Clock driven dynamic datapath chaining

Abstract

A system includes a plurality of datapaths, each having structural arithmetic elements to perform various arithmetic operations based, at least in part, on configuration data. The system also includes a configuration memory coupled to the datapaths, the configuration memory to provide the configuration data to the datapaths, which causes the datapaths to collaborate when performing the arithmetic operations.

Description

TECHNICAL FIELD

The present disclosure relates generally to programmable devices, and more particularly to dynamically chaining Universal Digital Blocks (UDB).

BACKGROUND

Field-programmable gate arrays (FPGAs) and programmable logic devices (PLDs) have been used in data communication and telecommunication systems. Conventional PLDs and FPGAs consist of an array of programmable elements, with the elements programmed to implement a fixed function or equation. Some currently-available Complex PLD (CPLD) products comprise arrays of logic cells. Conventional PLD devices have several drawbacks, such as high power consumption and large silicon area.

In developing complex electronic systems, there is often a need for additional peripheral units, such as operational and instrument amplifiers, filters, timers, digital logic circuits, analog to digital and digital to analog converters, etc. As a general rule, implementation of these extra peripherals creates additional difficulties: extra space for new components, additional attention during production of a printed circuit board, and increased power consumption. All of these factors can significantly affect the price and development cycle of the project.

The introduction of Programmable System on Chip (PSoC) chips feature digital and analog programmable blocks, which allow the implementation of a large number of peripherals. A programmable interconnect allows analog and digital blocks to be combined to form a wide variety of functional modules. The digital blocks consist of smaller programmable blocks and are configured to provide different digital functions. The analog blocks are used for development of analog elements, such as analog filters, comparators, inverting amplifiers, as well as analog to digital and digital to analog converters. Current PSoC architectures provide only a coarse grained digital programmability in which a few fixed functions with a small number of options are available.

SUMMARY

A system includes a plurality of datapaths, each having structural arithmetic elements to perform various arithmetic operations based, at least in part, on configuration data. The system also includes a configuration memory coupled to the datapaths, the configuration memory to provide the configuration data to the datapaths, which causes the datapaths to collaborate when performing the arithmetic operations.

The system includes a first datapath that receives a chained carry-in value from a second datapath for utilization in the arithmetic operations. The first datapath includes a multiplexor to at least receive a chained carry-in value from the second datapath, the multiplexor to provide the first datapath the chained carry-in value according to the configuration data from the configuration memory.

The datapaths receive the configuration data based on a clock cycle, and the datapaths are dynamically chained to collaborate on the arithmetic operations according to configuration data and the clock cycle. The datapaths are operable to perform time-division multiplexed arithmetic operations according to the configuration data, where the datapaths are dynamically chained to collaborate on the arithmetic operations during a first clock cycle and unchained to perform independent arithmetic operations during a second clock cycle.

A method includes receiving a chained carry-in value from a datapath and configuration data from a configuration memory, selecting the chained carry-in value according to the configuration data from the configuration memory, and collaborating with the datapath in the performance of arithmetic operations by utilizing the chained carry-in selected according to the configuration data when performing arithmetic operations.

The method includes multiplexing multiple input values according to the configuration data, where at least one of the input values is the chained carry-in value received from datapath. The method includes receiving the configuration data based on a clock cycle, and dynamically chaining to the datapath for collaboration on the arithmetic operations according to configuration data and the clock cycle. The method includes dynamically chaining to the datapath for collaborate on the arithmetic operations during a first clock cycle, and dynamically unchaining to perform independent arithmetic operations during a second clock cycle.

A device includes a structural arithmetic circuit to perform various arithmetic operations based, at least in part, on dynamic configuration data. The device also includes a selection circuit to receive a chained carry-in value from a datapath, to select the chained carry-in value according to the dynamic configuration data, where the structural arithmetic circuit performs arithmetic operations by utilizing the chained carry-in selected.

The selection circuit is operable to multiplex multiple input values according to the configuration data, where at least one of the input values is the chained carry-in value received from the datapath. The selection circuit is operable to receive the configuration data based on a clock cycle, and the structural arithmetic circuit to dynamically chain to the datapath for collaboration on the arithmetic operations according to configuration data and the clock cycle.

