WO2001028097A1

WO2001028097A1 - Heterogeneous interconnection architecture for programmable logic devices

Info

Publication number: WO2001028097A1
Application number: PCT/US2000/028189
Authority: WO
Inventors: Vaughn Betz; Jonathan Rose
Original assignee: Altera Corporation
Priority date: 1999-10-12
Filing date: 2000-10-12
Publication date: 2001-04-19
Also published as: JP2003511947A; JP4847663B2; JP5285718B2; WO2001028097A9; US6828824B2; EP1224737A1; JP2011087334A; US20040017222A1; US6590419B1

Abstract

This invention relates to a PLD comprising one or more function blocks; and a plurality of groups of interconnect resources each group of a selected type for programmable connection to one or more of the function blocks, and wherein a first number of at least one type of interconnect resources being optimized for a first operational parameter of the PLD and a second number of the same type being optimized for a second operational parameter. For example the operational parameters may include speed, area, power, reliability and flexibility.

Description

HETEROGENEOUS INTERCONNECTION ARCHITECTURE FOR PROGRAMMABLE LOGIC DEVICES This invention relates to programmable logic devices (PLDs) and more particularly to an improved interconnect architecture for such devices.

BACKGROUND OF THE INVENTION

Programmable Logic Devices (PLDs) are a widely used form of integrated circuit due to the flexibility provided by their customizable nature. In general PLDs include field programmable gate arrays (FPGAs), complex programmable logic devices (CPLDs), simple programmable logic devices and laser programmable devices. Architecturally a PLD includes logic blocks and input/output (I/O) blocks, which are connectable through a programmable interconnect structure or network.

A typical PLD is an integrated circuit chip that, wholly or in part, consists of an array of one or more logic blocks, I O and a programmable routing or interconnect network that can be programmed by a user to provide an interconnection between the logic blocks and I/O circuits to achieve a desired logic function. The PLD can be a standalone device or be embedded in a larger integrated circuit such as ASICs or the like. Exemplary forms of such devices are disclosed in US Patent No. 5,825,202 and US Patent No. 5,687,325.

The logic blocks may be comprised of a fixed logic function or may in turn also have programmable interconnections or functionality. The logic blocks may be further broken down as sub-blocks or grouped together as a cluster of blocks. The blocks may also include input/output circuits. Typically the I O circuits enable connection of the chip to external circuits or to other parts of the chip as in the case of embedded FPGAs. The I O blocks are typically arranged at the periphery of a chip. A PLD is normally arranged as a regular array of logic blocks each of which may be identical or may be of several different types such as RAM blocks, Look-Up-Table based blocks, P-term based blocks etc. The conductors of the programmable interconnect network array are typically arranged along rows and columns defined by the array of logic blocks as shown schematically in figure 1.

The interconnect structure of the PLD consists of pre-fabricated wires and pre- fabricated switches which can be programmed to electrically connect different logic blocks to provide a desired function. The connections between the conductors (wires) and the logic blocks and between different wire segments is implemented by means of programmable switches at predetermined interconnection points. The programmable switches may be implemented as pass- transistors, tri-state buffers, fuses, antifuses or combinations thereof. Laser programming of interconnections may also be achieved by burning off the metal conductors at desired locations. In some cases a switch state can be controlled by a Static or Dynamic Random Access Memory (SRAM or DRAM), Read Only Memory (ROM), Electrically Programmable Read Only Memory (EPROM), Electrically Erasable and Programmable Read Only Memories (EEPROMs), Flash memory or any other variation or combination of the above memory. Various types of switches are well known in the art and are described for example in US Patent No. 4,870,302 and Patent No. 5,955,751 to list a few.

There are a number of tradeoffs involved in the design of PLDs. A PLD designer or architect is constrained by various operational parameters such as speed of circuits implemented in the PLD, semiconductor or silicon area required for a given logic capacity, power dissipation of the PLD once it has been programmed by a user, reliability (e.g. low glitch probability) and routing flexibility. A PLD architect cannot know before hand all the many uses for which a user may program the device. Thus PLDs are designed to be as general as possible.

