AU723157B2 - Configurable logic array - Google Patents

Description

AUSTRALIA

PATENTS ACT 1990 COMPLETE SPECIFICATION FOR A STANDARD PATENT (Original) APPLICATION NO:

LODGED:

COMPLETE SPECIFICATION LODGED:

ACCEPTED:

PUBLISHED:

RELATED ART: NAME OF APPLICANT: ACTUAL INVENTOR(S): ADDRESS FOR SERVICE: GARETH JAMES JONES GORDON STIRLING WORK KELVIN LORD AND COMPANY, Patent Trade Mark Attorneys, of 4 Douro Place, West Perth, Western Australia, 6005, AUSTRALIA.

INVENTION TITLE: "CONFIGURABLE LOGIC ARRAY" ooeo• The following Statement is a full description of this invention including the best method of performing it known to me/us: 2 The present invention relates to semi-conductor integrated circuits of the type comprising configurable logic circuit arrays.

The invention is a development of the configurable logic circuit arrays disclosed in our British Patent Specification No. 2180382 (having a corresponding U.S. Patent No. 4935734) and U.S. Patent No. 5001368. In the former, the logic circuit array comprises a matrix of discrete sites or cells at each of which is a logic circuit which is adapted to perform a simple logic function. Typically the simple logic function is implemented by means of a two input NAND gate. As made each logic circuit has what may be referred to as a restricted signal translation system 15 by which each logic circuit has selectable direct connection paths to only a few of the other logic ""circuits. More particularly each direct connection path which is selectable as to its conduction state, extends, for each said logic circuit, from its output to inputs of a first set of some of other said logic circuits and from its inputs to outputs of a second set of some of .other said logic circuits, all of the sets (for all of the logic circuits) each being unique. Such a restricted signal translation system provides what can be conveniently referred to as local direct connection paths.

3 An array of this type is capable of being programmed in such a manner as to configure the various NAND gates, as required, to perform various and different logic functions. One such function is known as a latching function and in the 'logic array as disclosed in Specification No. 2180382, a latching function may be implemented using four NAND gates. This has the disadvantage that the greater the number of latching functions that may be required from any logic array, the fewer NAND gates remain for other required functions. This has the effect of reducing the overall effectiveness of the array.

The invention of U.S. 5001368 overcomes this disadvantage by providing an additional logic circuit for inclusion in each of the logic circuits at each i discrete site to enable each site to have a greater programmable facility and thereby increase the overall utilisation of the array. As with GB 2180382 the site/cell of each logic circuit has the aforesaid 20 restricted signal translation system. Each additional logic circuit is arranged within the logic circuit of the site or cell to be selectively controlled by control means to cause each logic circuit and additional logic circuit comprising each cell to operate as either a first or a second different simple logic function. More particularly, each cell is constructed to operate as a 4 NAND gate or as a latch circuit function.

Providing such local direct connection paths between physically neighbouring logic circuit sites facilitates the establishment of a desired particular circuit function in a localised configuration of logic circuits, ie. in part only of the chip area occupied by an appropriate number of discrete sites of the logic circuits and by the direct connection paths.

Interconnections between such localised configurations for overall circuit/system requirements can be either simply by said direct connection paths or by a further signal translation system by way of a direct connection bus directly connected to the logic circuits and extending throughout the array for example as a series 15 of rows and columns. Such connections might conveniently be referred to as global connections.

The so-called local direct connections and global connections constitute routingresources which are used to connect to discretely programmed logic functions.

These resources can be -combined by routing through a cell in order to complete the connection between functions. Once a logic circuit has been used in this matter to complete the routing that logic circuit can no •9o*o* longer be used for function. This reduces the overall effectiveness of the array. An object of one aspect of this invention is to overcome this disadvantage by re providing additional connection resources in an arrangement not previously contemplated.

A feature of the afore-described configurable logic arrays is that each discrete site or cell is identical thus in the case of GB 2180382 each discrete site comprises a NAND gate whilst in the case of US 5001368 each discrete cell comprises a NAND gate and an additional logic circuit which enables the cell to function exclusive as a NAND gate or exclusively as a latch circuit function. Implementing other circuit functions (be they logic functions or otherwise) requires the various cells to be configured as required to perform various and different circuit functions.

Other functions frequently required are XOR and D-flipflop. Again this has the disadvantage that the greater number of these functions that may be required from any logic array the fewer gates remain for other required functions which has the effectof reducing the overall effectiveness of the array.

S: 20 An object of another aspect of this invention is to overcome this disadvantage by providing a series of different types of cells (eg. providing different functions) and arranged in a particular manner in the eeeee: ".:array to thereby increase the overall utilisation of the array.