The structural arithmetic circuit is operable to dynamically chain to the datapath for collaborate on the arithmetic operations during a first clock cycle, and then dynamically unchaining to perform independent arithmetic operations during a second clock cycle. The structural arithmetic circuit is operable to dynamically chain and unchain to the datapath on alternate clock cycles to time-division multiplex performance of arithmetic operations.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic block diagram illustrating an example Programmable System on a Chip (PSoC) architecture that includes a Universal Digital Block (UDB) array.

FIG. 2 is a schematic block diagram illustrating one of the UDBs in FIG. 1 that includes both uncommitted PLD blocks and a structural dedicated datapath block.

FIG. 3 is a schematic block diagram illustrating the UDB in FIG. 2 in more detail.

FIG. 4 is a schematic block diagram also showing a datapath block in FIG. 2 in more detail.

FIG. 5 is a schematic block diagram illustrating example embodiments of the datapath block with chaining.

FIG. 6 is a block diagram illustrating example embodiments of datapath chaining with the datapath block shown in FIG. 5.

FIG. 7 is a schematic block diagram showing a multiple timer implementation utilizing datapath block chaining shown in FIG. 5-FIG. 6.

FIG. 8 is a flow diagram showing how a micro-controller or other Central Processing Unit (CPU) programs the UDBs.

DETAILED DESCRIPTION

A new Universal Digital Block (UDB) architecture combines PLDs and a datapath module in the same digital logic block to allow for the implementation of universal embedded digital functions. The new UDB architecture includes an integrated ALU that removes limitations associated with fixed functionality and provides users with the ability to customize digital operations to match system requirements.

FIG. 1 is a high level view of a Universal Digital Block (UDB) array 110 contained within a Programmable System on a Chip (PSoC) Integrated Circuit (IC) 100. The UDB array 110 includes a programmable interconnect matrix 130 that connects together different UDBs 120. The individual UDBs 120 each include a collection of uncommitted logic in the form of Programmable Logic Devices (PLDs) and structural dedicated logic elements that form a datapath 210 shown in more detail in below.

UDB Array

The UDB array 110 is arranged into UDB pairs 122 that may or may not be connected together through the interconnect matrix 130. The UDB pairs 122 each include two UDBs 120 that can be tightly coupled to a shared horizontal routing channel 132. The UDB pairs 122 can also be programmably connected to the horizontal routing channels 132 of other UDB pairs 122 either in the same horizontal row or in different rows through vertical routing channels 134. The horizontal and vertical routing channels and other switching elements are all collectively referred to as the interconnect matrix 130.

A Digital System Interconnect (DSI) routing interface 112 connects a micro-controller system 170 and other fixed function peripherals 105 to the UDB array 110. The micro-controller system 170 includes a micro-controller 102, an interrupt controller 106, and a Direct Memory Access (DMA) controller 108. The other peripherals 105 can be any digital or analog functional element that is preconfigured in PSoC 100. The DSI 112 is an extension of the interconnect matrix 130 at the top and bottom of the UDB array 110.

UDB

FIG. 2 is a top-level block diagram for one of the UDBs 120. The major blocks include a pair of Programmable Logic Devices (PLDs) 200. The PLDs 200 take inputs from the routing channel 130 and form registered or combinational sum-of-products logic to implement state machines, control for datapath operations, conditioning inputs and driving outputs.

The PLD blocks 200 implement state machines, perform input or output data conditioning, and create look-up tables. The PLDs 200 can also be configured to perform arithmetic functions, sequence datapath 210, and generate status. PLDs are generally known to those skilled in the art and are therefore not described in further detail.

The datapath block 210 contains highly structured dedicated logic that implements a dynamically programmable ALU, comparators, and condition generation. A status and control block 204 allows micro-controller firmware to interact and synchronize with the UDB 120 by writing to control inputs and reading status outputs.

A clock and reset control block 202 provides global clock selection, enabling, and reset selection. The clock and reset block 202 selects a clock for each of the PLD blocks 200, the datapath block 210, and status and control block 204 from available global system clocks or a bus clock. The clock and reset block 202 also supplies dynamic and firmware resets to the UDBs 120.

