Method and apparatus for saturated multiplication and accumulation in an application specific signal processor

[0002] This invention relates generally to digital signal processing devices. More particularly, the invention relates to multiply and accumulators (MACs).

[0003] Single chip digital signal processing devices (DSP) are relatively well known. DSPs generally are distinguished from general purpose microprocessors in that DSPs typically support accelerated arithmetic operations by including a dedicated multiplier and accumulator (MAC) for performing multiplication of digital numbers. The instruction set for a typical DSP device usually includes a MAC instruction for performing multiplication of new operands and addition with a prior accumulated value stored within an accumulator register. A MAC instruction is typically the only instruction provided in prior art digital signal processors where two DSP operations, multiply followed by add, are performed by the execution of one instruction. DSPs are often programmed in a loop to continuously perform a MAC instruction using different operands. An area where DSPs may be utilized is in telecommunication systems. One use of DSPs in telecommunication systems is digital filtering. In this case a DSP is typically programmed with instructions to implement some filter function in the digital or time domain. The mathematical algorithm for a typical finite impulse response (FIR) filter may look like the equation Y_n=h₀X₀+h₁X₁+h₂X₂+. . . +h_NX_Nwhere h_nare fixed filter coefficients numbering from 1 to N and X_nare the data samples. The equation Yn may be evaluated by using a software program. However in some applications, it is necessary that the equation be evaluated as fast as possible. One way to do this is to perform the computations using hardware components such as a DSP device programmed to compute the equation Yn. In order to further speed the process, it is desirable to vectorize the equation and distribute the computation amongst multiple DSPs such that the final result is obtained more quickly. The multiple DSPs operate in parallel to speed the computation process. In this case, the multiplication of terms is spread across the multipliers of the DSPs equally for simultaneous computations of terms. The adding of terms is similarly spread equally across the adders of the DSPs for simultaneous computations. In vectorized processing, the order of processing terms is unimportant since the combination is associative. If the processing order of the terms is altered, it has no effect on the final result expected in a vectorized processing of a function.

[0004] However, not all computations of equations can be vectorized. This is the case with some of the standardized compression/decompression algorithms used in telecommunication systems. In this case, the computation processes typically can not be performed in parallel by standard DSPs. A nonvectorizable equation usually must be processed in a specific manner due to some constraints. Exemplary of this is the voice compression and voice decompression ITU standards used in coder/decoders (CODECs). Specifically, this includes the ITU standards G.723.1, G.728, and G.729. In equations for these functional operations, sixteen bit fixed point arithmetic is utilized. In order to be compliant with the standards, operations found within the specification must be carried out exactly in order to maintain signal quality. Additionally, the bits must be exact and the saturation must be carried out appropriately. Four terms of a typical nonvectorizable equation where nonvectorized processing is required, are YOUTn=((((AC+_La₀X₀)+_La₁X₁)+_La₂X₂)+_La₃X₃) where “+_L” refers to a limiting of the addition to a range of values between a positive saturation value and a negative saturation value and AC is an accumulated value for a prior processing cycle. A typical DSP device needs to process each term separately from inner brackets to outer brackets. One disadvantage to processing in this fashion is that the extra processing steps require additional cycles. The additional cycles decrease the channel handling capability of a DSP device thereby requiring that additional DSP devices be used to increase the channel handling capability. Additionally, in a multiple DSP system, a number of DSPs may be sitting idle awaiting completion of the term presently being computed by the active DSP. This is an inefficient use of computing resources.

[0005] Additionally, the quality of service over a telephone system often relates to the processing speed of signals. That is particularly the case when a DSP is to provide voice processing, such as voice compression, voice decompression, and echo cancellation for multiple channels. More recently, processing speed has become even more important because of the desire to transmit voice aggregated with data in a packetized form for communication over packetized networks. Delays in processing the packetized voice signal tend to result in the degradation of signal quality on receiving ends.

[0006] It is desirable to provide improved processing of voice and data signals to enhance the quality of voice and data communication over packetized networks. It is desirable to improve the efficiency of using computing resources when performing signal processing functions.

[0007] Briefly, the present invention includes a method, apparatus and system as described in the claims. An application specific signal processor (ASSP) performs vectorized and nonvectorized operations. Nonvectorized operations may be performed using a saturated multiplication and accumulation operation. The ASSP includes a serial interface, a buffer memory, a core processor for performing digital signal processing which includes a reduced instruction set computer (RISC) processor and four signal processing units. The four signal processing units execute the digital signal processing algorithms in parallel including the execution of the saturated multiplication and accumulation operation. The ASSP is utilized in telecommunication interface devices such as a gateway. The ASSP is well suited to handling voice and data compression/decompression in telecommunication systems where a packetized network is used to transceive packetized data and voice.

[0008] FIG. 1A is a block diagram of a system utilizing the present invention.

[0009] FIG. 1B is a block diagram of a printed circuit board utilizing the present invention within the gateways of the system in FIG. 1A.

[0010] FIG. 2 is a block diagram of the Application Specific Signal Processor (ASSP) of the present invention.

[0011] FIG. 3 is a block diagram of an instance of the core processors within the ASSP of the present invention.

[0012] FIG. 4 is a block diagram of the RISC processing unit within the core processors of FIG. 3.

[0013] FIG. 5 is a block diagram of an instance of the signal processing units within the core processors of FIG. 3.

[0014] FIG. 6 is a chart illustrating the discussion tree of the possible partial products for the saturated multiplication and accumulation (SATMAC) operation of the ASSP of FIG. 2.

[0015] FIG. 7 is a block diagram of an instance of the adders within the signal processing unit of FIG. 5.