According to said one aspect of the invention a 6 semi-conductor integrated circuit comprises an area thereof formed with a plurality of logic circuits at discrete sites or cells respectively defining a matrix array of cells, and wherein the matrix array of cells is sub-divided at least into zones each comprising a matrix array of said cells and further comprising a porting arrangement for each zone, the integrated circuit as made having an hierarchical routing resource structure comprising:global connection paths having selectable connections with the porting arrangement of each zone, (ii) medium connection paths extending between the porting arrangement and at least some of the cells in a zone, and 15 (iii) local direct connection paths comprising for each cell a restricted signal translation system between oooo inputs and outputs of the cells each selectable as to its conduction state, those paths extending for each said logic circuit from its output to inputs of a first 20 set of some of other said logic circuits and from its inputs to outputs of a second set of some of other said logic circuits, each first set of logic circuits being different from any other first set of logic circuits, and each second set of logic circuits being different from any other second set of logic circuit.

In essence the present invention provides a semi- 7 conductor integrated circuit comprising an area thereof formed with a plurality of logic circuits at discrete sites with three levels of interconnect, namely global, medium and local, and are brought about by the existence of the porting arrangement that separates the array into zones. The global level interconnect can span the whole array but is separated from the medium and local level interconnects by the porting arrangement, the medium interconnect extend at a zone but can connect directly with port cells and core cells, whilst the local interconnect provides connections between only a few cells.

The zones are preferably arranged in a matrix array of zones, the plurality of zones defining a 15 quadrant and the integrated circuit may have a plurality of quadrants also conveniently arranged in a matrix array. In a preferred construction there are 10 x cells to a zone and 5 x 5 zones to a quadrant and 2 x 2 quadrants. The global connection paths conveniently 20 extend horizontally and vertically across a plurality of zones. Indeed it is preferable to have the global connection paths comprise lines which extend continuously across more than one zone and which are selectably connectable with at least some of the zones.

Preferably there are a plurality of global connection paths for each row and column of cells in a zone. More 8 preferably still there are four global connection paths for each row and column of cells.

As concerns the medium connection paths, it is preferred to have a plurality of these for each row and column of cells of a zone, and preferably four for each row and column. Each medium connection path is selectably connectable with an output and/or input of one or more cells.

Preferably the semi-conductor integrated circuit comprises further global routing resources comprising horizontal and vertical buses extending across a zone horizontally and vertically respectively, and which are connectable with the port cells and *with one another.

It is preferred that they are not connectable to the 15 cells. Conveniently there is one such additional global o o Boo connection bus for each row and column of cells, and sees preferably there is a connection between the horizontal

S•

and vertical buses for .each cell. Conveniently such additional global connection paths are referred to as x buses.

It is convenient to have horizontal and vertical •go• port cells. The porting arrangement permits resources to be routed from the global lines through.,to the medium

SSSSS*

lines and thereby into the cells without necessarily having to use cells for routing thereby leaving more cells available for function.

9 The use of port cells is also advantageous in that it allows the general interconnect structure (global and medium buses) to interconnect with special interconnect structure, for example clocks and tristates.

Accordingly, the circuit advantageously further comprises clock distribution channels, preferably comprising a vertical clock bus (with say 8 lines) extending between clock pads to top and bottom of the array, a horizontal clock distribution Spine (with say 8 lines), extending between clock pads to opposite sides of the array and intersecting with the vertical clock bus. Preferably there is a vertical clock distribution spine for each zone (conveniently tapped off from the 15 horizontal clock spine) by which clock and reset signals *fee are distributed to the vertical port cell of that zone.

Preferably alternate core cells have clock and reset signal lines. Buffers are used in the clock signal oo i distribution lines as appropriate.

20 Further advantageous of signal paths arise from dedicated connection lines with the horizontal port cells, specifically wired -OR buses (say 5) running S• across each zone and connecting directly with the wired -OR core cells. Preferably connection of the wired -OR 25 buses may be made to a certain (restricted) number of global buses, eg. by way of the horizontal port cells.

A further advantageous feature is to have wired -OR connections between horizontal x buses and say three horizontal global buses. Pull ups are advantageously provided for the WO buses, eg. in the horizontal port cells, and pull ups for the global buses are advantageously situated in the vertical inter-quadrant region.

The preferred connection provisions afforded by the porting arrangement will be described further hereinafter, along with other features of the integrated circuit.

According to said another aspect the invention provides a configurable -semi-conductor integrated circuit comprising a matrix array of core cells each of the cells having a first simple function in common and at least one subsidiary function, there being at least two different subsidiary functions, the core cells being grouped in tiles comprising a matrix array of the core o:oo cells and wherein each tile has at least one of each 20 different subsidiary functions.

The tiles comprise a matrix array of the core cells smaller than the whole array. The arrangements S"of subsidiary functions within the cells of the tile are substantially different. The resultant tile of core 25 cells is arranged so as to uniformly cover the array.

Said another aspect the invention has a 6.

11 particular application with the circuit configuration described in the aforesaid British Patent No. 2180302.