Routing channel 130 connects to UDB I/O through a programmable switch matrix and provides connections between the different elements of the UDBs in FIG. 2. A system bus interface 140 maps all registers and RAMs in the UDBs 120 into a system address space and are accessible by the micro-controller 102 shown in FIG. 1.

The PLDs 200 and the datapath 210 have chaining signals 212 and 214, respectively, that enable neighboring UDBs 120 to be linked to create higher precision functions. The PLD carry chain signals 212 are routed from the previous adjacent UDB 120 in the chain, and routed through each macrocell in both of the PLDs 200. The carry out is then routed to the next UDB 120 in the chain. A similar connectivity is provided for the set of conditional signals generated by the datapath chain 214 between datapath blocks 210 in adjacent UDBs 120. Embodiments of clock driven chaining of multiple datapaths 210 will be described below in greater detail corresponding to FIG. 5-FIG. 8.

Referring to FIG. 3, each UDB 120 comprises a combination of user defined control bits that are loaded by the micro-controller 102 into control register 250. The control register 250 is part of the control blocks 202 and 204 described above in FIG. 2. The control register 250 feeds uncommitted programmable logic 200 and control for datapath inputs. The same control blocks 202 and 204 described above in FIG. 2 also include associated status registers 256 that allow the micro-controller 102 to selectably read different internal states for both the uncommitted logic elements and structural arithmetic elements 254 within the datapath 210.

The datapath 210 comprises highly structured logic elements 254 that include a dynamically programmable ALU 304, conditional comparators 310, accumulators 302, and data buffers 300. The ALU 304 is configured to perform instructions on accumulators 302, and to perform arithmetic sequences as controlled by a sequence memory. The conditional comparators 310 can operate in parallel with the ALU 304. The datapath 210 is further optimized to implement typical embedded functions, such as timers, counters, pseudo random sequence generators, Cyclic Redundancy Checkers (CRC), Pulse Width Modulators (PWM), etc.

The combination of uncommitted PLDs 200 with a dedicated datapath module 210 allows the UDBs 120 to provide embedded digital functions with more silicon efficient processing. The dedicated committed structural arithmetic elements 254 more efficiently implement arithmetic sequencer operations, as well as other datapath functions. Since the datapath 210 is structural, fewer gates are needed to implement these structural elements 254 and fewer interconnections are needed to connect the structural elements 254 together into an arithmetic sequencer. Implementing the same datapath 210 with PLDs could require a much greater quantity of additional combinational logic and additional interconnections.

The structured logic in the datapath 210 is also highly programmable to provide a wide variety of different dynamically selectable arithmetic functions. Thus, the datapath 210 not only conserves space on the integrated circuit 100 (FIG. 1) but also is highly configurable similar to PLDs. It has an additional advantage of being dynamically configurable and reconfigurable.

The functionality of the datapath 210 may be controlled through writes to the control registers 250 allowing the micro-controller 102 to arbitrarily set the system state and selectively control different arithmetic functions. The status registers 256 allow the micro-controller 102 to also identify different states associated with different configured arithmetic operations. The flexible connectivity scheme provided by the routing channel 130 selectively interconnects the different functional element 250, 200, 254, and 256 together as well as programmably connecting these functional elements to other UDBs, I/O connections, and peripherals.

Thus, the combination of uncommitted logic 252, structural logic 254, and programmable routing channel 130 provide as much functionality and uses less integrated circuit space, while at the same time providing the potential for increased performance and substantially the same functional configurability.

Datapath

FIG. 4 shows one embodiment of the datapath 210 in more detail. The datapath 210 contains a single cycle ALU 304 and associated conditional logic comparators 310. The datapath 210 can be chained with neighboring datapaths to achieve single cycle functionality with additional bit widths. A RAM based control store 324 dynamically selects the operation and configuration performed in any given cycle.