For example, a PLD architect's objective may be to minimize the surface area of the PLD while maximizing the logic capacity. That is, maximize the number of and functionality of logic blocksand provide sufficient interconnect resources and flexibility to allow as broad a range of circuit implementations as possible. Another design objective may be to maximize the speed of circuits implemented in the PLD while minimizing the surface area, logic and interconnect resources required to implement the circuits. For example, a PLD architect knows that speed can be improved by reducing the number of programmable connection switches, but this comes at the expense of routing flexibility. Reliability, in terms of glitch prevention can be improved by spacing interconnect wires further apart, but this comes at the expense of area.

Consequently, one area of focus in PLD architecture design is the interconnect network architecture, where the objective is to minimize the surface area required for the interconnect resources while maximizing the speed and minimizing the PLD resources required to implement circuits on the PLD. The interconnect architecture here refers to interconnect conductors, the programmable switches and the interface between them and the logic blocks. The reader is referred to US Patent No. 5,907,248, which provides a background on various interconnection architectures and improvements thereto.

However, circuits that are implemented in PLDs tend to have different parts with different requirements. For example, some critical paths of the circuit have to be fast to meet timing requirements, some parts (e.g. clock signals) need to be glitch free and parts which are not speed sensitive should use as little silicon area or PLD resources as possible.

Most PLD architectures do not take into account this circuit heterogeneity. Most of the prior art use a homogenous interconnection architecture wherein parameters such as wire widths, spacing between wires, dimensions of the transistors used in the buffers and switches, interface circuits between the wires/switches and the logic blocks are constant throughout the PLD.

Furthermore homogenous architectures are easier for the CAD tools that are used to map circuit descriptions into the PLD architecture.

Current interconnect architectures have considered varying a single parameter in order to optimize the PLD. For example Actel's US Patent No. 5,073,729, Altera's US Patent No. 5,900,743 and Xilinx US Patent No.'s 5,801,546 and US 5,907,248, describe PLD architectures in which optimization is limited to using different lengths of interconnect wires. On the other hand, US Patent No. 5,942,913 describes an interconnect structure using a mixture of buffered and unbuffered interconnect lines. In both cases, optimization is limited to just one parameter.

However, these architectures are limited in that they do not offer a user sufficient flexibility when designing a circuit that requires different PLD interconnect resources for different parts of the circuit. For example in U.S. Patent No. 5,900,743 horizontal conductors of many different lengths makes it possible to make interconnections between horizontally aligned logic regions using conductors that are close to the appropriate lengths for making that connection. Accordingly there is a need for an improved PLD architecture that is optimised in terms of one or more selected operational parameters such as speed, power, area, flexibility and reliability, while minimizing the impact on the remaining parameters.

SUMMARY OF THE INVENTION An advantage of the present invention is the provision of a PLD architecture that is optimized for a selected one or more operational parameters, while minimizing the impact on the remaining parameters. For example the operational parameters may include speed, area, power, reliability and flexibility.

In accordance with this invention there is provided a PLD comprising: (a) one or more function blocks; and (b) a plurality of groups of interconnect resources each group of a selected type for programmable connection to one or more of the function blocks, and wherein a first number of at least one type of interconnect resources being optimized for a first operational parameter of the PLD and a second number of the same type being optimized for a second operational parameter. In a further embodiment of the invention the function blocks are logic blocks.

In another embodiment of the invention the function blocks are input/output circuits. In one embodiment the interconnect resource includes a plurality of interconnect conductors (or wires) wherein a first number of conductors is selected to have spacing between adjacent conductors that are less than that of the spacing between conductors of a second number of conductors. For example, portions of wires may be spaced at one unit and three units apart. The wires that are spaced further apart will have much lower coupling capacitances than the wires spaced closely together. The lower coupling capacitance reduces the signal delay through the wires, increasing the achievable speeds of circuits implemented in the PLD. The wires that are spaced closer together occupy less area, and are thus optimized for area. In a further embodiment the interconnect resources includes a plurality of interconnect conductors wherein a first number is selected to have conductor widths that are different compared to the conductors widths of a second number or conductors. The wires that have larger widths have reduced resistance, which leads to increased speed, particularly for long wires. The wires with the narrower widths occupy less area. In a further embodiment the interconnect resource includes a group of interconnect conductors (or wires) wherein a portion of the routing wires are connected by switches of a first size and a second portion of the wires are connected by switches of a second size. The switches with larger dimension have increased drive strength, which decreases the time required to charge and discharge the routing wire capacitance. The net result is increased speed. The smaller switches, although slower, occupy less area. In a preferred embodiment the switch is composed of pass transistors. In an alternate embodiment the switch is composed of tri-state buffers, and a speed advantage can be achieved through larger transistors or by changing the number of stages (buffer sub-blocks) in the tri-state buffer.