Accordingly in that application said another aspect may be defined by a configurable semi-conductor integrated circuit which comprises an area thereof formed with a plurality of logic circuits at discrete sites or cells respectively, each said logic circuit having in common a restricted simple logic function capability and itself only being capable of implementing a simple logic function, and the cell having a restricted signal translation system between inputs and outputs of the logic circuits affording (local) direct connection paths each selectable as to its conduction state, those paths extending, for each said logic circuit; from its output 15 to inputs of first set of some of other said logic o circuits and from its inputs to outputs of a second set of some of other said logic circuits, each first set of logic circuits being different from any other first set of logic circuits, and each second set of logic circuits 20 being different from any other second set of logic circuits, the integrated circuit further comprises at each discrete site or cell additional optionally •selectable circuit configurations selectively controlled by control means to enable each cell to operate in a 25 selected one of two or more ways, and wherein there are a plurality of different optional circuit configurations and wherein the cells are arranged in groups, hereinafter referred to as tiles, with each tile having at least one of each of the different circuit configurations.

The selected one of two or more ways comprises either the simple logic function common to each cell or a selected subsidiary function afforded by the optionally selectable circuit configurations. The arrangements of the subsidiary functions within the cells of the tile are substantially different, and the resultant tile of cells is arranged so as to uniformly cover the array.

According to this aspect, there is provided a uniform array with externally identical cells allowing a uniform interconnect structure to be used regardless of the function of the cell. The interconnect is identical for all cells within a tile. In use the autolayout tools will- know which cell supports which function but o• if the common function only is used then the array can S 20 be treated completely uniformly, ie. if the subsidiary functions are not used then the array appears as a uniform array of simple primary functions. There is a Ieeoi S"hierarchy of functions within the array. It is possible to implement most logic functions using the primary 25 function only, ie. the simple cells, eg. the AND gates with programmable inversions on the inputs but this is I I I I I 13 not very efficient. In order to improve the efficiency of the uniform base array it is overlaid with the subsidiary functions. The subsidiary functions represent optimised implementations of functions that can be created using the primary functions. By using this hierarchy the autolayout tools have a flexible uniform target for placement purposes for the more common functions with a slightly coarser target for the less commonly used secondary functions. The secondary functions could not be used on their own to complete a design. An hierarchy of function is preferred as opposed to an arbitrary distribution of different logic functions.

The additional or subsidiary functions are distributed throughout the array. A group of commonly used functions are -selected on the basis of their frequency of occurrences within gate array designs. The selected group of functions are distributed throughout the array as the tile, which is repeated in a regular 20 pattern over the array. The autolayout placement target for the primary function is a cell whilst the placement target for a secondary function is a tile. Although the tile is a coarser target it still represents a uniform resource throughout the array.

25 In one embodiment the common simple logic function is a NAND gate. Preferably the optionally 14 selectable circuit configurations afford optional functions of:- wired -OR output buffer wlich is to be viewed as a function for the purposes of this specification; XOR; D-type flip flop (with reset and enable) and latch function (with reset and enable).

Thus four different subsidiary functions are available, and the above recited function are the preferred subsidiary functions. As specified above each tile has at least one each of the available subsidiary functions.

Some cells may have additional circuit configurations to facilitate by virtue of the tile arrangement building up higher level functions, for example a 2-to-i multiplexer; or fast carry logic.

These subsidiary functions are chosen because they are functions which are commonly required and which would otherwise be arrived at by configuring several of the basic function cells. We have identified that the number of different subsidiary functions which need to S...i be employed in configuring of an integrated circuit to :eI 20 perform as desired results in some subsidiary function being required more frequently than others. Thus we e prefer to have the number of different subsidiary •o•functions available to a tile reflect this. Thus we prefer to have two XOR functions to a tile.

25 .Furthermore, by having one cell offer optionally two functions (preferably the D-flip flop and latch) an advantages construction can be arrived at with a smaller tile configuration.

Our preferred tile comprises four cells, preferably arranged as a matrix of 2 x 2 cells, and thus in the preferred embodiment whilst there is notionally one different subsidiary function for each cell, we prefer to have two cells offering the XOR functions, one cell offering the wired-OR function -and one cell offering the option of D flip flop or' latch function.

Each cell has available to it its basic function, ie. a NAND gate.

The preferred arrangement of 2 x 2 cells with the above preferred distribution of subsidiary functions is such as to naturally form efficient larger elements such as adders, counters and multiplexers.

As mentioned above for the preferred application oooe each cell has the restricted signal translation system providing direct local interconnections between only some of the cells. However, it will be appreciated that ooo• the tiling arrangement can be utilised in circuits without this specific restricted translation system.

999o Preferably the integrated circuit has additional connection resources as described above and hereinafter according to said one aspect of the invention.

