Microprocessor with non-aligned memory access

Abstract

A data processing system (1300) is provided with a digital signal processor (DSP) (1301) that has an instruction set architecture (ISA) that is optimized for intensive numeric algorithm processing. The DSP has dual load/store units (.D1, .D2) connected to dual memory ports (T1, T2) in a level one data cache memory controller (1720a). The DSP can execute two aligned data transfers each having a length of one byte, two bytes, four bytes, or eight bytes in parallel by executing two load/store instructions. The DSP can also execute a single non-aligned data transfer having a length of four bytes or eight bytes by executing a non-aligned load/store instruction that utilizes both memory target ports.

Description

NOTICE

(C) Copyright 2000 Texas Instruments Incorporated. A portion of the disclosure of this patent document contains material which is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent disclosure, as it appears in the Patent and Trademark Office patent file or records, but otherwise reserves all copyright rights whatsoever.

TECHNICAL FIELD OF THE INVENTION

This invention relates to data processing devices, electronic processing and control systems and methods of their manufacture and operation, and particularly relates to memory access schemes of microprocessors optimized for digital signal processing.

BACKGROUND OF THE INVENTION

Generally, a microprocessor is a circuit that combines the instruction-handling, arithmetic, and logical operations of a computer on a single semiconductor integrated circuit. Microprocessors can be grouped into two general classes, namely general-purpose microprocessors and special-purpose microprocessors. General-purpose microprocessors are designed to be programmable by the user to perform any of a wide range of tasks, and are therefore often used as the central processing unit (CPU) in equipment such as personal computers. Special-purpose microprocessors, in contrast, are designed to provide performance improvement for specific predetermined arithmetic and logical functions for which the user intends to use the microprocessor. By knowing the primary function of the microprocessor, the designer can structure the microprocessor architecture in such a manner that the performance of the specific function by the special-purpose microprocessor greatly exceeds the performance of the same function by a general-purpose microprocessor regardless of the program implemented by the user.

One such function that can be performed by a special-purpose microprocessor at a greatly improved rate is digital signal processing. Digital signal processing generally involves the representation, transmission, and manipulation of signals, using numerical techniques and a type of special-purpose microprocessor known as a digital signal processor (DSP). Digital signal processing typically requires the manipulation of large volumes of data, and a digital signal processor is optimized to efficiently perform the intensive computation and memory access operations associated with this data manipulation. For example, computations for performing Fast Fourier Transforms (FFTs) and for implementing digital filters consist to a large degree of repetitive operations such as multiply-and-add and multiple-bit-shift. DSPs can be specifically adapted for these repetitive functions, and provide a substantial performance improvement over general-purpose microprocessors in, for example, real-time applications such as image and speech processing.

DSPs are central to the operation of many of today's electronic products, such as high-speed modems, high-density disk drives, digital cellular phones, complex automotive systems, and video-conferencing equipment. DSPs will enable a wide variety of other digital systems in the future, such as video-phones, network processing, natural speech interfaces, and ultra-high speed modems. The demands placed upon DSPs in these and other applications continue to grow as consumers seek increased performance from their digital products, and as the convergence of the communications, computer and consumer industries creates completely new digital products.

Microprocessor designers have increasingly endeavored to exploit parallelism to improve performance. One parallel architecture that has found application in some modern microprocessors utilizes multiple instruction fetch packets and multiple instruction execution packets with multiple functional units, referred to as a Very Long Instruction Word (VLIW) architecture.

Digital systems designed on a single integrated circuit are referred to as an application specific integrated circuit (ASIC). MegaModules are being used in the design of ASICs to create complex digital systems a single chip. (MegaModule is a trademark of Texas Instruments Incorporated.) Types of MegaModules include SRAMs, FIFOs, register files, RAMs, ROMs, universal asynchronous receiver-transmitters (UARTs), programmable logic arrays and other such logic circuits. MegaModules are usually defined as integrated circuit modules of at least 500 gates in complexity and having a complex ASIC macro function. These MegaModules are predesigned and stored in an ASIC design library. The MegaModules can then be selected by a designer and placed within a certain area on a new IC chip.

Designers have succeeded in increasing the performance of DSPs, and microprocessors in general, by increasing clock speeds, by removing data processing bottlenecks in circuit architecture, by incorporating multiple execution units on a single processor circuit, and by developing optimizing compilers that schedule operations to be executed by the processor in an efficient manner. For example, non-aligned data access is provided on certain microprocessors. Complex instruction set computer (CISC) architectures (Intel, Motorola 68K) have thorough support for non-aligned data accesses; however, reduced instruction set computer (RISC) architectures usually do not support non-aligned accesses with a single load or store instruction. Some RISC architectures allow two data accesses per cycle, but they allow only two aligned accesses. Certain CISC machines now allow doing two memory accesses per cycle as two non-aligned accesses. A reason for this is that the dual access implementations are superscalar implementations that are running code compatible with earlier scalar implementations.

The increasing demands of technology and the marketplace make desirable even further structural and process improvements in processing devices, application systems and methods of operation and manufacture.

SUMMARY OF THE INVENTION

An illustrative embodiment of the present invention seeks to provide a microprocessor and a method for accessing memory by a microprocessor that improves digital signal processing performance. Aspects of the invention are specified in the claims.

In an embodiment of the present invention, each .D unit of a DSP can load and store data items up to double words (up to 64 bits) at aligned addresses in a two port memory subsystem. The .D units can also access 10 words and double words on any byte boundary, whether aligned or not. The address generation circuitry in the first .D unit has a first address output connected to the first memory port and a second address output selectively connected to the second memory port. The address generation circuitry can provide two addresses simultaneously to request two aligned data items. An extraction circuit is connected to the memory subsystem to provide a non-aligned data item extracted from two adjacent aligned data items requested by the .D unit.

In another embodiment of the present invention, the address generation circuitry in the .D units is operable to form an address for non-aligned double word instructions by combining a base address value and an offset value.

In another embodiment of the invention, one or more additional .D units have similar addressing circuitry for non-aligned accesses.

In another embodiment of the present invention, two .D units can simultaneously access aligned data items in the memory.

In another embodiment of the present invention, the memory subsystem has a plurality of memory banks connected to the extraction circuitry. Decode circuitry is connected to the first memory port and to the second memory port for receiving addresses. A plurality of address multiplexers are connected respectively to the plurality of memory banks such that the decode circuitry is operable to individually control each of the plurality of address multiplexers.

BRIEF DESCRIPTION OF THE DRAWINGS

Other features and advantages of the present invention will become apparent by reference to the following detailed description when considered in conjunction with the accompanying drawings in which the Figures relate to the processor of FIG. 1 unless otherwise stated, and in which:

FIG. 1 is a block diagram of a digital system with a digital signal processor (DSP), showing components thereof pertinent to an embodiment of the present invention;

FIG. 2 is a block diagram of the functional units, data paths and register files of the DSP;

FIG. 3A illustrates an opcode map for the load/store instructions which are executed in a .D unit of the DSP;

FIG. 3B illustrates an opcode map for the load/store non-aligned double word instruction which are executed in a .D unit of the DSP;

FIG. 3C illustrates an opcode map for Boolean instructions executed by the .D units;

FIG. 4 illustrates an addressing mode register (AMR) of the DSP;

FIGS. 5A , 5 B and 5 C illustrate aspects of non-aligned address formation and non-aligned data extraction from a circular buffer region, according to an aspect of the present invention;

FIG. 6 illustrates the basic format of a fetch packet of the DSP;

FIG. 7 is a memory map of a portion of the memory space of the DSP and illustrates various aligned and non-aligned memory accesses;

FIG. 8 is a block diagram illustrating D-unit address buses of the DSP in more detail and illustrating the two ports of the DSP memory;

FIG. 9 is a block diagram of the memory of FIG. 8 illustrating address decoding of the two address ports and byte selection circuitry to extract a non-aligned data item according to an embodiment of the present invention;

FIG. 10 is a block diagram illustrating the extraction circuitry of FIG. 9 in more detail;

FIG. 11 is a block diagram illustrating the store byte selection circuitry for storing non-aligned data items in the memory system FIG. 8 in more detail;

FIG. 12A is a more detailed block diagram of the D-unit of the DSP;

FIG. 12B is a more detailed block diagram of the circular buffer circuitry of FIG. 12A ;

FIG. 12C is a flow chart illustrating formation of circular buffer addresses for both aligned access instruction types and non-aligned access instruction types, according to an aspect of the present invention;

FIG. 13 is a block diagram of an alternative embodiment of the present invention in digital system having a DSP with a data cache;

FIGS. 14A and 14B together is a block diagram of the data cache of FIG. 13 ; and

FIG. 15 illustrates an exemplary implementation of a digital system that includes an embodiment of the present invention in a mobile telecommunications device.