The datapath 210 is optimized to implement typical embedded functions, such as timers, counters, Pulse Width Modulators (PVs), Pseudo Random Sequence (PRS) generators, Cyclic Redundancy Checks (CRC), shifters, dead band generators, etc. The addition of the add and subtract functions in ALU 304 also allow support for digital delta-sigma operations.

Internal connections 330 can be externally connected to either the system bus 140 and/or the routing channel 130. Different combinations of connections 330 are interconnected between different datapath components according to their related functions. Connections 330 are shown as a single bus in FIG. 4 for illustrative purposes only and there may or may not be certain connections that are shared by multiple different datapath components.

Dynamic configuration is the ability to change the datapath function and interconnect configuration on a cycle-by-cycle basis. This is implemented using the information in configuration RAM 324. The address 323 input to RAM 324 can be routed from any functional element connected to the routing channel 130, and most typically include the PLDs 200 (FIG. 2), I/O pins 104 (FIG. 1), micro-controller 102 (FIG. 6), or PLDs or datapaths from other UDBs 120.

The ALU 304 can perform different general-purpose functions such as increment, decrement, add, subtract, logical AND, OR, XOR, or PASS. In addition to these functions, hardware structures and connections are provided to implement a single cycle CRC operation. In addition to the ALU 304, an independent shifter 306 provides left, right, nibble swap operations. Another independent masking function 308 masks selectable bits output from the ALU 304.

Each datapath 210 includes conditional logic comparators 310 which can be configured to receive a variety of different datapath register inputs. The comparators 310 check for conditions such as zero detect, all one's detect, and overflow. These conditions produce datapath outputs that are selectively routed back through the same datapath 210 or routed through output multiplexer 326 and the routing channel 130 to other UDBs or peripheral components.

Each datapath 210 contains multiple FIFOs 312 that can be individually configured to operate as input buffers or output buffers. When operating as input buffers, the system bus 140 can write to the FIFOs 312 and datapath internal logic elements can read from the FIFOs 312. When operating as output buffers, datapath internal logic elements write to the FIFO 312 and the system bus 140 reads from the FIFO 312. The FIFOs 312 generate status that can be routed to interact with sequencers, interrupt, or DMA requests.

As described above in FIG. 2, the datapath 210 can be configured to chain conditions and signals with neighboring datapaths. The shift, carry, capture, and other conditional signals can also be chained to form higher precision arithmetic, shift, and CRC/PRS functions. For example, 16-bit functions in an 8-bit datapath can be provided by interconnecting two datapaths together, or CRC generation can be implemented between two datapaths 210 using data shifting.

In applications that are oversampled, or don't need the highest clock rates, the ALU block 304 can be efficiently shared with two sets of registers and condition generators. Selected outputs from the ALU 304 and shifter 306 are registered and can be used as inputs in subsequent cycles.

The datapath 210 receives configuration inputs, control inputs, and data inputs. At least some configuration data can be received over input 320 and used for selecting the current address 323 for configuration RAM 324. Input 320 can come from either to the system bus 140 and/or to the routing channel 130. Control inputs can come over the system bus 140 or the routing channel 130 and are used to load the data registers 314 and capture outputs from accumulators 302. Data inputs can also come from the system bus 140 and/or the routing channel 130 and can include shift in and carry in signals received over input multiplexer 322. Other data inputs include parallel data input and output ports 318 that can be programmably connected through the routing channel 130 to the ALU 304.

There are multiple conditional, data, and status signals that can be selectively output via output multiplexer 326. For maximum routing flexibility, any of the status or data output signals connected to output mux 326 can be programmably connected to the routing channel 130.

The datapath 210 has multiple working registers. These registers are readable and writable by the micro-controller 102 and DMA 108 in FIG. 1. The accumulators 302 can be a source for the ALU 304 and a destination of the ALU output. The accumulators 302 can also be loaded from an associated data register 314 or FIFO 312. The accumulators 302 contain the current value of the ALU function, for example, the count, CRC or shift.

Clock Driven Datapath Chaining

FIG. 5 is a schematic block diagram illustrating example embodiments of the datapath block 210 chaining. Referring to FIG. 5, the datapath 210 includes a carry-in selection circuit or multiplexor 510 to determine a carry-in value 514 to provide to the ALU 304. In some embodiments, the carry-in selection circuit 510 can select between a default carry-in value 502, a chained carry-in value 504, a carry-in value 506 routed from an interconnect 130, or a registered carry-in 508.

