APPARATUS AND METHOD FOR CODE-ENHANCED PERFORMANCE IN A DIGITAL SIGNAL PROCESSING UNIT

Abstract

An improved microprocessor is provided having a program control unit for storing and then decoding instructions, a program address generation unit for generating addresses used to obtain instructions, an address register arithmetic unit for generating addresses for data, an arithmetic logic unit for performing operations on data, a shifter unit for shifting data in response to a predetermined instruction, a multiplier unit for performing multiplication of two input values; and a plurality of registers of which at least a portion are individually selectively associated with one or more of said units as a function of an instruction.

Description

TECHNICAL FIELD OF THE INVENTION

[0002] This invention relates to microprocessors, and particularly relates to improved microprocessors.

BACKGROUND OF THE INVENTION

[0003] A microprocessor is a central processing unit (CPU) for a digital processor which is usually contained in a single semiconductor integrated circuit or “chip” fabricated by “MOS/LSI” technology, as shown in U.S. Pat. No. 3,757,306 issued to Gary W. Boone and assigned to Texas Instruments. The Boone patent shows a single-chip 8-bit CPU including a parallel ALU, registers for data and addresses, and instruction register and a control decoder, all interconnected using the Von Neumann architecture and employing a bidirectional parallel bus for data, address and instructions. U.S. Pat. No. 4,074,351, issued to Gary W. Boone and Michael J. Cochran, assigned to Texas Instruments, shows a single-chip “microcomputer” type device which contains a 4-bit parallel ALU and its control circuitry, with on-chip ROM for program storage and on-chip RAM for data storage, constructed in the Harvard architecture. The term microprocessor usually refers to a device employing external memory for program and data storage, while the term microcomputer refers to a device with on-chip ROM and RAM for program and data storage; the terms are also used interchangeably, however, and are not intended as restrictive as to some features of this invention.

[0004] Subsequent to 1971 when U.S. Pat. Nos. 3,757,306 and 4,074,351 were originally filed, many improvements have been made in microprocessors and microcomputers to increase the speed and capability of these devices and reduce the cost of manufacture, providing more circuitry and functions in less space, i.e., smaller chip size. Improved VLSI semiconductor processing and photolithographic techniques allow narrower line widths and higher resolution, providing added circuit density and higher speed, but circuit and system improvements also contribute to the goals of increased performance with smaller chip size. Some of these improvements in microcomputers are disclosed in the following U.S. Patents, all assigned to Texas Instruments; U.S. Pat. No. 3,991,305 issued to Edward R. Caudel and Joseph H. Raymond, Jr.; U.S. Pat. No. 4,156,927 issued to David J. McElroy and Graham S. Tubbs; U.S. Pat. No. 3,934,233 issued to R. J. Fisher and G. D. Rogers; U.S. Pat. No. 3,921,142 issued to J. D. Bryant and G. A. Hartsell; U.S. Pat. No. 3,900,722 issued to M. J. Cochran and C. P. Grant; U.S. Pat. No. 3,932,846 issued to C. W. Brixey et al; U.S. Pat. No. 3,939,335 issued to G. L. Brantingham, L. H. Phillips and L. T. Novak; U.S. Pat. No. 4,125,901 issued to S. P. Hamilton, L. L. Miles, et al; U.S. Pat No. 4,158,432 issued to M. G. VanBavel; U.S. Pat. Nos. 3,757,308 and 3,984,816. The devices described in these patents have been of the Harvard architecture and of the 4-bit type, particularly adapted for calculator or controller applications.

[0005] Additional examples of microprocessor and microcomputer devices in the evolution of this technology are described in publications. In Electronics, Sep. 25, 1972, pp. 31-32, a 4-bit P-channel MOS microcomputer with on-chip ROM and RAM is shown which is similar to U.S. Pat. No. 3,991,305. Two of the most widely used 8-bit microprocessors like that of U.S. Pat. No. 3,757,306 are described in Electronics, Apr. 18, 1974 at pp. 85-95 (the Motorola 6800) and pp. 95-100 (the Intel 8080). A microcomputer version of the 6800 is described in Electronics, Feb. 2, 1978 at pp 99-105. Likewise, a single-chip microcomputer version of the 8080 is shown in Electronics, Nov. 25, 1976 at pp. 99-105. Another single-chip microcomputer, the Mostek 3872, is shown in Electronics, May 11, 1978, at p. 105-110, and an improved version of the 6800 is disclosed in Electronics, Sep. 17, 1979 at pp. 122-125. Sixteen-bit microprocessors based on minicomputer instruction sets evolved such as the part number TMS9900 described in a book entitled “9900 Family Systems Design”, published in 1978 by Texas Instruments Incorporated, P.O. Box 1443, M/S 6404, Houston, Tex. 77001, Library of Congress Catalog No. 78-058005. The 8086, a 16-bit microprocessor evolving from the 8080, is described in Electronics, Feb. 16, 1978, pp. 99-104, while a 16-bit microprocessor identified as the 68000 (based on the 6800) is described in Electronic Design, Sep. 1, 1978 at pp. 100-107, and in IEEE Computer, Vol. 12. No. 2, pp. 43-52 (1979).

[0006] These prior 8-bit and 16-bit microprocessors and microcomputers have been general-purpose processors of the Von Neumann architecture with multiplexed address/data busses; some have been microcoded as described in U.S. patent application Ser. No. 209,915, filed Nov. 24, 1980 by Guttag, McDonough and Laws, or Ser. No. 253,624, filed Apr. 13, 1981 by Hayn, McDonough and Bellay, both assigned to Texas Instruments, and at pp. 28-34, IEEE Spectrum, March 1979, by McKevitt and Bayliss, or Proceedings 11th Annual Microprogramming Workshop, December, 1979 by Stintter and Tredenick. Microcoding, originally described by Wilkes in 1951, employs a control ROM to store microinstruction sequences entered by instruction words; the programmer works in a higher level machine code, so the number of assembly language code statements is supposedly reduced, and thus programming cost is reduced.

[0007] Some additional improvements in microcomputers are disclosed in the following U.S. Patents, all assigned to Texas Instruments; U.S. Pat. Nos. 4,577,282, 4,831,570 and 5,072,418. The devices described in these patents have been 16-bit modified Harvard architecture devices, particularly adapted for varying applications because of their programmability.

[0008] Microprocessor designers have increasingly endeavored to improve performance in various microprocessors by increasing clock speeds and adding parallelism. Large blocks of random access memory (RAM) are included within the microprocessor for data storage and for program storage in order to reduce memory access times. Various types of input/output devices are included within a microprocessor integrated circuit in order to reduce total system chip count and cost.

[0009] In contrast, features of the invention may be preferably employed in a special-purpose high-speed microcomputer device according to the embodiment described herein which departs from these contemporary microprocessor devices in several major respects in order to achieve substantial speed and performance advantages. This device is generally a non-microcoded processor of modified Harvard architecture.

[0010] An object of the present invention is to provide an improved microprocessor.

SUMMARY OF THE INVENTION

[0011] In one embodiment of the present invention an improved microprocessor is provided having a program control unit for storing and then decoding instructions, a program address generation unit for generating addresses used to obtain instructions, an address register arithmetic unit for generating addresses for data, an arithmetic logic unit for performing operations on data, a shifter unit for shifting data in response to a predetermined instruction, a multiplier unit for performing multiplication of two input values; and a plurality of registers of which at least a portion are individually selectively associated with one or more of said units as a function of an instruction.

[0012] These and other features of the invention that will be apparent to those skilled in the art from the following detailed description of the invention, taken together with the accompanying drawings.

DESCRIPTION OF THE DRAWINGS

[0013] FIG. 1 illustrates a high-level block diagram of a DSP core of the present invention;

[0014] FIG. 2 shows a high-level memory map for the DSP core of FIG. 1;

[0015] FIGS. 3A, 3B and 3C show possible maps for a portion of FIG. 2;

[0016] FIGS. 4A, 4B and 4C show possible maps for a portion of FIG. 2;

[0017] FIG. 5 is a block diagram of a CPU of FIG. 1;

[0018] FIG. 6 illustrates individually accessible portions of the accumulator of the CPU of FIG. 5;

[0019] FIG. 7 illustrates individually accessible halves of the P Register of the CPU of FIG. 5;

[0020] FIG. 8 illustrates pages of Data Memory;

[0021] FIG. 9 shows Address Reach of the Stack Pointer;

[0022] FIG. 10 shows Address Reach of the Auxiliary Registers;

[0023] FIG. 11 AR6 and AR7;

[0024] FIG. 12 shows the bit positions of Status Register ST0;

[0025] FIG. 13 shows the bit positions of Status Register ST1;

[0026] FIG. 14 shows Status Register ST0;

[0027] FIG. 15 shows Status Register ST1;

[0028] FIG. 16 illustrates a block Diagram of Components Involved in the multiplier of the CPU of FIG. 5;

[0029] FIG. 17 illustrates the symbols and their meanings for the Shift Operations depicted in FIGS. 18;

[0030] FIGS. 18A, 18B and 18C show Shift Operations;

[0031] FIG. 19 illustrates the bit positions for an Interrupt Flag Register (IFR);

[0032] FIG. 20 illustrates the bit positions for an Interrupt Enable Register (IER);

[0033] FIG. 21 illustrates the bit positions for a Debug Interrupt Enable Register (DBGIER);

[0034] FIG. 22 illustrates a Process for Maskable Interrupts;

[0035] FIG. 23 illustrates a Functional Flow Chart for an Interrupt Initiated by a TRAP Instruction;

[0036] FIGS. 24A and 24B illustrate pipeline phases and relationships between Address Counters FC, IC and PC;

[0037] FIG. 25 illustrates pipeline Phases During Which Address Registers Are Read From or Written To;

[0038] FIG. 26 shows an Accessing Memory in Immediate-Pointer Addressing Mode;

[0039] FIG. 27 illustrates Pages of Data Memory;

[0040] FIG. 28 illustrates Accessing Data Memory in Stack-Pointer Indirect Addressing Mode;

[0041] FIG. 29 illustrates Accessing Data Memory in PAGE0 Stack Addressing Mode;

[0042] FIG. 30 illustrates Accessing Data Memory with the Auxiliary Registers;

[0043] FIG. 31 shows a Circular Addressing Mode;

[0044] FIG. 32 illustrates Opcode Addressing Mode Information;

[0045] FIG. 33 shows an Opcode Format for Immediate Addressing With an 8-Bit Constant;

[0046] FIG. 34 shows an Opcode Format for Immediate Addressing With a 5-Bit Constant;

[0047] FIG. 35 shows an Opcode Format for Immediate Addressing With a 16-Bit Constant;

[0048] FIG. 36 shows an Opcode Format for Immediate Addressing With a 22-Bit Constant;

[0049] FIG. 37 shows an Opcode Format for an Instruction Using Register Addressing Mode;

[0050] FIG. 38 shows an Opcode Format for an Instruction Using DP Direct Addressing Mode;

[0051] FIG. 39 shows an Opcode Format for an Instruction Using Auxiliary-Register Indirect Addressing Mode;

[0052] FIG. 40 shows an Opcode Format for an Instruction Using Stack-Pointer Indirect Addressing Mode;

[0053] FIG. 41 illustrates a 32-Bit Read From Data Memory;

[0054] FIG. 42 illustrates a 32-Bit Write to Data Memory; and

[0055] FIG. 43 illustrates an example Opcode taken from the Assembly Language Instructions listed in Appendix B.

DETAILED DESCRIPTION OF THE INVENTION

Architectural Overview

[0056] A digital signal processor (DSP) chip contains a DSP core. The DSP core is responsible for the central control activities of a DSP, activities such as address generation, arithmetic operations, data transfers, and system monitoring. The following briefly describes the components of a DSP core of the present invention, its presently preferred associated memory map, and a presently preferred memory interface.

[0057] Preferably, the DSP core of the present invention is a 16-bit, fixed-point processor. As shown in FIG. 1, this DSP core contains:

[0058] a) a CPU for: generating data- and program-memory addresses, decoding and executing instructions, performing arithmetic, logical, and shift operations, and controlling data transfers among CPU registers, data memory, and program memory;

[0059] b) emulation logic for: monitoring and controlling various parts and functionalities of the DSP and for testing device operation; and

[0060] c) signals for: interfacing with memory and peripherals, clocking and controlling the CPU and the emulation circuitry, showing the status of the CPU and the emulation circuitry, and using interrupts.

[0061] The DSP core does not contain memory, a clock generator, or peripheral devices.

[0062] The CPU is discussed in more detail later herein, but the following is a list of its major features:

[0063] a) a protected pipeline—The CPU implements an 8-phase pipeline that prevents a read from and a write to the same location from occurring out of order;

[0064] b) independent register space—The CPU contains registers that are not mapped to data space. These registers function as system-control registers, math registers, and data pointers. The system-control registers are accessed by special instructions. The other registers are accessed by special instructions or by a special addressing mode (register addressing mode);

[0065] c) arithmetic logic unit (ALU)—The32-bit ALU performs 2s-complement arithmetic and Boolean logic operations;

[0066] d) address register arithmetic unit (ARAU)—The ARAU generates data-memory addresses and increments or decrements pointers in parallel with ALU operations;

[0067] e) barrel shifter—This shifter performs all left and right shifts of data. It can shift data to the left by up to 16 bits and to the right by up to 16 bits; and

[0068] f) multiplier—The multiplier performs 16-bit×16-bit 2s-complement multiplication with a 32-bit result. The multiplication can be performed with two signed numbers, two unsigned numbers, or one signed number and one unsigned number.

[0069] The emulation logic includes the following features:

[0070] a) debug and test direct memory access (DT-DMA)—The debug host can gain direct access to the content of registers and memory by taking control of the memory interface during unused cycles of the instruction pipeline;

[0071] b) data logging—The emulation logic enables application-initiated transfer of memory contents between the DSP and a debug host device;

[0072] c) 32-bit counter for performance benchmarking;

[0073] d) multiple debug events—Any of the following debug events can cause a break in program execution:

[0074] i) A breakpoint initiated by the ESTOP0 or ESTOP1 instruction;

[0075] ii) An access to a specified program or data location; and

[0076] iii) A request from the debug host or other hardware.

[0077] When a debug event causes the DSP to enter the debug state, the event is called a break event.

[0078] e) real-time mode of operation—When the DSP is in this mode and a break event occurs, the main body of program code comes to a halt, but time-critical interrupts can still be serviced; and

[0079] f) interrupt generation capabilities.

[0080] As shown in FIG. 1, the DSP core has four main types of signals:

[0081] a) memory-interface signals—These signals transfer data among the core, memory, and peripherals; indicate program-memory accesses and data-memory accesses; and differentiate between accesses of different sizes (16-bit or 32-bit).

[0082] b) clock and control signals—These signals provide clocking for the CPU and the emulation logic, and they are used to control and monitor the CPU.

[0083] c) reset and interrupt signals—These signals are used for generating a hardware re-set and interrupts, and for monitoring the status of interrupts.

[0084] d) emulation signals—These signals are used for testing and debugging.

[0085] The DSP core contains no memory, but can access memory elsewhere on-chip with the DSP or off chip. The DSP uses 22-bit addresses, supporting a total address reach of 4M words (1 word=16 bits) in both program space and data space. FIG. 2 shows a high-level memory map view of how addresses may be allocated between program space and data space. In most cases, hexadecimal numbers are shown with a subscript of 16. For example, the hexadecimal number 40 would be shown as 4016. The exception to this rule is a hexadecimal number in a code example; these hexadecimal numbers have the suffix h. For example, the number 40 in the following code is a hexadecimal 40. MOVB ARO,#40h

[0086] The memory map in FIG. 2 has been divided into the following segments:

[0087] a) on-chip program/data—Addresses 00 000016-00 03FF16 in program space and 00 040016-00 07FF16 in data space are for use with a single on-chip block of program/data memory. For example, the same location may be accessed with address 00 004116 in program space or address 00 044116 in data space. Within these addresses resides a block B0, which is 256, 512, or 1K words.

[0088] b) program—Addresses 00040016-3FFFFF16 in program space are avail-able for program memory. This additional memory may be on-chip and/or off-chip.

[0089] c) on-chip data—Addresses 00000016-0003FF16 in data space are for a block of data-memory on board the DSP. This block, called B1, may be 256, 512, or 1K words.

[0090] d) reserved—Addresses 00080016-000CFF16 in data space are reserved for special registers. Some of the addresses in the range 00080016-0009FF16 are used for accessing memory-mapped emulation registers.

[0091] e) other data—Addresses 000D006 -3FFFFF16 in data space are available for additional data memory. This additional memory may be on-chip and/or off-chip.

[0092] Sixty-four addresses in program space are set aside for a table of 32 interrupt vectors. The vectors can be mapped to the top or bottom of program space by using a VMAP bit. As shown in FIGS. 3A-C, if VMAP=0, the vectors are mapped to addresses 00 000016-00 003F16; they are also mapped to addresses 00 040016-00 043F16 in data space. When VMAP=1, the vectors are only mapped to program space at addresses 3F FFCO16 -3F FFFF16.

[0093] The possible maps for block B0 are also shown in FIGS. 3A-C. BO may be selected as 256, 512, or 1K words, as depicted in FIGS. 3A, 3B, and 3C, respectively. If B0 has fewer than 1024 (1K) words, some of the program/data addresses are reserved (not available for use).

[0094] The 1K block of memory shown in FIGS. 3A-C, whether all used for B0 or partially reserved, can be addresses in program space or data space.

[0095] For example, the location with address 00 004116 may be in program space or in data space.

[0096] FIGS. 4A-C show the possible maps for block B1, which reside only in data space. B1 may be selected to be 256, 512, or 1K words, as depicted in FIGS. 4A, 4B, and 4C, respectively. If B1 has 256 words, mirrors of B1 are in the address ranges 00 0100016-00 01FF16, 00 020016-00 02FF16, and 00 030016-00 03FF16 As an example of what this means, the first word of B1, may be accessed at any one of the following addresses: 00 000016, 00 010016, 00 020016, or 00 030016.

[0097] If B1 has 512 words, there is one mirror of B1 at addresses 00 020016-00 040016. If B1 has 1K words, it uses all the available addresses.

[0098] The DSP CPU supports long call and branch operations, and a 22-bit program counter (PC) enables access of program space as one linear memory space. The DSP also supports relative offset branching, which enables code to be relocated at run time.

[0099] The DSP CPU supports a 16-bit stack pointer (SP) that points to data memory. Using this register, a stack may be created within the first 64K of data space (addresses 00 000016-00 FFFF16). SP is incremented such that the stack grows from low memory to high memory and SP always points to the next empty location. After a hardware reset, SP points to address 00 000016.

[0100] Data space is supported by a rich set of addressing modes. Indirect addressing modes use 16-bit pointers (the SP and the auxiliary registers AR0-AR5) to access the low 64K of data space and 22-bit pointers (auxiliary registers XAR6 and XAR7) to access the full range of data space. Direct addressing modes access all of data space by concatenating a 16-bit data page number with a 6-bit offset that is embedded in an instruction. The addressing modes are described later herein.

[0101] Instructions are included to move data between program space and data space. The PREAD instruction can be used to transfer a program-memory value to data memory. The PWRITE instruction can transfer a data-memory value to program memory.

[0102] The DSP memory map is accessible outside the core by the memory interface, which connects the DSP core to memories, peripherals, or other interfaces. The memory interface includes separate buses for program space and data space. This means an instruction can be fetched from program memory while data memory is being accessed.

[0103] The interface also includes signals that indicate the type of read or write being requested by the CPU. These signals can select a specified memory block or peripheral for a given bus transaction. In addition to 16-bit and 32-bit accesses, the DSP supports special byte-access instructions which can access the least significant byte (LSByte) or most significant byte (MSByte) of an addressed 16-bit word. Strobe signals indicate when such an access is occurring on a data bus.

[0104] The memory interface has three 22-bit address buses:

[0105] 1) the program address bus (PAB) carries addresses for reads and writes from program space;

[0106] 2) the data-read address bus (DRAB) carries addresses for reads from data space; and

[0107] 3) the data-write address bus (DWAB) carries addresses for writes to data space.

[0108] The memory interface also has three 32-bit data buses:

[0109] 1) the program-read data bus (PRDB) carries instructions or data during reads from program space;

[0110] 2) the data-read data bus (DRDB) carries data during reads from data space;

[0111] 3 ) the data-/program-write data bus (DWDB) carries data during writes to data space or program space.

[0112] Table 1 below summarizes how these buses are used during accesses.

TABLE 1
Summary of Bus Use During
Data-Space and Program-Space Accesses
	Access Type	Address Bus	Data Bus
	Read from program space	PAB	PRDB
	Read from data space	DRAB	DRDB
	Write to program space	PAB	DWDB
	Write to data space	DWAB	DWDB

[0113] A program-space read and a program-space write cannot happen simultaneously because both use the PAB. Similarly, a program-space write and a data-space write cannot happen simultaneously because both use the DWDB.

[0114] Transactions that use different buses can happen simultaneously. For example, the CPU can read from program space (using PAB and PRDB), read from data space (using DRAB and DRDB), and write to data space (using DWAB and DWDB) at the same time.

[0115] Typically, PAB and PRDB are used only for reading instructions from program space, and DWDB is used only for writing data to data space. However, the instructions in Table 2 below are exceptions to this behavior.

TABLE 2
Special Bus Operations
Instruction Special Bus Operation
PREAD       This instruction reads a data value rather than an instruction
            from program space. It then transfers that value to data
            memory or a register. For the read from program space,
            the core places the source address on the program address
            bus (PAB), sets the appropriate program-space select signals,
            and reads the data value from the program-read data bus
            (PRDB).
WRITE       This instruction writes a data value to program space. The
            value is read from data space or a register. For the write to
            program space, the core places the destination address on the
            program address bus (PAB), sets the appropriate program -
            space select signals, and writes the data value to the
            data-/program-write data bus (DWDB).
MAC         As part of its operation, this instruction multiplies two data
            values, one of which is read from program space. For the read
            from program space, the core places the program-space source
            address on the program address bus (PAB), sets the
            appropriate program-space select signals, and reads the
            program data value from the program read data bus (PRDB).