Corresponding numerals and symbols in the different figures and tables refer to corresponding parts unless otherwise indicated.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

FIG. 1 is a block diagram of a microprocessor 1 that has an embodiment of the present invention. Microprocessor 1 is a RISC VLIW digital signal processor (“DSP”). In the interest of clarity, FIG. 1 only shows those portions of microprocessor 1 that are relevant to an understanding of an embodiment of the present invention. Details of general construction for DSPs are well known, and may be found readily elsewhere. For example, U.S. Pat. No. 5,072,418 issued to Frederick Boutaud, et al, describes a DSP in detail and is incorporated herein by reference. U.S. Pat. No. 5,329,471 issued to Gary Swoboda, et al, describes in detail how to test and emulate a DSP and is incorporated herein by reference. Details of portions of microprocessor 1 relevant to an embodiment of the present invention are explained in sufficient detail hereinbelow, so as to enable one of ordinary skill in the microprocessor art to make and use the invention.

In microprocessor 1 there are shown a central processing unit (CPU) 10 , data memory 22 , program memory/cache 23 , peripherals 60 and an external memory interface (EMIF) with a direct memory access (DMA) 61 . CPU 10 further has an instruction fetch/decode unit 10 a-c , a plurality of execution units, including an arithmetic and load/store unit D 1 , a multiplier M 1 , an ALU/shifter unit S 1 , an arithmetic logic unit (“ALU”) L 1 , a shared multi-port register file 20 a from which data are read and to which data are written. Instructions are fetched by fetch unit 10 a from instruction memory 23 over a set of busses 41 . Decoded instructions are provided from the instruction fetch/decode unit 10 a-c to the functional units D 1 , M 1 , S 1 , and L 1 over various sets of control lines which are not shown. Data are provided to/from the register file 20 a from/to to load/store units D 1 over a first set of busses 32 a , to multiplier M 1 over a second set of busses 34 a , to ALU/shifter unit S 1 over a third set of busses 36 a and to ALU L 1 over a fourth set of busses 38 a . Data are provided to/from the memory 22 from/to the load/store units D 1 via a fifth set of busses 40 a . Note that the entire data path described above is duplicated with register file 20 b and execution units D 2 , M 2 , S 2 , and L 2 . In this embodiment of the present invention, two unrelated aligned double word (64 bits) load/store transfers can be made in parallel between CPU 10 and data memory 22 on each clock cycle using bus set 40 a and bus set 40 b.

An aligned transfer is one in which the address of a datum modulo its size (in addressible units) is 0. For example, in a system where each 8-bit datum has an address, a 16-bit datum (2 addressible units) at address 0x24002576 is aligned because 0x24002576 mod 2 is 0. Converesly, a non-aligned transfer is one in which the address of a datum modulo its size (in addressible units) is non-zero. For example, in a system where each 8-bit datum has an address, a 32-bit datum (4 addressible units) at address 0x24F78006 is non-aligned because 0x24F78006 mod 4 is 2, not zero.

A single non-aligned double word load/store transfer is performed by scheduling a first .D unit resource and two load/store ports on memory 22 . Advantageously, an extraction circuit is connected to the memory subsystem to provide a non-aligned data item extracted from two aligned data items requested by the .D unit. Advantageously, a second .D unit can perform 32-bit logical or arithmetic instructions in addition to the .S and .L units while the address port of the second .D unit is being used to transmit one of two contiguous addresses provided by the first .D unit. Furthermore, a non-aligned access near the end of a circular buffer region in the target memory provides a non-aligned data item that wraps around to the other end of the circular buffer.

Emulation circuitry 50 provides access to the internal operation of integrated circuit 1 that can be controlled by an external test/development system (XDS) 51 . External test system 51 is representative of a variety of known test systems for debugging and emulating integrated circuits. One such system is described in U.S. Pat. No. 5,535,331 which is incorporated herein by reference. Test circuitry 52 contains control registers and parallel signature analysis circuitry for testing integrated circuit 1 .

Note that the memory 22 and memory 23 are shown in FIG. 1 to be a part of a microprocessor 1 integrated circuit, the extent of which is represented by the box 42 . The memories 22 - 23 could just as well be external to the microprocessor 1 integrated circuit 42 , or part of it could reside on the integrated circuit 42 and part of it be external to the integrated circuit 42 . These are matters of design choice. Also, the particular selection and number of execution units are a matter of design choice, and are not critical to the invention.

When microprocessor 1 is incorporated in a data processing system, additional memory or peripherals may be connected to microprocessor 1 , as illustrated in FIG. 1 . For example, Random Access Memory (RAM) 70 , a Read Only Memory (ROM) 71 and a Disk 72 are shown connected via an external bus 73 . Bus 73 is connected to the External Memory Interface (EMIF) which is part of functional block 61 within microprocessor 1 . A Direct Memory Access (DMA) controller is also included within block 61 . The DMA controller is generally used to move data between memory and peripherals within microprocessor 1 and memory and peripherals which are external to microprocessor 1 .

In the present embodiment, CPU core 10 is encapsulated as a MegaModule, however, other embodiments of the present invention may be in custom designed CPU's or mass market microprocessors, for example.

A detailed description of various architectural features of the microprocessor 1 of FIG. 1 is provided in coassigned U.S. Pat. No. 6,182,203 and is incorporated herein by reference. A description of enhanced architectural features and an extended instruction set not described herein for CPU 10 is provided in coassigned U.S. patent application Ser. No. 09/703,096 Microprocessor with Improved Instruction Set Architecture and is incorporated herein by reference.

FIG. 2 is a block diagram of the execution units and register files of the microprocessor of FIG. 1 and shows a more detailed view of the buses connecting the various functional blocks. In this figure, all data busses are 32 bits wide, unless otherwise noted. There are two general-purpose register files (A and B) in the processor's data paths. Each of these files contains 32 32-bit registers (A 0 -A 31 for file A and B 0 -B 31 for file B). The general-purpose registers can be used for data, data address pointers, or condition registers. Any number of reads of a given register can be performed in a given cycle.

The general-purpose register files support data ranging in size from packed 8-bit data through 64-bit fixed-point data. Values larger than 32 bits, such as 40-bit long and 64-bit double word quantities, are stored in register pairs, with the 32 LSBs of data placed in an even-numbered register and the remaining 8 or 32 MSBs in the next upper register (which is always an odd-numbered register). Packed data types store either four 8-bit values or two 16-bit values in a single 32-bit register.

There are 32 valid register pairs for 40-bit and 64-bit data, as shown in Table 1. In assembly language syntax, a colon between the register names denotes the register pairs and the odd numbered register is specified first.

Operations requiring a long input ignore the 24 MSBs of the odd register. Operations producing a long result zero-fill the 24 MSBs of the odd register. The even register is encoded in the opcode.

All eight of the functional units have access to the opposite side's register file via a cross path. The .M 1 , .M 2 , .S 1 , .S 2 , .D 1 and .D 2 units' src 2 inputs are selectable between the cross path and the same side register file by appropriate selection of multiplexers 213 , 214 and 215 , for example. In the case of the .L 1 and .L 2 both src 1 and src 2 inputs are also selectable between the cross path and the same-side register file by appropriate selection of multiplexers 211 , 212 , for example.

TABLE 1
40-Bit/64-Bit Register Pairs
Register Files
A       B
A1:A0   B1:B0
A3:A2   B3:B2
A5:A4   B5:B4
A7:A6   67:B6
A9:A8   B9:B8
A11:A10 B11:B10
A13:A12 B13:B12
A15:A14 B15:B14
A17:A16 B17:B16
A19:A18 B19:B18
A21:A20 B21:B20
A23:A22 B23:B22
A25:A24 B25:B24
A27:A26 B27:B26
A29:A28 B29:B28
A31:A30 B31:B30

Referring again to FIG. 2 , the eight functional units in processor 10 's data paths can be divided into two groups of four; each functional unit in one data path is almost identical to the corresponding unit in the other data path. The functional units are described in Table 2.