The chained carry-in 504 can be from a neighboring datapath 210. By selecting the chained carry-in value 504, the datapath is dynamically linked or chained to the neighboring datapath 210, for example, to perform higher precision functions or processing tasks, such as arithmetic, shift, and CRC/PRS functions. For example, 16-bit functions can be provided by interconnecting two 8-bit datapaths together, or 16-bit CRC generation can be implemented between two 8-bit datapaths 210 using data shifting.

The carry-in selection circuit 510 can determine which carry-in value 502-508 to supply to the ALU 304 as the carry-in value 514 responsive to a carry-in selection 512 from the dynamic configuration memory 324. The dynamic configuration memory 324 can provide the carry-in selection 512 to the carry-in selection circuit 510 dynamically in a given clock cycle. Thus, the carry-in selection 512 and the chaining of the datapath 210 to one or more neighboring datapaths is clock driven and can be dynamically performed.

The ALU 304 provides a carry-out 514 to a buffer or register 520 to store the carry-out for possible future processing. The carry-out 514 can be provided to a neighboring datapath as a chained carry-out 516 for use by the neighboring datapath as a chained carry-in value. The register 520 can provide the buffered carry-out to the carry-in selection circuit 510 as the registered carry-in value 508. In some embodiments, the register 520 can provide the buffered carry-out to the carry-in selection circuit 510 responsive to a clock signal, which can have the same or different cycle as the clock driving the carry-in selection 512. This allows the 8-bit datapath to compute the first 8-bits of a function in a given clock cycle with a default carry input, and then compute the next 8-bit of a function is the following clock cycle with the registered carry. This allows at 16-bit function to be computed in a single 8-bit datapath.

Another selection of the multiplexer allows the input of carry in values from routing array. This setting can be dynamically selected when the function calls for a special user generated, or computed carry in.

Dynamic clock driven datapath chaining grants system designers the flexibility to selectively chain any number of datapaths to perform higher resolution functions without having to incorporate larger arithmetic circuits, such as statically chained datapaths, into their designs. As described above, in applications that are oversampled, or don't need the highest clock rates, the ALU block 304 can be efficiently shared with two sets of registers and condition generators. This can enable a time-division multiplexing of the clock signal by the datapath 210 and an interleaving of the chained functions with various other functions on successive clock cycles. In other words, the datapath 210 can share ALU 304 operations on alternate clock cycles with varying precision. An example implementation of this time-division multiplexing will be described below corresponding to FIG. 7 in greater detail.

FIG. 6 is a block diagram illustrating example embodiments of datapath chaining with the datapath block 210 shown in FIG. 5. Referring to FIG. 6, multiple datapaths 602-608 are chained together to perform higher precision functions or processing tasks, such as arithmetic, shift, and CRC/PRS functions. In some embodiments, each datapath 602-608 can process 8-bits of data, and thus the chaining of four datapaths 602-608 together allows for up to 32-bit processing capability. Although FIG. 6 shows four datapaths 602-608 chained together, any number of datapaths can be reconfigurably and dynamically chained together depending on the specific processing requirement.

A datapath 602 can receive a carry-in 601, which can be selected according to the carry-in chain 512. This carry-in 601 can be from any source, and can be a default carry-in 502, chained carry-in 504, interconnect carry-in 506, or a registered carry-in 508. The other datapaths 604-608 receive a chained carry 603-607 from a neighboring datapath 602-606, and select the chained carry 603-607 according to the carry-in selection 512. For instance, datapath 604 can receive chained carry 603 from datapath 602. The chained carry 603 can be a carry-out of datapath 602 and a carry-in of datapath 604 when selected according to the carry-in selection 512. Datapath 604 provides a carry-out to datapath 606 as a chained carry 605, where the datapath 606 can utilize the chained carry 605 as a carry-in according to the carry-in selection 512.