The present invention will now be described further, by way of example only, with reference to the 16 accompanying drawings, in which:- Figure 1 is a diagrammatic representation of the hierarchical structure of a configurable logic array embodying the inventions hereof, Figure 2a shows schematically the basic function of a tile of four cells used in the array of Figure i, Figure 2b shows schematically the alternative functions for each cell of the tile of four cells of Figure 2a, Figure 3 illustrates diagrammatically one embodiment of local connections between cells, Figure 4 illustrates diagrammatically one series of medium connection paths to the four cells of a tile, Figure 5 illustrates diagrammatically a zone of 15 cells and the associated vertical and horizontal port 9 cells, .Figure 5a shows further details of the zone of i.g Figure Figure 5b illustrates a corner zone, 20 Figure 6 illustrates diagrammatically details of eeoc connections made via the horizontal port cells, ~Figure 7 illustrates diagrammatically details of oooeo connections made via the vertical port cells, Figure 7a illustrates diagrammatically details of 25 connections made via the vertical port cells at the top and bottom edges of the array, i.

17 Figure 8 illustrates diagrammatically details of connections between quadrants, Figure 9a illustrates diagrammatically the primary clock structure, Figure 9b illustrates further details of the clock structure of Figure 9a, and Figure 10 illustrates diagrammatically the hierarchy of interconnect provided according to the invention.

Referring firstly to Figure 1, this illustrates an overview of the hierarchical structure of the configurable logic array embodying the inventions hereof. The array of the illustrated embodiment consists of 10,000 core cells CC, all of which can be used as a simple NAND gate. For convenience the following description describes the array as a regular array of cells, comprising rows and columns. A matrix Sarray of 10 by 10 core cells CC with associated port cells (described further hereinafter with reference to Figures 6 and 7) constitutes a zone 11 (see Figure of which there are 100 in the illustrated embodiment. A matrix array of 5 by 5 zones constitutes a quadrant 13.

ioooo• In the illustrated embodiment the quadrants are disposed in a 2 by 2 matrix array. Inter-quadrant switches 25 (generally designated by reference SG see Figure 8) are provided between the adjacent quadrants. The-array is Ir.

18 also provided with user input output cells designated by blocks 17 (Figure 1) and the illustrated embodiment has 50 per side. Also provided are input/output multiplexers 18. The circuit also has a clock structure (including clock pads CP and clock bus CB) which is described further with reference to Figures 9a and 9b..

Referring now to Figures 2a and 2b, here there is illustrated diagrammatically a group of 4 core cells hereinafter referred to as a tile T and disposed in a 2 by 2 matrix array. Each cell comprises a two-input-NAND gate N1, multiplexer means Ml, 2, 3 and 4 on the input side and inverters II, 12 between respective multiplexers Ml, M3, M2, M4. Each cell also has an output multiplexer MO. In addition to the facility for S 15 each core cell to be used as a simple NAND gate as represented by circuity illustrated diagrammatically in Figure 2a, each core cell has an extra function box (fn) that can be selected during configuration. There are a plurality of different functions available and each tile S 20 contains at least one of the available functions. The 4 cells making up a tile are designated by the numbers 1, 2, 3 and 4 respectively (see Figures 2a, 2b) denoting different types of core cell (ie. CC-1, CC-2, CC-3, CCi In the illustrated embodiment the alternative 25 function available to the type 1 cell is a wired-OR configuration, the additional function available to the 19 type 2 cell is that of XOR as well having circuitry to provide half of a 2-to-i multiplexer, or fast carry logic. The CIN input is driven from the type 4 cell of the tile below. The COUT output is input to the type 4 cell above. The CIN and COUT signals form the first carry logic. The type 3 cell in the illustrated embodiment has two alternative available functions, namely a D flip flop or a latch (each with reset and enable). The type 4 cell also has as its alternative function the x-OR function and additionally has the other half of the circuitry providing a 2-to-l multiplexer or fast carry logic. The multiplexer and fast carry alternative functions require 2 core cells, thus the circuitry of the type 2 and type 4 cells are used together. In the context of the present application the wired-0R output driver of type 1 cell is to be regarded as an alternative function although S. strictly speaking it is not a true alternative function o9* in the sense of the other functions. Figure 2b shows circuitry representative of the functions available to o the 4 types of core cell.

The tiles are arranged into zones as mentioned •co• above and connections within zones are made using local interconnect described further with reference to Figure 25 3, or medium range interconnect to anywhere in the zone.

The latter comprise horizontal and vertical medium buses M which run the length and breadth of each zone (4 per row and column of cells). These are .designated as M1, M2, M3, M4. Port cells (VPC and HPC) (Figures 5, 6 and 7) at the edge of each zone make connections via the medium buses to neighbouring zones or to the global bus network. Horizontal and vertical x buses (referenced x) (Figures 5 and 5b) also run the length and breadth of each zone (1 per row and column of cells) and in conjunction with a switch (not illustrated) in each core cell they are used for making right angled turns on these buses. Global buses G (Figure 5a) run the length and breadth of each quadrant and in the illustrated embodiment there are 4 per row and column of core cells identified as Gl, G2, G3 and G4 (Figures 6 and Switches SG1...SG4 (Figure 8) between the quadrants allow global buses to run the whole length of the device if required.

It is envisaged that all devices will have 4 quadrants but the number of zones in each quadrant may 20 differ from one device to another. It is however preferred to have a matrix of 10 by 10 core cells to each zone.