[0116] The DSP core expects memory wrappers or peripheral-interface logic to align any 32-bit read or write to an even address. If the address-generation logic generates an odd address, the CPU must begin reading or writing at the previous even address. This alignment does not affect the address values generated by the address-generation logic.

[0117] Most instruction fetches from program space are performed as 32-bit read operations and are aligned accordingly. However, alignment of instruction fetches are effectively invisible to a programmer. When instructions are stored to program space, they do not have to be aligned to even addresses. Instruction boundaries are decoded within the core.

[0118] When using instructions that perform 32-bit reads from or writes to data space, alignment should be considered. Examples of alignment of data-memory accesses and the instructions that are subject to this alignment are provided later herein.

[0119] The central processing unit (CPU) is responsible for controlling the flow of a program and the processing of instructions. It performs arithmetic, Boolean-logic, multiply, and shift operations. FIG. 5 shows the major blocks and data paths of the CPU of the DSP core of the present invention. The shaded buses are memory-interface buses that are external to the CPU. The main blocks of the CPU are:

[0120] a) program control logic—This logic stores a queue of instructions that have been fetched from program memory by way of the program-read bus (PRDB). It also decodes these instructions and passes commands and constant data to other parts of the CPU.

[0121] b) program address generation logic—This logic generates the addresses used to fetch instructions or data from program memory and places each address on the program address bus, (PAB). As part of its operation, the program-address generation logic uses several counters, addresses supplied by the program control logic, and addresses from the extended auxiliary register XAR7.

[0122] c) address register arithmetic unit (ARAU)—The ARAU generates addresses for values that must be fetched from data memory. For a data read, it places the address on the data-read address bus (DRAB); for a data write, it loads the data-write address bus (DWAB). The ARAU also increments or decrements the stack pointer (SP) and the auxiliary registers (AR0, AR1, AR2, AR3, AR4, AR5, XAR6, and XAR7).

[0123] d) arithmetic logic unit (ALU)—The 32-bit ALU performs 2s-complement arithmetic and Boolean logic operations. Before doing its calculations, the ALU accepts data from registers, from data memory, or from the program control logic. Inputs come from the registers, the program control logic, and the data-read data bus (DRDB). Results that go to data memory are sent to the data-/program-write data bus (DWDB).

[0124] e) multiplier—The multiplier performs 16-bit×16-bit 2s-complement multiplication with a 32-bit result. In conjunction with the multiplier, the DSP uses the 16-bit multiplicand register (T), the 32-bit product register (P), and the 32-bit accumulator (ACC). The T register supplies one of the values to be multiplied. The result of the multiplication can be sent to the P register or to ACC.

[0125] f) barrel shifter—The shifter performs all the shifts for the DSP. It can shift data to the left by up to 16 bits and to the right by up to 16 bits. The amount and direction of the shift is determined by the particular instruction that causes the shift.

[0126] Table 3 lists the main CPU registers and their values after reset. The registers are described in more detail later herein.

TABLE 3
CPU Register Summary
Register	Size	Description	Value After Reset
ACC	32 bits	Accumulator	0000_000016
AH	16 bits	High half of ACC	000016
AL	16 bits	Low half of ACC	000016
AR0	16 bits	Auxiliary register	0 000016
AR1	16 bits	Auxiliary register 1	000016
AR2	16 bits	Auxiliary register 2	000016
AR3	16 bits	Auxiliary register 3	000016
AR4	16 bits	Auxiliary register 4	000016
AR5	16 bits	Auxiliary register 5	000016
XAR6	22 bits	Extended auxiliary	00 000016
		register 6
AR6	16 bits	16 LSBs of XAR6	000016
XAR7	22 bits	Extended auxiliary	00 000016
		register 7
AR7	16 bits	16 LSBs of XAR7	000016
DP	16 bits	Data page pointer	000016
IFR	16 bits	Interrupt flag register	000016
IER	16 bits	Interrupt enable	000016
		register
DBGIER	16 bits	Debug interrupt enable	000016
		register
P	32 bits	Product register	0000 000016
PH	16 bits	High half of P	000016
PL	16 bits	Low half of P	000016
PC†	22 bits	Program counter	If VMAP bit = 0
			PC = [00_000016]
			If VMAP bit = 1
			PC = [3F_FFC016]
SP	16 bits	Stack pointer	000016
ST0	16 bits	Status register	0 000016
ST1‡	16 bits	Status register 1	If VMAP input signal low
			ST1 = 000316
			If VMAP input signal high
			ST1 = 000B16
T	16 bits	Multiplicand register	000016
†The value in the PC after reset depends on the value in the VMAP bit at the time reset is initiated. The VMAP bit is bit 3 of ST1. The PC will be loaded with the reset vector at either address 00_000016 or address 3F_FFC016.
‡The value in ST1 after reset depends, in part, on the level on the VMAP input signal at the time reset is initiated. If the VMAP signal is low, bit 3 (VMAP) is cleared. If the VMAP signal is high bit 3 is set.

[0127] The accumulator (ACC) is the main working register for the CPU. It is the destination for all ALU operations except those which operate directly on memory and registers. ACC supports single-cycle move, add, subtract, and compare operations from 32-bit-wide data memory. It can also accept the 32-bit result of a multiplication operation. As depicted in FIG. 6, the halves and quarters of the ACC can also be individually accessed. ACC can be treated as two independent 16-bit registers: AH (high 16 bits) and AL (low 16 bits). The bytes within AH and AL can also be accessed independently. Special byte-move instructions load and store the most significant byte or least significant byte of AH or AL. This enables efficient byte packing and unpacking.

[0128] The following abbreviations are used in FIG. 6 and include the associated accumulator bit positions:

[0129] AH=ACC (31:16); AH.MSB=ACC (31:24); AH.LSB=ACC (23:16); AL=ACC (15:0); AL.MSB=ACC (15:8); and AL.LSB=ACC (7:0).

[0130] The accumulator has the following associated status bits. The details on these bits are described later herein with reference to a status register, ST0.

[0131] Overflow mode bit (OVM), Sign-extension mode bit (SXM), Test/control flag bit (TC), Carry bit (C), Zero flag bit (Z), Negative flag bit (N), Latched overflow flag bit (V), Overflow counter bits (OVC).

[0132] Table 4 below shows the ways to shift the content of AH, AL, or ACC.

TABLE 4
Available Operations for
Shifting Values in the Accumulator
Register	Shift Direction	Shift Type	Instruction
ACC	Left	Logical	LSL
		Rotation	ROL
	Right	Arithmetic	SFR with SXM = 1
		Logical	SFR with SXM = 0
		Rotation	ROR
AH or AL	Left	Logical	LSL
	Right	Arithmetic	ASR
		Logical	LSR

[0133] The multiplicand register (T register) is primarily used to store a 16-bit signed integer value prior to a multiply operation. During every multiplication, one of the multiplicands is taken from the T register. For some left and right barrel-shift operations, the shift amount (0 to 15) is determined by the four least significant bits of the T register. In these shift operations, the 12 most significant bits of T are ignored. The contents of T can be stored back to memory.

[0134] The product register (P register) (P/PH, PL) is typically used to hold the 32-bit result of a 16×16 multiplication. It can also be loaded directly from a 16- or 32-bit data-memory location, a 16-bit constant, the 32-bit ACC, or a 16-bit addressable CPU register. As shown in FIG. 7, the P register may be treated as a 32-bit register or as two independent 16-bit registers: PH (high 16 bits) and PL (low 16 bits).

[0135] When an instruction accesses P, PH, or PL, all 32-bits are copied to the ALU-shifter block, where the barrel shifter may perform a left shift, a right shift, or no shift. Then the entire 32 bits (P) or the upper or lower 16 bits (PH or PL) are used in an ALU calculation or stored. The action of the shifter is determined by the product shift mode (PM) bits in status register ST0. Table 5 below shows the possible PM values and the corresponding product shift modes. When the barrel shifter performs a left shift, the low order bits are filled with zeros. When the shifter performs a right shift, the P register value is sign extended. Instructions that use PH or PL as operands ignore the product shift mode.

TABLE 5
Product Shift Modes
PM Value Product Shift Mode
0002     Left shift by 1
0012     No shift
0102     Right shift by 1
0112     Right shift by 2
1002     Right shift by 3
1012     Right shift by 4
1102     Right shift by 5
1112     Right shift by 6

[0136] In the direct addressing modes, data memory is addressed in blocks of 64 words called data pages. The entire 4M words of data memory consists of 65 536 data pages labeled 0 through 65535, as shown in FIG. 8. In DP direct addressing mode, the 16-bit data page pointer (DP) holds the current data page number. The data page is changed by loading the DP with a new number.

[0137] The stack pointer (SP) enables the use of a software stack in data memory. As depicted in FIG. 9, the stack pointer has only 16 bits and can only address the low 64K of data space. When the SP is used, the upper six bits of the 22-bit address are forced to 0. (For information about addressing modes that use the SP). After reset, SP points to address 00 000016.

[0138] The operation of the stack is as follows: the stack grows from low memory to high memory; the SP always points to the next empty location in the stack; at reset, the SP is cleared, so that it points to address 00000016.

[0139] When 32-bit values are saved to the stack, the least significant 16 bits are saved first, and the most significant 16 bits are saved to the next higher address (little endian format).

[0140] When 32-bit operations read or write a 32-bit value, the DSP core expects the memory wrapper or peripheral-interface logic to align that read or write to an even address. For example, if the SP contains the odd address 00008316, a 32-bit read operation reads from addresses 00008216 and 00008316.

[0141] The SP will overflow if its value is increased beyond FFFF16 or decreased below 000016. When the SP increases past FFFF16,it counts forward from 000016. For example, if SP=FFFE16 and an instruction adds 3 to the SP, the result is 00016. When the SP decreases past 000016, it counts backward from FFFF16. For example, if SP=000216 and an instruction subtracts 4 from SP, the result is FFFE16.

[0142] When values are being saved to the stack, the SP is not forced to align with even or odd addresses. Alignment is forced by the memory wrapper or peripheral-interface logic.

[0143] The CPU provides eight auxiliary registers that can be used as pointers to memory. There are six 16-bit auxiliary registers (AR0, AR1, AR2, AR3, AR4, and AR5), and two 22-bit extended auxiliary registers (XAR6 and XAR7).

[0144] FIG. 10 shows the ranges of data memory accessible to the 16-bit and 22-bit auxiliary registers. Because AR0-AR5 are 16-bits wide, their range is limited to addresses 00 000016-00 FFFF16. When one of these six auxiliary registers is used, the content of the auxiliary register (or the content plus an offset) is concatenated with six leading zeros to form the full 22-bit address. XAR6 and XAR7, being 22 bits wide, can be used to access the full range of data memory. XAR7 can also be used by some instructions to point to any value in program memory.

[0145] Although the addressing modes use the full extended auxiliary registers, many instructions allow you to access the 16 least significant bits (LSBs) of XAR6 and XAR7. As shown in FIG. 11, the 16 LSBs of XAR6 are known as auxiliary register AR6, and the 16 LSBs of XAR7 are known as auxiliary register AR7.

[0146] When the pipeline is full, the 22-bit program counter (PC) always points to the instruction that is currently being processed—the instruction that has just reached the decode 2 phase of the pipeline. Once an instruction reaches this phase of the pipeline, it cannot be flushed from the pipeline by an interrupt. It is executed before the interrupt is taken.

[0147] The DSP has two status registers, ST0 and ST1, which contain various flag bits and control bits. These registers can be stored into and loaded from data memory, enabling the status of the machine to be saved and restored for sub-routines. FIG. 12 and FIG. 13 show the organization of status registers ST0 and ST1, respectively. The status bits have been organized according to when the bit value is modified in the pipeline. Bits in ST0 are modified in the execute phase of the pipeline; bits in ST1 are modified in the decode 2 phase. The status bits are described in detail later herein.

[0148] In FIGS. 12 and 13, R means a Read access and W means a Write access; the value following dash (−) is value after reset. Reserved bits are always 0s and are not affected by writes. In FIG. 13, the value of the VMAP bit after reset depends on the level of the VMAP input signal at reset. If the VMAP signal is low, the VMAP bit is 0 after reset; if the VMAP signal is high, the VMAP bit is 1 after reset.

[0149] The DSP has three registers dedicated to the control of interrupts: interrupt flag register (IFR); interrupt enable register (IER); and debug interrupt enable register (DBGIER).

[0150] The IFR contains flag bits for maskable interrupts (those that can be enabled and disabled with software). When one of these flags is set, by hardware or software, the corresponding interrupt will be serviced if it is enabled. A maskable interrupt is enabled or disabled by its corresponding bit in the IER. The DBGIER indicates the time-critical interrupts that will be serviced (if enabled) while the DSP is in real-time emulation mode and the CPU is halted. The DSP interrupts and the interrupt-control registers are described later herein.

[0151] FIG. 14 shows the bit fields of status register ST0. All of these bit fields are modified in the execute phase of the pipeline. Detailed descriptions of these bits follow the figure.

[0152] In FIG. 14, R means a Read access and W means a Write access; the value following dash (−) is value after reset.

[0153] In FIG. 14, bits 15-10 are the overflow counter (OVC). The overflow counter is a 6-bit signed counter with a range of −32 to 31. When overflow mode is on (OVM=1) and an overflow occurs in ACC, the CPU automatically fills ACC with a positive or negative saturation. When overflow mode is off (OVM=0 ), ACC overflows normally, and OVC can be used to in-crease the effective dynamic range of ACC to 38 bits.

[0154] When ACC overflows in the positive direction (from 7FFFF FFFF16 to 8000 000016) the OVC is incremented by 1. When ACC overflows in the negative direction (from 8000 000016 to 7FFF FFFF16) the OVC is decremented by 1. The increment or decrement is performed as the overflow affects the V flag.

[0155] If OVC increments past its most positive value, 31, the counter wraps around to −32. If OVC decrements past its most negative value, −32, the counter wraps around to 31. At reset, OVC is cleared. OVC is not affected by overflows in registers other than ACC and is not affected by compare instructions (CMP and CMPL). Table 6 below, explains how OVC may be affected by the saturate accumulator (SAT ACC) instruction.

TABLE 6
Operation Performed by SAT ACC Instruction
Condition Operation Performed by SAT ACC Instruction
OVC = 0   Leave ACC and OVC unchanged.
OVC > 0   Saturate ACC in the positive direction
          (fill ACC with 7FFF FFFF16), and clear OVC.
OVC < 0   Saturate ACC in the negative direction
          (fill ACC with 8000 000016), and clear OVC.

[0156] In FIG. 14, bits 9-7 are product shift mode bits (PM). This 3-bit value determines the shift mode for any output operation from the product (P) register. The shift modes are shown in Table 7. The output can be to the ALU or to memory. All instructions that are affected by the product shift mode will sign extend the P register value during a right shift operation. At reset, PM is cleared.

TABLE 7
Product Shift Mode
PM  Product Shift Mode
000 Left shift by 1. During the shift, the low-order bit is zero filled.
    At reset, this mode is selected.
001 No shift
010 Right shift by 1. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
011 Right shift by 2. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
100 Right shift by 3. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
101 Right shift by 4. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
110 Right shift by 5. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
111 Right shift by 6. During the shift, the lower bits are lost, and the
    shifted value is sign extended.
Note:
If performing unsigned arithmetic use a product shift of 0 (PM = 001) to avoid sign extension and generation of incorrect results.

[0157] In FIG. 14, bit 6 is the overflow flag (V). If the result of an operation causes an overflow in the register holding the result, V is set and latched. If no overflow occurs, V is not modified. Once V is latched, it remains set until it is cleared by reset or by a conditional branch instruction that tests V. Such a conditional branch clears V regardless of whether the tested condition (V=0 or V=1) is true.

[0158] An overflow occurs in ACC (and V is set) if the result of an addition or subtraction does not fit within the signed numerical range −231 to (+231−1), or 800 00016 to 7FFF FFFF16. An overflow occurs in AH, AL, or another 16-bit register or data-memory location if the result of an addition or subtraction does not fit within the signed numerical range −215 to (+215−1), or 800016to 7FFF16.

[0159] Compare instructions (CMP, CMPB and CMPL) do not affect the state of the flag. V can be summarized in Table 8 as follows:

TABLE 8
V Condition
V Condition
0 V has been cleared.
1 An overflow has been detected, or V has been set.

[0160] In FIG. 14, bit 5 is the negative flag (N). During certain operations, N is set if the result of the operation is a negative number or cleared if the result is a positive number. At reset, N is cleared. Results in ACC are tested for the negative condition. Bit 31 of ACC is the sign bit. If bit 31 is a 0, ACC is positive; if bit 31 is a 1, ACC is negative. N is set if a result in ACC is negative or cleared if a result is positive.

[0161] Results in AH, AL, and other 16-bit registers or data-memory locations are also tested for the negative condition. In these cases bit 15 of the value is the sign bit (1 indicates negative, 0 indicates positive). N is set if the value is negative or cleared if the value is positive.

[0162] The TEST ACC instruction sets N if the value in ACC is negative. Otherwise the instruction clears N. As shown in Table 9, for compare operations, the N flag is set differently under overflow conditions than for addition or subtraction operations. For addition or subtraction operations, the N flag is set to match the most significant bit of the truncated result. For compare operations, the N flag assumes infinite precision. This applies to operations whose result is loaded to ACC, AH, AL, another register, or a data-memory location.

TABLE 9
Overflow
A†	B†	(A - B)	Subtraction	Compare‡
Pos	Neg	Neg	N = 1	N = 0
		(due to overflow in
		positive direction)
Neg	Pos	Pos	N = 0	N = 1
		(due to overflow in
		negative direction)
†For 32-bit data:
Pos = Positive number from 0000 000016 to 7FFF FFFF16
Neg = Negative number from 8000 000016 to FFFF FFFF16
For 16-bit data:
Pos = Positive number from 000016 to 7FFF16
Neg = Negative number from 800016 to FFFF16
‡The compare instructions are CMP, CMPB, and CMPL.

[0163] N can be summarized as follows:

TABLE 10
N Condition
N Condition
0 The tested number is positive, or N has been cleared.
1 The tested number is negative, or N has been cleared

[0164] In FIG. 14, bit 4 is the zero flag (Z). Z is set if the result of certain operations is 0 or is cleared if the result is nonzero. This applies to results that are loaded into ACC, AH, AL, another register, or a data-memory location. At reset, Z is cleared. The TEST ACC instruction sets Z if the value in ACC is 0. Otherwise, it clears Z. Z can be summarized as follows:

TABLE 11
Z Condition
Z Condition
0 The tested number is nonzero, or Z has been cleared.
1 The tested number is zero, or Z has been cleared

[0165] In FIG. 14, bit 3 is the carry bit (C). This bit indicates when an addition or increment generates a carry or when a subtraction, compare, or decrement generates a borrow. It is also affected by rotate operations on ACC and barrel shifts on ACC, AH, and AL. During additions/increments, C is set if the addition generates a carry; otherwise C is cleared. There is one exception: when using the ADD instruction with a shift of 16, the ADD instruction can set C but cannot clear C.

[0166] During subtractions/decrements/compares, C is cleared if the subtraction generates a carry; otherwise C is set. There is one exception: when using the SUB instruction with a shift of 16, the SUB instruction can clear C but can-not set C. This bit can be individually set and cleared by the SETC C instruction and CLRC C instruction, respectively. At reset, C is cleared. C can be summarized as follows:

TABLE 12
C Condition
C Condition
0 A subtraction generated a borrow, an addition did not
  generate a carry, or C has been cleared. Exception:
  An ADD instruction with a shift of 16 cannot clear C.
1 An addition generated a carry, a subtraction did not
  generate a borrow, or C has been set. Exception:
  A SUB instruction with a shift of 16 cannot set C.

[0167] In FIG. 14, bit 2 is the test/control flag (TC). This bit shows the result of a test performed by either the TBIT (test bit) instruction or the NORM (normalize) instruction. The TBIT instruction tests a specified bit. When TBIT is executed, the TC bit is set if the tested bit is 1 or cleared if the tested bit is 0. When a NORM instruction is executed, TC is modified as follows: If ACC holds 0, TC is set. If ACC does not hold 0, the CPU calculates the exclusive- OR of ACC bits 31 and 30, and then loads TC with the result. This bit can be individually set and cleared by the SETC TC instruction and CLRC TC instruction, respectively. At reset, TC is cleared.

[0168] In FIG. 14, bit 1 is the overflow mode bit (OVM). When ACC accepts the result of an addition or subtraction and the result causes an overflow, OVM determines how the CPU handles the overflow:

TABLE 13
How ACC Overflow is Handled
OVM How ACC Overflow is Handled
0   Results overflow normally in ACC.
1   ACC is filled with either its most positive or
    most negative value as follows: If ACC overflows
    in the positive direction (from 7FFF FFFF16 to
    8000 000016), ACC is then filled with 7FFF
    FFFF16. If ACC overflows in the negative
    direction (from 8000 000016 to 7FFF FFFF16),
    ACC is then filled with 8000 000016.

[0169] This bit can be individually set and cleared by the SETC OVM instruction and CLRC OVM instruction, respectively. At reset, OVM is cleared.

[0170] In FIG. 14, bit 0 is the sign-extension mode bit (SXM). When data is shifted before it is loaded into, added to, or subtracted from the accumulator, SXM determines whether the data is sign extended during the shift:

TABLE 14
Sign Extension Mode
SXM Sign Extension Mode
0   Sign extension is suppressed.
1   Sign extension is enabled.

[0171] SXM affects the ADD, SUB, and MOV instructions that shift a 16-bit value before using it in an operation on the 32-bit ACC. SXM also determines whether ACC is sign extended when it is shifted right by the SFR instruction. SXM does not affect instructions that shift the product register value; all right shifts of the product register value use sign extension. This bit can be individually set and cleared by the SETC SXM instruction and CLRC SXM instruction, respectively. At reset, SXM is cleared.

[0172] FIG. 15 shows the bit fields of status register ST1. All of these bit fields are modified in the decode 2 phase of the pipeline. Detailed descriptions of these bits are provided later herein.