TABLE 2
Functional Units and Operations Performed
Functional Unit           Fixed-Point Operations
.L unit (.L1, .L2), 18a,b 32/40-bit arithmetic and compare operations
                          32-bit logical operations
                          Leftmost 1 or 0 counting for 32 bits
                          Normalization count for 32 and 40 bits
                          Byte shifts
                          Data packing/unpacking
                          5-bit constant generation
                          Paired 16-bit arithmetic operations
                          Quad 8-bit arithmetic operations
                          Paired 16-bit min/max operations
                          Quad 8-bit min/max operations
.S unit (.S1, .S2) 16a,b  32-bit arithmetic operations
                          32/40-bit shifts and 32-bit bit-field operations
                          32-bit logical operations
                          Branches
                          Constant generation
                          Register transfers to/from control register
                          file (.S2 only)
                          Byte shifts
                          Data packing/unpacking
                          Paired 16-bit compare operations
                          Quad 8-bit compare operations
                          Paired 16-bit shift operations
                          Paired 16-bit saturated arithmetic operations
                          Quad 8-bit saturated arithmetic operations
.M unit (.M1, .M2) 14a,b  16 × 16 multiply operations
                          16 × 32 multiply operations
                          Bit expansion
                          Bit interleaving/de-interleaving
                          Quad 8 × 8 multiply operations
                          Paired 16 × 16 multiply operations
                          Paired 16 × 16 multiply with add/subtract
                          operations
                          Quad 8 × 8 multiply with add operations
                          Variable shift operations
                          Rotation
                          Galois Field Multiply
.D unit (.D1, .D2) 12a,b  32-bit add, subtract, linear and circular
                          address calculation
                          Loads and stores with 5-bit constant offset
                          Loads and stores with 15-bit constant offset
                          (.D2 only)
                          Load and store double words with 5-bit
                          constant
                          Load and store non-aligned words and
                          double words
                          5-bit constant generation
                          32-bit logical operations

Most data lines in the CPU support 32-bit operands, and some support long (40-bit) and double word (64-bit) operands. Each functional unit has its own 32-bit write port into a general-purpose register file 20 a , 20 b (Refer to FIG. 2 ). All units ending in 1 (for example, .L 1 ) write to register file A 20 a and all units ending in 2 write to register file B 20 b . Each functional unit has two 32-bit read ports for source operands src 1 and src 2 . Four units (.L 1 , .L 2 , .S 1 , and .S 2 ) have an extra 8-bit-wide port (long-dst) for 40-bit long writes, as well as an 8-bit input (long-src) for 40-bit long reads. Because each unit has its own 32-bit write port dst, when performing 32 bit operations all eight units can be used in parallel every cycle. Since each multiplier can return up to a 64-bit result, two write ports (dst 1 and dst 2 ) are provided from the multipliers to the register file.

Memory, Load and Store Paths

Processor 10 supports double word loads and stores. There are four 32-bit paths for loading data from memory to the register file. For side A, LD 1 a is the load path for the 32 LSBs (least significant bits); LD 1 b is the load path for the 32 MSBs (most significant bits). For side B, LD 2 a is the load path for the 32 LSBs; LD 2 b is the load path for the 32 MSBs. There are also four 32-bit paths, for storing register values to memory from each register file. ST 1 a is the write path for the 32 LSBs on side A; ST 1 b is the write path for the 32 MSBs for side A. For side B, ST 2 a is the write path for the 32 LSBs; ST 2 b is the write path for the 32 MSBs.

The ports for long and double word operands are shared between the S and L functional units. This places a constraint on which long or double word operations can be scheduled on a datapath in the same execute packet.

Data Address Paths

Bus 40 a has an address bus DA 1 which is driven by mux 200 a . This allows an address generated by either load/store unit D 1 or D 2 to provide a memory address for loads or stores for register file 20 a . Data Bus LD 1 a,b loads data from an address in memory 22 specified by address bus DA 1 to a register in register file 20 a . Likewise, data bus ST 1 a,b stores data from register file 20 a to memory 22 . Load/store unit D 1 performs the following operations: 32-bit add, subtract, linear and circular address calculations. Load/store unit D 2 operates similarly to unit D 1 , with the assistance of mux 200 b for selecting an address.

The DA 1 and DA 2 address resources and their associated data paths are connected to target memory ports on memory 22 specified as T 1 and T 2 respectively. T 1 connects to the DA 1 address path and the LD 1 a , LD 1 b , ST 1 a and ST 1 b data paths. Similarly, T 2 connects to the DA 2 address path and the LD 2 a , LD 2 b , ST 2 a and ST 2 b data paths. The T 1 and T 2 designations appear in functional unit fields for load and store instructions.

For example, the following load instruction uses the .D 1 unit to generate the address but is using the LD 2 a path resource with DA 2 address bus connected to target port T 2 to place the data in the B register file:

LDW .D 1 T 2 *A 0 [ 3 ], B 1 .

The use of the DA 2 address resource is indicated with the T 2 designation.

Instruction Syntax

An instruction syntax is used to describe each instruction. An opcode map breaks down the various bit fields that make up each instruction. There are certain instructions that can be executed on more than one functional unit. The syntax specifies the functional unit and various resources used by an instruction, and typically has a form as follows: operation, .unit, src, dst.

src and dst indicate source and destination operands respectively. The .unit dictates which functional unit the instruction is mapped to (.L 1 , .L 2 , .S 1 , .S 2 , .M 1 , .M 2 , .D 1 , or .D 2 ). Several instructions have three opcode operand fields: src 1 , src 2 , and dst.

FIG. 3A illustrates an opcode map for the load/store instructions which are executed in a .D unit of the DSP. FIG. 3B illustrates an opcode map for non-aligned double word load/store instructions, and FIG. 3C illustrates an opcode map for Boolean instructions executed by the .D units. Table 3 lists the opcodes for various load store (LD/ST) instructions performed by the CPU of the present embodiment. Opcode field 510 and R-field 512 define the operation of the LD/ST instructions. An aspect of the present invention is that processor 10 performs non-aligned load and store instructions by using resources of one D unit and both target ports T 1 and T 2 , as will be described in more detail below. Advantageously, the second D unit is available to execute a Boolean or arithmetic instruction in parallel with the execution of a non-aligned load/store instruction.

The dst field of the LD/STNDW instruction selects a register pair, a consecutive even-numbered and odd-numbered register pair from the same register file. The instruction can be used to load a pair of 32-bit integers. The least significant 32 bits are loaded into the even-numbered register and the most significant 32 bits are loaded into the next register (which is always an odd-numbered register).

TABLE 3
Load/Store Instruction Opcodes
R-Opcode	LD/ST
extension	Op	Instruction	Size	Alignment
0	000	LDHU	Half word unsigned	Half word
0	001	LDBU	Byte unsigned	Byte
0	010	LDB	Byte	Byte
0	011	STB	Byte	Byte
0	100	LDH	Half word	Half
0	101	STH	Half word	Half word
0	110	LDW	Word	Word
0	111	STW	Word	Word
1	010	LDNDW	Double word	byte
				(non-aligned)
1	011	LDNW	Word	Byte
				(non-aligned)
1	100	STDW	Double word	Double word
1	101	STNW	Word	Byte
				(non-aligned)
1	110	LDDW	Double word	Double word
1	111	STNDW	Double word	Byte
				(non-aligned)

Addressing Modes

The addressing modes are linear, circular using block size field BK 0 , and circular using block size field BK 1 . Eight registers can perform circular addressing. A 4 -A 7 are used by the .D 1 unit and B 4 -B 7 are used by the .D 2 unit. No other units can perform circular addressing modes. For each of these registers, an addressing mode register (AMR) contained in control register file 102 specifies the addressing mode. The block size fields are also in the AMR.

Referring again to FIGS. 3A and 3B , linear mode addressing simply shifts the offsetR/cst operand 516 to the left by 3, 2, 1, or 0 for double word, word, half-word, or byte access respectively and then performs an add or subtract to baseR 514 , depending on the address mode specified. For the pre-increment, pre-decrement, positive offset, and negative offset address generation options, the result of the calculation is the address to be accessed in memory. For post-increment or post-decrement addressing, the value of baseR before the addition or subtraction is the address to be accessed from memory. Address modes are specified by mode field 500 and listed in Table 4. The increment/decrement mode controls whether the updated address is written back to the register file. Otherwise, it is rather similar to offset mode. The pre-increment and offset modes differ only in whether the result is written back to “base”. The post-increment mode is similar to pre-increment (e.g. the new address is written to “base”), but differs in that the old value of “base” is used as the address for the access. The same applies for negative offset vs. decrement mode.

TABLE 4
Address Generator Options
Mode
Field	Syntax	Modification Performed
0 1 0 1	*+R[ offsetR]	Positive offset; addr = base + offset * scale
0 1 0 0	*−R[ offsetR]	Negative offset; addr = base − offset * scale
1 1 0 1	*+ +R[ offsetR]	Preincrement; addr = base + offset * scale;
		base = addr
1 1 0 0	*− −R[ offsetR]	Predecrement; addr = base − offset * scale;
		base = addr
1 1 1 1	*R+ +[ offsetR]	Postincrement; addr = base; base = base +
		offset * scale
1 1 1 0	*R− −[ offsetR]	Postdecrement; addr = base; base = base −
		offset * scale
0 0 0 1	*+R[ ucst5]	Positive offset; addr = base + offset * scale
0 0 0 0	*−R[ ucst5]	Negative offset; addr = base − offset * scale
1 0 0 1	*+ +R[ ucst5]	Preincrement; addr = base + offset * scale;
		base = addr
1 0 0 0	*− −R[ ucst5]	Predecrement; addr = base − offset * scale;
		base = addr
1 0 1 1	*R+ +[ ucst5]	Postincrement; addr = base; base = base +
		offset * scale
1 0 1 0	*R− −[ ucst5]	Postdecrement; addr = base; base = base −
		offset * scale

FIG. 4 illustrates the addressing mode register, (AMR), which is included in control register file 102 , is accessible via a “move between control file and the register file” (MVC) instruction. Eight registers (A 4 -A 7 , B 4 -B 7 ) can perform circular addressing. For each of these registers, the AMR specifies the addressing mode. A 2-bit field for each register is used to select the address modification mode: linear (the default) or circular mode. With circular addressing, the field also specifies which BK (block size) field to use for a circular buffer. In this embodiment, the buffer must be aligned on a byte boundary equal to the block size. The mode select field encoding is shown in Table 5.