[0016] FIGS. 8 through 16 are block diagrams illustrating the configuration and operation of the ASSP of the present invention when performing a SATMAC operation.

[0017] Like reference numbers and designations in the drawings indicate like elements providing similar functionality. A letter after a reference designator number represents an instance of an element having the reference designator number.

[0018] In the following detailed description of the present invention, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be obvious to one skilled in the art that the present invention may be practiced without these specific details. In other instances well known methods, procedures, components, and circuits have not been described in detail so as not to unnecessarily obscure aspects of the present invention.

[0019] The present invention includes a method, apparatus and system for voice and data communication processing. Briefly, multiple application specific signal processors (ASSPs) are provided within gateways in communication systems to provide improved voice and data communication over a packetized network. A gateway is a device for connecting multiple networks and devices that use different protocols. Each ASSP includes four core processors in order to simultaneously process four channels of voice or data. Each core processor preferably includes four signal processing units (SPs) to simultaneously process four samples or terms of digital voice signals originally packetized for communication over IP, ATM, Frame Relay, SONET or other packetized network. Each of the SPs can perform vectorized operations normally associated with prior art DSP devices as well as non-vectorized operations including saturated multiplication and accumulation for voice compression standards. Each SP includes multiple adders for simultaneous processing of multiple voice and data communication signal samples. Each ASSP is flexible and can be programmed to perform many network functions or data/voice processing functions, including voice compression for CODECs, simply by altering the software program controlling the commands executed by the ASSP.

[0020] Referring now to FIG. 1A, a voice and data communication system 100 is illustrated. The system 100 includes a network 101 which is a packetized or packet-switched network, such as IP, ATM, frame relay, or SONET/SDL. The network 101 allows the communication of voice/speech and data between endpoints in the system 100, using packets. Data may be of any type including audio, video, email, and other generic forms of data. At each end of the system 100, the voice or data requires packetization when transceived across the network 101. The system 100 includes gateways 104A, 104B, and 104C in order to packetize the information received for transmission across the network 101. Voice and data information may be provided to a gateway 104 from a number of different sources in a variety of digital formats. In system 100, analog voice signals are transceived by a telephone 108. In system 100, digital voice signals are transceived at public branch exchanges (PBX) 112A and 112B which are coupled to multiple telephones, fax machines, or data modems. Digital voice signals are transceived between PBX 112A and PBX 112B with gateways 104A and 104C, respectively. Digital data signals may also be transceived directly between a digital modem 114 and a gateway 104A. Digital modem 114 may be a Digital Subscriber Line (DSL) modem or a cable modem. Data signals may also be coupled into system 100 by a wireless communication system by means of a mobile unit 118 transceiving digital signals or analog signals wirelessly to a base station 116. Base station 116 converts analog signals into digital signals or directly passes the digital signals to gateway 104B. Data may be transceived by means of modem signals over the plain old telephone system (POTS) 107B using a modem 110. Modem signals communicated over POTS 107B are traditionally analog in nature and are coupled into a switch 106B of the public switch telephone network (PSTN). At the switch 106B, analog signals from the POTS 107B are digitized and transceived to the gateway 104B by time division multiplexing (TDM) with each time slot representing a channel and one DSO input to gateway 104B. At each of the gateways 104A, 104B and 104C, incoming signals are packetized for transmission across the network 101. Signals received by the gateways 104A, 104B and 104C from the network 101 are depacketized for distribution to the appropriate destination.

[0021] Referring now to FIG. 1B, a network interface card (NIC) 130 of a gateway 104 is illustrated. The NIC 130 includes one or more application-specific signal processors (ASSPs) 150A-150N. The number of ASSPs within a gateway is expandable to handle additional channels. Network interface devices 131 of NIC 130 provide interfaces to various devices connected to the gateway, including the network 101. In interfacing to the network 101, the network interface devices packetize data for transmission out on the network 101 and depacketize data which is to be received by the ASSP devices. Network interface devices 131 process information received by the gateway on the receive bus 134 and provides it to the ASSP devices. Information from the ASSP devices 150 is communicated on the transmit bus 132 for transmission out of the gateway. A traditional network interface device is a multi-channel serial interface or a UTOPIA device. The NIC 130 couples to a gateway chassis bus 136 within the gateway 104. Bridge logic 138 transceives information between bus 136 and NIC 130. Bridge logic 138 transceives signals between the NIC 130 and the chassis bus 136 onto the control bus 139 for communication to either one or more of the ASSP devices 150A-150N, a host processor 140, or a host memory 142. Coupled to each of the one or more ASSP devices 150A through 15ON (generally referred to as ASSP 150) are local memory 145A through 145N (generally referred to as local memory 145), respectively. Digital data on the receive bus 134 and transmit bus 132 is preferably communicated in 8-byte wide fashion. Local memory 145 is used as a scratchpad memory by each of the ASSPs 150.

[0022] Each of the ASSPs 150 provide signal processing capability for the gateway. The type of signal processing provided is flexible because each ASSP may executed differing signal processing programs. Typical signal processing applications for an ASSP include (a) echo cancellation; (b) video, audio, and voice/speech compression/decompression (voice/speech coding and decoding); (d) delay handling (packets, frames, SONETs/SDLs); (e) loss handling; (f) connectivity (LAN and WAN); (g) security (encryption/decryption); (h) telephone connectivity; (i) protocol processing (reservation and transport protocols, RSVP, TCP/IP, RTP, UDP); (j) filtering; (k) suppression; (l) length handling (frames, packets, SONETs); and other digital signal processing functions associated with the communication of voice and data over a communication system. Each ASSP 150 can perform other functions in order to transmit voice and data to the various endpoints of the system 100 within a packet data stream over a packetized network.

