRDY# or STOP# until the current data phase completes. Neither the master nor the target can change its mind once it has committed to the data transfer.
At such time as the master intends to complete only one more data transfer (which could be immediately after the address phase), FRAME# is deasserted and IRDY# is asserted indicating the master is ready. After the target indicates the final data transfer (TRDY# is asserted), the interface returns to the IDLE state with both FRAME# and IRDY# deasserted.
Addressing
PCI defines three physical address spaces. The memory and I/O address spaces are customary. The configuration address space has been defined to support PCI hardware configuration.
Each agent is responsible for its own address decode. PCI supports two styles of address decoding: positive and subtractive. Positive decoding is faster since each device is looking for accesses in the address range(s) that it has been assigned. Subtractive decoding can be implemented by only one device on the bus, since it accepts all accesses not positively decoded by some other agent. This decode mechanism is slower since it must give an other bus agents a "first right of refusal" on the access. However, it is very useful for an agent like a standard expansion bus that must respond to address phase or data phase. RST#, IRQA#, IRQB#, IRQC#, and IRQD# are not qualified or synchronous.
The information contained in the two low order address bits (AD1::0!) varies by address space. In the I/O address space, all 32 AD lines are used to provide a full byte address. This allows an agent requiring byte level address resolution to complete address decode and claim the cycle without waiting an extra cycle for the byte enables (thus delaying all subtractive decode cycles by an extra clock). AD1::0! are used for the generation of DEVSEL# only and indicate the least significant valid byte involved in the transfer. For example, if BEO# were asserted then AD1::0! would be "00"; if only BE3# were asserted, then AD1::0! would be "11". Once a target has claimed an I/O access (using AD1::0!), it then determines if it can complete the entire access as indicated in the byte enables. If all the selected bytes are not in the selected target's address range, the entire access cannot be completed. In this case, the target does not transfer any data but terminates with a target-abort.
Bus Driving and Turnaround
A turnaround cycle is required on all signals that may be driven by more than one agent. This is indicated on the timing diagrams as two arrows pointing at each others' tail. This turnaround cycle occurs at different times for different signals. For instance, IRDY#, TRDY#, DEVSEL#, STOP#, and ACK64# use the address phase as their turnaround-cycle. FRAME#, REQ64#, C/BE3::0!#, C/BE7::4!, AD31::00!, and AD63::32! use the IDLE cycle between transactions as their turnaround cycle. An IDLE cycle is when both FRAME# and IRDY# are deasserted (e.g., clock 9 in FIG. B1).
Read Transaction
FIG. B1 illustrates a read transaction and starts with an address phase which occurs when FRAME# is asserted for the first time and occurs on clock 2. During the address phase AD31::00! contain a valid address and C/BE3::0!# contain a valid bus command.
The first clock of the first data phase is clock 3. During the data phase C/BE# indicate which byte lanes are involved in the current data phase. A data phase may consist of a data transfer and wait cycles. The C/BE# output buffers must remain enabled (for both read and writes) from the first clock of the data phase through the end of the transaction. This ensures C/BE# are not left floating for long intervals.
The first data phase on a read transaction requires a turnaround-cycle (enforced by the target via TRDY#). in this case the address is valid on clock 2 and then the master stops driving AD. The earliest the target can provide valid data is clock 4. The target must drive the AD lines following the turnaround cycle when DEVSEL# is asserted. Once enabled, the output buffers must stay enabled through the end of the transaction. (This ensures AD are not left floating for long intervals.)
A data phase completes when data is transferred, which occurs when both IRDY# and TRDY# are asserted on the same clock edge. (TRDY# cannot be driven until DEVSEL# is asserted.). When either is deasserted a wait cycle is inserted and no data is transferred. As noted in the diagram, data is successfully transferred on clocks 4, 6, and 8, and wait cycles are inserted on clocks 3, 5, and 7.
The master knows at clock 7 that the next data phase is the last. However, because the master is not ready to complete the last transfer (IRDY# is deasserted on clock 7), FRAME# stays asserted. Only when IRDY# is asserted can FRAME# be deasserted, which occurs on clock 8.
Write Transaction
FIG. B2 illustrates a write transaction. The transaction starts when FRAME# is asserted for the first time which occurs on clock 2. A write transaction is similar to a read transaction except no turnaround cycle is required following the address phase because the master provides both address and data. Data phases work the same for both read and write transactions.
Transaction Termination
Termination of a PCI transaction may be initiated by either the master or the target. While neither can actually stop the transaction unilaterally, the master remains in ultimate control, bringing all transactions to an orderly and systematic conclusion regardless of what caused the termination. All transactions are concluded when FRAME# and IRDY# are both deasserted, indicating an IDLE cycle (e.g., clock 9 in FIG. B2).
Master Initiated Termination
The mechanism used in master initiated termination is when FRAME# is deasserted and IRDY# is asserted. This signals the target that the final data phase is in progress. The final data transfer occurs when both IRDY# and TRDY# are asserted. The transaction reaches completion when both FRAME# and IRDY# are deasserted (IDLE bus condition).
A modified version of master-initiated termination mechanism allows the master to terminate the transaction when no target responds. This abnormal termination is referred to as master-abort. Although it may cause a fatal error for the application originally requesting the transaction, the transaction completes gracefully, thus preserving normal PCI operation for other agents.
Two examples of normal completion are shown in FIG. B3. The final data transfer is indicated when FRAME# is deasserted and when both IRDY# and TRDY# are asserted which occurs at clock 3. The bus reaches an IDLE condition when IRDY# is deasserted which occurs on clock 4. Because the transaction has completed, TRDY# is deasserted on clock 4 also. Note that TRDY# is not required to be asserted on clock 3, but could have delayed the final data transfer (and transaction termination) until it is ready by delaying the final assertion of TRDY#. If the target does that, the master is required to keep IRDY# asserted until the final data transfer occurs.
Both sides of FIG. B3 could have been caused by a timeout termination. On the left side. FRAME# is deasserted on clock 3 because the timer expires, GNT# is deasserted and the master is ready (IRDY# asserted) for the final transfer. Because GNT# was deasserted when the timer expired continued use of the bus is not allowed except when using the Memory Write and Invalidate command. which must be stopped at the cache line boundary. Termination then proceeds as normal. If TRDY# is deasserted on clock 2, that data phase continues until TRDY# is asserted. FRAME# and IRDY# must remain asserted until the data phase completes.
The right-hand example shows a timer expiring on clock 1. Because the master is not ready to transfer data (IRDY# is deasserted on clock 2) FRAME# is required to stay asserted. FRAME# is deasserted on clock 3 because the master is ready (IRDY# is asserted) to complete the transaction on clock 3. The master must be driving valid data (write) or be capable of receiving data (read) whenever IRDY# is asserted. This delay in termination should not be extended more than 2 or 3 cycles. Also note that the transaction need not be terminated after timer expiration unless GNT# is deasserted.
Master-abort termination, as shown in FIG. B4, is an abnormal case (except for configuration or Special Cycle commands) of master initiated termination. A master determines that there will be no response to a transaction if DEVSEL# remains deasserted on clock 6. The master must assume that the target of the access is incapable of dealing with the requested transaction or that the address was bad. Once the master has detected the missing DEVSEL# (clock 6 in this example), FRAME# is deasserted on clock 7 and IRDY# on clock 8. The earliest a master can terminate a transaction with master-abort is five clocks after FRAME# was first sampled asserted, which occurs when the master attempts a single data transfer. However, the master may take longer to deassert FRAME# and terminate the access. The master must support the FRAME#-IRDY# relationship on all transactions which includes master-abort. FRAME# cannot be deasserted before IRDY# is asserted and IRDY# must remain asserted for at least one clock after FRAME# is deasserted, even when the transaction is terminated with master-abort.