TABLE 5
Addressing Mode Field Encoding
Mode Description
00   Linear modification (default at reset)
01   Circular addressing using the BK0 field
10   Circular addressing using the BK1 field
11   Reserved

The block size fields, BK 0 and BK 1 , specify block sizes for circular addressing. The five bits in BK 0 and BK 1 specify the width. The formula for calculating the block size width is:

Block size (in bytes)=2 (N+1)

where N is the value in BK 1 or BK 0

Table 6 shows block size calculations for all 32 possibilities.

TABLE 6
Block Size Calculations
N     Block Size
00000 2
00001 4
00010 8
00011 16
00100 32
00101 64
00110 128
00111 256
01000 512
01001 1?024
01010 2?048
01011 4?096
01100 8?192
01101 16?384
01110 32?768
01111 65?536
10000 131,072
10001 262,144
10010 524,288
10011 1,048,576
10100 2,097,152
10101 4,194,304
10110 8,388,608
10111 16,777,216
11000 33,554,432
11001 67,108,864
11010 134,217,728
11011 268,435,456
11100 536,870,912
11101 1,073,741,824
11110 2,147,483,648
11111 4,294,967,296
Note: when N is 11111, the behavior is identical to linear addressing

Circular mode addressing uses the BK 0 and BK 1 fields in the AMR to specify block sizes for circular addressing. Circular mode addressing operates as follows with LD/ST Instructions: after shifting offsetR/cst to the left by 3, 2, 1, or 0 for LDW, LDH, or LDB respectively, and is then added to or subtracted from baseR to produce the final address. This add or subtract is performed by only allowing bits N through 0 of the result to be updated, leaving bits 31 through N+1 unchanged after address arithmetic. The resulting address is bounded to 2{circumflex over ( )}(N+1) range, regardless of the size of the offsetR/cst.

The circular buffer size in the AMR is not scaled; for example: a size of 8 is 8 bytes, not 8× size of (type). So, to perform circular addressing on an array of 8 words, a size of 32 should be specified, or N=4. Table 7 shows an example LDW instructions performed with register A 4 in circular mode, with BK 0 =4, so the buffer size is 32 bytes, 16 halfwords, or 8 words. The value put in the AMR for this example is 00040001h. In this example, an offset of “9” is specified. 9h (hexadecimal) words is 24h bytes. 24h bytes is 4 bytes beyond the 32-byte (20h) boundary 100h−11Fh; thus, it is wrapped around to (124h−20h=104h).

TABLE 7
LDW in Circular Mode
LDW .D1 *+ +A4[9],A1
Before LDW	1 cycle after LDW	5 cycles after LDW
A4	0000 0100h	A4	0000 0104h	A4	0000 0104h
A1	XXXX XXXXh	A1	XXXX XXXXh	A1	1234 5678h
Mem	1234 5678h	mem	1234 5678h	mem	1234 5678h
104h		104h		104h

Non-Aligned Memory Access Considerations

Circular addressing may be used with non-aligned accesses. When circular addressing is enabled, address updates and memory accesses occur in the same manner as for the equivalent sequence of byte accesses. The only restriction is that the circular buffer size be at least as large as the data size being accessed. Non-aligned access to circular buffers that are smaller than the data being read produce undefined results.

Non-aligned accesses to a circular buffer apply the circular addressing calculation to logically adjacent memory addresses. The result is that non-aligned accesses near the boundary of a circular buffer will correctly read data from both ends of the circular buffer, thus seamlessly causing the circular buffer to “wrap around” at the edges.

FIGS. 5A , 5 B and 5 C illustrate aspects of non-aligned address formation and non-aligned data extraction from a circular buffer region, according to an aspect of the present invention. Consider, for example, a circular buffer 500 that has a size of 16 bytes illustrated in FIG. 5A. A circular buffer of this size is specified by setting either BK 0 or BK 1 to “00011.” For example with register A 4 in circular mode and BK 0 =3, the buffer size is 16 bytes, 8 half words, or 4 words. The value put in the AMR for this example is 00030001h. The buffer starts at address 0x0020 ( 502 ) and ends at 0x002F ( 504 ). The register A 4 is initialized to the address 0x0028, for example; however, the buffer could be located at other places in the memory by setting more significant address bits in register A 4 . Below the buffer at address 0x1F ( 506 ) and above the buffer at address 0x30 ( 508 ) data can be stored that is not relevant to the buffer.

The effect of circular buffering is to make it so that memory accesses and address updates in the 0x20-0x2F range stay completely inside this range. Effectively, the memory map behaves as illustrated in FIG. 5 B. Executing a LDW instruction with an offset of 1 in post increment mode will provide an address of 0x0028 ( 511 ) and access word 510 , for example. Executing the instruction a second time will provide an address of 0x002C ( 513 ) and access word 512 at the end of the circular buffer. Executing the instruction a third time will provide an address of 0x0020 ( 502 a ) and access word 514 . Note that word 514 actually corresponds to the other end of the circular buffer, but was accessed by incrementing the address provided by the LDW instruction.

FIG. 5C illustrates the operation of an access into the circular buffer using a non-aligned load/store instruction. In this example, A 4 is initialized to the address 0x002A and a non-aligned double word load instruction (LDNDW) with a non-scaled offset of “1” in post increment mode. As discussed above, two addresses will be sent to the two ports on the memory. An address of 0x002A ( 534 ) will be sent on the first port, which results in accessing an aligned double word DW 1 from memory, aligned on address 0x0028 ( 530 ). A second address will be sent to the memory system that is incremented by the line size of the instruction. Since in this example the instruction is a double word instruction, the line size is two words, or eight bytes. Thus the second address in incremented by eight bytes to be 0x0032. However, according to an aspect of the present invention, this address is bounded to 0x0022 ( 536 ) by circular addressing circuitry to remain within the bounds of circular buffer region 500 . The memory system then accesses a second aligned double word DW 2 , aligned at address 0x0020 ( 532 ). Extraction circuitry then extracts non-aligned double word NADW 1 from the two logically adjacent double words DW 1 and DW 2 , even though they are actually physically from different ends of the circular buffer.

Still referring to FIG. 5C , executing the LDNDW instruction again results in sending a first address of 0x002B ( 538 ) incremented by a non-scaled offset of “1” and a second address of 0x0023 ( 540 ) incremented by the line size and bounded to remain within the circular buffer region to the memory system. The memory system will access the same two aligned double words DW 1 , DW 2 . However, the extraction circuitry now extracts non-aligned double word NADW 2 in response in response to incremented address 538 .

As another example, Table 8 shows an LDNW performed with register A 4 in circular mode and BK 0 =3, so the buffer size is 16 bytes, 8 half words, or 4 words. The value put in the AMR for this example is 00030001h. The buffer starts at address 0x0020 and ends at 0x002F. The register A 4 is initialized to the address 0x002A. In this example, on offset of “2” is specified. 2h words is 8h bytes. 8h bytes is 3 bytes beyond the 16 byte (10h) boundary starting at address 002Ah; thus, it is wrapped around to 0022h (002Ah+8h=0022h). In this example, the two address sent to the memory subsystem are contiguous; the first address is 0x0022 and the second address is incremented by a line size of 4h, to become 0x0026.

TABLE 8
LDNW in Circular Mode
LDNW .D1 *+ +A4[2],A1
Before LDW	1 cycle after LDW	5 cycles after LDW
A4	0000 002Ah	A4	0000 0022h	A4	0000 00022h
A1	XXXX XXXXh	A1	XXXX XXXXh	A1	5678 9ABCh
Mem 0022h	5678 9ABCh	mem 0022h	5678 9ABCh	mem 0022h	5678 9ABCh

FIG. 6 illustrates the basic format of a fetch packet of the DSP. In this embodiment, instructions are always fetched eight at a time. This constitutes a fetch packet. The execution grouping of the fetch packet is specified by the p-bit, bit zero, of each instruction. Fetch packets are 8-word aligned and can contain up to eight instructions. A p bit in each instruction controls the parallel execution of instructions. A set of instructions executing in parallel constitute an execute packet. An execute packet can contain up to eight instructions.

The p bit controls the parallel execution of instructions. The p bits are scanned from left to right (lower to higher address). If the p bit of instruction i is 1, then instruction i+1 is to be executed in parallel with (in the same cycle as) instruction i. If the p-bit of instruction i is 0, then instruction i+1 is executed in the cycle after instruction i. All instructions executing in parallel constitute an execute packet. An execute packet can contain up to eight instructions. All instructions in an execute packet must use a unique functional unit.