[0023] Referring now to FIG. 2, a block diagram of the ASSP 150 is illustrated. At the heart of the ASSP 150 are four core processors 200A-200D. Each of the core processors 200A-200D is respectively coupled to a data memory 202A-202D and a program memory 204A-204D. Each of the core processors 200A-200D communicates with outside channels through the multi-channel serial interface 206, the multi-channel memory movement engine 208, buffer memory 210, and data memory 202A-202D. The ASSP 150 further includes an external memory interface 212 to couple to external memory. The ASSP 150 includes an external host interface 214 for interfacing to the external host processor 140 of FIG. 1B. The external host interface 214 may be replaced with an internal host processor 140 within the ASSP 150. Further included within the ASSP 150 are timers 216, clock generators and a phase-lock loop 218, miscellaneous control logic 220, and a Joint Test Action Group (JTAG) test access port 222 for boundary scan testing. Multi-channel serial interface unit may be replaced with a UTOPIA parallel interface for some applications such as ATM. The ASSP 150 further includes an interrupt controller 224 to assist in the control of the ASSP 150.

[0024] Referring now to FIG. 3, a block diagram of the core processor 200 is illustrated coupled to its respective data memory 202 and program memory 204. Core processor 200 is the block diagram for each of the core processors 200A-200D. Data memory 202 and program memory 204 refers to a respective instance of data memory 202A-202D and program memory 204A-204D. The core processor 200 includes four signal processing units SP0300A, SP1300B, SP2300C and SP3300D. The core processor 200 further includes a reduced instruction set computer (RISC) unit 302 and a pipeline control unit 304. The signal processing units 300A-300D perform the signal processing tasks on data while the RISC control unit 302 and the pipeline control unit 304 perform control tasks related to the signal processing function performed by the SPs 300A-300D.

[0025] Referring now to FIG. 4, a block diagram of the RISC control unit 302 is illustrated. RISC control unit 302 includes a data aligner and formatter 402, a memory address generator 404, three adders 406A-406C, an arithmetic logic unit (ALU) 408, a multiplier 410, a barrel shifter 412, and a register file 413. The register file 413 points to a starting memory location from which memory address generator 404 can generate addresses into data memory 202.

[0026] Referring now to FIG. 5, a block diagram of a signal processing unit 300 is illustrated which represents an instance of the SPs 300A-300D. Each of the signal processing units 300 includes a data typer and aligner 502, a first multiplier M1504A, a compressor 506, a first adder A1510A, a second adder A2510B, an accumulator register 512, a third adder A3510C, and a second multiplier M2504B. Adders 510A-510C are similar in structure and are generally referred to as adder 510. Multipliers 504A and 504B are similar in structure and generally referred to as multiplier 504. Each of the multipliers 504A and 504B have a multiplexer 514A and 514B respectively at its input stage to multiplex different inputs from different busses into the multipliers. Each of the adders 510A, 510B, 510C also have a multiplexer 520A, 520B, and 520C respectively at its input stage to multiplex different inputs from different busses into the adders. These multiplexers and other control logic allow the adders, multipliers and other components within the signal processing units 300A-300C to be flexibly interconnected by proper selection of multiplexers. In the preferred embodiment, multiplier M1504A, compressor 506, adder A1510A, adder A2510B and accumulator 512 can receive inputs directly from external data buses through the data typer and aligner 502. In the preferred embodiment, adder 510C and multiplier M2504B receive inputs from the accumulator 512 or the outputs from the execution units multiplier M1504A, compressor 506, adder A1510A, and adder A2510B. The multiplexers used to configure the processing units 300A-300B are not shown in the following figures so as not to unnecessarily obscure aspects of the present invention.

[0027] Referring now to FIG. 6, a decision tree is illustrated for the saturation, multiplication and accumulation (SATMAC) operation of the ASSP 150. The decision tree in FIG. 6 illustrates additions and multiplication of terms necessary for a nonvectorized computation of four terms of addition and four terms of multiplication. The nonvectorized equation is YOUTn=((((AC+_La₀X₀)+_La₁X₁)+_La₂X₂)+_La₃X₃) where “+_L” refers to a limiting of the addition to a range of values between a positive saturation value and a negative saturation value. The decision tree of FIG. 6 minimizes the number of hardware adders necessary for carrying out the evaluation of the SATMAC equations. The decision tree in FIG. 6 provides for the computation of the four terms in the YOUTn equation within one clock cycle. This is equivalent to performing four multiplications and four additions in one step. In this manner, the processing speed of the signal processor is improved. If an additional set of four terms is desired to be computed for the equation YOUTn, additional clock cycles are utilized with the prior value becoming AC. If additional terms are desired to be computed in one clock cycle, a similar decision tree to that of FIG. 6 may be provided in order to minimize the number of adders and multiplication steps necessary for its computation.

[0028] In accordance with ITU standards for compression/decompression, the highest possible value obtained by these equations is the positive saturation value (PSAT) of 7FFF hex. In accordance with ITU standards, the smallest possible value as a result of each step of these equations is the negative saturation value (NSAT) of 8000 hex. The limiting addition in of each addition step assures this is the case which results in this being a nonvectorized equation. Referring to FIG. 6, in the first level of the decision tree when adding a₀X₀to AC there are three possible resultant values because of the limiting addition. The three possible resultant values, PSAT, NSAT, and (AC+a₀X₀), are illustrated by the decision paths 601-603. In case that the value of (AC+a₀X₀) is greater than or equal to PSAT (7FFF_h) as indicated by path 601, the result is limited to the PSAT value of 7FFF_h. In the case that (AC+a₀X₀) is less than or equal to NSAT (8000_h) as indicated by path 602, the result is limited to the NSAT value of 8000_h. In the case that the value of (AC+a₀X₀) does not meet or exceed either PSAT or NSAT, the resultant value of (AC+a₀X₀) may be computed as indicated by path 603.