[0173] In FIG. 15, R means a Read access and W=Write access; the value following dash (−) is value after reset; reserved bits are always 0s and are not affected by writes. In FIG. 15, the value of the VMAP bit after reset depends on the level of the VMAP input signal at reset. If the VMAP signal is low, the VMAP bit is 0 after reset; if the VMAP signal is high, the VMAP bit is 1 after reset.

[0174] In FIG. 15, bits 15-13 are the auxiliary register pointer (ARP). This 3-bit field points to the current auxiliary register. This could be one of the 16-bit auxiliary registers (AR0 to AR5) or one of the 22-bit extended auxiliary registers (XAR6 or XAR7). The mapping of ARP values to auxiliary registers is as follows:

TABLE 15
Selected Auxiliary Register
ARP Selected Auxiliary Register
000 AR0 (selected at reset)
001 AR1
010 AR2
011 AR3
100 AR4
101 AR5
110 XAR6
111 XAR7

[0175] In FIG. 15, bits 12-8 are reserved and writes to these bits have no effect.

[0176] In FIG. 15, bit 7 is the IDLE status bit (IDLESTAT). This read-only bit is set when the IDLE instruction is executed. It is cleared by any one of the following events: an interrupt is serviced; an interrupt is not serviced but takes the core out of the IDLE state; a valid instruction enters the instruction register (the register that holds the instruction currently being decoded); and a device reset occurs.

[0177] When the CPU services an interrupt, the current value of IDLESTAT is saved on the stack (when ST1 is saved on the stack), and then IDLESTAT is cleared. Upon return from the interrupt, IDLESTAT is not restored from the stack.

[0178] In FIG. 15, bit 6 is the emulation access enable bit (EALLOW). This bit, when set, enables access to emulation registers. EALLOW is set by the EALLOW instruction and cleared by the EDIS instruction. EALLOW is written to by using the POP ST1 instruction or the POP DP:ST1 instruction. When the CPU services an interrupt, the current value of EALLOW is saved on the stack (when ST1 is saved on the stack), and then EALLOW is cleared. Therefore, at the start of an interrupt service routine (ISR), access to emulation registers is disabled. If the ISR must access emulation registers, it must include an EALLOW instruction. At the end of the ISR, EALLOW can be restored by the IRET instruction.

[0179] In FIG. 15, bit 5 is the Loop instruction status bit (LOOP). LOOP is set when a loop instruction (LOOPNZ or LOOPZ) reaches the decode 2 phase of the pipeline. The loop instruction does not end until a specified condition is met. When the condition is met, LOOP is cleared. LOOP is a read-only bit; it is not affected by any instruction except a loop instruction. When the CPU services an interrupt, the current value of LOOP is saved on the stack (when ST1 is saved on the stack), and then LOOP is cleared. Upon return from the interrupt, LOOP is not restored from the stack.

[0180] In FIG. 15, bit 4 is the Stack pointer alignment bit (SPA). SPA indicates whether the CPU has previously aligned the stack pointer to an even address:

TABLE 16
SPA Condition
SPA Condition
0   The stack pointer has not been aligned to an even address.
1   The stack pointer has been aligned to an even address.

[0181] When the align stack pointer (ASP) instruction is executed, if the stack pointer (SP) points to an odd address, SP is incremented by 1 so that it points to an even address, and SPA is set. If SP already points to an even address, SP is not changed, but SPA is cleared. When the unalign stack pointer (NASP) instruction is executed, if SPA is 1, SP is decremented by 1, and SPA is cleared. At reset, SPA is cleared.

[0182] In FIG. 15, bit 3 is the Vector map bit (VMAP). VMAP determines whether the interrupt vectors (including the reset vector) are mapped to the lowest or highest addresses in program memory:

TABLE 17
VMAP Condition
VMAP Condition
0    Interrupt vectors are mapped to the bottom of program memory,
     addresses 00000016 - 00003F16.
1    Interrupt vectors are mapped to the top of program memory,
     addresses 3FFFC016 - 3FFFFF16.

[0183] The core input signal VMAP, is sampled on reset. If the signal is high, the VMAP bit is set. If the signal is low, the VMAP bit is cleared. The output core signal VMAPS reflects the status of the VMAP bit. This bit can be individually set and cleared by the SETC VMAP instruction and CLRC VMAP instruction, respectively.

[0184] In FIG. 15, bit 2 is the PAGE0 addressing modes configuration bit (PAGE0). PAGE0 selects between two mutually-exclusive addressing modes: PAGE0 direct addressing mode and PAGE0 stack addressing mode. Selection of the modes is as follows:

TABLE 18
Addressing Mode Enabled
PAGE0 Addressing Mode Enabled
0     PAGE0 stack addressing mode
1     PAGE0 direct addressing mode

[0185] This bit can be individually set and cleared by the SETC PAGE0 instruction and CLRC PAGE0 instruction, respectively. At reset, the PAGE0 bit is cleared (PAGE0 stack addressing mode is selected).

[0186] In FIG. 15, bit 1 is the Debug enable mask bit (DBGM). DBGM is primarily used in emulation to block debug events in time-critical portions of program code. DBGM enables or disables debug events as follows:

TABLE 19
DBGM Condition
DBGM Condition
0    Debug events are enabled.
1    Debug events are disabled.

[0187] When the CPU services an interrupt, the current value of DBGM is saved on the stack (when ST1 is saved on the stack), and then DBGM is set. Upon return from the interrupt, DBGM is restored from the stack. This bit can be individually set and cleared by the SETC DBGM instruction and CLRC DBGM instruction, respectively. DBGM is also set automatically during interrupt operations. At reset, DBGM is set. Executing the ABORTI (abort interrupt) instruction also sets DBGM.

[0188] In FIG. 15, bit 0 is the Interrupt global mask bit (NTM). This bit globally enables or disables all maskable interrupts (those that can be blocked by software):

TABLE 20
INTM Condition
INTM Condition
0    Maskable interrupts are globally enabled. In order to be
     acknowledged by the CPU, a maskable interrupt must also be
     locally enabled by the interrupt enable register (IER).
1    Maskable interrupts are globally disabled. Even if a maskable
     interrupt is locally enabled by the interrupt enable register
     (IER), it is not acknowledged by the CPU.

[0189] INTM has no effect on the nonmaskable interrupts, including a hardware re-set or the hardware interrupt NMI. The state of INTM is saved/restored whenever status register ST1 is pushed to/popped from the stack. When the CPU services an interrupt, the current value of INTM is saved on the stack (when ST1 is saved on the stack), and then INTM is set. Upon return from the interrupt, INTM is restored from the stack. This bit can be individually set and cleared by the SETC INTM instruction and CLRC INTM instruction, respectively. At reset, INTM is set. The value in INTM does not cause modification to the interrupt flag register (IFR), the interrupt enable register (IER), or the debug interrupt enable register (DBGIER). When the CPU is halted in real-time emulation mode, an interrupt enabled by the IER and the DBGIER will be serviced even if INTM is set to disable maskable interrupts.

[0190] The program control logic and program-address generation logic work together to provide proper program flow. Normally, the flow of a program is sequential: the CPU executes instructions at consecutive program-memory addresses. At times, a discontinuity is required; that is, a program must branch to a non-sequential address and then execute instructions sequentially at that new location. For this purpose, the DSP supports interrupts, branches, calls, returns, and repeats. Proper program flow also requires smooth flow at the instruction level. To meet this need, the DSP supports a fully protected pipeline and an instruction-fetch mechanism that attempts to keep the pipeline full.

[0191] Interrupts are hardware- or software-driven events that cause the CPU to suspend its current program sequence and execute a subroutine called an interrupt service routine. Interrupts are described in detail later herein.

[0192] Branches, calls, and returns break the sequential flow of instructions by transferring control to another location in program memory. A branch only transfers control to the new location. A call also saves the return address (the address of the instruction following the call). Every called subroutine or interrupt service routine is concluded with a return instruction, which pops the return address off the stack and back into the program counter (PC). The branch instructions B, SB, and BANZ are conditional. They are executed only if a certain specified or predefined condition is met.

[0193] The repeat (RPT) instruction allows the execution of a single instruction (N +1) times, where N is specified as an operand of the RPT instruction. The instruction is executed once and then repeated N times. When RPT is executed, the repeat counter (RPTC) is loaded with N. RPTC is then decremented every time the repeated instruction is executed, until RPTC equals 0.

[0194] Each instruction passes through eight independent phases that form an instruction pipeline. At any given time, up to eight instructions may be active, each in a different phase of completion. Not all reads and writes happen in the same phases, but a pipeline-protection mechanism stalls instructions as needed to ensure that reads and writes to the same location happen in the order in which they are programmed. To maximize pipeline efficiency, an instruction-fetch mechanism attempts to keep the pipeline full. Its role is to fill an instruction-fetch queue, which holds instructions in preparation for decoding and execution. The instruction-fetch mechanism fetches 32-bits at a time from program memory; it fetches one 32-bit instruction or two 16-bit instructions.

[0195] The instruction-fetch mechanism uses three program-address counters: the program counter (PC), the instruction counter (IC), and the fetch counter (FC). When the pipeline is full the PC will always point to the instruction in its decode 2 pipeline phase. The IC points to the next instruction to be processed. When the PC points to a 1-word instruction, IC=(PC+1); when the PC points to a 2-word instruction, IC=(PC+2). The value in the FC is the address from which the next fetch is to be made.

[0196] The DSP has a 16-bit×16-bit hardware multiplier that can produce a 32-bit signed or unsigned 32-bit product. FIG. 16 shows the CPU components involved in multiplication. The multiplier accepts two 16-bit inputs:

[0197] a) One input is always from the 16-bit multiplicand register (T). Most DSP multiply instructions require loading T from a data-memory location or a register before executing the instruction. However, the MAC instruction, one syntax of the MPY instruction, and one syntax of the MPYA instruction also load T before the multiplication.

[0198] b) The other input is from one of the following:

[0199] i) A data-memory location or a register (depending on which is specified in the multiply instruction).

[0200] ii) An instruction opcode. Some DSP multiply instructions allow including a constant as an operand.

[0201] After the T register value has been multiplied by the second value, the 32-bit result is stored in one of two places, depending on the particular multiply instruction: the 32-bit product register (P) or the 32-bit accumulator (ACC).

[0202] The shifter holds 32 bits and accepts either a 16-bit or 32-bit input value. When the input value has 16 bits, the value is loaded into the 16 least significant bits (LSBs) of the shifter. Depending on the instruction that uses the shifter, the output of the shifter may be all of its 32 bits or just its 16 LSBs. When a value is shifted right by an amount N, the N LSBs of the value are lost and the bits to the left of the value are filled with all 0s or all 1s. If sign extension is specified, the bits to the left are filled with copies of the sign bit. If sign extension is not specified, the bits to the left are filled with 0s, or zero filled.

[0203] When a value is shifted left by an amount N, the bits to the right of the shifted value are zero filled. If the value has 16 bits and sign extension is specified, the bits to the left are filled with copies of the sign bit. If the value has 16 bits and sign extension is not specified, the bits to the left are zero filled. If the value has 32 bits, the N most significant bits (MSBs) of the value are lost, and sign extension is irrelevant.

[0204] FIGS. 18A-18C list the instructions that use the shifter, provides an illustration of the corresponding shifter operation and uses the graphical symbols depicted in FIG. 17.

[0205] The following describes the available interrupts and how they are handled by the CPU. It also explains how to control those interrupts that can be controlled through software. Finally, it describes how a hardware reset affects the CPU.

[0206] Interrupts are hardware- or software-driven signals that cause the DSP to suspend its current program sequence and execute a subroutine. Typically, interrupts are generated by hardware devices that need to give data to or take data from the DSP (for example, A/D and D/A converters and other processors). Interrupts can also signal that a particular event has taken place (for example, a timer has finished counting).

[0207] On the DSP, interrupts can be triggered by software (the INTR, OR IFR, or TRAP instruction) or by hardware (a pin, a peripheral, or on-chip logic). If hardware interrupts are triggered at the same time, the DSP services them according to a set priority ranking. Each of the DSP interrupts, whether hardware or software, can be placed in one of the following two categories:

[0208] a) Maskable interrupts. These are interrupts that can be blocked (masked) or enabled (unmasked) through software.

[0209] b) Nonmaskable interrupts. These interrupts cannot be blocked. The DSP will immediately approve this type of interrupt and branch to the corresponding subroutine. All software-initiated interrupts are in this category.

[0210] The DSP handles interrupts in four main phases:

[0211] 1) Receive the interrupt request. Suspension of the current program sequence must be requested by a software interrupt (from program code) or a hardware interrupt (from a pin or an on-chip device).

[0212] 2) Approve the interrupt The DSP must approve the interrupt request. If the interrupt is maskable, certain conditions must be met in order for the DSP to approve it. For nonmaskable hardware interrupts and for software interrupts, approval is immediate.

[0213] 3) Prepare for the interrupt service routine and save registers. The main tasks performed in this phase are:

[0214] Complete execution of the current instruction and flushes from the pipeline any instructions that have not reached the decoded 2 phase;

[0215] Automatically save most of the current program context by saving the following registers to the stack: ST0, T, AL, AH, PL, PH, ARO, AR1, DP, ST1, DBGSTAT, and IER; and

[0216] Fetch the interrupt vector and load it into the PC.

[0217] 4) Execute the interrupt service routine. The DSP branches to its corresponding subroutine called an interrupt service routine (ISR). The DSP branches to the address (vector) stored at a predetermined vector location and executes the ISR.

[0218] The DSP supports 32 interrupt vectors, including the reset vector. Each vector is a 22-bit address that is the start address for the corresponding interrupt service routine (ISR). Each vector is stored in 32 bits at two consecutive addresses. The lower address holds the 16 least significant bits (LSBs) of the vector. The higher address holds the 6 most significant bits (MSB) right-justified. When an interrupt is approved, the 22-bit vector is fetched, and the 10 MSBs at the higher address are ignored.

[0219] Table 21 lists the available interrupt vectors and their locations. The addresses are shown in hexadecimal form. The table also shows the priority of each of the hardware interrupts.

TABLE 21
Interrupt Vectors and Priorities
	Absolute Address
	(hexadecimal)	Hardware
Vector	VMPA = 0	VMAP = 1	Priority	Description
RESET	000000	3FFFC0	1 (highest)	Reset
INT1	000002	3FFFC2	5	Maskable interrupt 1
INT2	000004	3FFFC4	6	Maskable interrupt 2
INT3	000006	3FFFC6	7	Maskable interrupt 3
INT4	000008	3FFFC8	8	Maskable interru t 4
INT5	00000A	3FFFCA	9	Maskable interrupt 5
INT6	000000	3FFFCC	10	Maskable interrupt 6
INT7	00000E	3FFFCE	11	Maskable interrupt 7
INT8	000010	3FFFD0	12	Maskable interrupt 8
INT9	000012	3FFFD2	13	Maskable interrupt 9
INT10	000014	3FFFD4	14	Maskable interrupt 10
INT11	000016	3FFFD6	15	Maskable interrupt 11
INT12	000018	3FFFD8	16	Maskable interrupt 12
INT13	00001A	3FFFDA	17	Maskable interrupt 13
INT14	000010	3FFFDC	18	Maskable interrupt 14
DLOGTNT†	00001E	3FFFDE	19 (lowest)	Maskable data log interrupt
RTOSINT‡	000020	3FFFE0	4	Maskable real-time operating
				system interrupt
Reserved	000022	3FFFE2	—	Reserved
NMI	000024	3FFFE4	3	Nonmaskable interrupt
ILLEGAL	000026	3FFFE6	—	illegal-instruction trap
USER1	000028	3FFFE8	—	User-defined software interrupt
USER2	00002A	3FFFEA	—	User-defined software interrupt
USER3	000020	3FFFEC	—	User-defined software interrupt
USER4	00002E	3FFFEE	—	User-defined software interrupt
USER5	000030	3FFFF0	—	User-defined software interrupt
USER6	000032	3FFFF2	—	User-defined software interrupt
USER7	000034	3FFFF4	—	User-defined software interrupt
USER8	000036	3FFFF6	—	User-defined software interrupt
USER9	000038	3FFFF8	—	User-defined software interrupt
USER10	00003A	3FFFFA	—	User-defined software interrupt
USER11	00003C	3FFFFC	—	User-defined software interrupt
USER12	00003E	3FFFFE	—	User-defined software interrupt
Interrupts DLOGINT and RTOSINT are generated by the emulation hardware internal to the core.

[0220] The vector table can be mapped to the top or bottom of program space, depending on the value of the vector map bit (VMAP) in status register ST1. If the VMAP bit is 0, the vectors are mapped beginning at address 00000016. If the VMAP bit is 1, the vectors are mapped beginning at address 3FFFCO16. Table 21 lists both address offsets and absolute addresses. The VMAP bit can be set by the SETC VMAP instruction and cleared by the CLRC VMAP instruction. The input core signal VMAP, sampled only at reset, determines the reset value of the VMAP bit. The state of the VMAP bit is also reflected at the output core signal VMAPS.

[0221] {overscore (INT1)}-{overscore (IN14)} are 14 general-purpose interrupts. DLOGINT (the data log interrupt) and RTOSINT (the real-time operating system interrupt) are available for emulation purposes. These interrupts are supported by three dedicated registers: the interrupt flag register (IFR), the interrupt enable register (IER), and the debug interrupt enable register (DBGIER).

[0222] The 16-bit interrupt flag register (IFR) contains flag bits that indicate which of the corresponding interrupts are pending (waiting approval from the CPU). The external input lines {overscore (INT1)}-{overscore (IN4)} are sampled at every CPU clock cycle. If an active interrupt is recognized, the corresponding bit in the IFR is set and latched. For DLOGINT or RTOSINT, a signal sent by the core's on-chip analysis logic causes the corresponding flag bits to be set and latched. One or more of the IFR bits may be set at the same time by using the OR IFR instruction.

[0223] The interrupt enable register (IER) and the debug interrupt enable register (DBGIER) each contain bits for individually enabling or disabling {overscore (INT1)}-{overscore (IN14)}, DLOGINT, and RTOSINT. To enable one of the interrupts in the IER, set the corresponding bit in the IER; to enable the same interrupt in the DBGIER, set the corresponding bit in the DBGIER. The DBGIER indicates which interrupts may be serviced when the core is in the real-time emulation mode. The IER and the DBGIER are discussed later herein.

[0224] {overscore (INT1)}-{overscore (IN14)}, DLOGINT, and RTOSINT also share bit 0 in status register ST1. This bit, the interrupt global mask bit (INTM), is used to globally enable or globally disable these interrupts. When INTM=0, these interrupts are globally enabled. When INTM=1, these interrupts are globally disabled. Set and clear INTM with the SETC and CLRC instructions, respectively.

[0225] After a flag has been latched in the IFR, the corresponding interrupt will not be serviced until it is appropriately enabled by two of the following: the IER, the DBGIER, and the INTM bit. As shown in Table 22, the requirements for enabling {overscore (INT1)}-{overscore (IN14)}, DLOGINT, and RTOSINT depend on the interrupt-handling process used. In the standard process, which occurs in most circumstances, the DBGIER is ignored. When the DSP is in real-time emulation mode and the CPU is halted, a different process is used. In this special case, the DBGIER is used and the INTM bit is ignored. (If the DSP is in real-time mode and the CPU is running, the standard interrupt-handling process applies.) Once an interrupt has been requested and properly enabled, the CPU prepares for and then executes the corresponding interrupt service routine.

TABLE 22
Requirements for Enabling One of the
Maskable Interrupts INT1-{overscore (IN14)}, DLOGINT, and RTOSINT
Interrupt-Handling Process Interrupt Enabled If . . .
Standard                   INTM = and bit in IER is 1
DSP in real-time mode and  Bit in IER is 1 and bit in
CPU halted                 DBGIER is 1

[0226] As an example of varying interrupt-enable requirements, it is desired that interrupt INTS be enabled. This corresponds to bit 4 in the IER and bit 4 in the DBGIER. Usually, INT5 is enabled if INTM=0 and IRE(4)=1. In real-time emulation mode with the CPU halted, INT5 is enabled if IRE(4)=1 and DBGIRE(4)=1.

[0227] FIG. 19 shows the Interrupt Flag Register (IFR). To identify pending interrupts, use the PUSH IFR instruction and then test the value on the stack. Use the OR IFR instruction to set IFR bits, and use the AND IFR instruction to clear pending interrupts. When the CPU approves a hardware interrupt request, the corresponding IFR bit is automatically cleared. An 25 INTR instruction also clears the corresponding IFR bit. All pending interrupts can be cleared by the AND IFR, #0 instruction or by a hardware reset.

[0228] When an interrupt is requested by the TRAP instruction, if the corresponding IFR bit is set, the CPU does not clear it automatically. If an application requires that the IFR bit be cleared, the bit must be cleared in the interrupt service routine.

[0229] In FIG. 19, R means a Read access and means a W=Write access; the value following dash (−) is value after reset.

[0230] Bits 15 and 14 of the IFR correspond to the interrupts RTOSINT and DLOGINT:

[0231] RTOSINT-Bit 15-Real-time operating system interrupt flag.

[0232] RTOSINT=0 RTOSINT is not pending.

[0233] RTOSINT=1 RTOSINT is pending.

[0234] DLOGINT-Bit 14-Data log interrupt flag.

[0235] DLOGINT=0 DLOGINT is not pending.

[0236] DLOGINT=1 DLOGINT is pending.

[0237] For bits INT1-INT14, the following general description applies:

[0238] INTx-Bit (x−1)-Interrupt x flag (x=1, 2, 3, . . . , or 14).

[0239] INTx=0 INTx is not pending.

[0240] INTx=1 INTx is pending.

[0241] FIG. 20 shows the Interrupt Enable Register (IER). To enable an interrupt, set its corresponding bit to 1. To disable an interrupt, clear its corresponding bit to 0. Two syntaxes of the MOV instruction allow reading from the IER and writing to the IER. In addition, the OR IER instruction enables setting IER bits, and the AND IER instruction enables clearing IER bits. When a hardware interrupt is serviced, or when an INTR instruction is executed, the corresponding IER bit is cleared. At reset, all the IER bits are cleared to 0, disabling all the corresponding interrupts. When an interrupt is requested by the TRAP instruction, if the corresponding IER bit is set, the CPU does not clear it automatically. If an application requires that the IER bit be cleared, the bit must be cleared in the interrupt service routine.