The global resources available to the array comprise the global interconnect lines G1...G4 and 25 associated switches SG1...SG4. The additional letters V and H are used throughout to denote vertical and horizontal connections, cells and switches etc. as the case may be. Global resources further comprise the above-mentioned x buses described further herein below and port cells that connect between global and zone resources. There are 10 port cells along the top of each zone (the vertical port cells VPC) and 10 along the right hand side (the horizontal port cells HPC).

Connections between routing resources within zones via the medium buses M1...M4 and the global resources are only possible through the port cells. Port cells also support connections between medium buses in adjacent zones and are used for clock and tristate net distribution described further hereinafter.

Referring now to Figure 7 here we illustrate diagrammatically how the zone and global *buses connect to the vertical port cells VPC. Port cells are arranged in pairs, aligned with the core cell tiling. Four Sroutes are possible through each port cellmultiplexers A and C support one route each, and 20 multiplexer B (shown as B1 and B2) supports two separate connections. Two of the zone medium buses (Ml and M2) are used for connections to the global and x buses through multiplexers A and C. In addition, M1 and M2 from the adjacent column in the tile also connect 25 to A and C. This arrangement allows a cross over connection between pairs of port cells. Multiplexers

A

and C also provide bufferred connections from M1/M2 or M1/M2 in the column from the adjacent column in the tile to or from M3/4 in the zone above. The other two zone medium buses (M3 and M4) connect to multiplexers B1 and B2. B1 and 2 allow two independent unbufferred links to M3 or M4 in the zone above. M3 and M4 can be interchanged using a programmable twist. These interzone connections can be used when a fast connection is required between zones.

Alternate port cells have either a clock (CLK) or a reset (RST) multiplexer. These select the source for the dedicated clock and reset lines to the D flip flop and latch core cells, ie. the type 3 cells. The clock or reset multiplexer provides programmable inversion of 15 clock and reset.

Referring now to Figure 7a, here will illustrate the connections at the top and bottom edges of the array. At the top edge connections to the vertical port cell (VPC) are with the input/output cells 17 by way of S 20 the input/output multiplexers 18. In the illustration there are two 7:1 multiplexers and one 2:1 multiplexer aee.

for each adjacent pair of port cells. It will be seen that the lines emerging from the two horizontal port cells correspond to those of Figure 7 and that each port 25 cell connects with a respective 7:1 multiplexer. In addition the two x buses from each port cell have a 23 branch into the 2:1 multiplexer which communicates with control circuitry.

At the bottom edge of the array, the lines passing over the zone boundary also connect with input/output cells 17 via multiplexers. For two adjacent cells also comprises a respective 7:1 multiplexer and a common 2:1 multiplexer. The lines comprise Global lines the x bus and medium interconnect lines M3, M4. The respective x buses are branched to the 2:1 multiplexer as at the top edge.

Note in each case one 7:1 multiplexer feeds to an output cell 17 OUT and one is fed from an input cell 17

IN.

In the illustrated embodiment, the input/output connection at the edges mirror the connections between zones. This is not to be taken as limiting, merely an example and the connection to the input/output multiplexers may be for. more extensive. For example other of the medium interconnect lines may be 20 substituted for M3, M4 or supplement them. In the illustrated embodiment the medium interconnect lines Ml, SM2 terminate at the lowermost cell of each zone, ie.

S" they do not pass over the zone boundary.

The connections to the input/output cells of the 25 right and left edges of the array according to one embodiment mirror the connections to the horizontal port n i 24 cells (HPC) as illustrated in Figure 6 in a corresponding manner to the arrangement of Figure 7b, utilising two 7:1 multiplexers for. each pair of adjacent cells and a corresponding 2:1 multiplexer for the x buses.

The primary clock structure will now be described. Primary clocks may originate external to the device via 8 special purpose-clock pads CP or from the array by routing on general resource to'the clock pads.

The clock pads CP are situated at the corner of each quadrant (see Figure 1) and illustrated in further detail with reference to Figure 9a which is a chip level diagram. The clock bus CB runs vertically between the top and bottom clock pads CP and connects with a 15 horizontal clock spine HCS (having 8 lines) running between the horizontal clock pads to opposite sides of the array. A central clock buffer CCB is disposed at the intersection of the vertical clock bus CB and the horizontal clock spine or bus (HCS). A total of 8 20 global signals (clocks or reset clocks) may be driven; these can be either internally or externally generated.

Any normal user I/O input signal may also be used as a primary clock by simply routing it internally to one of the clock pads. The horizontal clock spine is tapped off at various points to drive multiple vertical clock spines VCS which run between adjacent zones to provide one vertical channel of 8 global clock signals per column of array zones. In addition there are equivalent vertical channels for the I/O zones to the left and right hand sides of the device. The I/O zones to the top and bottom of the array connect t6 the vertical clock distribution channels driven up and down through the array. The primary clock and reset signals are input into the vertical port cells of each zone via the zone clock and reset cells. An 8-2 multiplexer (Figure 9b) distributes signals from the VCS into the vertical port cell VPC and as described with reference to Figure 7 alternate core cells CC have a clock switch or a reset switch whereby clock (CLK) or reset (RST) signals can be distributed vertically through the zone (see also Figure S. 15 9b).