In a further embodiment, the interconnect resource includes a plurality of multiplexers or tri-state buffer switches for coupling to the interconnect conductors (or wires) wherein a first number of the multiplexer or tri-state buffers have a different circuit topology than or a second number or multiplexer or tri-state buffers. In a preferred embodiment, the multiplexer or tri-state buffer switches have a circuit topology that yields the highest speed for some switches, and a topology that is most area-efficient for other switches. For example, a 16:1 multiplexer can be implemented as sixteen independent pass transistors (a single-stage 16:1 multiplexer) or as a set of 2: 1 multiplexers cascaded together in four (4) stages. The single-stage 16:1 multiplexer would have better speed, but requires a larger area.

In a further embodiment, the interconnect resource includes a plurality of programmable switches for connecting the interconnect conductors (or wires) to the logic block input or outputs or for connecting two different interconnect conductor segments, and wherein a first number of the wires have fewer routing switches connected to them (hence smaller capacitive load on them, hence faster) and a second number of the wires have a larger number of switches connected to them. Thus, by reducing the number of routing switches connecting to some of the wires, the capacitance load on the wires is reduced. For example, some wires may have twenty (20) programmable switches which can drive them, while others may have only ten (10). The wires which can be driven by fewer switches, will have higher speed, because they have less parasitic loading due to switches, and they will require less switch area per wire. By reducing the number of switches that can connect to only some of the routing wires, a gain in speed of more lightly loaded wires is made without reducing the flexibility of the PLD. As well, it is also often possible to gain area savings by reducing the number of switches on only some of the wires in this manner.

In a still further embodiment, the interconnect resources includes a plurality of routing wires with a portion of the wires having extra-fast paths into logic blocks or I O blocks and the second portion having more flexible routing. This may be implemented with multiplexers coupled between routing wires and logic block (or I/O block) input pins in an unbalanced tree topology. In a still further embodiment, the interconnect resource includes two types of input pins at the interface between the logic block input and the interconnect wires. One set of input pins is optimized for speed and the other for routing flexibility.. By having the extra fast wires connect to special input pins on the logic blocks, these input pins can be driven by a ^mailer number of routing wires, and hence have smaller (and faster) multiplexers between them and the routing wires. The second group of pins will be driven by a larger number of routing wires, hence allowing a larger number of wires to connect to the logic block, hence improving PLD flexibility.

In a still further embodiment, the interconnect resources includes two different types of buffers at the interface between the logic block or I/O block output and the interconnect wires. Each output pin is connected to two groups of buffers. One group of buffers has lower intrinsic delay by virtue of having fewer stages and lower overall delay by virtue of having fewer interconnect wires that can be connected to it. The second group has more programmable switches loading it, hence it is slower, but has higher routing flexibility. The net effect is that speed critical connections can use the faster, but less flexible, group of buffers, while the non- speed-critical connections can use the slower more flexible group of buffers. As a result, the speed of a circuit implemented in the PLD is enhanced while routing flexibility is maintained.

In another embodiment the PLD includes a number of wires having either increased spacing between the wires or including a grounded shield wire between them or by using differential signaling for reducing unwanted transitions or glitches on the wires. The other wires use a smaller spacing and no shield wires and hence are optimized for area.

In a further embodiment of the invention, a given type of interconnect resource may be optimized for two or more operational parameters.

In a further embodiment, two or more different types of interconnect resources may each be optimized for two or more operational parameters.

BRIEF DESCRIPTION OF THE DRAWINGS

An embodiment of the invention will now be described by way of example only, with reference to the following drawings in which Figure 1 is block diagram of a typical PLD architecture according to the prior art; Figure 2 is a block diagram of a PLD architecture in accordance vvi.li one embodiment of the present invention;

Figure 3 is a graph iiiu-.ιrating the correlation between the speed of a circuit and fraction of widely spaced tracks used; Figure 4(a) is a schematic diagram of a high speed multiplexer;

Figure 4(b) is a schematic diagram of a low speed multiplexer;

Figure 5 is a schematic diagram of PLD architecture with different numbers of connect wires;

Figure 6(a) is a schematic diagram of a balanced multiplexer; Figure 6(b) is a schematic diagram of an unbalanced multiplexer according to an embodiment of the present invention;

Figure 7 is a schematic diagram showing heterogeneous inputs to a logic block; and

Figure 8 is a schematic diagram of a heterogeneous interface between an output of a logic block and the routing wires.