[0242] In FIG. 20, R means a Read access and W means a Write access; the value following dash (−) is value after reset. When using the AND IER and OR IER instructions, do not modify the state of bit 15 (RTOSINT) unless a real-time operating system is present.

[0243] Bits 15 and 14 of the IER enable or disable the interrupts RTOSINT and DLOGINT:

[0244] RTOSINT-Bit 15-Real-time operating system interrupt enable bit.

[0245] RTOSINT=0 RTOSINT is disabled.

[0246] RTOSINT=1 RTOSINT is enabled.

[0247] DLOGINT-Bit 14 -Data log interrupt enable bit.

[0248] DLOGINT=0 DLOGINT is disabled.

[0249] DLOGINT=1 DLOGINT is enabled.

[0250] For bits INT1-INT14, the following general description applies:

[0251] INTx-Bit (x−1) -Interrupt x enable bit (x=1, 2, 3, . . . , or 14).

[0252] INTx=0 INTx is disabled.

[0253] INTx=1 INTx is enabled.

[0254] FIG. 21 shows the Debug Interrupt Enable Register (DBGIER), which is only used when the CPU is halted in real-time emulation mode. An interrupt enabled in the DBGIER is defined as a time-critical interrupt. When the CPU is halted in real-time mode, the only interrupts that are serviced are time-critical interrupts that are also enabled in the IER. If the CPU is running in real-time emulation mode, the standard interrupt handling process is used, and the DBGIER is ignored.

[0255] As with the IER, DBGIER may be read to identify enabled or disabled interrupts and write to the DBGIER to enable or disable interrupts. To enable an interrupt, set its corresponding bit to 1. To disable an interrupt, set its corresponding bit to 0. Use the PUSH DBGIER instruction to read from the DBGIER and the POP DBGIER instruction to write to the DBGIER. At reset, all the DBGIER bits are set to 0.

[0256] In FIG. 21, R means a Read access and W means a Write access; the value following dash (−) is value after reset. Bits 15 and 14 of the DBGIER enable or disable the interrupts RTOSINT and DLOGINT:

[0257] RTOSINT-Bit 15-Real-time operating system interrupt enable bit.

[0258] RTOSINT=0 RTOSINT is disabled.

[0259] RTOSINT=1 RTOSINT is enabled.

[0260] DLOGINT-Bit 14 -Data log interrupt enable bit.

[0261] DLOGINT=0 DLOGINT is disabled.

[0262] DLOGINT=1 DLOGINT is enabled.

[0263] For bits INT1-INT14, the following general description applies:

[0264] INTx-Bit (x−1) -Interrupt x enable bit (x=1, 2, 3, . . . , or 14).

[0265] INTx=0 {overscore (INTx)} is disabled.

[0266] INTx=1 {overscore (INTx)} is enabled.

[0267] The flow chart in FIG. 22 shows the standard process for handling interrupts. When more than one interrupt is requested at the same time, the DSP services them one after another according to their set priority ranking. See the priorities in Table 3-1.

[0268] FIG. 22 is not a precise representation of how an interrupt is handled; it is a conceptual model of the important events useful for discussing interrupt handling.

[0269] What follows are explanations for the steps shown in FIG. 22:

[0270] 1) Interrupt request sent to CPU. One of the following events occurs:

[0271] a) One of the pins INT1-INT14 is driven low.

[0272] b) The core emulation hardware sends to the CPU a signal for DLOGINT or RTOSINT.

[0273] One of the interrupts INT1-INT14, DLOGINT, and RTOSINT is initiated by way of the OR IFR instruction.

[0274] 2) Set corresponding IFR flag bit. When the CPU detects a valid interrupt in step 1, it sets and latches the corresponding flag in the interrupt flag register (IFR). This flag stays latched even if the interrupt is not approved by the CPU in step 3.

[0275] 3) Interrupt enabled in IER? Interrupt enabled by INTM bit? The CPU will only approve the interrupt if the following conditions are true:

[0276] a) The corresponding bit in the IER is 1.

[0277] b) The INTM bit in ST1 is 0.

[0278] Once an interrupt has been enabled and then approved by the CPU, no other interrupts can be serviced until the CPU has begun executing the interrupt service routine (step 13).

[0279] 4) Clear corresponding IFR bit. Immediately after the interrupt is approved, its IFR bit is cleared. If the interrupt signal is kept low, the IFR register bit will be set again. However, the interrupt will not be immediately triggered again. The CPU blocks new hardware interrupts until the interrupt service routine begins (ISR). In addition, the IER bit is cleared (in step 10) before the ISR begins; therefore, an interrupt from the same source cannot disturb the ISR until the IER bit is set again by the ISR.

[0280] 5 ) Empty the pipeline. The CPU completes any instructions that have reached or passed their decode 2 phase in the instruction pipeline. Any instructions that have not reached this phase are flushed from the pipeline.

[0281] 6) Increment and temporarily store PC. The PC is incremented by 1 or 2, depending on the size of the current instruction. The result is the return address, which is temporarily saved in an internal hold register. During the automatic context save (step 9), the return address is pushed onto the stack.

[0282] 7) Fetch interrupt vector. The PC is filled with the address of the appropriate interrupt vector, and the vector is fetched from that location.

[0283] 8) Increment stack pointer by 1. The stack pointer (SP) is incremented by 1 in preparation for the automatic context save (step 9). During the automatic context save, the CPU performs 32-bit accesses, and the core expects 32-bit accesses to be aligned to even addresses by the memory wrapper. Incrementing SP by 1 ensures that the first 32-bit access will not overwrite the previous stack value.

[0284] 9) Perform automatic context save. A number of register values are saved automatically to the stack. These registers are saved in pairs; each pair is saved in a single 32-bit operation. At the end of each 32-bit save operation, the SP is incremented by 2. Table 23 shows the register pairs and the order in which they are saved. The core expects all 32-bit saves to be even-word aligned by the memory wrapper. As shown in the table, the SP is not affected by this alignment.

TABLE 23
Register Pairs Saved and
SP Positions for Context Saves
		Bit 0 of Storage Address
Save	Register	SP Starts at	SP Starts at
Operation†	Pairs	Odd Address	Even Address
		1 <- SP position	1
		before save
1st	ST0	0	0 <- SP position
			before save
	T	1	1
2nd	AL	0	0
	AH	1	1
3rd	PL‡	0	0
	PH	1	1
4th	ARO	0	0
	AR1	1	1
5th	ST1	0	0
	DP	1	1
6th	IER	0	0
	DBGSTAT§	1	1
7th	Return Address	0	0
	(low half)
	Return Address
	(high half)	1	1
		0 <- SP position	0
		after save
		1	1 <- SP position
			aftersave
†All registers are saved as pairs, as shown.
‡The P register is saved with 0 shift (CPU ignores current state of the product shift mode bits, PM, in status register 0).
§The DBGSTAT register contains special emulation information.

[0285] 10) Clear corresponding IER bit. After the IER register is saved on the stack in step 9, the CPU clears the IER bit that corresponds to the interrupt being handled. This prevents re-entry into the same interrupt. To nest occurrences of the interrupt, have the ISR set that IER bit again.

[0286] 11) Set INTM and DBGM. Clear LOOP, EALLOW, and IDLESTAT. All these bits are in status register ST1. By setting INTM to 1, the CPU prevents maskable interrupts from disturbing the ISR. To nest interrupts, have the ISR clear the INTM bit. By setting DBGM to 1, the CPU prevents debug events from disturbing time-critical code in the ISR. To not block debug events, have the ISR clear DBGM. The CPU clears LOOP, EALLOW, and IDLESTAT so that the ISR operates within a new context.

[0287] 12) Load PC with fetched vector. The PC is loaded with the interrupt vector that was fetched in step 7. The vector forces program control to the ISR.

[0288] 13) Execute interrupt service routine. Here is where the CPU executes the program code prepared to handle the interrupt. A typical ISR is shown in TABLE 24. Although a number of register values are saved automatically in step 10, others may need to be saved (such as auxiliary registers AR2-AR5, XAR6, and XAR7) at the beginning of the ISR. These register values must then be restored before the return from the ISR. If the ISR informs external hardware that the interrupt is being serviced, the IACK instruction may be used to send an interrupt acknowledge signal. The IACK instruction accepts a 16-bit constant as an operand and drives this 16-bit value on the 16 least significant lines of the data-write bus, DWDB(15:0).

[0289] 14) Program continues. If the interrupt is not approved by the CPU, the interrupt will be ignored, and the program will continue uninterrupted. If the interrupt is approved, its interrupt service routine is executed, and then the program continues where it left off (at the return address).

TABLE 24
Typical ISR
INTX: .long	INTXvector	; Interrupt vector
INTXvector:		; Automatic context save already
		occurred.
IACK	#INTX	; Send out interrupt acknowledge
		(optional).
AND	IER,#16BitMask	; Disable lower priority interrupts.
CLRC	INTM	; Enable interrupts.
PUSH	AR3:AR2	; Save registers not saved by during
PUSH	AR5:AR4	; automatic context save.
PUSH	XAR6	;
PUSH	XAR7	;
.	;
.	;
.	;
; Interrupt-specific user code . . .
POP	XAR7	;
POP	XAR6	;
POP	AR5:AR4	;
POP	AR3:AR2	;
IRET		; Return and automatically restore
	register pairs

[0290] Nonmaskable Interrupts

[0291] Nonmaskable interrupts cannot be blocked by any of the enable bits (the INTM bit, the DBGM bit, and enable bits in the IFR, IER, or DBGIER). The DSP will immediately approve this type of interrupt and branch to the corresponding interrupt service routine. There is one exception to this rule: When the CPU is halted in stop mode (an emulation mode), no interrupts are serviced.

[0292] The DSP nonmaskable interrupts include: software-initiated interrupts (the INTR and TRAP instructions), hardware interrupt NMI; illegal-instruction trap, and hardware reset interrupt. The instructions and NMI are described in this section.

[0293] The INTR instruction may be used to initiate one of the following interrupts by name: INT1-INT14, DLOGINT, RTOSINT and NMI. For example, the interrupt service routine for INT1 can be executed by using the following instruction:

INTR INT1

[0294] Once an interrupt is initiated by the INTR instruction, how it is handled depends on which interrupt it is:

[0295] a) {overscore (INT1)}-{overscore (INT14)}, DLOGINT, and RTOSINT. These maskable interrupts have corresponding flag bits in the IFR. When a request for one of these interrupts is received at an external pin, the corresponding IFR bit is set and the interrupt must be enabled. In contrast, when one of these interrupts is initiated by the INTR instruction, the IFR flag is not set, and the interrupt is serviced regardless of the value of any enable bits. However, in other respects, the INTR instruction and the hardware request are the same. For example, both clear the IFR bit that corresponds to the requested interrupt.

[0296] 2) NMI. Because this interrupt is nonmaskable, a hardware request at a pin and a software request with the INTR instruction lead to the same events. These events are identical to those that take place during a TRAP instruction.

[0297] TRAP Instruction

[0298] The TRAP instruction may be used to initiate any interrupt, including one of the user defined software interrupts (see Table 21). The TRAP instruction refers to one of the 32 interrupts by a number from 0 to 31. For example, the interrupt service routine for INT1 may be executed by using the following instruction:

TRAP#1

[0299] Regardless of whether the interrupt has bits in the IFR and IER, neither the IFR nor the IER is affected by this instruction. FIG. 23 shows a functional flow chart for an interrupt initiated by the TRAP instruction. The TRAP #0 instruction does not initiate a full reset. It only forces execution of the interrupt service routine that corresponds to the RESET interrupt vector.

[0300] The steps shown in FIG. 23 are explained below:

[0301] 1) TRAP instruction fetched. The CPU fetches the TRAP instruction from program memory. The desired interrupt vector has been specified as an operand and is now encoded in the instruction word. At this stage, no other interrupts can be serviced until the CPU begins executing the interrupt service routine (step 9).

[0302] 2) Empty the pipeline. The CPU will complete any instructions that have reached or passed the decode 2 phase of the pipeline. Any instructions that have not reached this phase will be flushed from the pipeline.

[0303] 3) Increment and temporarily store PC. The PC will be incremented by 1. This value is the return address, which is temporarily saved in an internal hold register. During the automatic context save (step 6), the return address will be pushed onto the stack.

[0304] 4) Fetch interrupt vector. The PC is set to the appropriate vector location (based on VMAP bit and the interrupt source), and the vector located at the PC address is loaded into the PC.

[0305] 5) Increment stack pointer by 1. The stack pointer (SP) is incremented by 1 in preparation for the automatic context save (step 6). During the automatic context save, the CPU performs 32-bit accesses, which are aligned to even addresses. Incrementing SP by 1 ensures that the first 32-bit access will not overwrite the previous stack value.

[0306] b 6) Perform automatic context save. A number of register values are saved automatically to the stack. These registers are saved in pairs; each pair is saved in a single 32-bit operation. At the end of each 32-bit operation, the SP is incremented by 2. Table 25 shows the register pairs and the order in which they are saved. All 32-bit saves are even-word aligned. As shown in the table, the SP is not affected by this alignment.

TABLE 25
Register Pairs Saved and
SP Positions for Context Saves
		Bit 0 of Storage Address
Save	Register	SP Starts at	SP Starts at
Operation†	Pairs	Odd Address	Even Address
		1 <- SP position	1
		before save
1st	ST0	0	0 <- SP position
			before save
	T	1	1
2nd	AL	0	0
	AH	1	1
3rd	PL‡	0	0
	PH	1	1
4th	AR0	0	0
	AR1	1	1
5th	ST1	0	0
	DP	1	1
	IER	0	0
	DBGSTAT§	1	1
7th	Return Address	0	0
	(low half)
	Return Address
	(high half)	1	1
		0 <- SP position	0
		after save
		1	1 <- SP position
			after save
†All registers are saved as pairs, as shown.
‡The P register is saved with 0 shift (CPU ignores current state of the product shift mode bits, PM, in status register 0).
§The DBGSTAT register contains special emulation information.

[0307] 7) Set INTM and DBGM. Clear LOOP, EALLOW, and IDLESTAT. All these bits are in status register ST1. By setting INTM to 1, the CPU prevents maskable interrupts from disturbing the ISR. If you which to nest interrupts, have the ISR clear the INTM bit. By setting DBGM to 1, the CPU pre-vents debug events from disturbing time-critical code in the ISR. If you do not want debug events blocked, have the ISR clear DBGM. The CPU clears LOOP, EALLOW, and IDLESTAT so that the ISR operates within a new context.

[0308] 8) Load PC with fetched vector. The PC is loaded with the interrupt vector that was fetched in step 4. The vector forces program control to the ISR.

[0309] 9) Execute interrupt service routine. The CPU executes the program code prepared to handle the interrupt. It is possible to have the interrupt service routine (ISR) save register values in addition to those saved in step 6. A typical ISR is shown in TABLE 24. To have the ISR inform external hardware that the interrupt is being serviced, use the IACK instruction to send an interrupt acknowledge signal. The IACK instruction accepts a 16-bit constant as an operand and drives this 16-bit value on the 16 least significant lines of the data-write bus, DWDB(15:0).

[0310] 10) Program continues. After the interrupt service routine is completed, the program continues where it left off (at the return address).

[0311] An interrupt can be requested by way of the NMI input pin, which must be driven low. The pin can generate NMI even when the DSP is in real-time emulation mode and the CPU is halted. Once a valid request is detected on the NMI pin, the CPU handles the interrupt in the same manner as shown for the TRAP instruction.

[0312] The illegal-instruction trap is primarily a debug tool to help catch stray or run-away code conditions. Any of the following three events can cause an illegal instruction trap:

[0313] a) An invalid instruction is decoded (this includes invalid addressing modes).

[0314] The opcode value 000016 is decoded. This opcode corresponds to an ITRAP0 instruction.

[0315] c) The opcode value FFFF16 is decoded. This opcode corresponds to an ITRAP1 instruction.

[0316] Once initiated, an illegal-instruction trap will operate the same as a TRAP #19 instruction. The handling of an interrupt initiated by the TRAP instruction was described earlier herein. As part of its operation, the illegal-instruction trap saves the return address on the stack. Thus, to detect the offending address, examine this saved value.

[0317] The reset input signal (RS) will place the core into a known state. As part of a hardware reset, all current operations are aborted, the pipeline is flushed, and the CPU registers are reset as shown in Table 26. Then the RESET interrupt vector is fetched and the corresponding interrupt service routine is executed.

TABLE 26
Registers After Reset
Register	Bit(s)	Value After Reset	Comments
ACC	all	0000_000016
AR0	all	000016
AR1	all	000016
AR2	all	000016
AR3	all	000016
AR4	all	000016
AR5	all	000016
XAR6	all	00 000016
XAR7	all	00 000016
DP	all	000016	DP points to data page 0.
IFR	16 bits	000016	There are no pending interrupts.
			All interrupts pending at the
			time of reset have been cleared.
IER	16 bits	000016	Interrupts {overscore (INT1)}-{overscore (INT14)},
			DLOGINT, and RTOSINT are
			disabled in the IER.
DBGIER	all	000016	Interrupts {overscore (INT1)}-{overscore (INT14)},
			DLOGINT, and RTOSINT are
			disabled in the DBGIER.
P	all	0000 000016
PC	all	If VMAP bit = 0	PC is loaded with the reset
		PC = [000061]	interrupt vector at
		If VMAP bit = 1	program-space address
		PC = [3F FFC016]	00 000016 or 3F FFC016.
SP	all	000016	SP points to address 000016.
ST0	0: SXM	0	Sign-extension is suppressed.
	1: OVM	0	Overflow mode is off.
	2:TC	0
	3:C	0
	4:Z	0
	5:N	0
	6:V	0
	7-9: PM	0002	The product shift mode is
			set to left-shift-by-1.
	10-15: OVO	0000002
ST1‡	0:INTM 1		{overscore (INT1)}-{overscore (INT14)}, DLOGINT, and
			RTOSINT are globally disabled.
			They cannot be serviced unless
			the DSP is in real-time mode
			with the CPU halted.
	1: DBGM	1	Emulation accesses and
			events are disabled.
	2: PAGE0	0	PAGE0 stack addressing mode
			is enabled. PAGE0 direct
			addressing mode is disabled.
	3: VMAP	If the VMAY signal was	If the VMAP bit is 0,
		low at reset, the VMAP	the interrupt vectors are
		bit is 0. If the VMAP	mapped to program-memory
		signal was high at reset,	addresses 00 000016 - 00 003F16.
		the VMAP bit is 1.	If the VMAP bit is 1, the
			interrupt vectors are mapped
			to program-memory addresses
			3F FFC016 - 3F FFFF16.
	4:SPA	0
	5: LOOP	0
	6: EALLOW	0	Access to emulation
			registers is disabled.
	7:IDLESTAT	0
	8-12: Reserved	Undefined
	13-15: ARP	0002	ARP points to AR0.
T	all	000016

[0318] This portion describes the operation of the DSP instruction pipeline. The pipeline is protected; that is, pipeline hardware prevents reads and writes at the same location from happening out of order. However, the efficiency of programs may be increased by taking into account the operation of the pipeline. In addition, several operations are not protected by the pipeline.

[0319] When executing a program, the DSP CPU performs these basic operations: fetches instructions from program memory, decodes instructions, reads data values from memory or from CPU registers, executes instructions, and writes results to memory or to CPU registers.

[0320] For efficiency, the DSP performs these operations in eight independent phases. Reads from memory are designed to be pipelined in two stages, which correspond to the two pipeline phases used by the CPU for each memory-read operation. At any time, there can be up to eight instructions being carried out, each in a different phase of completion. What follows are descriptions of the eight phases in the order in which they occur:

[0321] Fetch 1—(F1)—In the fetch 1 (F1) phase of an instruction, the CPU drives a program- memory address on the 22-bit program address bus, PAB(21:0).

[0322] Fetch 2—(F2)—In the fetch 2 (F2) phase, the CPU reads from program memory by way of the program-read data bus, PRDB (31:0), and loads them into an instruction-fetch queue.

[0323] Decode 1—(D1)—The DSP supports both 32-bit and 16-bit instructions and an instruction can be aligned to an even or odd address. The decode 1 (D1) hardware identifies instruction boundaries in the instruction-fetch queue and determines the size of the next instruction to be executed. It also determines whether the instruction is a legal instruction.

[0324] Decode 2—(D2)—The decode 2 (D2) hardware requests an instruction from the instruction-fetch queue. The requested instruction is loaded into the instruction register, where decoding is completed. Once an instruction reaches the D2 phase, it runs to completion. In this pipeline phase, the following tasks are performed: if data is to be read from memory, the CPU generates the source address or addresses; if data is to be written to memory, the CPU generates the destination address; the address register arithmetic unit (ARAU) performs any required modifications to the stack pointer (SP) or to an auxiliary register and/or the auxiliary register pointer (ARP); and if a program-flow discontinuity (such as a branch or an illegal-instruction trap) is required, it is taken.

[0325] Read 1—(R1)—If data is to be read from memory, the read 1 (R1) hardware drives the address(es) on the appropriate address bus(es).

[0326] Read 2—(R2)—If data was addressed in the R1 phase, the read 2 (R2) hardware fetches that data by way of the appropriate data bus(es).

[0327] Execute—(E)—In the execute (E) phase, the CPU performs all multiplier, shifter, and ALU operations. This includes all the prime arithmetic and logic operations involving the accumulator and product register. Operations that involve reading a value, modifying it, and writing it back to the original location perform the modification (typically an arithmetic or a logic operation) in the E phase. Any CPU register values used by the multiplier, shifter, and ALU are read from the registers at the beginning of the E phase. A result that is to be written to a CPU register is written to the register at the end of the E phase.

[0328] Write—(W)—If a transferred value or result is to be written to memory, the write occurs in the write (W) phase. The CPU drives the destination address, the appropriate write strobes, and the data to be written. The actual storing, which takes at least one more clock cycle, is handled by memory wrappers or peripheral interface logic and is not visible as a part of the CPU pipeline.