[0029] In the next level of the decision tree, a determination is made on the outcomes of how a₁X₁is added to the three prior resultant values of (AC+a₀X₀). In the case that the prior resultant value is PSAT, if a₁X₁is a positive value, this would try to generate a value greater than PSAT, in which case the result is limited to PSAT along path 611 as provided for by the ITU specifications. If a₁X₁is a negative value which is added to PSAT, then the resultant output reduces PSAT by the value of a₁X₁as indicated by the path 612. Thus, when adding a negative value to PSAT, the negative value being one term which itself has been limited to the range of saturation values, such as when a₁X₁is a negative value, the result will not exceed the NSAT value of 8000 hex. Next, looking at the prior result being NSAT in the computation of (AC+a₀X₀), if a₁X₁is a negative value and added to NSAT, the resultant is limited to the NSAT value 8000 hex as indicated by the path 613. However, if a₁X₁is a positive value, when added to NSAT the resultant becomes (NSAT+a₁X₁) as indicated by path 614. Thus, the resultant addition of a₁X₁to NSAT does not exceed the positive saturation value PSAT. Next, when adding a₁X₁to the resultant value (AC+a₀X₀), three results are possible. In one case, the result may be greater than or equal to PSAT and limited thereto as indicated by the path 615. Alternatively, the resultant may be less than or equal to NSAT and limited thereto as indicated by path 616. If the prior result is (AC+a₀X₀) which neither meets or exceeds the PSAT or NSAT values, a₁X₁is added to (AC+a₀X₀) resulting in the equation (AC+a₀X₀+a₁X₁) as indicated by the path 617.

[0030] In the next level of the decision tree, the term a₂X₂is evaluated when added to the prior possible outcomes from the addition of the term a₁X₁. In adding a₂X₂to PSAT, there are two possible outcomes as indicated by path 621 and path 622. In the case of that a₂X₂is a positive number being added to PSAT, the result is greater than PSAT and limited thereto as indicated by the path 621. In the case that a₂X₂is a negative number, when added to PSAT the resultant becomes (PSAT+a₂X₂) as indicated by the path 622. When adding a₂X₂to the prior result of (PSAT+a₁X₁), there are two possible resultants. In the case that a₂X₂is a positive value having a magnitude greater than the magnitude of a negative value of a₁X₁, then when a₂X₂is added to (PSAT+a₁X₁) the resultant is PSAT as indicated by the path 623. In the case that a₂X₂is a negative value or a₂X₂is a positive value having a magnitude less than the magnitude of a negative value of a₁X₁, then the result becomes (PSAT+a₁X₁)+a₂X₂) as indicated by the path 624. There is no value of a₂X₂that would result in (PSAT+a₁X₁) reaching or exceeding NSAT, such that this result is not possible. When adding a₂X₂to the prior result of NSAT, there are two possible results along path 625 and 626. If a₂X₂is a negative number, when added to NSAT it becomes more negative than NSAT and thus requires limiting to NSAT as indicated by the path 625. In the case that a₂X₂is a positive number, when added to NSAT the resultant value becomes (NSAT+a₂X₂) as indicated by the path 626. Because the value of NSAT is of such negative magnitude, there is no value of a₂X₂that when added thereto would result in an outcome of PSAT, when added to NSAT. When adding a₂X₂to the prior result (NSAT+a₁X₁) there are two possible results. If a₂X₂is a negative value having a magnitude larger than the magnitude of a positive value for a₁X₁, then when added to (NSAT+a₁X₁) the result is NSAT as indicated by the path 627. In the case that a₂X₂is a positive value or a negative value having a magnitude less than the magnitude of a positive value of a₁X₁, then when added to (NSAT+a₁X₁) the resultant is (NSAT+a₁X₁+a₂X₂) as indicated by the path 628. Because of the limiting of terms and equation values, the value of (NSAT+a₁X₁) is a value nearer to zero such that there is no value of a₂X₂that when added to (NSAT+a₁X₁) would cause the result to reach or exceed PSAT. Only when there are three terms in the prior result is it possible for the result to reach or exceed either PSAT or NSAT. When adding a₂X₂to the prior result (AC+a₀X₀+a₁X₁), there are three possible results indicated by the paths 629-631. In the case that a₂X₂has a large positive value, when added to this prior result the resultant may be PSAT as illustrated by path 629. In the case that a₂X₂is a very large negative number, when added to this prior result, the resultant may be NSAT as indicated by the path 630. Alternatively, the resultant of the addition may be within the range of PSAT and NSAT such that when adding a₂X₂to the prior value result of (AC+a₀X₀+a₁X₁), the resultant is (AC+a₀X₀+a₁X₁+a₂X₂) as indicated by the path 631.