DETAILED DESCRIPTION OF A PREFERRRED EMBODIMENT

For convenience, like numerals in the description refer to like structures in the drawings.

Referring to Figure 1, a block diagram of a typical PLD architecture according to prior art as shown generally by numeral 10. The PLD comprises a number of logic blocks 12 arranged in a regular array with interconnect wires 14 arranged in vertical and horizontal channels between the logic blocks 12. The logic block 12 may perform one of many functions and the implementation of which is well known in the art. Typically a logic block will have input or output pins 15. Programmable switches 16 are used for connecting the interconnect wires 14 to the logic block input or output pins. Programmable switches 17 are also located at intersections of the interconnect wires 14 for connecting two different interconnect wires together.

The present invention is based on an observation that a detailed design or architecture of an interconnect network in a PLD has to take into account a very wide range of parameters, which affect the electrical characteristics of the interconnect resources. These characteristics include the interface between the logic block and the interconnect wires; the width and spacing between the interconnect wires; the type of programmable switch (pass-transistor or tri-state buffer) used to connect each interconnect wire to other wires; the dimensions of transistors used to build the programmable switches; the distribution of programmable switches in the interconnect network; the segmentation distribution; that is the different lengths of interconnect wire segments in the PLD, and the fraction of wires 14 in an interconnect channel. The choice of each involves balancing complex trade-offs and interactions with other parameters, a more detailed explanation of these tradeoff are described in "Architecture and CAD for Deep- Submicron FPGAs" by Betz et al, Chapter 7, and incorporated herein by reference.

Accordingly, referring to figure 2, a PLD architecture according to an embodiment of the invention is represented generally by the numeral 20. The logic blocks 12 are arranged in a regular array, with the interconnect resources shown schematically as rows and columns forming a network around the logic blocks 12. The interconnect resources include switches, wires and logic block interconnect interfaces. In the schematic diagram of figure 2, wires that are optimized for speed are indicated by a heavy black line while wires optimized for area density are indicated by a lighter black line. The high speed wires are spaced further apart than the low speed wires. The wider spaced wires 24 represent a fraction of the total wires and are spaced in order to improve the speed at which they operate. The narrowly spaced wires 26 are spaced at a minimum distance such that the circuit may operate properly and the surface area of the PLD is optimized.

Figure 3 shows the average speed achieved by a set of twenty (20) benchmark circuits when implemented in a realistic PLD architecture. The vertical axis represents the critical path delay. The horizontal axis represents the fraction of routing tracks laid out with a wide metal spacing. In a preferred embodiment, this width is greater than the minimum spacing allowed in the PLD manufacturing process typically 2.5 μm. The remaining tracks are laid out with the minimum metal spacing required for the circuit to be able to operate properly. In the preferred embodiment, the minimum width is equal to the minimum spacing allowed in the PLD manufacturing process typically 0.5 μm.

It is possible to ascertain from the graph that a PLD in which all the wires are widely spaced out is 15% faster than a PLD in which all of the wires are minimally spaced. However, if approximately 20% of the tracks are widely spaced out, it is still possible to obtain a 13% circuit speedup. Therefore, increasing the spacing of only 20% of the routing tracks minimizes the increase in metal area required, but still yields almost all of the achievable speedup. Although it appears as if 20% is the optimal percentage to use for maximizing the speed of an FPGA while minimizing its area, it need not necessarily be exactly 20%. The exact percentage depends on the PLD manufacturing process and on the PLD architect. It may be that the PLD architect decides that the space required to widen 20% of the wires is too great a trade off for the resulting speedup and therefore widens a smaller percentage of wires. Conversely, the architect may decide that the increase in area is worth the trade off for obtaining even a faster speed, and therefore widens a greater percentage of the wires. The PLD also maintains its flexibility for implementing a wide variety of circuits.