[0329] Although every instruction passes through the eight phases, not every phase is active for a given instruction. Some instructions complete their operations in the decode 2 phase, others in the execute phase, and still others in the write phase. For example, instructions that do not read from memory perform no operations in the read phases, and instructions that do not write to memory perform no operation in the write phase.

[0330] Because different instructions perform modifications to memory and registers during different phases of their completion, an unprotected pipeline could lead to reads and writes at the same location happening out of the intended order. The CPU automatically adds inactive cycles to ensure that these reads and writes happen as intended.

[0331] The fetch 1 through decode 1 (F1-D1) hardware acts independently of the decode 2 through write (D2-W) hardware. This allows the CPU to continue fetching instructions when the D2-W phases are halted. It also allows fetched instructions to continue through their D2-W phases when fetching of new instructions is delayed. Events that cause portions of the pipeline to halt are described later herein. Instructions in their fetch 1, fetch 2, and decode 1 phases are discarded if an interrupt or other program-flow discontinuity occurs. An instruction that reaches its decode 2 phase always runs to completion before any program-flow discontinuity is taken.

[0332] The instruction-fetch mechanism is the hardware for the F1 and F2 pipeline phases. During the F1 phase, the mechanism drives an address on the program address bus (PAB). During the F2 phase, it reads from the program-read data bus (PRDB). While an instruction is read from program memory in the F2 phase, the address for the next fetch is placed on the program address bus (during the next F1 phase). The instruction-fetch mechanism contains an instruction-fetch queue of four 32-bit registers. During the F2 phase, the fetched instruction is added to the queue, which behaves like a first-in, first-out (FIFO) buffer. The first instruction in the queue is the first to be executed. The instruction-fetch mechanism performs 32-bit fetches until the queue is full. When a program-flow discontinuity (such as a branch or an interrupt) occurs, the queue is emptied. When the instruction at the bottom of the queue reaches its D2 phase, that instruction is passed to the instruction register for further decoding.

[0333] Three program-address counters are involved in the fetching and execution of instructions:

[0334] Fetch counter (FC). The fetch counter contains the address that is driven on the program address bus (PAB) in the F1 pipeline phase. The CPU continually increments the FC until the queue is full or the queue is emptied by a program-flow discontinuity. Generally, the FC outputs an even address and is incremented by 2. The only exception to this is when the code after a discontinuity begins at an odd address; in this case, the FC will reflect the odd address.

[0335] Instruction counter (IC). After the D1 hardware determines the instruction size (16-bit or 32-bit), it fills the instruction counter (IC) with the address of the next instruction to undergo D2 decoding. On an interrupt or call operation, the IC value represents the return address, which is saved to the stack or to auxiliary register XAR7.

[0336] Program counter (PC). When a new address is loaded into the IC, the previous IC value is loaded into the PC. The program counter (PC) always contains the address of the instruction that has reached its D2 phase.

[0337] FIGS. 24A and 24B show the relationship between the pipeline and the address counters. Instruction 1 has reached its D2 phase (it has been passed to the instruction register). The PC points to the address from which instruction 1 was taken (00 005016). Instruction 2 has reached its D1 phase and will be executed next (assuming no program-flow discontinuity flushes the instruction-fetch queue). The IC points to the address from which instruction 2 was taken (00 005116). Instruction 3 is in its F2 phase. It has been transferred to the instruction-fetch queue but has not been decoded. Instructions 4 and 5 are in their F1 phase. The FC address (00 0054 16) is being driven on the PAB. During the next 32-bit fetch, Instructions 4 and 5 will be transferred from addresses 00 005416 and 00 005516 to the queue.

[0338] The remainder of this section refers almost exclusively to the PC. The FC and the IC are visible in only limited ways. For example, when a call is executed or an interrupt is initiated, the IC value is saved to the stack or to auxiliary register XAR7.

[0339] Referring to Table 28, which lists five instructions, I1-I5, and Table 29, which shows a diagram of the pipeline activity for those instructions. In Table 29, the F1 column shows addresses and the F2 column shows the instruction opcodes read at those addresses. During an instruction fetch, 32 bits are read from the specified address and the following address. The D1 column shows instructions being isolated in the instruction-fetch queue, and the D2 column indicates address generation and modification of address registers. The Instruction column shows the instructions that have reached the D2 phase. The R1 column shows addresses, and the R2 column shows the data values being read from those addresses. In the E column, the diagram shows results being written to the low half of the accumulator (AL). In the W column, address and a data values are driven simultaneously on the appropriate memory buses. For example, in the last active W phase of Table 29, the address 00 0205 16 is driven on the data-write address bus (DWAB), and the data value 1234 16 is driven on the data-write data bus (DWDB).

[0340] The highlighted blocks in Tables 28 and 29 indicate the path taken by the instruction ADD AL,*ARO++. That path can be summarized as follows in Table 27:

TABLE 27
Phase Activity Shown
Phase Activity Shown
F1    Drive address 00_004216 on the program
      address bus (PAB).
F2    Read the opcodes F347 and F348 from addresses 00 004216
      and 00 004316, respectively.
D1    Isolate F348 in the instruction-fetch queue.
D2    Use AR0 = 666616 to generate source address 00 666616 and
      then increment AR0 to 666716.
R1    Drive address 00 666616 on the data-read data bus (DRDB).
R2    Read the data value 1 from address 00666616.
E     Add 1 to content of AL (123016) and store result (123116)
      to AL.
W     No activity

[0341]

TABLE 28
Diagramming Pipeline Activity
Address	Opcode	Instruction	Initial Values
00 0040	F345	I1: MOV	;DP,#VarA	; DP =page that has VarA.	VarA address = 00 0203
00 0041	F346	I2: MOV	AL,@VarA	; Move content of VarA to AL.	VarA = 1230
00 0042	F347	I3: MOVB	AR0,#VarB	; ARO points to VarB.	VarB address = 00 0066
00 0043	F348	I4: ADD	AL,*AR0++	; Add content of VarB to	VarB = 0001
				; AL, and add 1 to ARO.	(VarB + 1) = 0003
00 0044	F349	I5: MOV	@VarC,AL	; Replace content of VarC	(VarB + 2) = 0005
				; with content of AL.	VarC address = 00 0204
00 0045	F34A	I6: ADD	AL,*AR0++	; Add content of (VarB +1)	VarD address = 00 0205
				; to AL, and add 1 to ARO.
00 0046	F34B	I7: MOV	@VarD,AL	; Replace content of VarD
				; with content of AL.
00 0047	F34C	I8: ADD	AL,*AR0;	; Add content of(VarB +2)
				; to AL.

[0342]

TABLE 29
Diagramming Pipeline Activity
F1	F2	D1	Instruction	D2	R1	R2	E	W
00 0040
	F346:F345
00 0042		F345
	F348:F347	F346	I1: MOV	DP = 8
			DP #VarA
00 0044		F347	I2: MOV	Generate	—
			AL, @VarA	VarA address
	F34A:F349	F348	I3: MOVB	AR0 = 66	00 0203	—
			AR0, #VarB
00 0046		F349	I4: ADD	AR0 = 67	—	1230	—
			AL, *AR0++
	F34C:F34B	F34A	I5: MOV	Generate	00 0066	—	AL = 1230	—
			@VarC, AL	VarC address
		F34B	I6: ADD	AR0 = 68	—	0001	—	—
			AL, *AR0++
		F34C	I7: MOV	Generate	00 0067	—	AL = 1231	—
			@VarD, AL	VarD address
			I8: ADD	AR0 = 68	—	0003	—	—
			AL, *AR0
					00 0068	—	AL = 1234	00 0204
								1231
						0005	—	—
							AL = 1239	00 0205
								1234
								—
Note:
The opcodes shown in the F2 and D1 columns are not the actual opcodes of the instructions shown.

[0343] The pipeline activity in Tables 28 and 29 can also be represented by the simplified diagram in Table 30. This type of diagram is useful to focus on the path of each instruction rather than on specific pipeline events. Notice that in cycle 8, the pipeline is full: there is an instruction in every pipeline phase. Notice also that the effective execution time for each of these instructions is 1 cycle. Some instructions finish their activity at the D2 phase, some at the E phase, and some at the W phase. However, using one phase as a reference, each instruction is in that phase for 1 cycle.

Example 30. Simplified Diagram of Pipeline Activity

[0344]

F1	F2	D1	D2	R1	R2	E	W	Cycle
I1								1
I2	I1							2
I3	I2	I1						3
I4	I3	I2	I1					4
I5	I4	I3	I2	I1				5
I6	I5	I4	I3	I2	I1			6
I7	I6	I5	I4	I3	I2	I1		7
I8	I7	I6	I5	I4	I3	I2	I1	8
	I8	I7	I6	I5	I4	I3	I2	9
		I8	I7	I6	I5	I4	I3	10
			I8	I7	I6	I5	I4	11
				I8	I7	I6	I5	12
					I8	I7	I6	13
						I8	I7	14
							I8	15

[0345] This section describes two causes for freezes in pipeline activity: wait states, and an instruction-not-available condition.

[0346] When the core requests a read from or write to a memory device or peripheral device, that device may take more time to finish the data transfer than the core allots by default. Each device must use one of the core's ready signals to insert wait states into the data transfer when it needs more time. The core has three independent sets of ready signals: one set for reads from and writes to program space, a second set for reads from data space, and a third set for writes to data space. Wait-state requests freeze a portion of the 20 pipeline if they are received during the F1, R1, or W phase of an instruction:

[0347] 1) Wait states in the F1 phase. The instruction-fetch mechanism halts until the wait states are completed. This halt effectively freezes activity for instructions in their F1, F2, and D1 phases. However, because the F1-D1 hardware and the D2-W hardware are decoupled, instructions that are in their D2-W phases continue to execute.

[0348] 2) Wait states in the R1 phase. All D2-W activities of the pipeline freeze. This is necessary because subsequent instructions can depend on the data-read taking place. Instruction fetching continues until the instruction-fetch queue is full or a wait-state request is received during an F1 phase.

[0349] 3) Wait states in the W phase. All D2-W activity in the pipeline freezes. This is necessary because subsequent instructions may depend on the write operation happening first. Instruction fetching continues until the instruction-fetch queue is full or a wait-state request is received during an F1 phase.

[0350] The D2 hardware requests an instruction from the instruction-fetch queue. If a new instruction has been fetched and has completed its D1 phase, the instruction is loaded into the instruction register for more decoding. However, if a new instruction is not waiting in the queue, an instruction-not-available condition exists. Activity in the F1 -D1 hardware continues. However, the activity in the D2-W hardware ceases until a new instruction is available.

[0351] Instructions are being executed in parallel in the pipeline, and different instructions perform modifications to memory and registers during different phases of completion. In an unprotected pipeline, this could lead to pipeline conflicts—reads and writes at the same location happening out of the intended order. However, the DSP pipeline has a mechanism that automatically protects against pipeline conflicts. There are two types of pipeline conflicts on the DSP: conflicts during reads and writes to the same data-space location, and register conflicts.

[0352] The pipeline prevents these conflicts by adding inactive cycles between instructions that would cause the conflicts. This section explains the circumstances under which these pipeline-protection cycles are added, so that the number of inactive cycles in a program may be reduced.

[0353] Consider two instructions, A and B. Instruction A writes a value to a memory location during its W phase. Instruction B must read that value from the same location during its R1 and R2 phases. Because the instructions are being executed in parallel, it is possible that the R1 phase of instruction B could occur before the W phase of instruction A. Without pipeline protection, instruction B could read too early and fetch the wrong value. The DSP pipeline prevents that read by holding instruction B in its D2 phase until instruction A is finished writing.

[0354] Table 31 shows a conflict between two instructions that are accessing the same data-memory location. The pipeline activity shown is for an unprotected pipeline. For convenience, the F1-D1 phases are not shown. I1 writes to VarA during cycle 5. Data memory completes the store in cycle 6. I2 should not read the data-memory location any sooner than cycle 7. However, 12 performs the read during cycle 4 (3 cycles too early). To prevent this kind of conflict, the pipe-line-protection mechanism would hold I2 in the D2 phase for 3 cycles. During these pipeline-protection cycles, no new operations occur.

TABLE 31
Conflict During Read From and
Write To Same Memory Location
D2	R1	R2	E	W	Cycle	Comments
I1					1
I2	I1				2
	I2	I1			3	I2 issues read request
		I2	I1		4	I2 reads in R2 phase
			I2	II	5	I1 writes in W phase
				I2	6	I2 should issue read request here
					7
					8
I1: MOV@VarA,AL ;Write AL to data-memory location
I2: MOVAH.@VarA ;Read same location, store value in AH

[0355] To reduce or eliminate these types of pipeline-protection cycles, insert other instructions between the instructions that conflict. For example, the code in Table 31 could be improved by inserting a CLRC instruction from elsewhere in the program:

I1:	MOV	@VarA,AL	; Write AL to data-memory location
	CLRC	SXM	; Clear sign-extension mode
I2:	MOV	AH,@VarA	; Read same location, store value in AH

[0356] Inserting the CLRC instruction between I1 and I2 would reduce the number of pipeline-protection cycles to two. Inserting two more instructions would remove the need for pipeline protection. (Of course, the inserted instructions must not cause conflicts of their own.) As a general rule, if a read operation occurs within three instructions from a write operation to the same memory location, the pipeline protection mechanism adds at least one inactive cycle.

[0357] All reads from and writes to CPU registers occur in either the D2 phase or the E phase of an instruction. A register conflict arises when an instruction attempts to read and/or modify the content of a register (in the D2 phase) before a previous instruction has written to that register (in the E phase).

[0358] Table 32 shows a register conflict involving auxiliary register AR0. The pipeline activity shown is for an unprotected pipeline, and for convenience, the F1-D1 phases are not shown. I1 writes to AR0 at the end of cycle 4. I2 should not attempt to read AR0 until cycle 5. However, I2 reads AR0 (to generate an address) during cycle 2. To prevent this conflict, the pipeline-protection mechanism would hold I2 in the D2 phase for three cycles. During these cycles, no new operations occur.

TABLE 32
Register Conflict
	D2	R1	R2	E	W	Cycle	Comments
	I1					1
	I2	I1				2	I2 reads in D2 phase
		I2	I1			3
			I2	I1		4	I1 writes mE phase
				I2	I1	5	I2 should read here
					I2	6
	I1: MOV AR0,@7 ;Load AR0 with the value addressed by ;the operand @7.
	12: MOV AH,*AR0 ;Load AH with the value pointed to by ;AR0.

[0359] To reduce or eliminate pipeline-protection cycles due to a register conflict, insert other instructions between the instructions that cause the conflict. For example, the code in Table 32 could be improved by moving two other instructions from elsewhere in the program:

I1:	MOV ARO,@7	; Load ARO with the value addressed by
		the operand @7.
	NOP	; Delay by 1 cycle.
	SETCSXM	; Enable sign-extension mode (SXM = 1)
I2:	MOV AH, *ARO	; Load All with the value pointed to by
		; ARO.

[0360] Inserting the NOP and SETC instructions would reduce the number of pipe-line-protection cycles to one. Inserting one more instruction would remove the need for pipeline protection. As a general rule, if a read operation occurs within three instructions from a write operation to the same register, the pipeline-protection mechanism adds at least one inactive cycle.

[0361] The pipeline-protection mechanism resolves register conflicts by holding the later instruction in its D2 phase for as many cycles as needed (one to three). Do not consider register conflicts except to achieve maximum pipeline efficiency. To reduce the number of pipeline-protection cycles, identify the pipeline phases in which registers are accessed and move conflicting instructions away from each other.

[0362] Generally, a register conflict involves one of the address registers: 16-bit auxiliary registers AR0-AR7, 22-bit auxiliary registers XAR6 and XAR7, 16-bit data page pointer (DP), or 16-bit stack pointer (SP).

[0363] For a given access to an address register, FIG. 25 indicates whether the access occurs in the D2 phase or the E phase. The examples in the following table illustrate how the Figure helps.

TABLE 33
Phases Instruction
Instruction	Comments
ADD	@AR2,#4	Add 4 to AR2. AR2 is referenced by register addressing mode. Therefore,
		the instruction writes to AR2 in the E phase.
MOVB	AR2,#6	Load AR2 with 6. AR2 is loaded with an immediate constant, and register
		addressing mode is not used. AR2 is modified in the D2 phase.
MOV	T,@AR2	Load T with content of AR2. AR2 is read in the D2 phase.
SUB	AH,*SP++	Subtract addressed value from AH. In the D2 phase, the instruction uses
		SP to generate an address and then increments SP.
POP	DP	Restore DP from the stack. Because the source is memory, DP is loaded in
		the E phase.

[0364] There are three operations that the pipeline-protection mechanism does not protect against. Generally, these conflicts are avoidable, and this section offers suggestions for avoiding them.

[0365] A discontinuity is a break in sequential program flow caused by a branch, a call, a return, or an interrupt. If the first instruction after the discontinuity is a 32-bit instruction and that instruction begins at an odd address, a pipeline delay occurs. To avoid the delay where possible, begin each block of code with one or two (preferably two) 16-bit instructions:

FunctionA:
16-bit instruction; First instruction
16-bit instruction; Second instruction
32-bit instruction; 32-bit instructions can start here
.
.
.

[0366] To use a 32-bit instruction as the first instruction of a function or subroutine, and prevent a pipeline delay, make sure the instruction begins at an even address.

[0367] The pipeline protects only register and data-space reads and writes. It does not protect the program-space reads done by the PREAD and MAC instructions or the program-space write done by the PWRITE instruction. Be careful with these instructions when using them to access a memory block that is shared by data space and program space.

[0368] As an example, suppose a memory location can be accessed at address 00 0D5016 in program space and address 00 0D5016 in data space. Consider the following lines of code:

;XAR7 = 000D50 in program space
;Datal = 000D50 in data space
ADD	@Data1,AH	; Store AH to data-memory location
		; Data1.
PREAD	@AR1,*XAR7	; Load AR1 from program-memory
		; location given by XAR7.

[0369] The operands @Datal and *XAR7 are referencing the same location, but the pipeline cannot interpret this fact. The PREAD instruction reads from the memory location (in the R2 phase) before the ADD writes to the memory location (in the W phase).

[0370] However, the PREAD is not necessary in this program. Because the location can be accessed by an instruction that reads from data space, use another instruction, such as a MOV instruction:

ADD @ Data1, AH ; Store AH to memory location Data1.
MOV AR1,*XAR7   ; Load AR1 from memory location
                ; given by XAR7.

[0371] If an access to one location affects another location, correct for a pipeline conflict. Consider the following example:

	MOV	@DataA,#4	; This write to DataA causes a
			; peripheral to clear bit 15 of DataB.
$10:	TBIT @DataB,#15	; Test bit 15 of DataB.
	SB	$10,NTC	; Loop until bit 15 is set.

[0372] This program causes a misread. The TBIT instruction reads bit 15 (in the R2 phase) before the MOV instruction writes to bit 15 (in the W phase). If the TBIT instruction reads a 1, the code prematurely ends the loop. Because DataA and DataB reference different data-memory locations, the pipeline does not identify this conflict.

[0373] However, correct this type of error by inserting two NOP (no operation) instructions between the conflicting instructions. For example, fix the previous code as follows:

	MOV	@DataA,#4	; This write to DataA causes a
			; peripheral to clear bit 15 of DataB.
	NOP		; Delay by 1 cycle.
	NOP		; Delay by 1 cycle.
$10:	TBIT@DataB,#15	; Test bit 15 of DataB.
	SB	$10,NTC	; Loop until bit 15 is set.

[0374] With these inserted cycles, the MOV instruction finishes writing before the TBIT instruction reads from DataB.

[0375] This portion describes the modes by which the DSP instruction set accepts constants and addresses memory locations. There are four main modes: immediate addressing, register addressing, direct addressing, or indirect addressing (this includes circular addressing).

[0376] To enter a constant as an operand, use immediate-constant addressing. To access data memory, use direct addressing, indirect addressing, or immediate-pointer addressing. Point to a program-memory address using XAR7 indirect addressing or immediate-pointer addressing. When direct or indirect addressing mode is used, the address is generated by the address register arithmetic unit (ARAU).

[0377] The DSP supports two immediate addressing modes: immediate-constant addressing, or immediate-pointer addressing.

[0378] As shown in the following examples, instructions that support immediate-constant addressing take a constant as an operand. The constant must be preceded by the symbol #.

	ADD	ACC,	#8	; Add 8 to accumulator.
	SUBB	SP,	#10	; Subtract 10 from stack pointer.
	MOV	DP,	#200	; Load data page pointer with 200.

[0379] The syntax determines the range of numbers that can be supplied in the operand. For example, this is the syntax for the MOV instruction shown above:

MOV DP, # 10 bit

[0380] The variable 10 bit represents a 10-bit unsigned constant (a value in the range 0 to 1023). The size of an immediate constant determines how it is encoded in the instruction. If the constant is fewer than 16 bits wide, the instruction requires one 16-bit word, and the constant is embedded in that word. If the constant is 16 bits wide, the instruction requires two words, and the constant is sent to the CPU as the second word.

[0381] As shown in FIG. 26, immediate-pointer addressing mode allows reference to a 22-bit data-memory or program-memory address by supplying the 16 least significant bits of the address as a constant. The six most significant bits of the address are filled with zeros. Because these six bits are always zeros, the address reach of this mode is limited to the address range 00 000016-00 FFFF16.

[0382] This mode is only used by the following syntaxes of the MOV and MAC instructions:

[0383] 1) MOV loc, *(0: 16 bit)—The22-bit immediate address is the value 0: 16 bit. The value supplied for 16 bits is concatenated with six leading zeros to form the 22-bit address. This form of the MOV instruction loads the value at data-memory address (0: 16 bit) to the data-memory or register referenced by loc. The instruction takes two 16-bit words, where the second word is the 16-bit constant.

[0384] 2) MOV *(0: 16 bit), loc-Again, the immediate address is the value 0: 16 bit. This form of the MOV instruction loads the data-memory or register value referenced by loc to data-memory address 0: 16 bit. The instruction takes two 16-bit words, where the second word is the 16-bit constant.