In some embodiments, a datapath 210 can chain to any other datapath 210 present in the system, i.e., the arrangement of a datapath chain can be programmable. In other embodiments, the datapaths 210 have a pre-specified chaining order by chaining with other adjacent datapaths 210. For instance, in some embodiments, a datapath 210 can provide its carry-out 514 to specified datapath(s) capable of chaining with the datapath 210, and receive a chained carry-in from one or more specified datapaths.

FIG. 7 is a schematic block diagram showing a multiple timer implementation utilizing datapath block chaining shown in FIG. 5-FIG. 6. Referring to FIG. 7, a pair of 8-bit datapaths 705A and 705B are chained together to implement a 16-bit timer 710. A carry-in selection circuit 510 receives a carry-out from the datapath 705B and provides it to the datapath 705A as a carry-in when implementing the 16-bit timer 710. The carry-in selection circuit 510 can provide the chained carry-in to the datapath 705A according to a carry-in selection 512. The 16-bit timer 710 outputs a compare result 712 and a zero detect result 714 from datapath 705A to indicate a state of the 16-bit timer 710.

The carry-in selection 512 is clock driven, enabling the carry-in selection circuit 510 to provide the chained carry-in to the datapath 705A according to the cycle of the clock. In some embodiments, the datapaths 705A and 705B can be time-division multiplexed, interleaving additional functions between the clock cycles implementing the chained 16-bit timer 710. For instance, both datapaths 705A and 705B can implement 8-bit timers 720 and 730, respectively, in between operational functions for the 16-bit timer 710. The datapath 705A implementing the 8-bit timer 720 outputs a compare result 722 and a zero detect result 724 to indicate a state of the 8-bit timer 720. The datapath 705B implementing the 8-bit timer 730 outputs a compare result 732 and a zero detect result 734 to indicate a state of the 8-bit timer 730. The dynamic clock driven chaining allows datapaths 705A and 705B to be chained according to the carry-in selection 512 in one clock cycle, and then in the next clock cycle operate unchained according to the carry-in selection 512.

FIG. 8 is a flow diagram showing how a micro-controller or other Central Processing Unit (CPU) programs the UDBs and how subsequent datapath processing operates. Referring to FIG. 8, the flow diagram begins at a block 805 where a microcontroller programs the configuration memory. Some of the configuration is static and some is dynamic which, when processing starts, is based on the datapath inputs in a given clock cycle. This configuration includes instructions provided to the multiplexor regarding the selection for dynamic carry in chaining.

In a block 807, the microcontroller initiates datapath chain in operations. In some embodiments, the microprocessor enables a configured function, or the function automatically starts in response to a global enable signal. Once enabled, block 810 receives a chained carry-in value from a datapath and configuration data from a configuration memory based on a clock cycle. In some embodiments, the chained carry-in value can be a carry-out from the datapath.

The flow diagram continues to block 820 and selects the chained carry-in value according to the configuration data and the clock cycle from the configuration memory. In some embodiments, a multiplexor selects the chained carry-in value according to the configuration data from the configuration memory. The multiplexor can also receive multiple other input values to select from, including a default carry-in value, a interconnect carry-in value, or a registered carry-in value.

The flow diagram continues to block 830 and dynamically chains to the datapath for collaboration with the datapath in the performance of arithmetic operations by utilizing the chained carry-in value. In some embodiments, the dynamic datapath chaining can allow multiple datapaths to collaborate to perform high precision functions. For instance, when each datapath can perform 8-bit arithmetic operations, chaining three datapaths together allows for performance of operations up to 24-bits. Furthermore, since the chaining (and unchaining) is performed relative to a clock, or is clock driven, datapaths can be dynamically reconfigured to perform chained and unchained operations on a clock cycle-by-cycle basis. Thus, the above described configuration can utilize time-division multiplexing for increased processing utilization and power.

The system described above can use dedicated processor systems, micro controllers, programmable logic devices, or microprocessors that perform some or all of the operations. Some of the operations described above can be implemented in software and other operations can be implemented in hardware.