In alternate embodiments, a similar result to that described in the preferred embodiment can be achieved. As previously mentioned, some of the factors influencing the speed of the circuit, in addition to the spacing between wires, are the wire width, the switch size, logic block size, circuit design of switches, the circuit design of the interface between the logic block and the interconnect, and the like. It is possible to improve the speed of the PLD with a minimal increase in area for all of these factors in a similar fashion as described above. In one scenario for example, a portion of one of the interconnect resources is optimized for speed and the remainder of that resource is optimized for area. The rest of the interconnect resources are implemented at a constant width, or spacing (depending on the resource) that is a compromise of speed and area. For example, it is possible to alter the size of a portion of the switches while maintaining equal wire widths and spacing between wires throughout the circuit. Then, as previously described the PLD will achieve a maximum speed for a minimum area while maintaining its flexibility to be able to implement a wide variety of circuits. Further, since all of the components may not have the exact same effect on the circuit, the fraction of components that need to be optimized for speed may vary from component to component.

In a further embodiment, more than one component is optimized at a time. For example, it is possible to increase both a portion of the switch sizes and a portion of the wire widths for optimizing the area used to get a majority of the possible speedup. This embodiment is not limited to combining only two of the components. Rather, any number of components up to, and including, all possibilities may be optimized at one time.

Referring back to figure 2 then the interconnect resource includes a plurality of interconnect conductors or wires wherein a first portion is selected to have spacing between adjacent conductors that are less than that of the spacing between conductors of the remaining portion. The wires that are ^<^?Γ° urth r apart will h^vc ^m^ch lower coupling capacitances than the wires spaced closely together. The lower coupling capacitance reduces the signal delay through the wires, increasing the achievable speeds of circuits implemented in the PLD. The cost is the increased metal area (decreased wiring density) caused by the larger metal pitch - this may increase the overall PLD area.

In a further embodiment the interconnect resource includes a plurality of interconnect conductors or wires of the first and second type wherein the first type is selected to have conductor widths that are different to that of the conductors widths of the second type The wires that have larger widths have reduced resistance, which leads to increased speed, particularly for long wires. The cost is the increased metal area (decreased wiring density) caused by the larger metal pitch - again this may increase the overall PLD area.

In a further embodiment the interconnect resource includes a plurality of interconnect conductors or wires, the routing wires being coupled by switches of one of the first or second type wherein the switches of the first type are different in size to the switches of the second type. This increases the drive strength of these switches, which decreases the time required to charge and discharge the routing wire capacitance. The net result is increased speed. The cost is the increase in active area necessitated by these larger switches - again this may increase the overall PLD area.

In a further embodiment as shown schematically in figures 4(a) and (b) the interconnect resource includes a plurality of multiplexers or tri-state buffer switches for coupling to the interconnect conductors or wires wherein the multiplexer or tri-state buffers of the first type have a different switch topology than the multiplexer or tri-state buffers of the second type. In a preferred embodiment the multiplexer or tri-state buffer switches have a circuit topology that yields the highest speed for some switches, and a topology that is most area-efficient for other switches. For example, one could implement a 16:1 multiplexer as 16 independent pass transistors (a single-stage 16: 1 multiplexer) or as a set of 2:1 multiplexers cascaded together in four (4) stages. The single-stage 16:1 multiplexer would have better speed, but requires a larger area.

In a further embodiment, illustrated in figure 5, the interconnect resource includes a plurality of routing switches for coupling to the interconnect conductors or wires wherein the switches are arranged in groups of the first or second type and which are each different in number. Thus by reducing the number of routing switches connecting to some of the routing wires. For example, some routing wires may have twenty (20) routing switches, which can drive them, while others may have only ten (10). The wires, which can be driven by fewer switches, will have higher speed, because they have less parastic loading due to switches, and they will require less switch area per wire. The cost is that the wires that connect to fewer switches have reduced routing flexibility — they cannot be joined in as many different configurations as routing wires that connect to a larger number of routing switches. If too many wires have reduced routing flexibility, the routability of the PLD will be significantly impacted — either many application circuits will not route (i.e. will not work) on this PLD, or more routing wires will have to be added to the PLD to compensate for the reduced flexibility. Adding more routing wires to the PLD to compensate for the reduced flexibility will increase the PLD area. By reducing the number of switches that can connect to only some of the routing wires, we can gain the speed advantage of more lightly loaded wires without overly reducing the flexibility of the PLD. As well, it is also often possible to gain an area savings by reducing the number of switches on only some of the wires in this manner.