[0385] 3) MAC P, loc, 0: 16 bit—Unlike the immediate address in the MOV instructions, this address is neither preceded by an asterisk (*) nor contained within parentheses. In addition, this pointer is to program memory, not data memory. As part of its operation, this MAC instruction multiplies the value at program-memory address 0: 16 bit by the value referenced by loc. The instruction takes two 16-bit words, where the second word is the 16-bit constant.

[0386] Register addressing mode allows access of certain registers by name instead of by address. In addition, this mode allows transfer of data from one register to another without having to store the data to memory first. Table 34 explains the available operands for this mode. The operand @XARn is only supported by one of these instructions: ADDL, CMPL, MOVL, or SUBL. In addition, none of the other register-addressing operands can be used with ADDL, CMPL, MOVL, or SUBL. Here are two examples that use register addressing:

	PUSH	@AL		; Save low accumulator to stack.
	MOV	AR2,	@SP	; Store stack pointer to auxiliary
				; register 2.

[0387] In this MOV instruction, @SP is a register-operand, but AR2 is not. AR2 is simply a part of the particular syntax of the instruction.

TABLE 34
Operands for Register Addressing Mode
Operand	Description	Example
@AH	High word of accumulator	MOV @AH, #0 loads the high word of the
		accumulator with 0.
@AL	Low word of accumulator	MOV AH, @AL loads the low word of the
		accumulator into the high word of the accumulator.
@PH	High word of product register	MOV AR4, @PH loads AR4 with the high word of
		the P register.
@PL	Low word of product register	MOV T, @PL loads the low word of the P register
		into the T register.
@T	Multiplicand register	ADD @T, #5 adds 5 to the T register value.
@SP	Stack pointer	ADD @SP, AL adds the content of AL to SP.
@ARx	Auxiliary register x x = 0, 1, 2,	ADD ACC, @AR0 adds the content of AR0 to ACC.
	3, 4, 5, 6, or 7
@XARn	Extended auxiliary register n	SUBL ACC, @XAR6 subtracts the content of XAR6
	n = 6 or 7	from the accumulator.

[0388] In the direct addressing modes, data memory is seen as blocks of 64 words called data pages. As shown in FIG. 27, the entire 4M words of data memory consists of 65 536 data pages labeled 0 through 65535.

[0389] The data page is reflected in the 16 most significant bits (MSBs) of an address. For example, if the 16 MSBs are 0s, the address is in data page 0; if the 16 MSBs are all 1, the address is in data page 65535. The 6 least significant bits of an address are known as the offset. For example, an offset of 0 (00 00002) indicates the first word on a page; an offset of 63 (11 111116) indicates the last word on a page. There are two direct addressing modes: DP direct addressing mode, or PAGE0 direct addressing mode.

[0390] When DP direct addressing is used, the data page is determined by the value in the 16-bit data page pointer (DP), an independent CPU register. For example, if the DP value is 000016, the data page is 0. If the DP value is 000216, the data page is 2. To indicate which word on a page is accessed, supply a 6-bit offset preceded by the symbol @. In the following example, the operand @1 indicates an offset of 1.

ADD ACC, @1

[0391] To form a complete 22-bit address, the address register arithmetic unit (ARAU) concatenates the DP value and the 6-bit offset. Suppose DP is 1 (00 00 0000 0000 00012) and the offset is 63 (11 11112). When concatenated, they form the address of the last word on data page 1 (00 0000 0000 0000 0111 11112).

[0392] Before using the DP direct addressing mode, make sure the DP contains the desired page number (from 0 to 65 535). For example:

MOVW	DP,	#32	; Set data page to 32
ADD	ACC,	@1	; Add second word on page 32 to ACC.

[0393] There is no need to set the data page prior to every instruction that uses direct addressing. If all the instructions in a block of code access the same data page, simply load the DP at the front of the block.

[0394] Instructions that use PAGE0 direct addressing mode always access data page 0 (addresses 00 000016-00 003F16), regardless of the DP value. Indicate a particular word by supplying an operand of the form @0; 6 bit, where 6 bit is the 6-bit offset. The following example shows an offset of 1. To form a complete 22-bit address, the processor concatenates 16 leading zeros and the six bits of the offset.

ADD ACC, @0:1

[0395] PAGE0 direct addressing mode is only available when the PAGE0 bit in status register ST1 is 1. Before using this mode, make sure PAGE0=1. For example:

SETC	PAGE0	; Turn PAGE0 direct addressing on.
ADD	ACC, @0:1	; Add second word on page 0 to ACC.

[0396] There are two mutually-exclusive PAGE0 addressing modes: PAGE0 direct addressing, or PAGE0 stack addressing.

[0397] When PAGE0=1, PAGE0 stack addressing cannot be used. Likewise, when PAGE0=0, PAGE0 direct addressing cannot be used. Switch freely between the modes using the SETC and CLRC instructions. The output core signal PAGE0S indicates which PAGE0 mode is currently active.

[0398] The stack pointer (SP) always points to the next empty location in the software stack in data memory. Two indirect addressing modes use the stack pointer when referencing data memory: stack-pointer indirect addressing mode, or PAGE stack addressing mode.

[0399] Stack-pointer indirect addressing mode is used to access the data-memory location that is pointed to by the SP. PAGE0 stack addressing is used to access a data-memory location relative to the position of the SP. Because the stack pointer has only 16 bits, the reach of these modes is limited to the first 64K of data memory (addresses 00 000016-00 FFFF16).

[0400] In stack-pointer indirect addressing mode, the stack pointer is used to point to the data-memory address where a read or write is to occur. As shown in FIG. 28, the 22-bit address is the concatenation of six leading zeros and the 16 bits of the SP. Because the six most significant bits are always 0s in this mode, its reach is limited to the address range 00 000016-00 FFFF16.

[0401] Table 35 explains the two operands available for stack-pointer indirect addressing mode. The increment or decrement of SP is performed by the address register arithmetic unit (ARAU) in the decode 2 phase of the pipeline.

TABLE 35
Operands for Stack-Pointer Indirect Addressing Mode
Operand	Description	Example
*SP++	Increment SP after the data-	MOV *SP++, AH loads the high
	memory access. If (16-bit	word of the accumulator to the
	operation) SP = SP +1 If	16-bit location pointed to by SP
	(32-bit operation) SP =	and increments the content of
	SP +2	SP. Because the load is a 16-bit
		operation, SP is incremented
		by 1.
*−−SP	Decrement SP before the	MOVL ACC, *−−SP decrements
	data-memory access. If (16-	the content of SP and loads ACC
	bit operation) SP = SP − 1	with the 32-bit value referenced
	If (32-bit operation) SP =	by SP. Because the load is a 32-
	SP − 2	bit operation, SP is decremented
		by 2.

[0402] As shown in FIG. 29, PAGE0 stack addressing allows an instruction to access a data-memory location relative to the position of the SP. In the instruction, specify a positive 6-bit offset (from 0 to 63), and the address register arithmetic unit (ARAU) subtracts the offset from the SP. The result, (SP-offset), is concatenated with six leading zeros to form the full 22-bit address. Because the six most significant bits are always 0s in this mode, the reach of the mode is limited to the address range 00000016-00FFFF16.

[0403] PAGE0 stack addressing mode is only available when the PAGE0 bit in status register ST1 is 0. The general form for an operand in this mode is *—SP [6 bit], where 6 bit is the 6-bit offset. Here is an example of using this mode:

CLRC	PAGE0	; Select PAGE0 stack addressing mode.
SUB	ACC, *--SP[4]	; Subtract data-memory value from ACC.
		; Value is 4 locations above the
		; address in SP.

[0404] There are two mutually-exclusive PAGE0 addressing modes: PAGE0 direct addressing, or PAGE0 stack addressing. When PAGE0=1, PAGE0 stack addressing cannot be used. Likewise, when PAGE0=0, PAGE0 direct addressing cannot be used. Switch freely between the modes using the SETC and CLRC instructions. The output core signal PAGE0S indicates which PAGE0 mode is currently active.

[0405] For indirect addressing modes other than stack-pointer indirect addressing, the CPU provides eight auxiliary registers as pointers to memory. There are six 16-bit auxiliary registers (AR0, AR1, AR2, AR3, AR4, and AR5) and two 22-bit extended auxiliary registers (XAR6 and XAR7). FIG. 30 shows the ranges of data memory accessible to the auxiliary registers. XAR7 can also be used by some instructions to point to any value in program memory.

[0406] There are four indirect addressing modes that use auxiliary registers: auxiliary-register indirect addressing mode, ARP indirect addressing mode, XAR7 indirect addressing mode, or circular addressing mode.

[0407] Auxiliary-register indirect addressing mode enables pointing to addresses 00 000016-00 FFFF16 using AR0-AR5 and pointing to addresses 00 000016-3F FFFF16 using XAR6 and XAR7. Add an offset to an auxiliary register to access a location relative to the current position of that auxiliary register. Specify one of three sources for the offset:

[0408] 1) Auxiliary register AR0. Load AR0 with a value from 0 to 65 535 (000016 to FFFF16), and specify AR0 as the offset in the instruction. For example:

ADD AL,*+AR3[AR0] ; Concatenate sum (AR3 + AR0) with
                  ; siz zeros to form address
                  ; 0:(AR3 + AR0). Add content of that
                  ; address to the low accumulator.

[0409] 2) Auxiliary register AR1. Load AR1 with a value from 0 to 65 535, and specify AR1 as the offset in the instruction. For example:

ADD AL,*+XAR6[AR1] ; Calculate address (XAR6 + AR1).
                   ; Add content of that
                   ; address to the low accumulator.

[0410] 3) A constant. In an operand, specify a value from 0 to 7. For

example: ADD AL,*+AR2[4] ; Concatenate sum (AR2 + 4) with
                         ; six zeros to form address
                         ; 0:(AR2 + 4). Add content of that
                         ; address to the low accumulator.

[0411] When an operand specifies an offset, the auxiliary register is not modified. The offset is only used to generate the address. Some syntaxes of the MOVB instruction use the offset to indicate relative byte positions rather than relative word positions.

[0412] Table 36 lists and explains the operands that are available for auxiliary-register indirect addressing mode. The increments and decrements are performed by the address register arithmetic unit (ARAU) in the decode 2 phase of the pipeline. The auxiliary register pointer (ARP) in status register ST1 is updated by all these operands. This update also happens in the decode 2 phase of the pipeline.

TABLE 36
Operands for Auxiliary-Register
Indirect Addressing Mode
Operand	Description	Example
*ARx++ or	Increment the specified auxiliary	MOV *AR2++, AH loads the high
*XARn++	register after the data-memory access.	word of the accumulator to the
	Also, set ARP to x or n. If (16-bit	address 0:AR2, increments the
	operation) ARx/XARn = ARx/XARn + 1	content of AR2, and sets ARP = 2.
	If (32-bit operation) ARx/XARn =
	ARx/XARn + 2
*−−ARx or	Decrement the specified auxiliary	MOV *−−XAR6, AH decrements the
*−−XARn	register before the data-memory	content of XAR6, loads the high word
	access. Also, set ARP to x or n. If (16-	of the accumulator to the address
	bit operation) ARx/XARn = ARx/XARn −	pointed to by XAR6, and sets ARP =
	1 If (32-bit operation) ARx/XARn =	6.
	ARx/XARn − 2
*+ARx[AR0] or	Use the address 0:(ARx + AR0) or	MOV *+AR4[AR0], AH loads the high
*+XARn[AR0]	0:(XARn + AR0) for the data-memory	word of the accumulator to the 22-bit
	access. Also, set ARP to x or n. Do not	address 0:(AR4 + AR0) and sets ARP =
	modify ARx or XARn.	4.
*+ARx[AR1] or	Use the address (ARx + AR1) or (XARn +	MOV *+XAR7[AR1], AH loads the
*+XARn[AR1]	AR1) for the data-memory access.	high word of the accumulator to the
	Also, set ARP to x or n. Do not modify	22-bit address (XAR7 + AR1) and
	ARx or XARn.	sets ARP = 7. XAR7 is not modified.
*+ARx[3bit] or	Use the address (ARx+3bit) or	MOV *+AR5[6], AH loads the high
*+XARn[3bit]	(XARn+3bit) for the data-memory	word of the accumulator to the 16-bit
	access. Also, set ARP to x or n. Do not	location pointed to by 0; (AR5+6) and
	modify ARx or XARn.	sets ARP=5. AR5 is not modified
*ARx or	Equivalent to *+ARx[0] or *+XARn[0].	MOV *AR5, AH loads the high word
*XARn	Use the address in ARx or XARn for	of the accumulator to the 16-bit
	the data-memory access. Also, set ARP	location pointed to by AR5 and sets
	to x or n. Do not modify ARx or XARn.	ARP=5. AR5 is not modified.
Notes:
1) ARx = AR0, AR1, AR2, AR3, AR4, or AR5
2) XARn = XAR6 or XAR7
3) 3bit is a 3-bit constant (a value from 0 to 7).
4) The notation 0:ARx or 0:(ARx + n) indicates that when a 16-bit auxiliary register is used, the 6 MSBs of the data-memory address are zeros.

[0413] In ARP indirect addressing, the name of the auxiliary register is not specified in the instruction; rather, the instruction uses the auxiliary register pointed to by the 3-bit auxiliary register pointer (ARP) of status register ST1. Before using ARP indirect addressing mode, make sure ARP holds the appropriate value from 0 to 7 (0 is for AR0, 1 is for AR1, 2 is for AR2, and so on). The register pointed to by the ARP is referred to as the current auxiliary register.

[0414] Add an offset to access a location relative to the position of an auxiliary register. The source for the offset is AR0. After loading AR0 with a value from 0 to 65 535 (000016 to FFFF16), use one of two operands: *0++or *0−−. The operand *0++adds AR0 to the current auxiliary register; *0−−subtracts ARO from the current auxiliary register. Here is an example:

	ADD	AL,*0++	; Add content of current auxiliary
			; register to low half of
			; accumulator. Then add content of
			; AR0 to current auxiliary register.

[0415] Table 37 lists and explains the operands that are available for ARP indirect addressing mode. The increments and decrements are performed by the address register arithmetic unit (ARAU) in the decode 2 phase of the pipeline.

TABLE 37
Operands for ARP Indirect Addressing Mode
Operand	Description	Example
*ARPx	Set ARP to x	MOV *ARP2, AH loads the high word of the
	x	Selects	x	Selects	accumulator to the 16-bit location pointed to
	0	AR0	4	AR4	by the current auxiliary register and then
	1	AR1	5	AR5	sets ARP to 2.
	2	AR2	6	XAR6
	3	AR3	7	XAR7
*	No increment or decrement of	MOV *, AH loads the high word of the
	current auxiliary register.	accumulator to the 16-bit location pointed to
		by the current auxiliary register and does
		nothing to the content of the current AR.
*++	Increment current auxiliary register	MOV *++, AH loads the high word of the
	after the data-memory access.	accumulator to the 16-bit location pointed to
	If (16-bit operation)	by the current auxiliary register and
	ARx/XARn = ARx/XARn + 1	increments the content of the current AR by
	If (32-bit operation)	1.
	ARx/XARn = ARx/XARn + 2
*−−	Decrement current auxiliary register	MOV *−−, AH loads the high word of the
	after the data-memory access,	accumulator to the 16-bit location pointed to
	If (16-bit operation)	by the current auxiliary register and
	ARx/XARn = ARx/XARn − 1	decrements the content of the current AR by
	If (32-bit operation)	1.
	ARx/XARn = ARx/XARn − 2
*0++	Increment current auxiliary register	MOV *0++, AH loads the high word of the
	by index amount after the data-	accumulator to the 16-bit location pointed to
	memory access. The index amount is	by the current auxiliary register and adds
	the value in AR0.	the content of AR0 to the content of the
		current auxiliary register.
*0−−	Decrement current auxiliary register	MOV *0−−, AH loads the high word of the
	by index amount after the data-	accumulator to the 16-bit location pointed to
	memory access. The index amount is	by the current auxiliary register and
	the value in AR0.	subtracts the content of AR0 from the
		content of the current auxiliary register.

[0416] XAR7 indirect addressing mode allows use of the extended auxiliary register XAR7 to point to an address in program memory. This mode has a single operand, *XAR7, and is only used for the following instructions. This operand does not affect the auxiliary register pointer (ARP).

[0417] PREAD operand1, *XAR7—where operand1 represents a data-memory address. This instruction moves a 16-bit value from the program-memory location referenced by XAR7 to the data-memory location referenced by operand1.

[0418] PWRITE *XAR7, operand2 where operand2 represents a data-memory address. This instruction moves a 16-bit value from the data-memory location referenced by operand2 to the program-memory location referenced by XAR7.

[0419] LB *XAR7—This instruction forces a branch to the program-memory location referenced by XAR7.

[0420] CALL *XAR7—This instruction calls a subroutine at the program-memory location referenced by XAR7.

[0421] Circular addressing is used to access a user-defined buffer in data memory. FIG. 31 illustrates these characteristics of the buffer:

[0422] The buffer size is determined by the 8 least significant bits (LSBs) of auxiliary register AR1, or AR1(7:0). Specifically, the buffer size is AR1(7:0) plus 1. When AR1(7:0) is 255, the buffer is at its maximum size of 256.

[0423] Auxiliary register XAR6 points to the current address The top of the buffer must be at an address whose 8 LSBs are zeros. When the 8 LSBs of XAR6 match the 8 LSBs of AR1, XAR6 is pointing to the bottom of the buffer.

[0424] During a series of circular-addressing operations, the value in the 16 LSBs of XAR6 (AR6) is incremented until the 8 LSBs of AR6 match the 8 LSBs of AR1. When the values match, the 8 LSBs of AR6 are cleared, so that XAR6 points to the top of the buffer. The value in the upper 6 bits of XAR6 is not modified during any of the above operations.

[0425] Circular addressing mode has a single operand, *AR6% ++. The operation performed when this operand is used:

[0426] 1) Set the auxiliary register pointer (ARP) to 6, and perform the read from or write to data memory. The address in XAR6 identifies the data-memory location where the access takes place.

[0427] 2) If the 8 LSBs of AR6 match the 8 LSBs of AR1 (the bottom of the buffer has been reached), clear AR6(7:0). If the 8-bit values do not match, increment AR6 as follows: if the instruction accessed one 16-bit location, increment AR6 by 1; or if the instruction accessed two 16-bit locations, increment AR6 by 2.

[0428] The increment of AR6 is performed by the address register arithmetic unit (ARAU) in the decode 2 phase of the pipeline. The 6 MSBs of XAR6 are not affected. Here is an example of using circular addressing mode:

; AR6(7:0) is not equal to
AR1(7:0)
MOV  *AR6%++,AH ; Load high word of the accumulator to
                ; the address referenced by XAR6,
                ; increment AR6 by 1, and set ARP = 6.

[0429] Because this MOV instruction writes to only one 16-bit data-memory location, AR6 is incremented by 1. In the following example, the ADDL instruction performs a 32-bit operation; that is, it accesses two data-memory locations. As a result, AR6 is incremented by 2.

; AR6(7:0) is not equal to
AR1(7:0)
ADDL ACC,
*AR6%++ ; Read the 32-bit value at the addresses
        ; referenced by XAR6. Add the 32-bit
        ; value to ACC. Increment AR6 by 2.
        ; Set ARP to 6.

[0430] If at least one of the instructions accessing a circular buffer performs a 32-bit operation, make sure XAR6 and AR1 are both even before the buffer is accessed. This ensures that AR6(7:0) and AR1(7:0) will match at the bottom of the buffer every time XAR6 cycles through the buffer. As an example of the problems that could arise if XAR6 and AR1 are not both even, consider the following:

[0431] 1) XAR6 is loaded with the even address 01 FF0016 and AR1 is loaded with the odd value 00 000716. The value in AR1 indicates that the bottom of the buffer is at address 01 FF0716.

[0432] 2) The next six instructions perform 16-bit operations, and each causes AR6 to increment by 1. At this point, XAR6 points to 01 FF0616.

[0433] 3) Then a seventh instruction performs a 32-bit operation and increments AR6 by 2. Now XAR6 points to 01 FF0816, which is past the bottom of the buffer. Because AR6(7:0) and AR1(7:0) never match, XAR6 never gets returned to the top of the buffer.

[0434] Table 38 summarizes the operands for all indirect addressing modes except XAR7 indirect addressing mode. XAR7 indirect addressing mode is only used in four instructions.