Alteratively, IRDY# could be deasserted on clock 7, if FRAME# was deasserted as in the case of a transaction with a single data phase. The master will normally not retry this access. Note that if DEVSEL# had been asserted on clocks 3. 4, 5, or 6 of this example, it would indicate the request had been acknowledged by an agent, and master-abort termination would not be permissible.
The host bus bridge, in PC compatible systems, must return all l's on a read transaction and discard data on a write transaction when terminated with master-abort. The bridge is required to set the master-abort detected bit in the status register. Other master devices may report this condition as an error by signaling SERR# when the master cannot report the error through its device driver.
In summary, the following general rules govern FRAME# and IRDY# in all PCI transactions.
1. FRAME# and its corresponding IRDY# define the busy/IDLE state of the bus; when either is asserted, the bus is busy; when both are deasserted. the bus is IDLE.
2. Once FRAME# has been deasserted, it cannot be reasserted during the same transaction.
3. FRAME# cannot be deasserted unless IRDY# is asserted. (IRDY# must always be asserted on the first clock edge that FRAME# is deasserted.)
4. Once a master has asserted IRDY#, it cannot change IRDY# or FRAME# until the current data phase completes.
Target Initiated Termination
The mechanism used in target initiated termination is the STOP# signal. The target asserts STOP# to request that the master terminate the transaction. A modified version of this mechanism allows the target to terminate a transaction in which a fatal error has occurred, or to which the target will never be able to respond. This abnormal termination is referred to as target-abort.
In summary, the following general rules govern FRAME#, IRDY#, TRDY#, and STOP# in all PCI transactions.
Whenever STOP# is asserted, FRAME# must be deasserted as soon as possible pursuant to the rules for the deassertion of FRAME# (i.e., IRDY# must be asserted). The deassertion of FRAME# should occur as soon after STOP# is asserted as possible, preferably within two or three cycles. The target must not assume any timing relationship between STOP# assertion and FRAME# deassertion, but must Keep STOP# asserted until FRAME# is deasserted. When the master samples STOP# asserted, it must deassert FRAME# on the first cycle thereafter in which IRDY# is asserted. This assertion of IRDY# (and therefore FRAME# deassertion) may occur as a consequence of the normal IRDY# behavior of the master (had the Current transaction not been target terminated), and be delayed zero or more cycles depending on when the master is prepared to complete a data transfer. Alteratively, the master may assert IRDY# immediately (even without being prepared to complete a data transfer) if TRDY# is deasserted, thus indicating there will be no further data transfer.
Once asserted, STOP# must remain asserted until FRAME# is deasserted, whereupon, STOP# must be deasserted.
During the final data phase of a transaction (FRAME# deasserted and IRDY# asserted), any clock edge on which either STOP# or TRDY# is asserted becomes the last cycle of the transaction, and IRDY# is deasserted on the following clock edge. (This creates an IDLE cycle and defines the end of the transaction.)
The master must retry an access that was target terminated (except target-abort) with the address of the next untransferred data if it intends to complete the access. If the deice was prefetching, it may elect not to retry the access.
Once a target has asserted TRDY# or STOP#. it cannot change DEVSEL#, TRDY# or STOP# until the current data phase completes.
Device Selection
DEVSEL# is driven by the target of the current transaction as shown in FIG. B5. DEVSEL# may be driven one, two or three clocks following the address phase. Each PCI-bus device supports a Device Status register which indicates, among other things, the slowest time (fast, medium or slow) in which it will assert DEVSEL# for any bus command except Configuration Read and Configuration Write. DEVSEL# must be asserted with or prior to the edge at which the target enables its TRDY#, STOP#, or data (read). In other words, a target must assert DEVSEL# (claim the transaction) before it is allowed to issue any other target response. In all cases except one, once a target asserts DEVSEL# it must not deassert DEVSEL# until FRAME# is deasserted (IRDY# is asserted) and the last data phase has completed. With normal master termination, DEVSEL# deassertion must be coincident with the deassertion of TRDY#. The exception is the target-abort.
If no agent asserts DEVSEL# within three clocks of FRAME#, the agent doing subtractive decode may claim and assert DEVSEL#. If the system does not have a subtractive decode agent, the master never sees DEVSEL# asserted and terminates the transaction per the master-abort mechanism.
A target must do a full decode before driving/asserting DEVSEL#, or any other target response signal It is illegal to drive DEVSEL# prior to a complete decode and then let the decode combinationally resolve on the bus. (This could cause contention.) A target must qualify the AD lines with FRAME# before DEVSEL# can be asserted on commands other than configuration. A target must qualify IDSEL with FRAME# and AD1::0! before DEVSEL# can be asserted on a Configuration command.
It's expected that most (perhaps all) target devices will be able to complete a decode and assert DEVSEL# within one or two clocks of FRAME# being asserted (fast and medium in the figure). Accordingly, the subtractive decode agent may provide an optional device dependent configuration register that can be programmed to pull in by one or two clocks the edge at-which it samples DEVSEL#, allowing faster access to the expansion bus. Use of such an option is limited by the slowest positive decode agent on the bus. The edge at which it samples DEVSEL# is referred to herein as the PCI-bus device claiming deadline.
Once DEVSEL# has been asserted. it cannot be deasserted until the last data phase has completed, except to signal target-abort.
If the first access maps into the target's address range, it asserts DEVSEL# to claim the access. But if the master attempts to continue the burst access across the resource boundary, the target is required to signal disconnect.
When a target claims an I/O access, and the byte enables indicate one or more bytes of the access are outside the target's address range, it must signal target-abort. To deal with this type of I/O access problem, a subtractive decode device (expansion bus bridge) may do one of the following: (1) do positive decode (by including a byte map) on addresses for which different devices share common DWORDs, additionally using byte enables to detect this problem and signal target-abort; or (2) pass the full access to the expansion bus, where the portion of the access that cannot be serviced will quietly drop on the floor. (This occurs only when first addressed target resides on the expansion bus and the other is on PCI.)
APPEND)IX C OPTi.RTM. 82C802
82C802G OVERVIEW
The OPTi.RTM. 82C802G provides a highly integrated solution for fully compatible, high performance PC/AT platforms. This chipset will support 486SX/DX/DX2 and P24T microprocessors in the most cost effective and power efficient designs available today. For power users, this chipset offers optimum performance for systems running up to 50 MHz.
Based fundamentally on the proven 82C801 and 82C802 design architectures, the 82C802G adds additional memory configurations and power management for the Processor and other motherboard components. As an upgrade from the OPTi 82C802, the 82C802G adds extensive power management control.
The 82C802G supports the latest in Writeback Processor designs from Intel, AMD and Cyrix, as well as supporting the AT bus and VESA local bus for compatibility and performance. It also includes an integrated '206 IPC controller, all in a single 208-pin PFP package for low cost.
FUNCTIONAL DESCRIPTION
System Clock Generation
The 82C802G has a single high frequency clock input, CLK. CLK is a master single phase clock which is used to drive all host CPU synchronous signals and all of the 82C802G's internal state machines. This clocking scheme provides operation to support platforms at system speeds up to 50 MHz.