For the sake of convenience, the operations are described as various interconnected functional blocks or distinct software modules. This is not necessary, however, and there can be cases where these functional blocks or modules are equivalently aggregated into a single logic device, program or operation with unclear boundaries. In any event, the functional blocks and software modules or features of the flexible interface can be implemented by themselves, or in combination with other operations in either hardware or software.

Having described and illustrated the principles of the invention in a preferred embodiment thereof, it should be apparent that the invention can be modified in arrangement and detail without departing from such principles. Claim is made to all modifications and variation coming within the spirit and scope of the following claims.

Claims

1. An apparatus, comprising: a plurality of datapaths, each having structural arithmetic elements to perform various arithmetic operations based, at least in part, on configuration data; anda configuration memory coupled to the datapaths, the configuration memory to provide the configuration data to the datapaths, which causes the datapaths to collaborate when performing the arithmetic operations, where the datapaths receive the configuration data based on a clock cycle, and the datapaths are dynamically chained to collaborate on the arithmetic operations according to configuration data and the clock cycle.

2. The apparatus according to claim 1, where a first datapath receives a chained carry-in value from a second datapath for utilization in the arithmetic operations.

3. The apparatus according to claim 2, where the first datapath includes a multiplexor to at least receive a chained carry-in value from the second datapath, the multiplexor to provide the first datapath the chained carry-in value according to the configuration data from the configuration memory.

4. The apparatus according to claim 3, where the multiplexor is operable to receive the chained carry-in value and one or more of a default carry-in value, a registered carry-in value, a carry-in value received from an interconnect routing.

5. The apparatus according to claim 2, where the chained carry-in value received by the first datapath is a carry-out value of the second datapath.

6. The apparatus according to claim 1, where the datapaths are operable to perform time-division multiplexed arithmetic operations according to the configuration data, where the datapaths are dynamically chained to collaborate on the arithmetic operations during a first clock cycle and unchained to perform independent arithmetic operations during a second clock cycle.

7. A method comprising: receiving a chained carry-in value from a datapath and receiving configuration data from a configuration memory based on a clock cycle;selecting the chained carry-in value according to the configuration data from the configuration memory; anddynamically chaining to the datapath according to the configuration data and the clock cycle to collaborate with the datapath in the performance of arithmetic operations by utilizing the chained carry-in selected according to the configuration data when performing arithmetic operations.

8. The method according to claim 7, includes multiplexing multiple input values according to the configuration data, where at least one of the input values is the chained carry-in value received from the datapath.

9. The method according to claim 8, where the input values include one or more of a default carry-in value, a registered carry-in value, a carry-in value received from an interconnect routing.

10. The method according to claim 7, where the chained carry-in value is a carry-out value of the datapath.

11. The method according to claim 7, includes: dynamically chaining to the datapath for collaborate on the arithmetic operations during a first clock cycle; anddynamically unchaining to perform independent arithmetic operations during a second clock cycle.

12. A device comprising: a structural arithmetic circuit to perform various arithmetic operations based, at least in part, on configuration data; anda selection circuit to receive a chained carry-in value from a datapath, and select the chained carry-in value according to the configuration data, where the structural arithmetic circuit to perform arithmetic operations by utilizing the chained carry-in selected, where the selection circuit is operable to receive the configuration data based on a clock cycle, and the structural arithmetic circuit to dynamically chain to the datapath for collaboration on the arithmetic operations according to configuration data and the clock cycle.

13. The device according to claim 12, where the selection circuit is operable to multiplex multiple input values according to the configuration data, where at least one of the input values is the chained carry-in value received from the datapath.

14. The device according to claim 13, where the input values include one or more of a default carry-in value, a registered carry-in value, a carry-in value received from an interconnect routing.

15. The device according to claim 12, where the chained carry-in value is a carry-out value of the datapath.

16. The device according to claim 12, where the structural arithmetic circuit is operable to dynamically chain to the datapath for collaborate on the arithmetic operations during a first clock cycle, and then dynamically unchaining to perform independent arithmetic operations during a second clock cycle.

17. The device according to claim 12, where the structural arithmetic circuit is operable to dynamically chain and unchain to the datapath on alternate clock cycles to time-division multiplex performance of arithmetic operations.