TABLE 38
Indirect-Addressing Operands
Operand	Description	Example
*SP++	Increment SP after the data-memory	MOV *SP++, AH leads the high word of the
	access.	accumulator to the 16-bit location pointed to by
	If (16-bit operation) SP = SP + 1	SP and increments the content of SP. Because the
	If (32-bit operation) SP = SP + 2	load is a 16-bit operation, SP is incremented by 1.
*−−SP	Decrement SP before the data-memory	MOVL ACC, *−−SP decrements the content of SP
	access.	and loads ACC with the 32-bit value referance by
	If (16-bit operation) SP = SP − 1	SP. Because the load is a 32-bit operation, SP is
	If (32-bit operation) SP = SP − 2	decremented by 2.
*SP[6bit]	Decrement SP by amount 6bit before the	SUB ACC, *−SP[4] decrements the content of SP
	data-memory access. This operand is only	by 4 and subtracts the value referenced by SP
	available when the PAGE0 bit of ST1 is 0.	from ACC.
*ARx++ or	Increment the specified auxiliary register	MOV *AR2++, AH loads the high word of the
*XARn++	after the data-memory access. Also, set	accumulator to the address 0:AR2, increments the
	ARP to x or n.	content of AR2, and sets AP = 2.
	If (16-bit operation)
	ARx/XARn = ARx/XARn + 1
	If (32-bit operation)
	ARx/XARn = ARx/XARn + 2
*−−ARx or	Decrement the specified auxiliary register	MOV *−−XAR6, AH decrements the content of
*−−XARn	before the data-memory access. Also, set	XAR6, loads the high word of the accumulator to
	ARP to x or n.	the address pointed to by XAR6, and sets ARP = 6.
	If (16-bit operation)
	ARx/XARn = AXx/XARn − 1
	If (32-bit operation)
	ARx/XARn = ARx/XARn − 2
*+ARx[AR0] or	Use the address 0:(ARx + AR0) or 0:(XARn	MOV *+AR4[AR0], AH loads the high word of
*+XARRn[AR0]	+ AR0) for the data-memory access. Also,	the accumulator to the 22-bit address 0:(AR4 +
	set ARP to x or n.	AR0) and sets ARP = 4.
*+ARx[AR1] or	Use the address (ARx + AR1) or (XARn +	MOV *+XAR7[AR1], AH loads the high word of
*+XARn[AR1]	AR1) for the data-memory access. Also, set	the accumulator to the 22-bit address (XAR7 +
	ARP to x or n.	AR1) and sets ARP = 7.
*+ARx[3bit] or	Use the address (ARx + 3bit) or (XARn +	MOV *+AR5[6], AH loads the high word of the
*+XARn[ 3bit]	3bit) for the data-memory access. Also, set	accumulator to the 16-bit location pointed to by
	ARP to x or n.	0:(AR5 + 6) and sets ARP = 5.
*APx or *XARn	Equivalent to *+ARx[0] or *+XARn[0]. Use	MOV *AR5, AH loads the high word of the
	the address in ARx or XARn for the data-	accumulator to the 16-bit location pointed to by
	memory access. Also, set ARP to x or n.	AR5 and sets ARP = 5.
*ARPx	Set ARP to x	MOV *ARP2, AH loads the high word of the
	x	Selects	x	Selects	accumulator to the 16-bit location pointed to by
	0	AR0	4	AR4	the current auxiliary register and then sets ARP
	1	AR1	5	AR5	to 2.
	2	AR2	6	XAR6
	3	AR3	7	XAR7
*	No increment or decrement of current	MOV *, AH loads the high word of the
	auxiliary register.	accumulator to the 16-bit location pointed to by
		the current auxiliary register and does nothing to
		the content of the current AR.
*++	Increment current auxiliary register after	MOV *++, AH loads the high word of the
	the data-memory access,	accumulator to the 16-bit location pointed to by
	If (16-bit operation)	the current auxiliary register and increments the
	ARx/XARn = ARx/XARn + 1	content of the current AR by 1.
	If (32-bit operation)
	ARx/XARn = ARx/XARn + 2
*−−	Decrement current auxiliary register	MOV *−−, AH loads the high word of the
	after the data-memory access.	accumulator to the 16-bit location pointed to by
	If (16-bit operation)	the current auxiliary register and decrements the
	ARx/XARn=ARx/XARn − 1	content of the current AR by 1.
	If (32-bit operation)
	ARx/XARn = ARx/XARn − 2
*0++	Increment current auxiliary register by	MOV *0++, AH loads the high word of the
	index amount after the data-memory	accumulator to the 16-bit location pointed to by
	access. The index amount is the value in	the current auxiliary register and adds the
	AR0.	content of AR0 to the content of the current
		auxiliary register.
*0−−	Decrement current auxiliary register by	MOV *0−−, AH loads the high word of the
	index amount after the data-memory	accumulator to the 16-bit location pointed to by
	access. The index amount is the value in	the current auxiliary register and subtracts the
	AR0.	content of AR0 from the content of the current
		auxiliary register.
*AR6%++	After the data-memory access, if AR6(7:0)	ADDL ACC, *AR6%++ reads the 32-bit value at
	equals AR1(7:0), clear AR6(7:0). Otherwise	the addresses referenced by XAR6, adds that
	increment AR6. If (16-bit operation) AR6 =	value to ACC, and sets ARP to 6. If AR6(7:0)
	AR6 + 1 If (32-bit operation) AR6 = AR6 +2	equals AR1(7:0), AR6(7:0) is cleared. Otherwise,
		AR6 is incremented by 2.
Notes:
1) ARx = AR0, AR1, AR2, AR3, AR4, or AR5
2) XARn = XAR6 or XAR7
3) 3bit is a 3-bit constant (a value from 0 to 7).
4) The notation 0:ARx or 0:(ARx + n) indicates that when a 16-bit auxiliary register is used, the 6 MSBs of the data-memory address are zeros.

[0435] This portion explains how each of the addressing modes is encoded in the instruction opcodes of instructions. One of the ADD instructions is depicted in FIG. 32.

[0436] As shown in FIG. 32, the second word (16 BitSigned) is an embedded constant from an immediate addressing mode. Bits 7 through 0 of the first word encode information about a register-, direct-, or indirect-addressing operand. When a 2-word opcode is stored in program memory, the least significant word must be stored first.

[0437] When an immediate addressing mode is used, the constant entered is embedded directly in the instruction opcode. If the instruction requires a constant of fewer than 16 bits, only one word is needed for the opcode, and the constant is embedded in the lower portion of the opcode. Consider FIG. 33 and FIG. 34, which both use immediate-constant addressing mode. In FIG. 33, an 8-bit constant is embedded in bits 7 through 0 of the opcode. Bits 15 through 8 contain the code for the chosen syntax of the ADDB instruction. In FIG. 34, a 5-bit constant is embedded in bits 4 through 0, and bits 15 through 5 contain the code for the single syntax of the TRAP instruction.

[0438] When an instruction uses a constant of 16 or more bits, the instruction requires two 16-bit words for its opcode. This is shown in FIG. 35 and FIG. 36. In FIG. 35, the first word of the opcode contains the opcode for the MOVW instruction, and the second word contains the 16-bit constant. FIG. 36 uses a 22-bit constant. The 6 most significant bits of the constant are embedded in bits 5 through 0 of the first word of the opcode. The 16 least significant bits of the constant form the second word.

[0439] Many instructions use register addressing mode or one of the direct or indirect addressing modes to access data memory. For these instructions, the addressing-mode information is encoded in bits 7 through 0 of the first or only word of the opcode. Table 39 summarizes how these eight bits are filled for each of the modes.

TABLE 39
Encoding of Register-, Direct-, and
Indirect-Addressing Operands
Addressing	Bits 7-0 of
Mode	Opcode	Operand	Comments
DP direct	0 0 CCC CCC	@6bit	The 6-bit constant given in the operand is
			placed in the op-code where the six Cs
			are shown.
PAGE0 direct	0 1 CCC CCC	@0:6bit	The 6-bit constant given in the operand is
			placed in the op-code where the six Cs
			are shown. PAGE0 direct addressing
			mode and PAGE0 stack ad-dressing mode
			are encoded the same way because only
			one can be active at a time (selected by
			the PAGE0 bit).
PAGE0 stack	0 1 CCC CCC	*Sp[6bit]	The 6-bit offset given in the operand is
			placed in the opcode where the six Cs are
			shown. PAGE0 direct addressing mode
			and PAGE0 stack addressing mode are
			encoded the same way because only one
			can be active at a time (selected by the
			PAGE0 bit).
Auxiliary-	1 0 000 AAA	*ARx++ or XARn++	In the operands, x is a number from 0 to
Register	1 0 001 AAA	*−−ARx or *−−XARn	5, n is 6 or 7, and 3bit is a 3-bit constant.
Indirect	1 0 010 AAA	*+ARx[AR0] or *+XARn[AR0]	In the opcode, three As indicate where
	1 0 011 AAA	*+ARx[AR1] or *+XARn[AR1]	the 3-bit auxiliary register number (from
	1 1 CCC AAA	*+AR x[3bit] or *+XARn[3bit]	0 to 7) is embedded. Three Cs indicate
			where the 3-bit constant is embedded.
Register	1 0 100 AAA	@ARx or @XARn++	In the operands, x is a number from 0 to
	1 0 101 000	@AH	7, and n is 6 or 7. The three As in the
	1 0 101 001	@AL	first register-addressing code indicate
	1 0 101 010	@PH	where the 3-bit auxiliary register number
	1 0 101 011	@PL	(from 0 to 7) is em-bedded.
	1 0 101 100	@T
	1 0 101 101	@SP
ARP indirect	1 0 110 AAA	*ARP z	The variable z is a 3-bit number from 0 to
	1 0 111 000	*	7, indicating which auxiliary register is
	1 0 111 001	*++	selected. This 3-bit number is embedded
	1 0 111 010	*−−	in the opcode where the three As are
	1 0 111 011	*0++	shown.
	1 0 111 100	*0−−
Stack-pointer	1 0 111 101	*SP++
indirect	1 0 111 110	*−−SP
Circular	1 0 111 111	*AR6%++

[0440] FIGS. 37 through 40 show how bits 7 through 0 of opcodes are filled for particular cases of register, direct, and indirect addressing modes. The instruction in FIG. 37 uses register addressing to access the T register. This is indicated by the code in bits 7-0 of the opcode.

[0441] In FIG. 38, bits 7 and 6 of the opcode are both 0, indicating DP direct addressing mode. Bits 5 through 0 contain the offset of 5.

[0442] FIG. 39 uses auxiliary-register indirect addressing to access data-memory. This is indicated by bits 7 through 0 of the first word of the opcode. Because XAR6 is used, bits 2 through 0 contain 6 (1102). The example also uses immediate-pointer addressing to access program memory. The constant given in the third operand (5) indicates the 16 LSBs of address 00 000516. This 16-bit constant forms the second word of the opcode.

[0443] The instruction in FIG. 40 uses stack-pointer indirect addressing to point to a location in data memory. In addition, it uses immediate-constant addressing to indicate a 16-bit signed constant. This constant forms the second word of the opcode.

[0444] This portion presents a summary of the instructions, explains how some instructions are affected by alignment of 32-bit accesses to even addresses, defines special symbols and notations used and describes each instruction in detail in alphabetical order.

[0445] An alphabetical summary of the DSP assembly language instructions is provided as Appendix B. An additional section provides a summary of instruction syntaxes by operation type.

[0446] Table 40 lists the instruction set in alphabetical order by mnemonic, a brief description of the instruction, and comments about the instruction.

TABLE 40
Instruction Set Summary in Alphabetical Order
Mnemonic	Description	Comments
ABORTI	Abort interrupt	Used when an interrupt will not be returned from in the
		normal manner
ABS	Absolute value of	Finds the absolute value of the full 32-bit accumulator
	accumulator
ADD	Add value to specified	Adds a 16-bit value to a data-memory location or a register.
	location	This value can be from data-memory or a register, or can be a
		constant.
ADDB	Add short value to	Adds a 7-bit constant to an auxiliary register or the stack
	specified register	pointer, or adds an 8-bit constant to AH, AL, or ACC
ADDCU	Add unsigned value plus	Unsigned value is a 16-bit data-memory or register value
	carry bit to accumulator
ADDL	Add long value to	Adds a 32-bit data-memory value or a 22-bit auxiliary register
	accumulator	value to ACC
ADDU	Add unsigned value to	Unsigned value is a 16-bit data-memory or register value
	accumulator
ADRK	Add to current auxiliary	Adds an 8-bit constant to an auxiliary register
	register
AND	Bitwise AND	Performs an AND operation on two 16-bit values Can be used
		to clear values in the interrupt enable register (IER) and the
		interrupt flag register (IFR).
ANDB	Bitwise AND with short	Performs an AND operation on an 8-bit value and an
	value	accumulator half (AH or AL)
ASP	Align stack pointer	If the stack pointer (SP) points to an odd address, ASP
		increments SP by 1. This aligns SP to an even address.
ASR	Arithmetic shift right	Shifts AH or AL right by the amount you specify. During the
		shift, the value is sign extended.
B	Branch	Uses a 16-bit offset to perform a relative branch
BANZ	Branch if auxiliary	Uses a 16-bit offset to perform a relative branch
	register not equal to zero
CALL	Call	Stores the return address to the stack and then performs an
		absolute branch to a specified 22-bit address
CLRC	Clear Status Bits	Clears one or more of certain bits in status registers STO and
		ST1
CMP	Compare	Uses a subtraction to compare two values. The result is not
		stored, but is reflected by flag bits.
CMPB	Compare with short value	Uses a subtraction to compare AH or AL with an 8-bit
		constant. The result is not stored, but is reflected by flag bits.
CMPL	Compare with long value	Uses a subtraction to compare ACC with a 32-bit data-memory
		value or a 22-bit auxiliary register value. The result is not
		stored, but is reflected by flag bits.
DEC	Decrement specified	Acts on a 16-bit data-memory location or register
	value by 1
EALLOW	Allow access to emulation	Enables access to the memory-mapped emulation registers
	registers
EDIS	Disallow access to	Disables access to the memory-mapped emulation registers
	emulation registers
ESTOP0	Emulator software	Used to create a software breakpoint
	breakpoint
ESTOP1	Embedded software	Used to create an embedded software breakpoint
	breakpoint
FFC	Fast function call	Stores the return address to auxiliary register XAR7 and then
		performs an absolute branch to a specified 22-bit address
IACK	Interrupt acknowledge	Drives a specified 16-bit value on the low 16 bits of the data
		write data bus, DWDB(15:0). Can be used in an interrupt
		service routine to inform external hardware that a certain
		interrupt is being serviced by the CPU.
IDLE	Idle until interrupt	Places CPU in a dormant state until it is awakened by an
		enabled or nonmaskable hardware interrupt
INC	Increment specified value	Acts on a 16-bit data-memory location or register
	by 1
INTR	Software interrupt	Can be used to initiate the following interrupts: INT1-INT14,
		DLOGINT, RTOSINT, EMUINT, and NMI. For those
		interrupts with bits in the interrupt enable register (IER) and
		the interrupt flag register (IER), the bits are cleared.
IRET	Return from interrupt	Restores the program counter (PC) value and other register
	and restore register pairs	values that were saved to the stack by an interrupt operation
ITRAP0	Instruction trap 0	Causes an illegal instruction trap, as with a TRAP #19
		instruction
ITRAP1	Instruction trap 1	Causes an illegal instruction trap, as with a TRAP #19
		instruction
LB	Long branch	Performs an absolute branch to a specified 22-hit address
LOOPNZ	Loop while not zero	Repeats until a specified test results in zero
LOOPZ	Loop while zero	Repeats until a specified test results in a nonzero value
LSL	Logical shift left	Shifts AR, AL, or ACC left by the amount you specify
LSR	Logical shift right	Shifts AR or AL right by the amount you specify. During the
		shift, the value is not sign extended.
MAC	Multiply and accumulate	Adds the P register value to ACC, loads T register and then
	with preload to T register	multiplies T register value by value from program memory
MOV	Move value	Copies a value from one location to another, or loads a location
		with a specified value
MOVA	Load T register and add	Loads the T register and adds the P register value to ACC
	previous product to
	accumulator
MOVB	Move short value	Loads a register with an 8-bit constant or moves a specified
		byte from one location to another
MOVH	Store high word	Stores the high word of the P register or ACC to data-memory
		or to a register
MOVL	Move long value	Enables loads and stores that use ACC and 32-bit data-memory
		locations
MOVP	Load T register and load	Loads the T register and stores the P register value to ACC
	previous product to
	accumulator
MOVS	Load T register and	Loads the T register and subtracts the P register value from
	subtract previous product	ACC
	from accumulator
MOVU	Load accumulator with	Loads AL with an unsigned 16-bit value and clears AH
	unsigned word
MOVW	Load entire data page	Loads the entire data page pointer (DP) with a 16-bit constant.
	pointer	One syntax of the MOV instruction enables you to load only
		the 10 LSBs of the DP.
MPY	Multiply	Multiplies a 16-bit value by another 16-bit value
MPYA	Multiply and accumulate	Adds the P register value to ACC and then multiplies two 16-
	previous product	bit values
MPYB	Multiply signed value by	Multiplies a signed 16-bit value by an unsigned 8-bit value
	unsigned short value
MPYS	Multiply and subtract	Subtracts the P register value from ACC and then multiplies a
	previous product	16-bit value by another 16-bit value
MPYU	Unsigned multiply	Multiplies an unsigned 16-bit value by another unsigned 16-bit
		value
MPYXU	Multiply signed value by	Multiplies a signed 16-bit value by an unsigned 16-hit value
	unsigned value
NASP	Unalign stack pointer	Undoes a previous alignment of SP performed by the ASP
		instruction.
NEG	Negative of accumulator	Finds the negative of the value in AH, AL, or ACC
	value
NOP	No operation	Can be used to purposely create inactive cycles. Can also be
		used to increment an auxiliary register or the stack pointer
		without performing any other task.
NORM	Normalize accumulator	Can be used to remove extra sign bits from a value in ACC
NOT	Complement of	Finds complement of AH, AL, or ACC 6-179
	accumulator value
OR	Bitwise OR	Performs an OR operation on two 16-bit values Can be used to
		set values in the interrupt enable register (IER) and the
		interrupt flag register (IFR). If the AND instruction sets an
		IFR and the interrupt is enabled, the interrupt will be
		serviced.
ORB	Bitwise OR with short	Performs an OR operation on an 8-bit value and an
	value	accumulator half (AH or AL)
POP	Restore from stack	Copies a 16-bit value or a 32-bit register pair from the stack to
		a data-memory location or a register
PREAD	Read from program	Loads a 16-bit data-memory location or register with a value
	memory	from program memory
PUSH	Save value on stack	Copies a 16-bit value or a 32-bit register pair to the stack
PWRITE	Write to program	Loads a program-memory location with a value from a 16-bit
	memory	data-memory location or register
RET	Return	Loads the program counter (PC) from the stack
RETE	Return with interrupts	Loads the program counter (PC) from the stack and clears the
	enabled	interrupt global mask bit (INTM). By clearing INTM, it
		enables interrupts INT1-1NT14, DLOGINT, and RTOSINT.
ROL	Rotate accumulator left	Can be seen as the left rotation of a 33-bit value that is the
		concatenation of the carry bit (C) and ACC.
ROR	Rotate accumulator right	Can be seen as the right rotation of a 33-bit value that is the
		concatenation of the carry bit (C) and ACC.
RPT	Repeat next instruction	Causes the following instruction to repeat a specified number
		of times
SAT	Saturate accumulator	Fills ACC with a saturation value that reflects the net overflow
		represented in the overflow counter (OVC)
SB	Short branch	Uses an 8-bit offset to perform a relative branch
SBBU	Subtract unsigned value	Unsigned value is a 16-bit data-memory or register value
	plus inverse borrow from
	accumulator
SBRK	Subtract from specified	Subtracts an 8-bit constant from an auxiliary register
	auxiliary register
SETC	Set status bits	Sets one or more of specified bits in status registers ST0 and
		ST1
SFR	Shift accumulator right	Shifts ACC right by the amount you specify. During the shift,
		the value is sign extended if the sign-extension mode bit (SXM)
		is 1 or is not sign extended if SXM is 0.
SPM	Set product shift mode	Sets the product shift mode bits (PM), which determine how
		certain instructions shift the P register value
SUB	Subtract value from	Subtracts a 16-bit value from a data-memory location or a
	specified location	register. This value can be from data-memory or a register, or
		can be a constant.
SUBB	Subtract short value from	Subtracts a 7-bit constant from an auxiliary register or the
	specified register	stack pointer, or subtracts an 8-bit constant from ACC
SUBCU	Conditional subtraction	Used for 16-bit division
SUBL	Subtract long value from	Subtracts a 32-bit data-memory value or a 22-bit auxiliary
	accumulator	register value from ACC
SUBU	Subtract unsigned value	Unsigned value is a 16-bit data-memory or register value
	from accumulator
SXTB	Sign extend least	Sign extends the least significant byte of AH or AL
	significant byte of
	accumulator half
TBIT	Test specified bit	Tests a specified bit of a 16-bit data-memory location or
		register. The value of the bit is reflected by the test/control flag
		bit (TC).
TEST	Test for accumulator	Uses a subtraction to compare ACC with 0. The result is not
	equal to zero	stored, but is reflected by flag bits.
TRAP	Software trap	Can be used to initiate any interrupt. Unlike the INTR
		instruction, the TRAP instruction never affects bits in the IER
		and IFR.
XOR	Bitwise exclusive OR	Performs an exclusive OR operation on two 16-bit values
XORB	Bitwise exclusive OR	Performs an exclusive OR operation on an 8-bit value and an
	with short value	accumulator half (AH or AL)

[0447] Tables 41 through 53 provide a summary of the instruction syntaxes according to the following functional groups:

[0448] 1) Address Register Operations (AR0-AR7, XAR6, XAR7, DP, SP) (see Table 42)

[0449] 2) Push and Pop Stack Operations (see Table 43)

[0450] 3) AX (AH, AL) Operations (see Table 44)

[0451] 4) AX (AH, AL) Byte Operations (see Table 45)

[0452] 5) ACC Operations (see Table 46)

[0453] 6) ACC 32-Bit Operations (see Table 47)

[0454] 7) Operations on Memory or Register (see Table 48)

[0455] 8) Data Move Operations (see Table 49)

[0456] 9) Program Flow Operations (see Table 50)

[0457] 10) Math Operations (see Table 51)

[0458] 11) Control Operations (see Table 52)

[0459] 12) Emulation Operations (see Table 53)

[0460] The DSP uses the following syntax formats. When a destination is specified, it is always the left most operand.

TABLE 41
Summary by Operation Type
Format	Example
Mnemonic destination, source 1, source 2	AND AH,@7,#16	;Read 2 source values.
		;Write logical AND of
		;those values to destination.
Mnemonic destination, source 1	MOV ACC,@7	;Read from source. Write
		;value to destination.
Mnemonic source 1, source 2	CMPL ACC,@7	;Compare two source values.
Mnemonic source	PUSH ST0	;Read from source. Stack is
		;destination.
Mnemonic destination	POP ST1	;Source is stack. Write to
		;destination.