The 82C802G generates the AT bus clock (ATCLK) from an internal division of CLK. The ATCLK frequency is programmable and can be set to any of four synchronous mode clock division options by programming Register 25h1:0!. In addition to the CLK source, there is an asynchronous mode available by clearing Register 27h bit 1, which generates ATCLK by dividing the ATBUS OSC oscillator by 2 (OSC/2). This asynchronous mode is important when entering GREEN mode, where the CPU clock rate can change and thereby generate unsuitable ATCLK frequencies if left in synchronous mode. This allows the system designer to tailor the AT bus clock frequency to support a wide range of system designs and performance platforms, as well as to function reliably during power savings mode.
CPU Burst-Mode Control
The 82C802G chipset fully supports 486 burst cycles. The 82C802G cache and DRAM controllers insure that data is burst into the CPU whenever the 486 requests a burst line fill. The secondary cache provides data on read-hits and the DRAM supplies the data during cache read-misses.
For the cache read-hit cycle, BRDY# (Burst Ready) is asserted during the first T2 state when a 2-1-1-1 (zero wait state) cache burst cycle is chosen, otherwise it is asserted during the second T2 state when one wait state is required. If a read-miss occurs, the DRAM controller will burst new data into both the cache memory and CPU simultaneously. BRDY# will be asserted for each double-word during these cache read-miss update cycles. For a zero wait state cache burst read cycle, 2-11-1, BRDY# will be asserted during the first T2 and remain active until BLST# (Burst Last) from the CPU is detected. For a 3-2-2-2 cache burst read cycle, BRDY# will be toggled active on an every-other-clock basis to accommodate the more relaxed data access timing required above 33 MHz operation. BRDY# is never active during DMA or MASTER cycles.
The 82C802G contains separate burst counters to support DRAM and external cache burst cycles. The Read/Write DRAM burst counter performs the cache read-miss line fill (DRAM to external cache/CPU) and the cache burst counter supports the 486 burst line fill (external cache to the 486 CPU). The access order of the burst counter exactly matches the double-word address sequencing expected by the 486 CPU. The DRAM burst counter is used for cache read-miss cycles and dirty line fill write operations.
L1 Writeback Timing Description
Level 1 Write Back Support:
The L1 cache can contain modified data that is not contained in the L2 cache or DRAM. The CPU will not allow external devices to access its internal cache. The 82C802G will execute an inquire cycle to the L1 cache for all master accesses to the system memory area. Master devices, whether local or on the ISA bus must snoop the L1 cache during every access to system memory. If valid information is in the L1 cache and this information has been modified without being updated to the system memory, the HITM# signal will be generated. A write back cycle must be generated whenever a modified line was hit.
VESA Local Master Cycles
The L1 Cache Inquire Cycle begins with the CPU relinquishing the bus with the assertion of HLDA on sampling HLDA, the local bus card will generate ADS#. EADS# will be generated by the 82C802G for one clock following the ADS# generation. If the CPU does not respond with assertion of HITM#, the 82C802G will complete the cycle from the L2 cache or the system memory. If HITM# is asserted, the 82C802G will expect a castout cycle from the L1 cache. HITM# is connected to the WBACK# signal on the VL bus which will abort the VL cycle and allow the CPU to perform its castout cycle. The 82C802G will release hold to the CPU and generate a RDY to terminate the local bus cycle. Next, the CPU will write back its L1 contents to cache/system memory.
Master/DMA Write Cycle
HOLD will then be generated to the CPU in response to an ISA master or DMA cycle. The CPU will relinquish the bus with the assertion of HLDA. The 82C802G will issue AHOLD to the CPU to tristate the CPU's address bus. At this time, the DMA or master device drives the address onto the CPU bus. IOCHRDY# will be released. EADS# is generated by the 82C802G. HITM# will be generated if the address is a modified line in the cache. The CPU will then perform its castout cycle always starting at the address 0X0 of the 16 byte line. After the castout cycle, the CPU will deassert HITM# and issue HLDA. The ISA master or DMA device can then finish its cycle.
Cache Subsystem
The integrated cache controller, which uses a direct-mapped, bank-interleaved scheme dramatically boosts the overall performance of the local memory subsystem by caching writes as well as reads (write-back mode). Cache memory can be configured as one or two banks, and sizes of 64K, 128K, 256K, and 512K are supported. Provisions for two programmable non-cacheable regions are provided. The cache controller operates in non-pipeline mode, with a fixed 16-byte line size (optimized to match a 486 burst line fill) in order to simplify the motherboard design without increasing cost or degrading system performance. For 486 systems, the secondary cache operates independently and in addition to the CPU's internal cache.
Local DRAM Control Subsystem
The 82C802G supports up to 8 banks of page-mode local DRAM memory for configurations of up to 64 Mb. 256 Kb, 1 Mb, 4 Mb or 16 Mb page-mode DRAM devices may be used. The DRAM configuration is programmable through configuration register 24h. DRAM performance features are programmable through configuration register 25h.
AT-Bus State Machine
The AT Bus state machine gains control when the 82C802G's decoding logic detects a non-local memory cycle. It monitors status signals M16#,IO16#, CHRDY and NOWS# and performs the necessary synchronization of control and status signals between the AT Bus and the microprocessor. The 82C802G supports 8 and 16 bit memory and I/O devices located on the AT bus.
An AT bus cycle is initiated by asserting ALE in AT-TS I state. On the trailing edge of ALE, M 16# is sampled for a memory cycle to determine the bus size. It then enters AT-TC state and provides the command signal. For an I/O cycle, IO16# is sampled after the trailing edge of ALE until the end of the command.
Typically, the wait state for an AT 8/16 bit transaction is 5/1 respectively. The command cycle is extended when CHRDY is detected inactive, or the cycle is terminated when zero wait state request signal (NOWS#) from the AT bus is active. Upon expiration of the wait states, the AT State Machine terminates itself and passes internal READY to the CPU State Machine for outputting synchronous RDY# to the CPU. Index register 20h/bit 2 allows for the addition of an AT cycle wait state; bit 3 of this same register allows for the generation of a single ALE instead of multiple ALEs during bus conversion cycles. The AT bus state machine also routes data and address when an AT bus master or DMA controller accesses memory.
Bus Arbitration Logic
The 82C802G provides arbitration between the CPU, DMA controller, AT Bus masters, and the refresh logic. During DMA, AT bus master, and conventional refresh cycles, the 82C802G asserts HOLD to the CPU. The CPU will respond to an active HOLD signal by generating HLDA (after completing its current bus cycle) and placing most of its output and I/O pins in a high impedance state. After the CPU relinquishes the bus, the 82C802G responds by issuing RFSH# (refresh cycle) or generating the appropriate DRQ (AT bus master or DMA cycle), depending on the requesting device. During hidden refresh, HOLD remains negated and the CPU continues it's current program execution as long as it services internal requests or achieves cache hits (refer to the refresh section for additional information). The AT bus controller in the 82C802G arbitrates between DMA/Master and refresh requests, deciding which will own the bus once the CPU relinquishes control with the HLDA signal. The arbitration between refresh and DMA/Master is based on, FIFO (first in-first out) priority. However, a refresh request (RFSH#) will be internally latched and serviced immediately after the DMA or master finishes its term if queued after. DRQ's must remain active to be serviced if a refresh request come first. The "MASTER#" signal from the AT bus indicates an active AT bus master cycle.