[0031] In the next level of the decision tree, the addition of the term a₃X₃is evaluated when added to the prior possible results. There is a possibility of nine possible results after the term a₃X₃is added. These nine possible results are PSAT, (PSAT+a₃X₃), (PSAT+a₂X₂+a₃X₃), (PSAT+a₁X₁+a₂X₂+a₃X₃), NSAT, (NSAT+a₃X₃), (NSAT+a₂X₂+a₃X₃), (NSAT+a₁X₁+a₂X₂+a₃X₃) and (AC+a₀X₀+a₁X₁+a₂X₂+a₃X₃). When adding a₃X₃to the prior resultant value PSAT, there are two possible results indicated by paths 641 and 642. If a₃X₃is a positive value then the resultant is greater than PSAT and limited thereto as indicated by the path 641. There is no negative value for a₃X₃that would result in NSAT when it is added to the prior result PSAT. In the case that a₃X₃is a negative value, the resultant output is (PSAT+a₃X₃) as indicated by the path 642. When adding a₃X₃to the prior result of (PSAT+a₂X₂), there are two possible results. In the case that a₃X₃is a large positive value, when added to (PSAT+a₂X₂) the resultant is greater than or equal to PSAT and limited thereto as indicated by the path 643. Otherwise, when a₃X₃is added to (PSAT+a₂X₂), the resultant is (PSAT+a₂X₂+a₃X₃) as indicated by the path 644. There is no value of a₃X₃that when added to (PSAT+a₂X₂) results in the value of NSAT. When adding a₃X₃to the prior result of (PSAT+a₁X₁+a₂X₂), there are three possible results indicated by paths 645-647. In the case that a₃X₃is a large positive value, when added to (PSAT+a₁X₁+a₂X₂) it results in a value of greater than or equal to PSAT which is limited thereto as indicated by the path 645. In the case that a₃X₃is a large negative value, when added to (PSAT+a₁X₁+a₂X₂) it results in a value of less than or equal to NSAT which is limited to NSAT as indicated by the path 647. Otherwise, for a result that is within the range of PSAT and NSAT, when a₃X₃is added to (PSAT+a₁X₁+a₂X₂), the result is (PSAT+a₁X₁+a₂X₂+a₃X₃) as indicated by the path 646. When adding a₃X₃to the prior result of NSAT, there are two possible resultants as indicated by paths 648-649. In the case that a₃X₃is a negative value, the result of it being added to results in a value less than or equal to NSAT which is limited to NSAT as indicated by the path 648. In the case that a₃X₃is a positive value, the result of it being added to NSAT is (NSAT+a₃X₃) as indicated by the path 649. When adding a₃X₃to the prior result of (NSAT+a₂X₂), there are two possible results indicated by the paths 650-651. In the case that a₃X₃is a large negative value, the result of it being added to (NSAT+a₂X₂) is a value less than or equal to NSAT which is limited thereto as indicated by the path 650. There is no value of a₃X₃that when added to (NSAT+a₂X₂) would result in PSAT. If the result is within the range of PSAT and NSAT, alternative result when a₃X₃is added to (NSAT+a₂X₂) is (NSAT+a₂X₂+a₃X₃) as indicated by the path 651. When a₃X₃is added to the prior result of (NSAT+a₁X₁+a₂X₂), there are three possible resultants indicated by the paths 652-655. In the case that a₃X₃is a very large negative number, when a₃X₃is added to (NSAT+a₁X₁+a₂X₂) the result is less than or equal to NSAT and limited thereto as indicated by the path 653. In the case that a₃X₃is a very large positive number, when a₃X₃is added to (NSAT+a₁X₁+a₂X₂) a result may be a value greater than or equal to PSAT which would be limited to PSAT as indicated by the path 652. Otherwise, if the resultant value is between the range of PSAT and NSAT, when a₃X₃is added to (NSAT+a₁X₁+a₂X₂) the result is (NSAT+a₁X₁+a₂X₂+a₃X₃) as indicated by the path 654. When a₃X₃is added to the prior result (AC+a₀X₀+a₁X₁+a₂X₂), there are three possible resultant values indicated by paths 655-657. In the case that a₃X₃is a very large positive value, when added to the prior result (AC+a₀X₀+a₁X₁+a₂X₂), the result may be a value greater than or equal to PSAT which is limited thereto as indicated by the path 655. In the case that a₃X₃is a very large negative number, when added to the prior result (AC+a₀X₀+a₁X₁+a₂X₂), the result may be a value less than or equal to NSAT which is limited thereto as indicated by the path 656. If the resultant value is between the range of PSAT and NSAT, the resultant of adding a₃X₃to the prior result (AC+a₀X₀+a₁X₁+a₂X₂) is (AC+a₀X₀+a₁X₁+a₂X₂+a₃X₃) as indicated by the path 657. After the addition of a₃X₃is evaluated, one of the nine final results after the addition of a₃X₃is selected as the value of YOUTn. The selected term is a function of the saved carryout bits from the equations that are evaluated.

[0032] Not all additions in the decision tree need be performed or saved. In many instances only the sign bit of the result (the MSB of the sum output) needs to be saved to determine the next path along the decision tree which needs to be evaluated. Additionally, because NSAT and PSAT differ by only the value of 1, only one set of the final resultant equations having either PSAT or NSAT needs to be evaluated and its result saved for selection as YOUTn. In the case that the PSAT equations are utilized, the value of 1 need only to be added in order to obtain the NSAT equations. In the case that the NSAT equations are evaluated, the value of 1 need only be subtracted therefrom in order to obtain the PSAT equations. In the description of FIGS. 8 through 11 that follow, the PSAT equations are evaluated. In order to convert them into the NSAT equations, the value of 1 is provided as an input into the carry input path of the adders 510 in order to provide the NSAT equations. Therefore, when evaluating four terms of a nonvectorized equation in one clock cycle, only seven of the adders 510 are required. The seven adders add (PSAT+a₃X₃), (PSAT+a₂X₂+a₃X₃), (PSAT+a₁X₁+a₂X₂+a₃X₃), (AC+a₀X₀+a₁X₁), (AC+a₀X₀+a₁X₁+a₂X₂) and (AC+a₀X₀+a₁X₁+a₂X₂+a₃X₃).As previously discussed, the three PSAT equations are easily converted into NSAT equations by simply adding one into the carry in path of the full adders 510. In order to represent that either PSAT or NSAT can be evaluated, the equations with PSAT are referred to as (NSAT/PSAT+a₃X₃), (NSAT/PSAT+a₂X₂+a₃X₃), and (NSAT/PSAT+a₁X₁+a₂X₂+a₃X₃) it being understood that when NSAT is to be evaluated a logical one is asserted as the Carry In bit.