Secondary clocks can also be provided using conventional routing resource of the circuit. A routing comb is created by the software consisting of the horizontal spine and vertical teeth. The spine and 20 teeth are routed on global interconnect and connected via x bus switches. The global interconnect may be extended across quadrant switches. The clocks are input into zones via the zones vertical port cells (which also provide programmable clock inversion), the clock can only be connected to flip flops on the same column as the secondary clock. Tertiary clocks may also be provided using conventional routing resource and our input into zones via the zones vertical port cells and can be routed on any level of interconnect.

Referring now to Figure 6 which illustrates detailing of the horizontal port cell HPC. These provide the same medium, global and x bus connectivity as the vertical port cell described above. Since the primary clock and reset signals are distributed only from the vertical core cells down core cell columns, the horizonal port cell contains no clock/reset logic.

Instead, it provides support for tristate buses.

Tristates are implemented using a dedicated horizontal bus within the zone (the WO bus) plus the horizonal global interconnect. The wired -OR output from the 15 type 1 cell subsidiary function is connected to the ofl.

horizontal port cell via the dedicated horizontal WO bus. The WO bus connects to multiplexer C in the horizontal port cells. The internal multiplexers for normal interconnect are the same as those for the oo 20 vertical port cells.

The functional tiling within the zones means that only half the core cell rows contain core cells with q ae wired-OR drivers. Therefore, there are only five wired- OR buses per zone. This means that there have to be two types of port cell, one with zone wired-OR bus connecting to it (type and one without (type 2).

The zone wired-OR bus from the type 1 horizontal port cell is fed to the type 2 cell as indicated by line F.

This means that a WO bus can drive into horizontal global buses in every row. The x bus can also make a wired-OR connection bnto the global buses, allowing vertical steps between wired-OR buses. This connection can be inverted as required to maintain the sense of signals.

With regard to the global interconnections, horizonal and vertical global buses run across each quadrant connecting to each zone via its port cells, see Figures 5, 5a, 6 and 7. The global buses connect to the I/O cells at the periphery of the device.

There are 200 user configurable I/O cells 17 and 15 these are arranged so that they pitch match 1 to every two core cells around the edges of the device. Global, medium and x buses M1...M2 or M3...M4, X) from one core cell row or column are used as inputs and M1...M2 or M3...M4 and X from the adjacent row 20 or column are used as outputs. An 8 bit peripheral bus runs around the whole device which is accessed by the I/O cells. Each I/O cell can read or write any of the 8 bits. Wired-OR buffers can optionally be selected when writing to the peripheral bus.

Referring again to Figure 5, the basic structure of a zone is illustrated showing the vertical port cells I 28 VPC and the horizontal port cells HPC described with reference to Figures 7 and 6 respectively. Also illustrated is the zone CLK/RST multiplexer. Referring to Figure 3, regarding the local interconnect of each cell, the upper and lower input multiplexers (A and B) of each core cell can be connected to the outputs of nearby cells using the network of local interconnect.

This provides the fastest connections between cells.

Thus in the illustrated embodiment the multiplexer A provides inputs to the core cell C from the outputs of cells U, LL, F, FB and FF, whilst the multiplexer B provides inputs to the core cell C from the outputs of cells UU, L, F, FB and FBB. Thus in the illustrated embodiment each cell connects with its eight nearest 15 perpendicular neighbours. Each cell C will have a set of local connection possibilities. le. each cell connects with only some of the other cells in a manner described in our BritiSh patent 2180382.

Referring now to Figure 4, this illustrates how 20 medium interconnect is used for connections within a zone that are not possible with local interconnect.

.i Medium buses are also used for interzone connections since they connect to the port cells. There are four horizontal and four vertical medium buses per row and column of core cells. All medium buses connect to the zones port cells, but only two can be used for 29 connection to the global routing network in the illustrated embodiment and these are referred to as Ml and M2 or the external medium buses. M3 and M4 "the internal" medium buses, are used for connections through the port cells to all four adjacent zones as will be apparent from reference to Figure 6 and 7.

Each of the two input multiplexers in a core cell provides connections from two medium buses. The core cell output multiplexers can connect' to four medium buses. Therefore each individual core cell can only access half the 8 medium buses that cross each cell. To compensate for this. there is a tiling of bus connections, using two different sets of connections, repeated in a 2 by 2 tile similar to the core cell 15 functional tiling. Thus, for the illustrated embodiment each core cell in the tile connects to its horizontal and vertical medium buses as set out in table 1 below.