Local Bus Interface
The 82C802G allows peripheral devices to share the "local bus" with the CPU. The performance of these devices (which may include the video subsystem, hard disk adapters, LAN and other PC/AT controllers) will dramatically increase when allowed to operate in this high-speed environment. These devices are responsible for their own address and bus cycle decode and must be able to operate compatibly at the elevated frequencies required for operation on the local CPU bus.
The LDEV# input signal to the 82C802G indicates that a local device is intercepting the current cycle. If this signal is sampled at the end of the first T2 clock cycle (end of the second T2 at 50 Mhz), then the 82C802G will allow the responding local device to assume responsibility for the current local cycle. When the device has completed its operation, it must terminate the cycle by asserting RDY# or BRDY# to the CPU. RDY# and BRDY# are bi-directional pins on the 82C802G and may be driven by a local bus peripheral or the chipset to terminate their respective cycles.
Data Bus Conversion/Data Path Control Logic
The 82C802G performs data bus conversion when the CPU accesses 16 or 8 bit devices through 16 or 32 bit instructions. lt also handles DMA and AT master cycles that transfer data between local DRAM or cache memory and locations on the AT bus. The 82C802G provides all of the signals to control external bi-directional data buffers.
Special Cycles
The 486 microprocessor provides special bus cycles to indicate that certain instructions have been executed, or certain conditions have occurred internally. Special cycles such as SHUTDOWN and HALT cycles are covered by dedicated handling logic in the 82C802G. This logic decodes the CPU bus status signals M/IO#, D/C# and W/R# and executes the appropriate action.
82C802G Register Description
The 82C802G emulates the Port 60h and 64h registers of the keyboard controller, allowing the gent A20 signal. The sequence here is BIOS transparent, and there is no need for the modification of thee sequence involves writing data Dlh to port 64h, then writing data 02h to port 60h.
82C802G I/O port 64h emulates the register inside the keyboard controller by generating a fast reset pulse. Writing data F1 to port 64h asserts the reset pulse. The pulse is generated immediately after the I/O write if bit 1 of Index register 20h is set.
This is a read/write register which is used to store configuration information.
APPENDIX D OPTi.RTM. 82C602
82C601/82C602 OVERVIEW
Introduction
The OPTi.RTM. 82C601 and 82C602 buffer chips are 100-pin devices designed to replace a number of discrete buffer components normally found on PC/AT motherboards. Pin compatible with each other, the 82C602 is an upgrade from the 82C601 with additional functionality. These chips replaces approximately 9 TTL components on a typical motherboard design, offering significant cost savings and a substantially simplified PCB layout.
The 82C601/82C602 are closely coupled with OPTi's 82CS01/802/802G, 82C499 and 82C495SLC/XLC chipsets to allow higher integration and lower cost.
The 82C602 is a pin-compatible upgrade to the 82C601 that adds Real Time Clock (RTC) and Power Management functionality.
82C601/82C602 Functional Description
SD0-7 to XD0-7 Data Bus Control
The 82C80X, 82C495SLC/XLC, and 82C499 families drive the SD bus directly. The 82C601 provides a bi directional buffer which controls the direction of the XD to SD bus. The XD bus is designed for devices which are not intended to drive the SD bus. These items include the RTC, Keyboard Controller and BIOS ROM. The direction of the internal buffer is controlled by the system logic chipset signal XDIR.
CD (23:16) to SD(7:0) and/or CD(31:24) to SD(15:8) Data Bus Buffering
On OPTi 82C80X and 82C499 chipsets, the third data byte is sourced from the CPU, and so is not output by the chipset. In order to align the byte properly for the 16 bit ISA bus, control signals are generated by the chipset. The control for this alignment is done by HLBOE1#, HLBOE2# and HLBTH#. On the 82C495SLC/XLC chipsets, the third and fourth byte is sourced from the CPU and not output by the chipset. This is controlled by the HLHDEN#, HDLEN# and HDDIR# of the 82C495SLC/XLC.
CA(15:2) to SA(15:2) Address Buffering
The 82C601 buffers all of the local bus address lines to the ISA bus. These address lines normally drive the ISA bus, and are inputs from the host during ISA master operations. During refresh, these buffers are disabled, allowing the refresh address to be broadcast to the ISA bus.
Refresh Counter and Refresh Address Logic
The 82C601 has its own built-in refresh counter. This refresh counter will broadcast the refresh address to the ISA bus during refresh cycles.
AEN Decode Signal
AEN is decoded by REF, MASTER# and HLDA#. This signal is normally low and is driven high during DMA accesses.
AT Command Buffer Signals MEMR# an MEMW# signals are buffered to the ISA bus.
Real Time Clock Control
Addresses 70h and 71h are decoded by the 82C601. These decoded addresses translate to RTCCS# and RTCAS strobes. This decode is needed for 82C802 designs. On 82C801 designs, the decode is done by the chipset.
Keyboard Clock Buffer
The clock signal from the keyboard controller to the keyboard can be buffered through the 82C601. This clock is tri-stated through the buffer chip and sent out to the keyboard.
Reset Circuit
SYSRST# from the 82C801 is buffered through the 82C601. RSTDRV to the ISA bus and LRESET# to the VESA local bus are generated.
82C602 Functional Description
The 82C602 has all of the functionality of the 82C601 buffer chip, adding an embedded RTC and GREEN power port.
Internal
RTC The internal RTC of the 82C602 is functionally compatible with the DS 1285/MC146818B. Additional features to the RTC are 242 bytes of general nonvolatile storage and an internal power sense monitor.
GREEN Power Port
The GREEN port on the 82C602, when coupled with the 82C802G, provides a means of controlling devices via output pins on the device. During the power saving mode, this port can be written to by the system software or can latch data from the 82C802G's port.
APPENDIX E OPTi.RTM. 82C822
2.0 Overview
OPTi.RTM.'s 82C822 VESA local bus to PCI Bridge (PCIB) chip is a high integration 208-pin PQFP device designed to work with VESA VL bus compatible core logic chipsets. The 82C822 PCIB provides interface to the high performance PCI bus and is fully compliant to the PCI Version 2.0 specification. The 82C822 requires no glue logic to implement the PCI bus interface and hence it allows designers to have a highly integrated motherboard with both VESA local bus and PCI local bus support. The PCIB chip offers premium performance and flexibility for VESA VL-based desktop systems running up to 50 MHz. The 82C822 PCIB can be used with OPTi's 82C802G core logic and 82C602 buffer chipsets to build a low cost and power efficient 486-based desktop solution It also works with OPTi 82C546/547 chipset to build a high performance PCI/VL solution based on the Intel P54C (Pentium) processor.
The 82C822 PCIB provides all of the control address and data paths to access the PCI bus from the VESA Local bus (VL bus). The 82C822 provides a complete solution including data buffering, latching, steering, arbitration, DMA and master functions between the 32-bit VL bus and the 32-bit PCI bus.
The PCIB works seamlessly with the motherboard chipset bus arbiter to handle all requests of the host CPU and PCI bus masters, DMA masters, I/O relocation and refresh. Extensive register and timer support are designed into the 82C822 to implement the PCI specification.