[0033] Referring now t o FIG. 7, adder 510 is illustrated which is representative of each of the adders 510A-510C in the signal processing unit 300. Adder 510 includes a 3-to-2 compressor 702, a carry select out full adder 704, and a multiplexer 706. Multiplexer 706 selects between the A+B output from the full adder 704 on bus 710 and A+B+1 output from the full adder 704 on bus 711 as its output, the selected sum 712, based on the carry input Cin 714. If Cin is a value of one, the A+B+1 output 711 is selected as the selected sum 712. If Cin is a value of zero, the A+B output 710 is selected as the selected sum 712. The 3-to-2 compressor 702 compresses the three inputs IN₁721, IN₂722, and IN₃723 into two outputs A 730 and B 731. Outputs A 730 and B 731 are provided as inputs into the carry select out full adder 704. Carry select out full adder 704 receives A 730 and B 731 from the 3-to-2 compressor 702 to generate (A+B) 701, (A+B+1) 711 and the carryout Cout 716 in accordance with the values of A and B. The (A+B) 710 is an evaluation of the equation containing the PSAT term while the (A+B+1) value is an e valuation of the equation containing the NSAT term.

[0034] Referring now to FIG. 8, a first configuration of block diagram elements within a core processor 200 for performing the saturated multiplication and accumulation is illustrated. While there are a number of ways to perform multiplication to evaluate terms of a_nX_n, the present invention preferably performs it as follows. In FIG. 8, a multiplier 504nand a compressor 506nare configured to evaluate one term of a_nX_nwhere a_nis multiplied together with X_n. Multiplier 504nrepresents an instance of four multipliers 504 one each from the SPs 300A-300D and compressor 506nrepresents an instance of four compressors 506 one each from the SPs 300A-300D. The term a_nX_nis preferably performed across the four SPs 300A-300D simultaneously such that four terms of a_nX_nare formed with n in the initial loop varying from 0 through 3. Multiplier 504nis configured to have one of its booth decoders 800A be coupled to the selection input of one of its booth multiplexers 800B. Booth decoder 800A receives as an input the sixteen bit word of a_n. Booth multiplexer 800B receives as an input the sixteen bit word of X_n. The booth decoder causes the booth multiplexer to generate eight partial product outputs. Of the eight partial product outputs, four are coupled into the 4-to-2 compressor 800C and the other four are coupled into the 4-to-2 compressor 800D of compressor 506n. The two outputs from the 4-to-2 compressor 800C and the two outputs from the 4-to-2 compressor 800D are each coupled into the 4-to-2 compressor 800E. The two outputs from the 4-to-2 compressor 800E are the partial products PPS_nand PPC_nof a_nX_n. In this manner, SP0300A performs the multiplication of a₀X₀to form a partial product output PPS₀and PPC₀. Signal processing unit 1 SP1300B performs the multiplication of a times X₁to form the partial products PPS₁and PPC₁. Signal processing unit 2 SP2300C performs the multiplication of a₂times X₂to form the partial product output PPS₂and FPC₂. The signal processing unit 3 SP3300D performs the multiplication of a₃times X₃to form the partial product outputs PPS₃and PPC₃. Each of the four signal processing units 300A-300D has a multiplier 504 and a compressor 506 configured to compute the a_nX_nterm. Each simultaneously performs the multiplication to obtain four a_nX_nterms.

[0035] Referring now to FIG. 9, the first of three saturation equations, YSAT₁is determined. YSAT₁is evaluated using one of the adders 510 within one of the signal processing units 300A-300D. In FIG. 9, the equation YSAT₁=NSAT/PSAT+a₃X₃is evaluated. The term NSAT/PSAT represents alternative terms and not the mathematical operation of division. Input into adder 510 is PSAT, the partial products PPS₃and PPC₃of a₂X₃, and an NSAT/PSAT selection signal. The NSAT/PSAT selection signal selects whether NSAT or PSAT is to be evaluated. The NSAT/PSAT selection signal is input into the adder 510 as the carry-in Cin bit. Full adder 704 receives the NSAT/PSAT selection signal in order to perform the appropriate operation. The three input values of PSAT, PPS₃, and PPC₃are input into the 3-to-2 compressor 702, generating the intermediate values 730 and 731, which are then input into the full adder 704 to generate the selected sum output of YSAT₁. In the case that NSAT is selected where the NSAT/PSAT selection signal is 1, Cin is 1 and the multiplexer 706 selects the (A+B+1) output from the full adder such that the equation YSAT₁=NSAT+a₃X₃. In the case that NSAT/PSAT selection signal is zero, Cin is zero and the multiplexer 706 selects the (A+B) output from the full adder 704 as the selected sum output such that the equation YSAT₁=PSAT+a₃X₃is evaluated.