There are three types of p-bit patterns for fetch packets which result in the following execution sequences for the eight instructions: fully serial; fully parallel, or partially parallel. As discussed above, in this embodiment of the invention, a non-aligned fetch or store instruction uses only one .D unit, so that the other .D unit is available for parallel execution of an instruction that does not access a load/store port on memory 22 . The processor can access words and double words at any byte boundary using non-aligned loads and stores. As a result, word and double word data does not always need alignment to 32-bit or 64-bit boundaries.

As an example of parallel execution of load/store instructions, the following execute packet is invalid in the present embodiment because there are two memory operations and one of them is non-aligned:

LDNW .D 2 T 2 *B 2 [B 12 ],B 13 ; | | LDB .D 1 T 1 *A 2 ,A 14 ;

However, the following execute packet is valid in the present embodiment because there is a non-memory Boolean/arithmetic instruction being executed on .D 1 in parallel with a non-aligned load instruction on .D 2 :

LDNW .D 2 T 2 *B 2 [B 12 ], A 13 ; | | ADD .D 1 x A 12 , B 13 , A 14 ;

Pipeline Operation

The instruction execution pipeline of DSP 1 has several key features which improve performance, decrease cost, and simplify programming, including: increased pipelining eliminates traditional architectural bottlenecks in program fetch, data access, and multiply operations; control of the pipeline is simplified by eliminating pipeline interlocks; the pipeline can dispatch eight parallel instructions every cycle; parallel instructions proceed simultaneously through the same pipeline phases; sequential instructions proceed with the same relative pipeline phase difference; and load and store addresses appear on the CPU boundary during the same pipeline phase, eliminating read-after-write memory conflicts.

A multi-stage memory pipeline is present for both data accesses in memory 22 and program fetches in memory 23 . This allows use of high-speed synchronous memories both on-chip and off-chip, and allows infinitely nestable zero-overhead looping with branches in parallel with other instructions.

There are no internal interlocks in the execution cycles of the pipeline, so a new execute packet enters execution every CPU cycle. Therefore, the number of CPU cycles for a particular algorithm with particular input data is fixed. If during program execution, there are no memory stalls, the number of CPU cycles equals the number of clock cycles for a program to execute.

Performance can be inhibited by stalls from the memory system, stalls for cross path dependencies, or interrupts. The reasons for memory stalls are determined by the memory architecture. Cross path stalls are described in detail in U.S. patent Ser. No. 09/702,453, to Steiss, et al and is incorporated herein by reference. To fully understand how to optimize a program for speed, the sequence of program fetch, data store, and data load requests the program makes, and how they might stall the CPU should be understood.

The pipeline operation, from a functional point of view, is based on CPU cycles. A CPU cycle is the period during which a particular execute packet is in a particular pipeline stage. CPU cycle boundaries always occur at clock cycle boundaries; however, memory stalls can cause CPU cycles to extend over multiple clock cycles. To understand the machine state at CPU cycle boundaries, one must be concerned only with the execution phases (E 1 -E 5 ) of the pipeline. The phases of the pipeline are described in Table 9.

TABLE 9
Pipeline Phase Description
				Instruction
	Pipeline			Types
Pipeline	Phase	Symbol	During This Phase	Completed
Pro-	Program	PG	Address of the fetch
gram	Address		packet is determined.
Fetch	Generate
	Program	PS	Address of fetch packet
	Address		is sent to memory.
	Send
	Program	PW	Program memory access
	Wait		is performed.
	Program	PR	Fetch packet is expected
	Data		at CPU boundary.
	Receive
Pro-	Dispatch	DP	Next execute packet in fetch
gram			packet determined and sent
Decode			to the appropriate functional
			units to be decoded.
	Decode	DC	Instructions are decoded
			at functional units.
Execute	Execute 1	E1	For all instruction types,	Single-
			conditions for instructions are	cycle
			evaluated and operands read.
			Load and store instructions:
			address generation is
			computed and address
			modifications written to
			register file †
			Branch instructions: affects
			branch fetch packet in PG
			phase †
			Single-cycle instructions:
			results are written to a
			register file †
	Execute 2	E2	Load instructions: address	Stores
			is sent to memory †	STP
			Store instructions and STP:
			address and data are sent
			to memory †
			Single-cycle instructions that	Multiplies
			saturate results set the SAT bit
			in the Control Status Register
			(CSR) if saturation occurs. †
			Multiply instructions: results
			are written to a register file †
	Execute 3	E3	Data memory accesses are
			performed. Any multiply
			instruction that saturates
			results sets the SAT bit in
			the Control Status Register
			(CSR) if saturation occurs. †
	Execute 4	E4	Load instructions: data is
			brought to CPU boundary †
	Execute 5	E5	Load instructions: data is	Loads
			loaded into register †
† This assumes that the conditions for the instructions are evaluated as true. If the condition is evaluated as false, the instruction will not write any results or have any pipeline operation after E1.

The pipeline operation of the instructions can be categorized into seven types shown in Table 10. The delay slots for each instruction type are listed in the second column.

TABLE 10
Delay Slot Summary
	Delay
Instruction Type	Slots	Execute Stages Used
Branch (The cycle when	5	E1-branch target E1
the target enters E1)
Load (LD) (Incoming Data)	4	E1-E5
Load (LD) (Address Modification)	0	E1
Multiply	1	E1-E2
Single-cycle	0	E1
Store	0	E1
NOP (no execution pipeline operation)	—	—
STP (no CPU internal results written)	—	—

The execution of instructions can be defined in terms of delay slots (Table 10). A delay slot is a CPU cycle that occurs after the first execution phase (E 1 ) of an instruction in which results from the instruction are not available. For example, a multiply instruction has 1 delay slot, this means that there is 1 CPU cycle before another instruction can use the results from the multiply instruction.

Single cycle instructions execute during the E 1 phase of the pipeline. The operand is read, operation is performed and the results are written to a register all during E 1 . These instructions have no delay slots.

Load instructions have two results: data loaded from memory and address pointer modification.

Data loads complete their operations during the E 5 phase of the pipeline. In the E 1 phase, the address of the data is computed. In the E 2 phase, the data address is sent to data memory. In the E 3 phase, a memory read is performed. In the E 4 stage, the data is received at the CPU core boundary. Finally, in the E 5 phase, the data is loaded into a register. Because data is not written to the register until E 5 , these instructions have 4 delay slots. Because pointer results are written to the register in E 1 , there are no delay slots associated with the address modification.

Store instructions complete their operations during the E 3 phase of the pipeline. In the E 1 phase, the address of the data is computed. In the E 2 phase, the data address is sent to data memory. In the E 3 phase, a memory write is performed. The address modification is performed in the E 1 stage of the pipeline. Even though stores finish their execution in the E 3 phase of the pipeline, they have no delay slots and follow the following rules (i=cycle):

1) When a load is executed before a store, the old value is loaded and the new value is stored.

2) When a store is executed before a load, the new value is stored and the new value is loaded.

3) When the instructions are in are in parallel, the old value is loaded and the new value is stored.

As discussed earlier, in this embodiment of the present invention non-aligned load and store instructions are performed by using resources of one D unit and both target ports T 1 and T 2 , as will be described in more detail below. Advantageously, the second D unit is available to execute a Boolean or arithmetic instruction in parallel with the execution of a non-aligned load/store instruction. Aspects of non-aligned memory accesses will now be described in more detail.

FIG. 7 is a memory map of a portion of the memory space of the DSP 1 and illustrates various aligned and non-aligned memory accesses. This portion of memory can be at any address YYYYYNNXh, but only the portion of the address represented by NNXh will be referred to herein, for convenience. Furthermore, the addresses used in the following discussion are only for example and are not intended to limit the invention in any manner.

DSP 1 can access both target ports T 1 , T 2 of data memory 22 by executing two aligned load or store instructions in parallel, as discussed above. For example, a double word 700 at address 700h and a double word 708 at address 708h can be accessed by two load double word (LDDW) instructions executed in parallel using .D 1 and .D 2 and target ports T 1 and T 2 . Likewise, word 780 and half word 786 can be accessed by executing a load word (LDW) instruction and a load half word (LDH) instruction in parallel using .D 1 and .D 2 and target ports T 1 and T 2 .

Advantageously, this embodiment of the present invention utilizes the two target ports and two address buses DA 1 , DA 2 to perform a non-aligned access. For example, double word 721 at address 721h is non-aligned by one byte. Double word 74 D a- 74 D b at address 74Dh is located in two different rows of the memory. Single word 7 B 7 located at address 7B7h is non-aligned by three bytes with respect to word aligned address 7B4h. Advantageously, each non-aligned access is performed in the same amount of time as each aligned access, unless the data word is not present in memory 22 and must be retrieved from secondary memory storage, such as off-chip memory 70 of FIG. 1 .