The 82C822 is a true VESA to PCI bridge. It has the highest priority on CPU accesses after cache and system memory. It generates LDEV# automatically and then compares the addresses with its internal registers to determine whether the current cycle is a PCI cycle. When a cycle is identified as PCI cycle, the 82C822 will take over the cycle and then return RDY# to the CPU. If not, the 82C822 will give up the cycle to the local device or, in the case of an ISA slave, generate a BOFF# cycle to the CPU. This action will abort the cycle and allow the CPU to rerun the cycle.
The 82C822 includes registers to determine shadow memory space, hole locations and sizes to allow the 82C822 to determine which memory space should be local and which is located on the ISA bus. Upon access to memory, the 82C822 can determine whether or not the cycle is a PCI access by comparing the cycle with its internal registers.
4.0 Functional Description
The following sub-sections will explain the various cycle operations the 82C822 PCIB performs to determine which memory space should be local and which is located on the ISA bus.
4.1 CPU Cycles
Cache/DRAM Accesses
The CPU begins the cycle. The 82C822 asserts LDEVO# on the next clock in response to LADS# (VESA local bus ADS#). In the OPTi Pentium chipsets, LADS# will not be generated for a local memory cycle and therefore, LDEVO# will not be generated by the 82C822. In the OPTi 486 chipsets, the CPU will generate LADS# (LADS# and CPU ADS# are the same in the 486 implementation) . The 82C822 will generate LDEVO#, but this will be ignored by the OPTi 486 chipsets during a local memory cycle. In this case, the 82C822 will deassert LDEVO# after the last BRDY#. The 82C822 is also programmable to mask LDEVO# during a local memory cycle.
VESA Slave Accesses
The CPU begins the cycle. The 82C822 asserts LDEVO# on the next clock in response to ADS#. The VESA slave will generate LDEV# to the 82C822 at the end of the first T2 or second T2 (depending on the VL speed) . At this point, the 82C822 will keep asserting LDEVO# until the VESA slave returns LRDY#/LBRDY#.
PCI Slave Accesses
The CPU begins the cycle. The 82C822 asserts LDEVO# on the next clock. It then determines that the cycle is a PCI slave cycle and starts the PCI cycle after synchronization. Data is latched into the 82C822 for read cycles. The 82C822 completes the cycle, returns LRDY#, and deasserts LDEVO#.
ISA Slave Accesses
The CPU begins the cycle. The 82C822 asserts LDEVO# on the next clock. It then determines that the cycle is not a local memory or VESA slave cycle and starts the PCI cycle after synchronization. The 82C822 receives master-abort termination and asserts BOFF# after synchronization. It then asserts LRDY# either at the same time or one or two clock cycles thereafter, programmably. The CPU restarts the aborted cycle. The 82C822 ignores the restarted cycle by keeping LDEVO# inactive. The OPTi VL chipset returns RDY# to the CPU at the end of the AT cycle.
4.2 VESA Master Cycles
Cache/DRAM Accesses
The VL bus master begins the cycle. The 82C822 asserts LDEVO# on the next clock. The OPTi VL chipset determines that the cycle is a local memory cycle and ignores LDEVO#. (The 82C822 is programmable to mask LDEVO# as well on local memory cycles). The OPTi VL chipset will then return BRDY# after the data is transferred. The 82C822 also determines that the cycle is a local memory cycle and takes no action other than to deassert LDEVO# after the last BRDY#.
VESA Slave Accesses
The VL bus master begins the cycle. The 82C822 asserts LDEVO# on the next clock. It then samples LDEVI# and, determining that the cycle is a VESA cycle, takes no further action. The VESA slave will then return LRDY#/LBRDY# after the data is transferred. RDYRTN# will be generated by the OPTi VL chipset. The 82C822 deasserts LDEVO# after the last BRDY#/RDYRTN#.
PCI Slave Accesses
The VL bus master begins the cycle. The 82C822 asserts LDEVO# on the next clock. It then determines that the cycle is a PCI slave cycle and starts the PCI cycle after synchronization. Data is latched into the 82C822 for read cycles. The 82C822 returns LRDY# to the OPTi VL chipset and deasserts LDEVO#.
ISA Slave Accesses
The VL bus master begins the cycle. The 82C822 asserts LDEVO# on the next clock. It then determines that the cycle is not a local memory or VESA slave cycle and starts the PCI cycle after synchronization. The 82C822 receives a master-abort termination and asserts BOFF# after synchronization. It then asserts LRDY# either at the same time or one or two clock cycles thereafter, programmably. The CPU restarts the aborted cycle. The 82C822 ignores the restarted cycle by keeping LDEVO# inactive. The OPTi VL chipset returns RDY# to the CPU at the end of the AT cycle.
4.3 PCI Master Cycles
Cache/DRAM Accesses
The PCI master begins the cycle by asserting PCI REQ# to the 82C822. The 82C822 asserts LREQ# to the OPTi VL chipset after arbitration of the bus requests from the PCI bus masters. The 82C822 receives LGNT# from the OPTi VL chipset and generates the appropriate VL bus signals according to the PCI command, address, and status lines. The OPTi VL chipset determines that the cycle is a local memory cycle. The 82C822 also determines that the cycle is a local memory cycle and asserts DEVSEL# to claim the cycle. Data is latched into the 82C822 for read cycles. The OPTi VL chipset returns RDY# to the 82C822. The 82C822 returns TRDY# to the PCI bus master after synchronization. The 82C822 deasserts DEVSELS after the last transfer.
VESA Slave Accesses
The PCI master begins the cycle by asserting PCI REQ# to the 82C822. The 82C822 asserts LREQ# to the OPTi VL chipset after arbitration of the bus requests from the PCI bus masters. The 82C822 receives LGNT# from the OPTi VL chipset and generates the appropriate VL bus signals according to the PCI command, address, and status lines. It then determines that the cycle is a VESA slave cycle and asserts DEVSEL# to claim the cycle. Data is latched into the 82C822 for read cycles. The OPTi VL chipset returns RDY# to the 82C822. The 82C822 returns TRDY# to the PCI bus master after synchronization. The 82C822 deasserts DEVSEL# after the last transfer.
PCI Slave Accesses
The PCI master begins the cycle by asserting PCI REQ# to the 82C822. The 82C822 asserts LREQ# to the OPTi VL chipset after arbitration of the bus requests from the PCI bus masters. The 82C822 receives LGNT# from the OPTi VL chipset and generates the appropriate VL bus signals according to the PCI command, address, and status lines. It then determines that the cycle is a PCI slave cycle and waits for DEVSEL# from the PCI slave. The 82C822 takes no action.
ISA Slave Accesses
The PCI master begins the cycle by asserting PCI REQ# to the 82C822. The 82C822 assets LREQ# to the OPTi VL chipset after arbitration of the bus requests from the PCI bus masters. The 82C822 receives LGNT# from the OPTi VL chipset and generates the appropriate VL bus signals according to the PCI command, address, and status lines. It then determines that the cycle is a not a local memory or VESA slave cycle and, detecting no DEVSEL# from the PCI slave, realizes the cycle is an ISA slave cycle. The OPTi VL chipset returns LRDY# to the 82C822 at the end of the AT cycle. The data is latched into the 82C822 for the read cycle when it returns RDY# to the OPTi VL chipset. The 82C822 returns TRDY# to the PCI master after synchronization. The 82C822 deasserts DEVSEL# after the data is transferred.