[0036] Referring now to FIG. 10, the evaluation of the equation YSAT₂is illustrated. One of the SPs 300A-300D is configured as shown in FIG. 10 to evaluate the equation of YSAT₂=NSAT/PSAT+a₂X₂+a₃X₃. A 4-to-2 compressor 800A within compressor 506 is configured to accept the partial products , PPS₂and PPC₂of a₂X₂and PPS₃and PPC₃of a₃X₃. The partial product PPS₂₃and PPC₂₃output from the 4-to-2 compressor 800A are input into the 3-to-2 compressor 702 of the adder 510. The third term input into the 3-to-2 compressor 702 is PSAT. The partial product output from the 3-to-2 compressor 702 are added by the full adder 704. The output of from the full adder 704 is selected by multiplexer 706 on its output, selected sum 712. The multiplexer selects the output from the full adder 704 in response to the NSAT/PSAT selection signal. In the case that NSAT/PSAT selection signal is one on the carry-in Cin bit, representing the selection of NSAT equations, then the YSAT₂equation evaluated is YSAT₂=NSAT+a₃X₃+a₂X₂. In the case that NSAT/PSAT selection signal is zero on the carry-in Cin bit, representing the selection of the PSAT equation, the multiplexer 706 selects the (A+B) output from the full adder 504 such that the equation YSAT₂=PSAT+a₃X₃+a₂X₂is evaluated.

[0037] Referring now to FIG. 11, the evaluation of the third saturation equation YSAT₃is illustrated. One of the SPs 300A-300D is configured as illustrated in FIG. 11 to evaluate the equation of YSAT₃=NSAT/PSAT+a₁X₁+a₂X₂+a₃X₃. The 4-to-2 compressors 800A and 800C within the compressor 506 and the adder 510 within the same signal processing unit are configured as shown in FIG. 11. Input into the signal processing unit are a₁X₁, a₂X₂a₃X₃, PSAT and the NSAT/PSAT selection signal. The NSAT/PSAT selection signal is coupled to the carry in Cin of adder 704 and the selection control signal of the two-to-one multiplexer 704. The 4-to-2 compressor 800A evaluates the partial products PPS₁and PPC₁of a₁X₁and the partial products PPS₂and PPC₂of a₂X₂. The 4-to-2 compressor 800C evaluates the partial products PPS₃and PPC₃of a₃X₃and the resultant partial product outputs from the evaluation made by the 4-to-2 compressor 800A. The outputs from the 4-to-2 compressor 800C are coupled into the inputs of the 3-to-2 compressor 702 within the adder 510. Outputs from the 3-to-2 compressor 702 are input into the full adder 704 and summed together to generate the (A+B) and the (A+B+1) outputs of the full adder 704. Multiplexer 706 selects between these two outputs in response to the value of the NSAT/PSAT selection signal. In the case that NSAT/PSAT selection signal is one on the carry-in Cin bit, representing the selection of NSAT equations, then the YSAT₂equation evaluated is YSAT₃=NSAT+a₁X₁+a₂X₂+a₃X₃. In the case that NSAT/PSAT selection signal is zero on the carry-in Cin bit, representing the selection of the PSAT equation, the multiplexer 706 selects the (A+B) output from the full adder 504 such that the equation YSAT₃=PSAT+a₁X₁+a₂X₂+a₃X₃is evaluated.

[0038] Referring now to FIG. 12, the first of four accumulation equations Y₀−Y₃is illustrated as being evaluated. One of the signal processing units 300A-300D has one of its adders 510 configured as illustrated in FIG. 12 to evaluate the equation Y₀=AC+a₀X₀. As illustrated in FIG. 12, the 3-to-2 compressor 702 of the adder 510 receives as inputs the term AC and the partial products PPS₀and PPC₀of a₀X₀. The partial product outputs from the 3-to-2 compressor are fed into the A and B inputs of the full adder 704. Carry-in Cin to the full adder 704 is set to zero. The (A+B) output 710 from the full adder 704 is selected by the multiplexer 706 through its control signal being set to zero in order to generate equation Y₀on the selected sum output 712 of adder 510.

[0039] Referring now to FIG. 13, a second accumulation equation Y₁is illustrated as being evaluated. One of the signal processing units 300A-300D has its component blocks configured as illustrated in FIG. 13 to evaluate equation Y₁=AC+a₀X₀+a₁X₁. As illustrated in FIG. 13, the 4-to-2 compressor 800A of the compressor block 506 receives as input the partial products PPS₀and PPC₀of a₀X₀and the partial products PPS₁and PPC₁of a₁X₁. The two resultant partial product outputs from the compressor 800A are input into two inputs of the 3-to-2 compressor 702 of adder 510. The third input term input into the 3-to-2 compressor 702 is AC. The partial products output from the 3-to-2 compressor 702 are fed into the A and B inputs of the full adder 704. Carry-in Cin to the full adder 704 and control input into the multiplexer 704 are both set to zero. The (A+B) output 710 from the full adder 704 is selected by the multiplexer 706 through its control signal being set to zero in order to generate equation Y₁equals AC+a₀X₀+a₁X₁on the selected sum output 712 of adder 510.

[0040] Referring now to FIG. 14, a third accumulation equation Y₂is illustrated as being evaluated. One of the signal processing units 300A-300D has its component blocks configured as illustrated in FIG. 14 to evaluate the equation Y₂=AC+a₀X₀+a₁X₁+a₂X₂. The signal processor 300 accepts as input partial products of a₀X₀, a₁X₁, a₂X₂and a₃X₃, and the accumulation value of AC. The 4-to-2 compressor 800A receives as inputs the partial products PPS₀and PPC₀of a₀X₀, the partial products PPS₁and PPC₁of a₁X₁. The 4-to-2 compressor 800C receives as inputs the partial products PPS₂and PPC₂of a₂X₂. The two additional inputs into the 4-to-2 compressor 800C are the resultant partial product outputs from the 4-to-2 compressor 800A. The partial product outputs from the 4-to-2 compressor 800C are coupled into two inputs of the 3-to-2 compressor 702 of the adder 510. The third input into the 3-to-2 compressor 702 is the term AC. The partial products output from the 3-to-2 compressor 702 are fed into the A and B inputs of the full adder 704. Carry-in Cin to the full adder 704 and control input into the multiplexer 704 are both set to zero. The (A+B) output 710 from the full adder 704 is selected by the multiplexer 706 through its control signal being set to zero in order to generate the equation Y₂=AC+a₀X₀+a₁X₁+a₂X₂on the selected sum output 712 of adder 510.