[0461]

TABLE 42
Address Register Operations
(AR0-AR7, XAR6, XAR7, DP, SP)
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
ADDB	aux, #7bit	1	1	—	—
ADDB	SP, #7bit	1	1	—	—
ADRK	#8bit	1	1	ARP	—
MOV	AR x, loc	1	1	—	—
MOV	XAR n, locLong	1	1	—	—
MOV	XAR n, #22bit	2	1	—	—
MOV	DP, # 10bit	1	1	—	—
MOV	loc, ARx	1	1	—	N,Z
MOV	locLong, XARn	1	1	—	—
MOVB	ARx, #8bit	1	1	—	—
MOVW	DP, #16bit	2	1	—	—
SBRK	#8bit	1	1	ARP	—
SUBB	aux, #7bit	1	1	—	—
SUBB	SP, #7bit	1	1	—	—

[0462]

TABLE 43
Push and Pop Stack Operations
	Status Bits
				Affected
Instruction	Syntax	Words	Cycles	By	Affects
ADDB	aux, #7bit	1	1	—	—
ADDB	SP, #7bit	1	1	—	—
ADRK	#8bit	1	1	ARP	—
MOV	AR x,loc	1	1	—	—
MOV	XAR n,locLong	1	1	—	—
MOV	XAR n, #22bit	2	1	—	—
MOV	DP, #10bit	1	1	—	—
MOV	loc, ARx	1	1	—N, Z
MOV	locLong, XARn	1	1	—	—
MOVB	ARx, #8bit	1	1	—	—
MOVW	DP, #16bit	2	1	—	—
SBRK	# 8bit	1	1	ARP	—
SUBB	aux, #7bit	1	1	—	—
SUBB	SP, #7bit	1	1	—	—
POP	loc	1	2	—	N, Z
POP	DBGIER	1	5	—	—
POP	DP	1	1	—	—
POP	ST0	1	1	—	C, N, OVM, OVC, PM, SXM,
					TC, V, Z
POP	ST1	1	5	—	ARP, DBGM, EALLOW,
					INTM, PAGE0, SPA, VMAP
POP	T:ST0	1	1	—	C, N, OVM, OVC, PM, SXM,
					TC, V, Z
POP	DP:ST1	1	5	—	ARP, DBGM, EALLOW,
					INTM, PAGE0, SPA, VMAP
POP	PH:PL	1	1	—	—
POP	AR1:AR0	1	1	—	—
POP	AR3:AR2	1	1	—	—
POP	AR5:AR4	1	1	—	—
PUSH	loc	1	2	—	—
PUSH	DBGIER	1	1	—	—
PUSH	DP	1	1	—	—
PUSH	IFR	1	1	—	—
PUSH	ST0	1	1	—	—
PUSH	ST1	1	1	—	—
PUSH	T:ST0	1	1	—	—
PUSH	DP:ST1	1	1	—	—
PUSH	PH:PL	1	1	—	—
PUSH	AR1:AR0	1	1	—	—
PUSH	AR3:AR2	1	1	—	—
PUSH	AR5:AR4	1	1	—	—

[0463]

TABLE 44
AX (Ah, AL) Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
ADD	AX,loc	1	1	—	C, N, V, Z
ADDB	AX,	1	1	—	C, N, V, Z
	#8BitSigned
AND	AX, loc	1	1	—	N, Z
AND	AX, loc,	2	1	—	N, Z
	#16BitMask
ANDB	AX,	1	1	—	N, Z
	#8BitMask
ASR	AX, shift	1	1	—	C, N, Z
ASR	AX,T	1	1	—	C, N, Z
CMP	AX, loc	1	1	—	C, N, Z
CMPB	AX, #8bit	1	1	—	C, N, Z
LSL	AX, shift	1	1	—	C, N, Z
LSL	AX,T	1	1	—	C, N, Z
LSR	AX, shift	1	1	—	C, N, Z
LSR	AX, T	1	1	—	C, N, Z
MOV	AX, loc	1	1	—	N, Z
MOV	Loc, AX	1	1	—	N, Z
MOVB	AX, #8bit	1	1	—	N, Z
NEG	AX	1	1	—	C, N, V, Z
NOT	AX	1	1	N, Z
OR	AX, loc	1	1	—	N, Z
ORB	AX,	1	1	—	N, Z
	#8BitMask
SUB	AX, loc	1	1	—	C, N, V, Z
XOR	AX, loc	1	1	—	N, Z
XORB	AX,	1	1	—	N, Z
	#8BitMask

[0464]

TABLE 45
AX (AH, AL) Byte Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
MOVB	AX.LSB, loc	1	1	—	N, Z
MOVB	loc, AX.LSB	1	1	—	N, Z
MOVB	AX.MSB, loc	1	1	—	N, Z
MOVB	loc, AX.MSB	1	1	—	N, Z
SXTB	AX	1	1	—	N, Z

[0465]

TABLE 46
ACC Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
ABS	ACC	1	1	OVM	C, N, V, Z
ADD	ACC, loc, << shift	1	1	OVM, SXM	C, N, V, Z
ADD	ACC, #16BitSU<< shift	2	1	OVM, SXM	C, N, V, Z
ADD	ACC, P	1	1	OVM, PM	C, N, V, Z
ADDB	ACC, #8bit	1	1	OVM	C, N, V, Z
ADDCU	ACC, loc	1	1	OVM	C, N, V, Z
ADDU	ACC, loc	1	1	OVM	C, N, V, Z
LSL	ACC, shift	1	1	—	C, N, Z
LSL	ACC, T	1	1	—	C, N, Z
MOV	ACC, loc << shift	1	1	SXM	N, Z
MOV	ACC, #16BitSU<<shift	2	1	SXM	N, Z
MOV	ACC, P	1	1	PM, SXM	N, Z
MOV	loc, ACC << shift	1	1	—	N, Z
MOVB	ACC, #8bit	1	1	—	N, Z
MOVH	loc, ACC << shift	1	1	—	N, Z
MOVU	ACC, loc	1	1	—	N, Z
NEG	ACC	1	1	OVM	C, N, V, Z
NORM	ACC, aux++	1	4†	—	N, TC, Z
NORM	ACC, aux−−	1	4†	—	N, TC, Z
NOT	ACC	1	1	—	N, Z
ROL	ACC	1	1†	—	C, N, Z
ROR	ACO	1	1†	—	C, N, Z
SAT	ACC	1	1	OVC	C, N, OVC, V, Z
SBBU	ACC, loc	1	1	OVM	C, N, V, Z
SFR	ACC, shift	1	1	SXM	C, N, Z
SFR	ACC, T	1	1	SXM	C, N, Z
SUB	ACC, loc<< shift	1	1	OVM, SXM	C, N, V, Z
SUB	ACC, #16BitSU<< shift	2	1	OVM, SXM	C, N, V, Z
SUB	ACC, P	1	1	OVM, PM	C, N, V, Z
SUBB	ACC, #8bit	1	1	OVM	C, N, V, Z
SUBCU	ACC, loc	1	1†	—	C, N, OVC, V, Z
SUBU	ACC, loc	1	1	OVM	C, N, V, Z
TEST	ACC	1	1	N, Z
†This instruction is repeatable. The number of cycles given is for execution of a single instruction.

[0466]

TABLE 47
ACC 32-Bit Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
ADDL	ACC, locLong	1	1	OVM	C, N, V, Z
CMPL	ACC, locLong	1	1	—	C, N, Z
MOVL	ACC, locLong	1	1	—	N, Z
MOVL	locLong, ACC	1	1	—	—
SUBL	ACC, locLong	1	1	OVM	C, N, V, Z

[0467]

TABLE 48
Operations on Memory or Register
	Status Bits
				Affected
Instruction	Syntax	Words	Cycles	By	Affects
ADD	loc, AX	1	1	—	C, N, V, Z
ADD	loc,	2	1	—	C, N, V, Z
	#16BitSigned
AND	loc, AX	1	1	—	N, Z
AND	loc,	2	1	—	N, Z
	#16BitMask
CMP	loc,	2	1	—	C, N, Z
	#16BitSigned
DEC	Loc	1	1	—	C, N, V, Z
INC	Loc	1	1	—	C, N, V, Z
LOOPNZ	loc,	2	5	—	LOOP, N, Z
	#16BitMask
LOOPZ	loc,	2	5	—	LOOP, N, Z
	#16BitMask
OR	loc, AX	1	1	—	N, Z
OR	loc,	2	1	—	N, Z
	#16BitMask
SUB	loc, AX	1	1	—	C, N, V, Z
TBIT	loc,	1	1	—	TC
	#BitNumber
XOR	loc, AX	1	1	—	N, Z
XOR	loc,	2	1	—	N, Z
	#16BitMask

[0468]

TABLE 49
Data Move Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
MOV	*(0: 16bit), loc	2	2†	—	—
MOV	loc, #0	1	1†	—	N, Z
MOV	loc, #16bit	2	1†	—	N, Z
MOV	loc, *();16bit)	2	2†	—	N, Z
PREAD	loc, *XAR7	1	2†	—	N, Z
PWRITE	*XAR7, loc	1	5†	—	—
†This instruction is repeatable. The number of cycles given is for execution of a single instruction.

[0469]

TABLE 50
Program Flow Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
B	16BitOffset, cond	2	4/7†	—	V (if tested)
BANZ	16BitOffset, ARx−−	2	2/4†	—	—
CALL	22BitAddress	2	4	—	—
CALL *XAR7	1	4	—	—	—
FFC	XAR7, 22BitAddress	2	4	—	—
IRET		1	8	—	ARP, C, DBGM, INTM,
					N, OVC, OVM, PAGE0,
					PM, SPA, SXM, TC, V,
					VMAP, Z
LB	22BitAddress	2	4	—	—
LB	*XAR7	1	4	—	—
RET		1	8	—	—
RETE	1	8	—	INTM
SB	SBitOffset, cond	1	4/7†	V (if tested)
†X/Y: X cycles are required if the branch is not taken. Y cycles are required if the branch is taken.

[0470]

TABLE 51
Math Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
CMP	ACC, P	1	1	PM	C, N, Z
MAC	P, loc, 0:	2	1†	OVM, PM	C, N, V, Z
	pmem
MOV	PH, loc	1	1	—	—
MOV	PL, loc	1	1	—	—
MOV	P, ACC	1	1	—	—
MOV	T, loc	1	1	—	—
MOV	loc, P	1	1	PM	N, Z
MOV	loc, T	1	1	—	N, Z
MOVA	T, loc	1	1	OVM, PM	C, N, V, Z
MOVH	loc, P	1	1	PM	N, Z
MOVP	T, loc	1	1	PM	N, Z
MOVS	T, loc	1	1	OVM, PM	C, N, V, Z
MPY	ACC, loc,	2	1	N, Z
	#16BitSigned
MPY	ACC, T, loc	1	1	—	N, Z
MPY	P, T, loc	1	1	—	—
MPYA	P, loc,	2	1	OVM, PM	C, N, V, Z
	#16BitSigned
MPYA	P, T, loc	1	1	OVM, PM	C, N, V, Z
MPYB	ACC, T, #8bit	1	1	—	N, Z
MPYB	P, T, #8bit	1	1	—	—
MPYS	P, T, loc	1	1	OVM, PM	C, N, V, Z
MPYU	ACC, T, loc	1	1	—	N, Z
MPYU	P, T, loc	1	1	—	—
MPYXU	ACC, T, loc	1	1	—	N, Z
MPYXU	P, T, loc	1	1	—	—
†This instruction is repeatable. The number of cycles given is for execution of a single instruction.

[0471]

TABLE 52
Control Operations
	Status Bits
Instruction	Syntax	Words	Cycles	Affected By	Affects
AND	IER, #16BitMask	2	2	—	—
AND	IFR, #16BitMask	2	2	—	—
ASP		1	1	SPA	SPA
CLRC	BitName1	1	1 or 2‡	—	Specified status bit(s): C,
	{, BitName2}				DBGM, INTM, OVM,
					PAGE0, SXM, TC, VMAP
CLRC	8BitMask	1	1 or 2{	—	Specified status bit(s): C,
					DBGM, INTM, OVM,
					PAGE0, SXM, TC, VMAP
LACK	# VectorValue	2	1	—	—
IDLE		1	5	—	IDLESTAT
INTR	INTi	1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM, LOOP
INTR	DLOGINT	1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM, LOOP
INTR	RTOSINT	1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM, LOOP
INTR	NMI	1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM, LOOP
MOV	IER, loc	1	5	—	—
MOV	loc, IER	1	1	—	—
NASP		1	1	SPA	SPA
NOP		1	1†	—	—
NOP	*ind	1	1†	—	—
OR	IER, #16BitMask	2	2	—	—
OR	IFR, #16BitMask	2	2	—	—
RPT	loc	1	4	—	—
RPT	#8bit	1	1	—	—
SETC	BitName1	1	1 or 2‡	—	Specified status bit(s): C,
	{, BitName2}				DBGM, INTM, OVM,
					PAGE0, SXM, TC, VMAP
SETC	8BitMask	1	1 or 2‡	—	Specified status bit(s): C,
					DBGM, INTM, OVM,
					PAGE0, SXM, TC, VMAP
SPM	ShiftMode	1	1	—	PM
TRAP	# VectorNumber	1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM, LOOP
†This instruction is repeatable. The number of cycles given is for execution of a single instruction.
‡If CLRC or SETC is to affect the INTM bit and/or the DBGM bit, the instruction requires 2 cycles; otherwise the instruction requires 1 cycle.

[0472]

TABLE 53
Emulation Operations
			Status Bits
				Affected-
Instruction	Syntax	Words	Cycles	By	Affects
ABORTI		1	2	—	DBGM, HPI, IDS
EALLOW		1	4	—	EALLOW
EDIS		1	4	—	EALLOW
ESTOP0		1	1	—	ESTOP0
ESTOP1		1	1	—	ESTOP1
ITRAP0		1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM,
					LOOP
ITRAP1		1	8	—	DBGM, EALLOW,
					IDLESTAT, INTM,
					LOOP

[0473] The DSP core expects memory wrappers and peripheral-interface logic to align any 32-bit data read or write to an even address. If the address-generation logic generates an odd address, the memory wrapper or peripheral-inter-face must begin reading or writing at the previous even address. This alignment does not affect the address values generated by the address-generation logic.

[0474] Consider FIG. 41, which shows a 32-bit read from data memory. In this example, DP direct addressing mode is used. The data page pointer and the offset point to an odd address, 00 008516. However, the CPU must receive the first word from an even address. It reads from addresses 00 008416 and 00 008516. Then it loads the word at 00 008416 to the low half of ACC and the word at 00 008516 to the high half of ACC.

[0475] FIG. 42 shows a 32-bit write. AR0 points to the odd address 00 008516. Due to alignment, the low half of ACC is saved at address 00 008416 and the high half of ACC is saved at address 00 008516. AR0 is not modified; it still holds 85 after the operation.

[0476] Table 54 describes the mechanisms on the DSP that generate 32-bit-wide data accesses.

TABLE 54
Mechanisms That Generate
32-Bit-Wide Data Accesses
Mechanism                        32-Bit Operation(s)
Interrupt initiated by hardware or Multiple register pairs are
software (OR, INTR, or TRAP      automatically saved on the stack.
instruction).                    Each pair is saved in a single 32-bit
                                 store operation.
IRET instruction                 Multiple register pairs are
                                 automatically restored from the stack.
                                 Each pair is re-stored in a single
                                 32-bit load operation.
Any of the following syntaxes of A register pair is saved on the stack
the PUSH instruction:            in a single 32-bit store operation
PUSH T:ST0  PUSH AR1:AR0
PUSH DP:ST1 PUSH AR3:AR2
PUSH PH:PL  PUSH AR5:AR4
Any of the following syntaxes of A register pair is restored from the
the POP instruction:             stack in a single 32-bit load
POP T:ST0  POP AR1:AR0           operation.
POP DP:ST1 POP AR8:AR2
POP PH:PL  POP AR5:AR4
Either of the following syntaxes of When XAR6 or XAR7 is the source,
the MOV instruction:             this 22-bit value is stored in a
MOV locLong, XAR n               single 32-bit store operation. The
MOV XAR n, locLong               6 MSBs of the value are stored as 0s.
                                 When XAR6 or XAR7 is loaded, 32
                                 bits are read from data memory. The
                                 22 LSBs are loaded to the auxiliary
                                 register.
MOVL instruction                 ACC is loaded or stored using a
                                 single 32-bit operation.
ADDL, CMPL, or                   A 32-bit value is read from data
SUBL instruction                 memory before being added or
                                 subtracted from ACC.

[0477] Appendix B contains detailed information on the instruction set. The description for each instruction presents the following categories of information: Syntax, Operands, Description, Execution, Status Bits, Repeat and Words. Some descriptions also include a category that shows one or more examples.

[0478] Each instruction begins with a list of the available assembler syntax expressions. For example, the description for the ADD instruction begins with:

1:	ADD	AX, loc
2:	ADD	loc,AX
3:	ADD	ACC,loc << shift3}
4:	ADD	ACC, #16BitSU << shift2}
5:	ADD	ACC, P
6:	ADD	loc, #16BitSigned

[0479] Each syntax is preceded by a reference number that is used throughout the rest of the instruction description. These are the notations used in the syntax expressions:

[0480] Italic symbols

[0481] and

[0482] boldface characters

[0483] Italic symbols in an instruction syntax represent variables.

[0484] Boldface characters in an instruction syntax must be typed as shown.

[0485] Example: ADD A X, loc

[0486] The X indicates that AH or AL may be used for the first operand. A variety of values for loc may be used. ADD, the following A, and the comma are boldface to indicate that they are always present when using this syntax. Samples with this syntax follow:

[0487] ADD AH, @T

[0488] ADD AL, *AR2++

[0489] {x} Operand x is optional.

[0490] Example: ADD ACC, loc<<shift3}

[0491] Supply ACC and loc, as in this instruction:

[0492] ADD ACC, @2

[0493] The option of adding a shift3 value is as in this instruction:

[0494] ADD ACC, @2<<4

[0495] # The symbol # is a prefix for constants used in immediate-constant addressing mode.

[0496] Example: ADD loc, #16 BitSigned

[0497] The variable 16 BitSigned is preceded by the symbol # to indicate that a constant must be entered. Similarly, because loc is not preceded by the symbol #, it cannot be a constant. It must be a reference to a memory location or a register.

[0498] Finally, consider this code example:

StoreResult MOV	*,	AL	;Store low word of accumulator
			;to location referenced by the
			;current auxiliary resister.

[0499] An optional label, StoreResult, is used as a reference in front of the instruction mnemonic. Place labels either before the instruction mnemonic on the same line or on the preceding line in the first column. (Be sure there are no spaces in the labels.) An optional comment field can conclude the syntax expression. At least one space is required between fields (label, mnemonic, operand, and comment).

[0500] Operands can be constants, or assembly-time expressions referring to memory, registers, shift counts, and a variety of other values. The operands category for each instruction description defines the variables and register names used for and/or within operands in the syntax expressions. For example, for the ADD instruction, the syntax category gives these syntax expressions:

[0501] ADD A X, loc

[0502] ADD loc, A X

[0503] ADD ACC, loc<<shift3}

[0504] ADD ACC, #16 BitSU<<shift2}

[0505] ADD ACC, P

[0506] ADD loc, #16 BitSigned

[0507] The operands category defines 16 BitSigned, 16 BitSU, ACC, A X, loc, shift3, and shift2.

[0508] The opcode category breaks down the various bit fields that make up each instruction word. When one of the fields contains a constant value derived directly from an operand, it will have the same name as that operand. For example, FIG. 43 shows one of the opcodes given for the ADD instruction.

[0509] The field 16 BitSigned relates directly to an operand defined in the operands category. Bits 7 through 0 of the first word depend on the addressing mode used.

[0510] The description category explains what happens during instruction execution and its effect on the rest of the processor or on memory contents. It also discusses any constraints on the operands imposed by the processor or the assembler. This description parallels and supplements the information given in the execution category.

[0511] The execution category presents an instruction operation sequence that describes the processing that takes place when the instruction is executed. Here are notations used in the execution category:

[0512] [r] The content of register or location r.

[0513] Example: [ACC] represents the value in the accumulator.

[0514] [x]→ y The content of x is assigned to register or location y.

[0515] Example: [addressed location] “ACC

[0516] The content of the specified data-memory location or register is put into the accumulator.

[0517] [x]<< z The content of x is shifted left by z bit positions.

[0518] Example: [ACC] <<16

[0519] The content of ACC is shifted left by 16.

[0520] [x] >>z The content of x is shifted right by z bit positions.

[0521] Example: [ACC] >>1

[0522] The content of ACC is shifted right by 1.

[0523] r(n) Bit n of register or location r.

[0524] Example: ACC(31) represents bit 31 of the accumulator.

[0525] r(n:m) Bits n through m of register or location r.

[0526] Example: ACC(15:0) represents bits 15 through 0 of the accumulator.

[0527] [r(n:m)] The content of bits n through m of register or location r.

[0528] Example: [ACC(31:16)] represents the content of bits 31 through 16 of the accumulator.

[0529] r1:r2 Registers r1 and r2 have been concatenated into one 32-bit data value. r2 is the low word; r1 is the high word.

[0530] Example: T:ST0 represents the concatenation of the T register and status register ST0. T forms the high word, and ST0 forms the low word.

[0531] The bits in status registers ST0 and ST1 affect the operation of certain instructions and are affected by certain instructions. The bits in the debug status register (DBGSTAT) are also affected by certain instructions. The status bits category of each instruction description states how certain bits (if any) affect the execution of the instruction and how certain bits (if any) are affected by the instruction.

[0532] Some of the instructions are repeatable. That is, if they are preceded by the RPT instruction, the CPU repeats them the number of times specified in the RPT instruction. The repeat category indicates whether the instruction is repeatable. If an instruction is repeatable, the repeat category provides tips for using this feature.

[0533] The words category specifies the number of words (one or two) required to store the instruction in program memory. When the number of words depends on the chosen syntax, the words category specifies which syntaxes require one word and which require two words. This category, when present, provides examples that show the effects of the instruction on memory and/or registers. Program code is shown in a special typeface. The sample code is then followed by a verbal or graphical description of the effect of that code.

Other Embodiments

[0534] Although the present invention and its advantages have been described in detail, it should be understood that various changes, substitutions and alterations can be made herein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A data processing device, comprising: program control unit for storing and then decoding instructions; program address generation unit for generating addresses used to obtain instructions; address register arithmetic unit for generating addresses for data; arithmetic logic unit for performing operations on data; shifter unit for shifting data in response to a predetermined instruction; multiplier unit for performing multiplication of two input values; and a plurality of registers of which at least a portion are individually selectively associated with one or more of said units as a function of an instruction.