4.4 DMA/ISA Master Cycles
Cache/DRAM Accesses
The OPTi VL chipset determines that the DMA/VISA master cycle is accessing local memory by detecting LMEM# asserted and LDEVI# unasserted at the sample point when the command is asserted. The OPTi VL chipset asserts MEMCS16# if this is an ISA master cycle. LADS# will not be generated by the OPTi VL chipset. No action is taken by the 82c822.
VESA Slave Accesses
The OPTi VL chipset determines that the DMA/ISA master cycle is a not a local memory cycle once the command is asserted. The OPTi VL chipset will generate the VESA local bus command lines. In response to LADS#, the 82C822 will assert LDEVO#. The 82C822 will receive LDEVI# by the next or subsequent clock. The OPTi chipset pulls IOCHRDY low to stall the command. The OPTi VL chipset samples LDEVO# active. If the cycle is a read cycle, the OPTi VL chipset will start driving the SD bus. LRDY# is asserted by the VESA slave. The OPTi VL chipset releases IOCHRDY and asserts RDY# at the end of the command after synchronization. The 82C822 takes no action other than to deassert LDEVO# as soon as LDEVI# goes inactive. The OPTi VL chipset deasserts MEMCS16# when LDEVO# goes inactive.
PCI Slave Accesses
The OPTi VL chipset determines that the DMA/ISA master cycle is a not a local memory cycle once the command is asserted. The OPTi VL chipset will generate the VESA local bus command lines. In response to LADS#, the 82C822 will assert LDEVO#. The OPTi VL chipset pulls IOCHRDY low to stall the command. The OPTi VL chipset samples LDEVO# active and pulls MEMCS16# low if the command is not an I/O cycle. The 82C822 starts the PCI cycle after synchronization, only if the ISA/DMA-to-PCI bit is set. Data is latched into the 82C822 for read cycles when both TRDY# and IRDY# are active. The 82C822 asserts LRDY# after synchronization. The OPTi VL chipset releases IOCHRDY and asserts RDY# at the end of the command after synchronization. The 82C822 deasserts LDEVO# at the end of the bus cycle. The OPTi chipset deasserts MEMCS16# when LDEVO# goes inactive. In the case of a master abort, the 82C822 still asserts LRDY# to release IOCHRDY and drive FFFFFFFFh onto the VL-bus data lines for a read cycle.
ISA Slave Accesses
The DMA/ISA master can access an ISA slave only if the ISA/DMA-to-PCI bit is low. Otherwise the 82C822 will claim this cycle. The OPTi VL chipset determines that the DMA/ISA master cycle is a not a local memory cycle once the command is asserted. The OPTi VL chipset will generate the VESA local bus command lines. In response to LADS#, the 82C822 does not assert LDEVO#. The OPTi VL chipset pulls IOCHRDY low to stall the command. The OPTi VL chipset samples LDEVO# inactive and keeps SD lines tri-stated for a read cycle. It also negates IOCHRDY. The chipset asserts RDY# at the end of command after synchronization. No action is taken by the 82C822.
4.5 Guidelines to Program the 82C822
The following briefly describes how to access the 82C822 and the PCI devices on the slots. The 82C822 uses the PCI Configuration Mechanism #1 to access the configuration spaces. Two double-word I/O locations are used in this mechanism. The first double-word location (CF8h) references a read/write register that is named CONFIG.sub.-- ADDRESS. The second double-word address (CFCh) references a register name CONFIG.sub.-- DATA. The general mechanism for accessing the configuration space is to write a value into CONFIG.sub.-- ADDRESS that specifies the PCI bus, the device on that bus, and the configuration register in that device being accessed. A read or write to CONFIG.sub.-- DATA will then cause the 82C822 to translate that CONFIG.sub.-- ADDRESS value to the requested configuration cycle on the PCI bus.
4.6 Asynchronous and Synchronous Modes of Operation
The 82C822 supports both synchronous and asynchronous modes of operation. In the synchronous mode, the PCI bus interfaces runs at the same speed as the VL bus. In the asynchronous mode, the PCI bus can run at a different speed than the VL bus. This facilitates the VL bus to run at 50 MHz with the PCI running at 33 MHz. The synchronous mode of operation is recommended when the VL bus is operating below 33 MHz. The asynchronous mode of operation is recommended when the VL bus is running at 50 MHz or to implement PCI interface in the riser card.
5.0 Configuration Registers
5.1 PCI Configuration Register Space
Note: In the address offsets given below, the most significant bit (MSB) corresponds to the upper address offset. Registers not pertinent to an understanding of the invention are omitted.
Claims
We claim:
1. A method for bridging a first bus (VL-bus) cycle to a second bus (PCI-bus), in a system including a first bus (VL-bus) having first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#) and a first bus device claiming signal line (LDEVO#), said system further including said second bus (PCI-bus) and a third bus (system bus), said second bus having cycle definition lines (e.g. C/BE#, FRAME#) and a second bus device claiming line (DEVSEL#), for use with a first bus (VL-bus) host (102) defining said first bus cycle by asserting first bus cycle definition signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), said first bus host further having a back-off input line (BOFF#) and repeating said first bus cycle after detecting assertion of a back-off signal on said back-off input line, for use further with a first-bus/third-bus (VL-bus/system-bus) bridge coupled to receive said first bus cycle definition signals, said first-bus/third-bus bridge defining a cycle on said third bus in response to certain combinations of said first bus cycle definition signals only if no first bus device claiming signal (LDEVO#) is asserted before a predetermined first bus device claiming deadline following detection by said first-bus/third-bus bridge of an asserted first bus cycle start signal (ADS#) (e.g. R/W#, M/IO#, D/C#, VA, BE#), said method comprising the steps of: asserting said first bus device claiming signal (LDEVO#) prior to said first bus device claiming deadline, in response to a first assertion of said first bus cycle start signal (ADS#), at the time of which the signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), being a first valid first bus cycle definition, are within a predetermined set of first bus cycle definitions; defining a cycle on said second bus (PCI-bus) in response to said first assertion of said first bus cycle start signal (ADS#) by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#); and asserting said back-off signal (BOFF#) if no second bus device claiming signal (DEVSEL#) is asserted before a predetermined second bus device claiming deadline following said assertion of second bus cycle definition signals on said second bus cycle definition lines.
2. A method according to claim 1, wherein said first bus is a VL-bus, said second bus is a PCI-bus and said third bus is an IBM PC/AT-compatible system bus.
3. A method according to claim 2, wherein said system bus comprises in combination an X-bus and a bus selected from the group consisting of an ISA-bus and an EISA-bus.
4. A method according to claim 2, wherein said first bus cycle definition lines include R/W#, M/IO#, D/C#, VA and BE# signal lines, and wherein said second bus cycle definition lines include C/BE# and FRAME#.
5. A method according to claim 1, wherein said first bus host comprises at least one element selected from the group consisting of a CPU, a cache memory controller and a bridge bridging said first bus (VL-bus) with a fourth bus (host-bus) different from said first, second and third buses.
6. A method according to claim 1, for use with at least one first bus device each coupled to receive said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#) and having a respective first bus device claiming output (LDEV<x>#), further comprising the step of asserting said first bus device claiming signal (LDEVO#) in response to assertion of one of said first bus device claiming outputs (LDEV(x)#).