Uniform access time is important for software programs that operate in real time, such as are commonly executed on DSPs. The problem for real time comes when a loop walks a data structure by a stride related to the cache/SRAM line size. If the structure starts at an offset such that the unaligned access doesn't require access outside of the single line, the loop runs quickly since every access runs without the stall. If the starting offset is such that the nonaligned load crosses the line boundary, there is a stall on every access. The same loop might run twice as long this time. If a real-time system is designed for the longer loop time, then twice as much performance is being sacrificed most of the time.

FIG. 8 is a block diagram illustrating D-unit address buses of DSP 1 in more detail and illustrating two target ports T 1 , T 2 of DSP memory 22 . An aspect of the embodiment of the present invention is that load/store unit .D 1 can generate an address for a non-aligned access and provide it on address bus DA 1 via address signals 800 and multiplexer 200 a , and simultaneously generate a contiguous address that is greater by the data size and provide it on address bus DA 2 via address signals 801 and multiplexer 200 b , as depicted in Table 11. In this embodiment of the invention, load/store unit .D 2 can also generate an address for a non-aligned access and simultaneously generate a contiguous address incremented by the data size and provide them to address buses DA 1 and DA 2 via address signal lines 810 and 811 and multiplexors 200 a and 200 b , respectively. However, in an alternative embodiment, only one load/store unit may be so equipped. In yet another embodiment, there may be more than two D units so equipped, for example.

In this embodiment of the invention, DSP 1 supports non-aligned memory loads and stores for words and doublewords. Only one non-aligned access can be performed in a single cycle because both target ports T 1 , T 2 are used to load/store part of the data. From the memory designer's perspective, the two memory operations due to a non-aligned access are indistinguishable from two memory operations resulting from two instructions executed in parallel and in both cases the same memory ordering properties apply. The DSP simply requests an aligned access to each target port T 1 , T 2 and byte strobes accompany data that must be written. Table 11 shows the accesses that are performed as a result of non-aligned accesses. Alternative embodiments of the present invention may support other data sizes for non-aligned access. An alternative embodiment of the present invention may provide the addresses in another form, such as a byte address without being truncated to the nearest word address, for example. Advantageously, memory 22 bank conflicts do not occur during non-aligned access.

TABLE 11
Non-Aligned Memory Access
	Non-Aligned
Request Size	Byte Offset	DA2 Byte Address	DA1 Byte Address
Word	0x1 to 0x3	0x4	0x0
Word	0x5 to 0x7	0x8	0x4
Word	0x9 to 0xB	0xC	0x8
Word	0xD to 0xF	0x10	0xC
Word	0x11 to 0x13	0x14	0x10
Word	0x15 to 0x17	0x18	0x14
Word	0x19 to 0x1B	0x1C	0x18
Word	0x1D to 0x1F	0x20	0x1C
Doubleword	0x1 to 0x7	0x8	0x0
Doubleword	0x9 to 0xF	0x10	0x8
Doubleword	0x11 to 0x17	0x18	0x10
Doubleword	0x19 to 0x1F	0x20	0x18

FIG. 9 is a block diagram of the memory of FIG. 8 illustrating address decoding of the two target ports T 1 , T 2 and byte selection circuitry to extract a non-aligned data item according to an aspect of the present invention. Byte selection circuitry 910 selects data from a set of memory banks 940 - 947 and provides the selected data to load data signals 901 and to load data signals 902 that are connected respectively to load data buses LD 1 a,b and LD 2 a,b . In this embodiment of the present invention, there are eight memory banks 940 - 947 that each store sixteen bits of data, so that two sets of 64 bit data can be selected and provided on load data signals 901 , 902 . Address ports 921 and 922 each receive an address from address buses DA 1 and DA 2 , respectively and provide a portion of the address to separate inputs on address multiplexers 950 - 957 that provide addresses to the memory banks. Decode circuitry 930 decodes a portion of the address MSBs to determine that a memory request is intended for memory 22 . Decode signals 932 are formed by decoder 930 and sent to address multiplexors 950 - 957 to select which address is provided to each memory bank.

Decode circuitry 930 also receives a set of control signals 931 from instruction decode circuitry 10 c of DSP 1 to identify if a non-aligned access is being processed by memory 22 . In response to control signals 931 and four LSB address bits from each address bus DA 1 , DA 2 , decode circuitry 930 forms byte selection signals 933 that are sent to byte selection circuitry 910 . When one or two aligned load requests are being executed, byte selection circuitry places the requested byte, half word, word or double word on the appropriate set of load data signals 901 , 902 in a right aligned manner in response to byte selection signals 933 .

When a non-aligned load request is being executed, byte selection circuitry 910 places the selected word or double word on the appropriate set of load data signals 901 or 902 in response to byte selection signals 933 . For example, referring back to FIG. 7 , for non-aligned double word access 74 D, memory banks 946 and 947 are accessed at aligned address 748h provided on address bus DA 1 and three bytes are selected corresponding to byte addresses 74Dh-74Fh. Memory banks 940 , 941 , and 942 are accessed at contiguous aligned address 750h provided on address bus DA 2 and five bytes are selected corresponding to byte addresses 750h-754h. Note that the address provided on DA 2 is a value of 8h greater than the aligned address on DA 1 , corresponding to the eight byte size of the requested non-aligned data item. These eight bytes are then right aligned and provided on load data signals 901 if register file A 20 a is the specified destination of the transfer or on load signals 902 if register file 20 b is the specified destination of the transfer. In this embodiment, the load data bus LDx that is not associated with the specified destination register file remains free so that an associated .S unit can use the shared register file write port.

FIG. 10 is a block diagram illustrating load byte selection circuitry 910 , also referred to as extraction circuitry, of FIG. 9 in more detail. For simplicity, only byte select multiplexors 1000 - 1007 connected to load data byte lanes 901 ( 0 )- 901 ( 7 ) are shown for simplicity. Another similar set of multiplexors is connected to load data signals 902 . Selected ones of byte selection signals 933 are connected to each multiplexor to select the appropriate one of sixteen bytes provided by the memory bank array.

FIG. 11 is a block diagram illustrating the store byte selection circuitry of the memory system FIG. 8 in more detail. For purposes of this document, the store byte selection circuitry is also referred to as insertion circuitry for storing a non-aligned data item into the memory subsystem. Pipe 1 store data signals 1121 provide store data from store data buses ST 1 a,b to byte selection multiplexors 1100 - 1115 . Likewise, Pipe 2 store data signals 1122 provide store data from store data buses ST 2 a,b to byte selection multiplexors 1100 - 1115 . Control signals (not shown) provided to each byte multiplexor from decode circuitry 930 selects the appropriate one of sixteen bytes and presents each selected byte to the respective memory bank 940 - 947 . Write signals byte 0 -byte 15 are asserted as appropriate to cause a selected byte to be written into the respective memory bank.

In this embodiment of the present invention, the load byte selection circuitry and the store byte selection circuitry is required to support the various aligned accesses available via each of the target ports T 1 , T 2 . Advantageously, a single non-aligned access can be supported with only minor changes to the byte selection circuitry. Advantageously, all of the memory address decoding circuitry and memory banks do not need any modification and execute a non-aligned access simply as two aligned accesses in response to the two addresses provided on address buses DA 1 and DA 2 .

FIG. 12A is a block diagram of a load/store .D unit, which executes the load/store instructions and performs address calculations. The .D unit receives a base address via first source input src 1 . An offset value can be selected from either a second source input src 2 or from a field in the instruction opcode, indicated at 1200 . An address is provided on address output 1202 that is in turn connected to at least one of address multiplexors 200 a,b . Additionally, an augmented address is provided on address output 1204 for non-aligned accesses. The augmented address is incremented by a byte address value of either four or eight as selected by multiplexer 1210 in response to the line size of the instruction being executed: four is selected for a word instruction and eight is selected for a double word instruction. Adder 1212 increments an address on signal lines 1213 by the amount selected by multiplexer 1210 to form the augmented address that is provided on signal lines 1214 . This contiguous address is provided on address output 1204 for a non-aligned access and is connected to the other address multiplexor 200 a,b , as discussed previously. A calculated address value is also provided to the output dst to update a selected base address register value in the register file when an increment or decrement address mode is selected. According to an aspect of the present invention, the address on signal lines 1213 and the augmented address on signal lines 1214 are passed through circular buffer circuitry 1230 prior to being output on 1202 , 1204 so that they can be bounded to remain within a circular buffer region.