In a still further embodiment as shown schematically in figures 6(a) and (b), the interconnect resource includes a plurality of routing wires which have extra-fast paths into logic blocks or I O blocks. This can be achieved by making the multiplexers between routing wires and logic block (or I/O block) input pins have an unbalanced tree topology. The extra-fast wires connect to multiplexer inputs that go through fewer stages of multiplexing than the regular wires do. The net result is that some fraction of the wires have an extra-fast path to the logic block input pins. Only a fraction of the routing wires can use this extra-fast path to the logic block input pins, however - it is simply not possible to make more than a fraction (which is always less than 1/2) of the inputs to a multiplexer extra-fast by unbalancing the tree, (see Fig. 5) In a still further embodiment the PLD includes groups of routing wires having extra-fast paths into logic blocks and wherein the logic blocks have dedicated inputs for these extra-fast paths. By having these wires connect to special input pins on the logic blocks; these input pins can be driven by a smaller number of routing wires, and hence have smaller (and faster) multiplexers between them and the routing wires. The number of such extra-fast input pins is kept limited (most input pins have higher flexibility, and hence larger and slower multiplexers) in order to keep the decrease in PLD routing flexibility small as shown in figure 7. In a further embodiment the PLD includes wires that can be driven by logic block or I O block output pin with extra-low delay. This can be achieved by using a separate output pin driver (buffer) with lower intrinsic delay because it has fewer stages. Because it has fewer stages, this output buffer has less drive strength than a standard output pin driver, so it is important that it drive only a subset of the wires that a logic block driver would ordinarily drive. A separate driver with more stages can be used to drive the regular routing wires to avoid a loss in routing flexibility, as shown in figure 8.

In a further embodiment, the heterogeneity can be expanded further by optimizing some wires for "glitch prevention." In this embodiment the PLD includes at number of wires having either increased spacing between the wires or including a grounded shield wire between them or by using differential signaling for reducing unwanted transitions or glitches on the wires. For example, it is often necessary to reduce the crosstalk between adjacent wires in a PLD below some threshold in order to guarantee that a switching transition on one wire would not lead to unwanted transition on an adjacent wire due to parasitic coupling. This unwanted transition or "glitch" would result in an incorrect operation of a circuit implemented in the PLD. If only a fraction of interconnect wires are designed to be glitch free (typically by increasing the spacing between the wires or by adding a grounded shield wire between them or by using differential signaling), the PLD can offer high reliability operation without significantly increasing the area requirements. To use this feature, the CAD tool used to place and route circuits into the PLD must be smart enough to be aware of which routing wires are "glitch-free" and be able to intelligently route (either through user input or through analysis of the circuit netlist) critical signals (such as clock signals or asynchronous control signals) only on the "glitch-free" wires.

Although the invention has been described with reference to certain specific embodiments, various modifications thereof will be apparent to those skilled in the art without departing from the spirit and scope of the invention as outlined in the claims appended hereto.

Claims

THE EMBODIMENTS OF THE INVENTION IN WHICH AN EXCLUSIVE PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:

1. A PLD comprising: (a) one or more function blocks; and

(b) a plurality of groups of interconnect resources each group of a selected type for programmable connection to one or more of the function blocks, and wherein a first number of at least one type of interconnect resources being optimized for a first operational parameter of the PLD and a second number of the same type being optimized for a second operational parameter.

2. A PLD as defined in claim 1, said first operational parameter being speed and said second parameter being area.

3. A PLD as defined in claim 1, said first operational parameter being speed and said second parameter being flexibility.

4. A PLD as defined in claim 1, said first operational parameter being speed and said second parameter being power.

5. A PLD architecture comprising:

(a) a plurality of logic blocks; and

(b) an electrically heterogeneous interconnect network for coupling the logic blocks whereby the network consists of one or more interconnect resources having portions of each resource optimized for one operational parameter and the other portion optimized for another operational parameter.

6. A method for designing a PLD, comprising the steps of: (a) selecting one or more logic blocks; (b) arranging said logic blocks in an array; (c) selecting a first portion of an interconnect resource, said first portion being optimized for a first operation parameter of said PLD;

(d) selecting a second portion of said resource, said second portion being optimized for a second operation parameter of said PLD; and

(e) arranging said first and second portions for connection to said logic blocks.