7. A method according to claim 1, wherein said predetermined set of first bus cycle definitions excludes all first bus cycles which match a programmable indication of first bus cycles intended for devices on said first bus (VL-bus), further comprising, prior to said step of asserting said first bus device claiming signal (LDEVO#), the step of comparing said first valid first bus cycle definition signals with said programmable indication of first bus cycles intended for devices on said first bus.
8. A method according to claim 1, wherein said predetermined set of first bus cycle definitions excludes all first bus cycles which match a programmable indication of first bus cycles intended for devices on said third bus (system bus), further comprising, prior to said step of asserting said first bus device claiming signal (LDEVO#), the step of comparing said first bus cycle definition signals with said programmable indication of first bus cycles intended for devices on said third bus.
9. A method according to claim 1, wherein said predetermined set of first bus cycle definitions excludes all first bus cycles which match a predefined indication of first bus cycles not intended for devices on said second bus (PCI-bus), further comprising, prior to said step of asserting said first bus device claiming signal (LDEVO#), the step of comparing said first valid first bus cycle definition signals with said indication of first bus cycles not intended for devices on said second bus.
10. A method according to claim 1, wherein said first bus host comprises a first-bus/fourth-bus bridge bridging said first bus with a fourth bus (host bus) different from said first, second and third buses, said first-bus/fourth-bus bridge defining a plurality of said first bus (VL-bus) cycles in response to each fourth bus (host-bus) cycle detected by said first-bus/fourth-bus bridge on said fourth bus, further comprising the step of asserting said first bus device claiming signal (LDEVO#) prior to said first bus device claiming deadline, in response to each assertion of said first-bus cycle start signal prior to said first assertion of said cycle start signal, and at the time of which the signals on said first bus (VL-bus) cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), are within a predetermined set of first bus cycle definitions.
11. A method according to claim 1, further comprising the steps of: after assertion of said back-off signal in said step of asserting said back-off signal, in response to a second assertion of said first bus cycle start signal (ADS#), determining whether a second valid first bus cycle definition then asserted on said first bus (VL-bus) cycle definition lines is a repetition of said first valid first bus cycle definition; asserting said first bus device claiming signal (LDEVO#) prior to said first bus device claiming deadline, in response to said second assertion of said first bus cycle start signal, only if said second valid first bus cycle definition is not a repetition of said first valid first bus cycle definition; and defining a cycle on said second bus (PCI-bus) in response to said second assertion of first bus cycle start signal by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#), only if said second valid first bus cycle definition is not a repetition of said first valid first bus cycle definition and only if said second valid first bus cycle definition is within said predetermined set of first bus cycle definitions.
12. A method according to claim 11, wherein said first bus host comprises a first-bus/fourth-bus bridge bridging said first bus with a fourth bus (host bus) different from said first, second and third buses, said first-bus/fourth-bus bridge defining a plurality of said first bus (VL-bus) cycles in response to each fourth bus (host-bus) cycle detected by said first-bus/fourth-bus bridge on said fourth bus, wherein said step of asserting said first bus device claiming signal (LDEVO#) in response to said second assertion of said first bus cycle start signal, asserts said first bus device claiming signal (LDEVO#) in response to said second assertion of said first bus cycle start signal only if said second valid first bus cycle definition is not one of the plurality of first bus (VL-bus) cycles defined by said first-bus/fourth-bus bridge which includes said repetition of said first valid first bus cycle definition.
13. A method according to claim 1, wherein said first bus (VL-bus) host (102) after asserting a first bus cycle start signal, withholds further assertions of said first bus cycle start signal until after detection by said first bus host (102) of a first bus (VL-bus) ready return signal (e.g. RDYRTN#), said first-bus/third-bus (VL-bus/system-bus) bridge asserting said ready return signal (e.g. RDYRTN#) in response to detection by said first-bus/third-bus (VL-bus/system-bus) bridge of a first bus (VL-bus) device ready signal (e.g. LRDY#), further comprising the step of asserting said first bus device ready signal (e.g. LRDY#) if no second bus device claiming signal (DEVSEL#) is asserted before said predetermined second bus device claiming deadline following said assertion of second bus cycle definition signals on said second bus cycle definition lines.
14. A method according to claim 13, wherein said back-off signal (BOFF#) is asserted in said step of asserting said back-off signal, before said first bus device ready signal (e.g. LRDY#) is asserted in said step of asserting said first bus device ready signal, further comprising the steps of: unasserting said first bus device ready signal (LRDY#); and subsequently unasserting said back-off signal (BOFF#).
15. A method according to claim 14, further comprising the step of unasserting said first bus device claiming signal (LDEVO#) after detection of said first bus ready return signal (e.g. RDYRTN#) asserted.
16. A method for responding to a first bus (VL-bus) cycle, for use with a first bus (VL-bus) host (102) defining said first bus cycle by asserting first bus cycle definition signals on first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#) and asserting a first bus cycle start signal (ADS#), said first bus host withholding any further assertion of said first bus cycle start signal until detecting a first bus (VL-bus) ready signal (RDYRTN#) subsequent to said assertion of first bus cycle start signal, or a first bus back-off (BOFF#) signal, whichever occurs earlier, said first bus host repeating said first bus cycle after detecting said back-off signal (BOFF#), for use further with a first bus controller which asserts said first bus ready signal (RDYRTN#) in response to detecting assertion of a local device ready signal (LRDY#), comprising, after assertion by said first bus host of a given first bus cycle start signal and no later than any assertion of said first bus start signal subsequent to said given first bus cycle start signal, the steps of: asserting said back-off signal (BOFF#) in response to a first clock pulse; asserting said local device ready signal (LRDY#) in response to a second clock pulse after said first clock pulse; negating said local device ready signal (LRDY#) in response to a third clock pulse after said second clock pulse; and negating said back-off signal (BOFF#) in response to a fourth clock pulse after said third clock pulse.
17. A method according to claim 16, further comprising the step of detecting said first bus ready signal (RDYRTN#) no later than said step of negating said back-off signal (BOFF#).
18. A method according to claim 16, for use further with a first-bus/third-bus (VL-bus/system-bus) bridge coupled to receive said first bus cycle definition signals, said first-bus/third-bus bridge defining a cycle on said third bus in response to certain combinations of said first bus cycle definition signals only if said first-bus/third-bus bridge fails to detect a first bus device claiming signal (LDEVO#) before a predetermined first bus device claiming deadline following detection by said first-bus/third-bus bridge of an asserted first bus cycle start signal(ADS#), further comprising the steps of: asserting said first bus device claiming signal (LDEVO#) before said first bus device claiming deadline in response to detecting said given first bus cycle start signal, and before said step of asserting said back-off signal (BOFF#); and negating said first bus device claiming signal (LDEVO#) no earlier than said step of negating said local device ready signal (LRDY#).
19. A method for bridging a first bus (VL-bus) cycle to a second bus (PCI-bus), in a system including a first bus (VL-bus) having first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), said system further including said second bus (PCI-bus), said second bus having second bus cycle definition lines (e.g. C/BE#, FRAME#), for use with a first bus (VL-bus) master (102) defining said first bus cycle by asserting first bus cycle definition signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), said first bus master further having a back-off input line (BOFF#) which causes said first bus master to repeat said first bus cycle after detecting assertion of a back-off signal on said back-off input line, said method comprising the steps of: defining a cycle on said second bus (PCI-bus) in response to a first assertion of said first bus cycle start signal (ADS#) by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#); and asserting said back-off signal (BOFF#) in response to a predefined event occurring on said second bus (PCI-bus) in response to said assertion of second bus cycle definition signals on said second bus cycle definition lines.