[0041] Referring now to FIG. 15, a fourth accumulation equation Y₃is illustrated as being evaluated. One of the signal processing units 300A-300D has its component blocks configured as illustrated in FIG. 15 to evaluate the equation Y₃=AC+a₀X₀+a₁X₁+a₂X₂+a₃X₃. The 4-to-2 compressors 800A, 800B and 800C of the compressor 506 are configured as shown in FIG. 15. The 4-to-2 compressor 800A receives as inputs the partial products PPS₀and PPC₀of a₀X₀and the partial products PPS₁and PPC₁of a₁X₁. The 4-to-2 compressor 800B receives as inputs the partial products PPS₂and PPC₂of a₂X₂and the partial products PPS₃and PPC₃of a₃X₃. The outputs from the 4-to-2 compressor 800A and the 4-to-2 compressor 800B are coupled into the inputs of the 4-to-2 compressor 800C. The resultant partial product outputs from the 4-to-2 compressor 800C are coupled into two of the inputs of the 3-to-2 compressor 702 of the adder 510. The third input into the 3-to-2 compressor 702 is the accumulation value AC. The partial products output from the 3-to-2 compressor 702 are fed into the A and B inputs of the full adder 704. Carry-in Cin to the full adder 704 and control input into the multiplexer 704 are both set to zero. The (A+B) output 710 from the full adder 704 is selected by the multiplexer 706 through its control signal being set to zero in order to generate equation Y₃=AC+a₀X₀+a₁X₁+a₂X₂+a₃X₃on the selected sum output 712 of adder 510. Note that the formation of these possible resultant outputs for YOUTn requires at least seven full adders. The four signal processing units 300A-300D have a total of twelve full adders such that performing these additions to obtain the possible resultants increases the efficiency of the available components.

[0042] Referring now to FIG. 16, the multiplexer blocks 508 of the four signal processing units SPs 300-300D are configured and controlled as illustrated and shown in FIG. 16. Multiplexer blocks 508 within each of the SPs 300A-300D are used to form the 9-to-1 multiplexer 1600 and to perform the nine to one multiplexing function to generate YOUTn. The 9-to-1 multiplexer 1600 receives at its nine inputs PSAT, YSAT₃, YSAT₂, YSAT₁, Y₃, Y₂, Y₁, Y₀and NSAT. Multiplexer control logic 1602 within the core processor 200 generates the multiplexer control signals 1604 to control the selection of the output of the 9-to-1 multiplexer 1600. The multiplexer control logic 1602 receives the sign bits from equations illustrated in FIG. 6 and other control signals in order to properly select one of the nine terms input into the 9-to-1 multiplexer 1600 as the multiplexer output. Some of the equations in FIG. 6 that do not need to be completely evaluated because they are not input into the 9-to-1 multiplexer 1600. Only the sign bit of the result, typically the MSB of the result, needs to be determined in these cases. In which case, simple combinatorial logic only need be used to determine the sign of the result. For example, in multiplying two terms together, only the sign bits of the terms need be evaluated to determine the sign of the result. If both terms are negative or both positive, the result is positive. If only one of the terms is negative, then the result is a negative number. Similar logic is used in evaluating the addition of terms and knowing the saturation values of PSAT and NSAT and what values can be added or subtracted therefrom to change the sign bit. In these cases it is easy to predict underflow or overflow of the result. The manner in which four terms of the saturated multiplication and accumulation equation YOUTn=((((AC+_La₀X₀)+_La₁X₁)+_La₂X₂)+_La₃X₃)are evaluated has been described. If additional terms are to be evaluated, YOUTn is saved as the new accumulation value AC and input with the evaluation of the next four terms on the next clock cycle. This can continue on indefinitely but is typically limited to a number of terms in order to receive a desired accuracy.

[0043] As those of ordinary skill will recognize, the present invention has many advantages. One advantage of the present invention is that for nonvectorized processing, multiple samples are simultaneously computed thereby reducing the delay. Another advantage of the present invention is that the number of full additions is minimized in the saturated multiplication and accumulation such that the present invention is more efficient and uses fewer adders to perform the addition. Another advantage of the present invention is that it reduces the number of cycles to process the saturated multiplication and accumulation such that the channel handling capability increases to aggregate more communication channels to one device. The preferred embodiments of the present invention are thus described. While the present invention has been described in particular embodiments, it may be implemented in hardware, software, firmware or a combination thereof and utilized in systems, subsystems, components or sub-components thereof. When implemented in software, the elements of the present invention are essentially the code segments to perform the necessary tasks. The program or code segments can be stored in a processor readable medium or transmitted by a computer data signal embodied in a carrier wave over a transmission medium or communication link. The “processor readable medium” may include any medium that can store or transfer information. Examples of the processor readable medium include an electronic circuit, a semiconductor memory device, a ROM, a flash memory, an erasable ROM (EROM), a floppy diskette, a CD-ROM, an optical disk, a hard disk, a fiber optic medium, a radio frequency (RF) link, etc. The computer data signal may include any signal that can propagate over a transmission medium such as electronic network channels, optical fibers, air, electromagnetic, RF links, etc. The code segments may be downloaded via computer networks such as the Internet, Intranet, etc. In any case, the present invention should not be construed as limited by such embodiments, but rather construed according to the claims that follow below.