'TABLE 1 Core Mux Connections from: Connections to: 20 cell 2/3 A M3(Vert),M2(Horiz) 2/3 B Ml(Vert),M4(Horiz) 2/3 OP M2,M3(Vert),M1,M4(Horiz) 1/4 A M2(Vert),M3(Horiz) S 25 1/4 B M4(Vert),Ml(Horiz) 1/4 OP M1,M4(Vert),M2,M3(Horiz) Figure 10 illustrates diagrammatically for illustrative purposes the hierarchy of interconnect and shows three interconnect structures, namely firstly, I 4.

the local interconnect structures which only have the scope of a few cells, and cannot connect to port cells, secondly the medium interconnect structures which only have the scope of a single zone but can directly connect with port cells and core cell, and thirdly the global interconnect which can span the whole array but cannot connect to core cells It will be apparent that these three levels of interconnect will be brought about by the existence of the porting arrangement that separates the array into zones. The global level of interconnect is separated from the medium and local levels of interconnect by the porting arrangement.

0* .o o

Claims (35)

1. A configurable semi-conductor integrated circuit comprising a matrix array of core cells each of the cells having a first simple function in common and at least one subsidiary function, there being at least two different subsidiary functions, the core cells being grouped in tiles comprising a matrix array of the core cells smaller that the whole array and wherein each tile has at least one of each different subsidiary functions, and wherein the tiles of core cells are arranged so as to uniformly cover the array.

2. A configurable semi-conductor integrated circuit claimed in claim 1 in which the arrangement of the subsidiary functions within the cells of the tile are substantially different.

3. A configurable semi-conductor integrated circuit which comprises an area thereof formed with a plurality of logic circuits at discrete sites or cells respectively, each said logic circuit having in common a restricted simple logic function capability and itself only being capable of implementing a simple logic function, and the cell having a restricted signal translation system between inputs and outputs of the logic circuits affording (local) direct connection paths each selectable as to its conduction state, those paths extending, for each said logic circuit, from its output to inputs of first set of some of other said logic circuits and from its inputs to outputs of a second set of some of other said logic circuits, each first set of logic circuits being different from any other first set of logic circuits, and each second set of logic circuits being different from any other second set of logic circuits, the integrated circuit further comprises at each discrete site or cell additional optionally selectable circuit configurations selectively controlled by control means to enable each cell to operate in a selected one of two or more ways, and wherein there are a plurality of different optional circuit 32 configurations and wherein the cells are arranged in groups, hereinafter referred to as tiles smaller than the whole array, with each tile having at least one of each of the different circuit configurations, and wherein the tiles of cells are arranged so as to uniformly cover the array.

4. A configurable semi-conductor integrated circuit as claimed in claim 3 in which the selected one of two or more ways comprises either the simple logic function common to each cell or a selected subsidiary function afforded by the optionally selectable circuit configurations. A configurable semi-conductor integrated circuit as claimed in claims 1 to 4 in which the common simple logic function is a NAND gate.

6. A configurable semi-conductor integrated circuit as claimed in claims 1 to 5 in which the optionally selectable circuit configurations (different subsidiary functions) afford optional functions of: wired -OR output buffer; XOR; D-type flip flop (with reset and enable) and latch function (with reset and enable).

7. A configurable semi-conductor integrated circuit as claimed in anyone of claims 1 to 6 in which some cells may have additional circuit configurations to facilitate by virtue of the tile arrangement building up higher level functions.

8. A configurable semi-conductor integrated circuit as claimed in claim 6 in which there are two XOR functions to a tile.

9. A configurable semi-conductor integrated circuit as claimed in claim 6 or 7 in which one cell offers optionally two functions. A configurable semi-conductor integrated circuit as claimed in claim 9 in which the two functions are D-flip flip and latch.

11. A configurable semi-conductor integrated circuit as claimed in anyone of claims 1 to 10 in which the tile comprises four cells. 33

12. A configurable semi-conductor integrated circuit as claimed in claim 11 in which the four cells are arranged in a matrix of 2 x 2 cells.

13. A configurable semi-conductor integrated circuit as claimed in claim 3 in which the arrangement of the subsidiary functions within the cells of the tile are substantially different.

14. A configurable semi-conductor integrated circuit as claimed in claims 1 or 2 or any one of claims 5 to 12 when appendant to claims 1 or 2 in which an area thereof is formed with a plurality of logic circuits at discrete sites or cells respectively defining said matrix array of cells, and wherein the matrix array of cells is sub-divided at least into zones each comprising a matrix array of said cells and further comprising a porting arrangement for each zone, the integrated circuit as made having an hierarchical routing resource structure comprising:- global connection paths which extend continuously across more than one zone and have selectable connections with the porting arrangement of at least some of the zones, (ii) medium connection paths extending from the porting arrangement and selectably connectable with at least some of the cells in a zone, and (iii) local direct connection paths comprising for each cell a restricted signal translation system between inputs and outputs of the cells each selectable as to its conduction state, those paths extending for each said logic circuit from its output to inputs of a first set of some of other said logic circuits and from its inputs to outputs of a second set of some of other said logic circuits, and each second set of logic circuits being different from any other second set of logic circuit. A configurable semi-conductor integrated circuit as claimed in claim 3 or any one of claims 4 to 12 when appendant to claim 3 in which the plurality of logic circuits 34 at discrete sites or cells respectively defining a matrix array of cells, and wherein the matrix array of cells is sub-divided at least into zones each comprising a matrix array of said cells and further comprising a porting arrangement for each zone, the integrated circuit as made having an hierarchical routing resource structure comprising:- global connection paths which extend continuously across more than one zone and have selectable connections with the porting arrangement of at least some of the zones, (ii) medium connection paths extending from the porting arrangement and selectably connectable with at least some of the cells in a zone, and (iii) local direct connection paths.