20. A method according to claim 19, wherein said predefined event is defined at least in part by expiration of a predefined time period that begins in response to said assertion of second bus cycle definition signals on said second bus cycle definition lines.
21. A method according to claim 20, wherein said predefined event is defined further by non-detection of a predefined signal before expiration of said predefined time period.
22. A method according to claim 19, wherein said predefined event occurs on said second bus without any intervening event occurring on said second bus in response to said assertion of second bus cycle definition signals on said second bus cycle definition lines.
23. A method according to claim 19, wherein said second bus further has a second bus device claiming signal line (DEVSEL#), and wherein said predefined event is defined at least in part by non-assertion of a second bus device claiming signal on said second bus device claiming signal line (DEVSEL#) before a predetermined second bus device claiming deadline following said assertion of second bus cycle definition signals on said second bus cycle definition lines.
24. A method according to claim 23, wherein said first bus further has a first bus device claiming signal line (LDEVO#), and wherein at the time of said first assertion of said first bus cycle start signal (ADS#) the signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#) are within a predetermined set of first bus cycle definitions, for use further with a third bus (system bus) and a first-bus/third-bus (VL-bus/system-bus) bridge coupled to receive said first bus cycle definition signals, said first-bus/third-bus bridge defining a cycle on said third bus in response to certain combinations of said first bus cycle definition signals only if no first bus device claiming signal (LDEVO#) is asserted before a predetermined first bus device claiming deadline following detection by said first-bus/third-bus bridge of an asserted first bus cycle start signal (ADS#) (e.g. R/W#, M/IO#, D/C#, VA, BE#), further comprising the step of asserting said first bus device claiming signal (LDEVO#) prior to said first bus device claiming deadline, in response to said first assertion of said first bus cycle start signal (ADS#).
25. A method according to claim 19, wherein the signals on said first bus cycle definition lines at the time of said first assertion of said first bus cycle start signal (ADS#) define a first valid first bus cycle, further comprising the steps of: after assertion of said back-off signal in said step of asserting said back-off signal, in response to a second assertion of said first bus cycle start signal (ADS#), determining whether a second valid first bus cycle defined by signals then asserted on said first bus (VL-bus) cycle definition lines is a repetition of said first valid first bus cycle; and defining a cycle on said second bus (PCI-bus) in response to said second assertion of first bus cycle start signal by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#), only if said second valid first bus cycle is not a repetition of said first valid first bus cycle.
26. A method according to claim 19, for use further with a third bus (ISA bus) having third bus cycle definition lines, further comprising a step of defining a cycle on said third bus (ISA-bus) in dependence upon said first bus cycle, after said step of defining a cycle on said second bus (PCI-bus).
27. A method according to claim 26, wherein said step of defining a cycle on said third bus (ISA-bus) occurs after said step of asserting said back-off signal (BOFF#).
28. Apparatus for bridging a first bus (VL-bus) cycle to a second bus (PCI-bus), in a system including a first bus (VL-bus) having first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), said system further including said second bus (PCI-bus), said second bus having second bus cycle definition lines (e.g. C/BE#, FRAME#), for use with a first bus (VL-bus) master (102) defining said first bus cycle by asserting first bus cycle definition signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#), said first bus master further having a back-off input line (BOFF#) and repeating said first bus cycle after detecting assertion of a back-off signal on said back-off input line, comprising: second bus interface logic which defines a cycle on said second bus (PCI-bus) in response to a first assertion of said first bus cycle start signal (ADS#) by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#); and first bus interface logic which asserts said back-off signal (BOFF#) in response to a predefined event occurring on said second bus (PCI-bus) in response to said assertion of second bus cycle definition signals on said second bus cycle definition lines.
29. Apparatus according to claim 28, wherein said second bus further has a second bus device claiming signal line (DEVSEL#), and wherein said predefined event is defined at least in part by non-assertion of a second bus device claiming signal on said second bus device claiming signal line (DEVSEL#) before a predetermined second bus device claiming deadline following said assertion of second bus cycle definition signals on said second bus cycle definition lines.
30. Apparatus according to claim 29, wherein said first bus further has a first bus device claiming signal line (LDEVO#), and wherein at the time of said first assertion of said first bus cycle start signal (ADS#) the signals on said first bus cycle definition lines (e.g. R/W#, M/IO#, D/C#, VA, BE#) are within a predetermined set of first bus cycle definitions, for use further with a third bus (system bus) and a first-bus/third-bus (VL-bus/system-bus) bridge coupled to receive said first bus cycle definition signals, said first-bus/third-bus bridge defining a cycle on said third bus in response to certain combinations of said first bus cycle definition signals only if no first bus device claiming signal (LDEVO#) is asserted before a predetermined first bus device claiming deadline following detection by said first-bus/third-bus bridge of an asserted first bus cycle start signal (ADS#) (e.g. R/W#, M/IO#, D/C#, VA, BE#), further comprising circuitry which asserts said first bus device claiming signal (LDEVO#) prior to said first bus device claiming deadline, in response to said first assertion of said first bus cycle start signal (ADS#).
31. Apparatus according to claim 28, wherein the signals on said first bus cycle definition lines at the time of said first assertion of said first bus cycle start signal (ADS#) define a first valid first bus cycle, further comprising a repetition detector which, after assertion of said back-off signal by said first bus interface logic in response to said predefined event, in response to a second assertion of said first bus cycle start signal (ADS#), determines whether a second valid first bus cycle defined by signals then asserted on said first bus (VL-bus) cycle definition lines is a repetition of said first valid first bus cycle, and wherein said second bus interface logic further defines a cycle on said second bus (PCI-bus) in response to said second assertion of first bus cycle start signal by asserting second bus cycle definition signals on said second bus cycle definition lines (e.g. C/BE#, FRAME#), only if said repetition detector determines that said second valid first bus cycle is not a repetition of said first valid first bus cycle.
32. Apparatus according to claim 28, for use further with a third bus (ISA bus) having third bus cycle definition lines, further comprising third bus interface logic which defines a cycle on said third bus (ISA-bus) in dependence upon said first bus cycle, after said second bus interface logic defines a cycle on said second bus (PCI-bus).
33. Apparatus according to claim 32, wherein said third bus interface logic defines said cycle on said third bus (ISA-bus) after said first bus interface logic asserts said back-off signal (BOFF#).
Patent Citations (7)
Non-Patent Literature (6)
- Acer Laboratories, Inc., "M1429G/M1431/M1435", Datasheet (Oct. 1993).
- Acer Laboratories, Inc., "M1435 PCI-VL Bus Bridge", Preliminary Data Sheet (Sep. 20, 1993), pp.1-7, 13-14.
- Acer Laboratories, Inc., "Pentium Solution of ALI", ALI Pentium Seminar (1994).
- Slater, Michael, et al., "Local Buses Poised to Enter PC Mainstream -- Intel's PCI and VESA's VL-Bus Vie for Vendors' Attention", Microprocessor Report (Jul. 2, 1992), pp. 7-13.
- Intel Corporation, "i486.TM. Microprocessor", Databook (Nov. 1989).
- Intel Corporation, "82434LX/82434NX PCI, Cache and Memory Controller (PCMC)", Advance Information (Mar. 1994), pp. 1-30, 70-104.