In this embodiment, Load and Store instructions operate on data sizes from 8 bits to 64 bits. Addressing modes supported by the .D unit are basic addressing, offset addressing, scaled addressing, auto-increment/auto-decrement, long-immediate addressing, and circular addressing, as defined by mode field 500 . In basic addressing mode, the content of a selected base register is used as a memory address. In offset addressing mode, the memory address is determined by two values, a base value and an offset that is either added or subtracted from the base. Referring again to FIG. 3 A and FIG. 3B , the base value always comes from a base register specified by a field 514 “base R” that is any of the registers in the associated register file 20 a or 20 b , whereas the offset value may come from either a register specified by an “offset R” field 516 or a 5-bit unsigned constant UCST 5 contained in field 516 of the instruction via signals 1200 . Certain load/store instructions have a long immediate address mode that uses a 15-bit unsigned constant contained in the instruction (not shown in FIG. 3 ). A selected offset is provided on signal lines 1218 to shifter 1220 . Scaled addressing mode functions the same as offset addressing mode, except that the offset is interpreted as an index into a table of bytes, half-words, words or double-words, as indicated by the data size of the load or store operation, and the offset is shifted accordingly by shifter 1220 in response to control signals 1226 which are derived by decoding opcode field 510 , 512 of the LD/ST instructions.

In this embodiment of the present invention, an SC bit 520 in load/store non-aligned double word (LDNDW/STNDW) instruction controls shifter 1220 so that an offset can be used directly, referred to as unscaled, or shifted by an amount corresponding to the type of instruction, referred to as scaled. Scaled/unscaled control signal 1224 is derived by decoding the SC field 520 of LDNDW/STNDW instructions. If SC field 520 is a logical 0, then the offset is not scaled and signal 1224 is deasserted. If SC field 520 is a logical 1, then the offset is scaled and signal 1224 is asserted. In this embodiment, for instructions other than LDNDW/STNDW, signal 1224 is asserted so that scaling will be performed according to data size control signals 1226 .

In auto-increment/decrement addressing mode, the base register is incremented/decremented after the execution of the load/store instruction by inc/dec unit 1222 . There are two sub-modes, pre-increment/decrement, where the new value in the base register is used as the load/store address, and post-increment/decrement where the original value in the register is used as the load/store address. In long-immediate addressing mode, a 15-bit unsigned constant is added to a base register to determine the memory address. In circular addressing mode, the base register along with a block size define a region in memory. To access a memory location in that region, a new index value is generated from the original index modulo the block size in circular addressing unit 1230 .

In this embodiment of the invention, a Boolean unit 1240 is provided and can be used for execution of logical instructions when the .D unit is not being used to generate an address.

FIG. 12B is a more detailed block diagram of circular buffer circuitry 1230 of FIG. 12 A. As explained earlier, circular mode addressing operates as follows with LD/ST Instructions: after shifting offsetR/cst to the left by 3, 2, 1, or 0 for LDDW, LDW, LDH, or LDB respectively, it is then added to or subtracted from baseR to produce the final address. This add or subtract is performed by only allowing bits N through 0 of the result to be updated, leaving bits 31 through N+1 unchanged after address arithmetic. The resulting address is bounded to 2{circumflex over ( )}(N+1) range, regardless of the size of the offsetR/cst. Bounding can be performed in a number of ways, such as by interrupting a carry bit at the appropriate place of adder 1222 . However, in order to support non-aligned accesses, both the address and the augmented address must be bounded separately.

In the present embodiment, bounding circuitry 1250 bounds the address provided on signal lines 1213 , while bounding circuitry 1260 bounds the augmented address provided on signal lines 1214 . Mask generation circuit 1232 forms a right extended mask (R-mask) in response to a selected block size from the AMR register, as described earlier, and provides it on bus 1234 . A right extended mask has a “1” in every bit position corresponding to an address bit within the bounds of the 2{circumflex over ( )}(N+1) range, and a “0” in every more significant address bit beyond this range.

The R-mask is bit-wise ANDed with the address on bus 1213 in AND block 1252 to form a least significant portion of the address bounded within the 2{circumflex over ( )}(N+1) range. An inverted R-mask is bit-wise ANDed with the original base address on bus 1216 in AND block 1254 to form a most significant portion of the address above the 2{circumflex over ( )}(N+1) range. The most significant address portion and the bounded least significant address portion are bit-wise combined in OR block 1256 to form the final address that is output on bus 1215 . The augmented address on bus 1214 is likewise bounded using AND blocks 1262 , 1264 and OR block 1266 and then output on bus 1217 .

Advantageously, by having two bounding circuits 1250 , 1260 both address are formed in a parallel manner so that a non-aligned access to a circular buffer region is performed in the same amount of time as an aligned access to a circular buffer region.

FIG. 12C is a flow chart illustrating formation of scaled and non-scaled addresses for accessing a linear region or a circular buffer region with either aligned or non-aligned accesses, according to an aspect of the present invention. In step 600 , a circular buffer region is setup in memory subsystem 22 by initializing the AMR register and an associated base register, as discussed above.

In step 602 , an instruction is fetched for execution. In this embodiment of the present invention, instructions are fetched in fetch packets of eight instructions simultaneously during instruction execution pipeline phases P/G, PS, PW and PR. Other embodiments of the present invention may fetch instructions singly or doubly, for example, in a different number of phases.

In step 610 , the instruction is decoded to form a plurality of fields. In this embodiment, decoding is performed in two phases of the instruction execution pipeline, but in other embodiments of the present invention decoding may be performed on one or three or more phases.

In step 620 , a base-offset address for accessing a data item for the instruction is formed by combining in 627 a base address value and an offset value, such that the offset value is selectively scaled or not scaled. Step 627 may include post or pre-incrementing or decrementing, for example, as indicated by mode field 500 . In 621 or 622 , for a non-aligned double word load or store instruction (LD/STNDW) the offset value is scaled by shifting left three bits only if the SC field 520 has a value of 1. If SC field 520 has a value of 0, then the offset value is not scaled and is therefore treated as a byte offset. If the instruction is a load or store double, then the offset is scaled by left shifting three bits in step 623 to form a double word offset. If the instruction is a LD/ST word, then the offset is scaled by shifting left two bits in step 624 to form a word offset. If the instruction is a half word LD/ST instruction, then the offset is scaled by shifting left one bit in step 625 to form a half word offset. If the instruction is a byte LD/ST instruction, then the offset is scaled by shifting zero bits in step 626 to form a byte offset. In the present embodiment, the scaling amount is determined by opcode field 510 , 512 that specifies the type of LD/ST instruction. In another embodiment, there may be a field to specify operand size, for example. In the present embodiment, step 620 is performed during the E 1 pipeline phase.

In step 630 , if the instruction fetched in step 602 is an aligned type instruction, then the base-offset address from step 627 is concatenated to stay within the boundary of the circular buffer region specified in step 602 , if circular addressing is specified by the AMR for the base register selected by the instruction. In step 632 , the resultant address is sent to the memory subsystem during pipeline phase E 2 . If circular addressing is not selected, then the base-offset from step 627 is used to access memory during pipeline phase E 2 .

In step 640 , if the instruction fetched in step 602 is a non-aligned type instruction, then a line size is added to the base-offset address from step 627 to form an augmented address. The line size is determined by the instruction type decoded in step 610 . For a double word instruction type, the line size is eight bytes. For a word instruction type, the line size is four bytes.

In step 650 , the base-offset address from step 627 is concatenated to stay within the boundary of the circular buffer region specified in step 602 , if circular addressing is specified by the AMR for the base register selected by the instruction. In step 652 , the resultant address is sent to the first port of the memory subsystem during pipeline phase E 2 . If circular addressing is not selected, then the base-offset from step 627 is used to access memory during pipeline phase E 2 . Likewise, in steps 651 and 653 , the augmented is selectively bounded if circular addressing is selected and a second address is sent to the second port of the memory subsystem during pipeline phase E 2 .

During step 654 , the requested non-aligned data item is extracted from the two aligned data items accessed in steps 652 , 653 .

An assembler which supports this embodiment of the invention defaults increments and decrements to 1 and offsets to 0 if an offset register or constant is not specified. Loads that do not modify to the baseR can use the assembler syntax *R. Square brackets, [ ], indicate that the ucst 5 offset is left-shifted by 3 for double word loads. Parentheses, ( ), are be used to tell the assembler that the offset is a non-scaled offset. For example, LDNDW (.unit) *+baseR ( 14 ), dst represents an offset of 14 bytes and the assembler writes out the instruction with offsetC=14 and sc=0. Likewise, LDNDW (.unit) *+baseR [ 16 ] dst represents an offset of 16 double words, or 128 bytes, and the assembler writes out the instruction with offsetC=16 and sc=1.

In this embodiment, LD/STDW instructions do not include an SC field, but LDNDW and STNDW instruction do include an SC field. However, parentheses, ( ), are used to tell the assembler that the offset is a non-scaled, constant offset. The assembler right shifts the constant by 3 bits for double word stores before using it for the ucst 5 field. After scaling by the STDW instruction, this results in the same constant offset as the assembler source if the least significant three bits are zeros. For example, STDW (.unit) src, *+baseR ( 16 ) represents an offset of 16 bytes (2 double words), and the assembler writes out the instruction with ucst 5 =2. STDW (.unit) src, *+baseR [ 16 ] represents an offset of 16 double words, or 128 bytes, and the assembler writes out the instruction with ucst 5 =16.