16. A configurable semi-conductor integrated circuit as claimed in claim 14 or in which the zones are arranged in a matrix array of zones.

17. A configurable semi-conductor integrated circuit as claimed in claim 14, 15 or 16 in which the plurality of zones define a quadrant and the integrated circuit has a plurality of quadrants.

18. A configurable semi-conductor integrated circuit as claimed in claim 17 in which the quadrants are arranged in a matrix array.

19. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 18 in which there are 10 x 10 cells to a zone, 5 x 5 zones to a quadrant and 2 x 2 quadrants. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 19 in which the global connection paths extend horizontally and vertically across a plurality of zones.

21. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 20 in which there are a plurality of global connection paths for each row and column of cells in a zone.

22. A configurable semi-conductor integrated circuit as claimed in claim 21 in which there are four global connection paths for each row and column of cells.

23. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 22 in which there are a plurality of medium connection paths for each row and column of cells of a zone.

24. A configurable semi-conductor integrated circuit as claimed in claim 23 in which there are four medium connection paths for each row and column. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 24 in which each medium connection path is selectably connectable with an output and/or input of one or more cells.

26. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 25 and comprising further global routing resources comprising horizontal and vertical buses extending across a zone horizontally and vertically respectively, and which are connectable with the port cells and with one another.

27. A configurable semi-conductor integrated circuit as claimed in claim 26 in which the further horizontal and vertical buses are not connectable to the cells.

28. A configurable semi-conductor integrated circuit as claimed in claim 26 or 27 in which there is one such further global connection bus for each row and column of cells.

29. A configurable semi-conductor integrated circuit as claimed in claim 28 in which there is a connection between the horizontal and vertical buses for each cell. 36 A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 29 comprising horizontal and vertical port cells, the porting arrangement of which permits resources to be routed from the global lines through to the medium lines and thereby into the cells.

31. A configurable semi-conductor integrated circuit as claimed in any one claims 14 to 30 in which the general interconnect structure (global and medium buses) are interconnectable with special interconnect structure, for example clocks and tristates, by way of the port cells.

32. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 31 further comprising clock distribution channels, comprising a vertical clock bus extending between clock pads to top and bottom of the array, a horizontal clock distribution spine extending between clock pads to opposite sides of the array and intersecting with the vertical clock bus.

33. A configurable semi-conductor integrated circuit as claimed in claim 32 in which the clock bus and the horizontal clock distribution spine each have 8 lines.

34. A configurable semi-conductor integrated circuit as claimed in claim 32 or 33 in which there is a vertical clock distribution spine for each zone by which clock and reset signals are distributed to the vertical port cell of that zone. A configurable semi-conductor integrated circuit as claimed in claim 34 in which the vertical clock distribution spine is tapped off from the horizontal clock spine.

36. A configurable semi-conductor integrated circuit as claimed in any one of claims 32 to 35 in which alternate cells have clock and reset signal lines.

37. A configurable semi-conductor integrated circuit as claimed in any one of claims 14 to 36 in which each cell has a common logic function and at least one E" 37 subsidiary function, there being at least two different subsidiary functions, the cells being grouped in tiles comprising a matrix array of the core cells and wherein each tile has at least one of each different subsidiary functions.

38. A configurable semi-conductor integrated circuit as claimed in claim 37 in which the different subsidiary functions comprise:- wired-OR; XOR; D-type flip flop and latch functions.

39. A configurable semi-conductor integrated circuit as claimed in claims 37 or 38 when appended to claim 30 comprising dedicated connection lines with the horizontal port cell comprising a plurality of wired -OR buses running across each zone and connecting directly with the wired -OR core cells. A configurable semi-conductor integrated circuit as claimed in claim 39 in which there are 5 wires -OR buses running across each zone.

41. A configurable semi-conductor integrated circuit as claimed in claim 39 or in which connection of the wired -OR buses is made to a certain (restricted) number of global buses, for example by way of the horizontal port cells.

42. A configurable semi-conductor integrated circuit as claimed in claim 31 when appendant to claim 26 in which there are wire -OR connections between the further global horizontal buses and a plurality of the horizontal global buses.

43. A configurable semi-conductor integrated circuit constructed and arranged substantially as hereinbefore described with reference to and as illustrated in the accompanying drawings. DATED THIS 2 2 ND DAY OF JUNE 2000 PILKINGTON MICRO-ELECTRONICS LIMITED By their Patent Attorneys LORD COMPANY PERTH, WESTERN AUSTRALIA