Referring again to step 620 of FIG. 6 , the SC bit (scale or not scaled) affects pre/post incrementing. If a pre or post increment/decrement is specified, then the increment/decrement amount is controlled by the SC bit. In non-scaled mode, the increment/decrement corresponds to a number of bytes. In a ssembly code , t hi s would be written as shown in Table 12, example 1 and 2. In both of these cases, reg 1 ends up with the value “reg1+reg2”.

In scaled mode, the increment/decrement corresponds to a number of double-words. The assembly syntax for this is shown in Table 12, examples 3 and 4. In both of these cases, reg 1 ends up with the value “reg1+8*reg2”. That is, reg 2 is “scaled” by the size of the access.

These comments also apply to the integer offset modes as well, as illustrated in Table 12, examples 5-8. Likewise, similar examples apply to the pre/post decrement instructions.

TABLE 12
Examples of Instructions With Various Pre/Post Increment,
Scaled and Non-Scaled Addressing Modes
example	Instruction syntax	operation
1	LDNDW *+ +reg1(reg2), reg3	pre-increment, non-scaled
2	LDNDW *reg1+ +(reg2), reg3	post-increment, non-scaled
8	LDNDW *+ +reg1[reg2], reg3	pre-increment, scaled
4	LDNDW *reg1+ +[reg2], reg3	post-increment, scaled
5	LDNDW *+ +reg1(cst5), reg2	pre-increment, non-scaled
6	LDNDW *reg1+ +(cst5), reg2	post-increment, non-scaled
7	LDNDW *+ +reg1[cst5], reg2	pre-increment, scaled
8	LDNDW *reg1+ +[cst5], reg2	post-increment, scaled

An advantage of scaled vs. non-scaled for the integer offset modes is that scaled provides a larger range of access whereas non-scaled provides finer granularity of access. Typically, when large offsets are used, they're multiples of the access size already. When small offsets are used, they're typically not, since typically a short moving distance is desired.

Scaled vs. non-scaled in register-offset modes is advantageous as well, but for different reasons. In scaled mode, the register offset usually corresponds to an array index of some sort. In non-scaled mode, the register offset may correspond to an image width or other stride parameter that isn't a multiple of the access width. For instance, accessing a 2 dimensional array whose row width is not a multiple of 8.

FIG. 13 is a block diagram of an alternative embodiment of a digital system 1300 with processor core 1301 similar to CPU 10 of FIG. 1. A direct mapped program cache 1710 , having 16 Kbytes capacity in memory 1710 b , is controlled by L 1 Program (L 1 P) controller 1710 a and connected thereby to the instruction fetch stage 10 a . A 2-way set associative data cache 1720 , having a 16 Kbyte capacity in memory 1720 b , is controlled by L 1 Data (L 1 D) controller 1720 a and connected thereby to data units D 1 and D 2 . An L 2 memory 1730 having four banks of memory, 128 Kbytes total, is connected to L 1 P 1710 a and to L 1 D 1720 a to provide storage for data and programs. External memory interface (EMIF) 1750 provides a 64-bit data path to external memory, not shown, which provides memory data to L 2 memory 1730 via extended direct memory access (DMA) controller 1740 .

EMIF 1752 provides a 16-bit interface for access to external peripherals, not shown. Expansion bus 1770 provides host and I/O support similarly to host port 60 / 80 of FIG. 1 .

Three multi-channel buffered serial ports (McBSP) 1760 , 1762 , 1764 are connected to DMA controller 1740 . A detailed description of a McBSP is provided in U.S. Pat. No. 6,167,466 (Seshan, et al) and is incorporated herein by reference.

Advantageously, non-aligned accesses to a data cache 1720 is performed in the same amount of time as an aligned access to data cache 1720 , as long as a miss does not occur. Likewise, advantageously, non-aligned accesses to a circular buffer region in data cache 1720 is performed in the same amount of time as an aligned access to a circular buffer region in data cache 1720 , as long as a miss does not occur.

FIGS. 14A and 14B together is a block diagram of data cache L 1 D 1720 b of FIG. 13 . L 1 D is a 16K byte 2-way associative cache that has 128 sets and a line size of 32 bytes. The data bus interface from L 2 to L 1 D is 32 bytes wide (32 total, not 32 for A side and 32 for B side)—it takes one clock cycle to send a line from L 2 to L 1 D. The store path from L 1 D to L 2 is 16 bytes wide (16 total, not 16 for A side and 16 for B side), resulting in a 4 clock cycle line eviction. L 1 D operates solely as a cache and is not memory-mapped. Table 13 summarizes the features of cache L 1 D.

TABLE 13
L1D Cache Features
Size                         16K bytes
Associativity                2-way
Line Size                    32 bytes
L2 To L1D Load Bus Width     32 bytes
L1D To L2 Store Bus Width    16 bytes
Accesses Per Clock Cycle     2
Cycles For L1D Fill          1
Cycles For L1D Line Eviction 4
Interleave/Bank Size         16 bit
Number Of Interleaves/Banks  16
Tag Array SRAM Size          4 − 128 × 19
Data Array SRAM Size         16 −  512 × 16
Write Hit Policy             Write back w/write-through tags
Write Miss Policy            no allocate
Replacement Strategy         LRU

L 1 D must carry four status bits for each cache line, including a valid bit, a modified bit, an NW (no-write) bit, and an LRU bit. The valid bits, modified (dirty) bits, and NW may reside in the tag RAM. The LRU bits reside in registers for single-cycle read-modify-write operation.

Read misses in LID cause the corresponding line to be brought into the cache. The line originates from L 2 if it is contained there; otherwise, the I/O subsystem transfers the line to L 2 from another memory-mapped location, and the line is forwarded to L 1 D. L 1 D must keep track of the modification state of each of its lines. If an L 1 D line has been modified, it must be written out to L 2 when it is replaced (or when requested to do so by an L 2 control register write).

L 1 D uses an LRU replacement strategy. A given line can reside at the same address offset within either way 0 or way 1 of the cache. Each line has an LRU bit that keeps track of which way was most recently used for the corresponding line. When a new line must replace a line that is already stored in the cache, it replaces the line that was least recently used and modifies the LRU status bit accordingly. If one of the two candidates for replacement is invalid, it is replaced without regard to LRU status, and the LRU bit indicates that way holding the new line is most recently used.

The L 1 D controller 1720 a must not use an unitialized value of the LRU bit that is present following power-up/reset. Invalidate operations do not affect LRU status.

Read hits allow the DSP to continue execution without stalling. The LRU status might need be updated.

Write misses do not cause allocation in LID; the write data is sent to a write buffer to await transfer to L 2 . Provided that the write buffer is not full, the DSP does not need to stall on write misses to L 1 D. The LRU status is unaffected by write misses.

On a write hit, data is written into the cache, and the DSP continues to execute without stalling. If the L 1 D line has not yet been modified, L 1 D must also perform a tag writethrough to L 2 , sending the tag value of the corresponding line to L 2 . L 2 stores a copy of the tag for each line in L 1 D along with a dirty bit for the line, and it must be informed that the L 1 D line has been modified. With this mechanism L 2 monitors the status of L 1 D data without the complexity or conflicts caused by snoop traffic. Tag writethrough should have a single cycle throughput, assuming no conflicts. The LRU status might need to be updated due to a write hit.

L 1 D adheres to the pipeline timing described in Table 14.

TABLE 14
Pipeline Description
Stage	Module	Action
E1	DSP	Register file read and address generate
E2	DSP	Send address and data (if a store) to L1D
E2	L1D	Receive data and perform tag lookup to initiate an
		address compare
E3	L1D	If address matches and tag is valid, initiate data RAM
		read or write. Else send data to write buffer for
		stores or stall DSP for loads
E4	DSP	Receive load data from L1D
E4	L1D	Send load data to DSP
E5	DSP	Write load data into register file

FIG. 14 a is a block diagram illustrating an implementation of the L 1 D tag array. Logically, 4 tag RAM's 1401 - 1404 are required for L 1 D. Both ways of the cache require a tag RAM, and the RAM's are replicated to allow both D units of the DSP to simultaneously perform tag lookups on different addresses. A first address is provided by .D 1 on a first address bus 1410 and accesses tag RAMs 1401 - 1402 . A second address is provided by .D 2 on bus 1411 and accesses tag RAMs 1403 - 1404 . When a line is replaced, the tag RAM's associated with both D units must be updated to maintain duplicate copies of the tags. A three-input mux 1420 , 1421 is provided to allow tag comparisons to occur one cycle earlier.

The individual tag RAM's are 128×20. One bit of the 20 bits is used to store a physical address attribute map (PAAM) NW (no-write) bit. Aspects of a PAAM are described in detail in co-assigned U.S. patent application Ser. No. 09/702,477 entitled System Address Properties Control Cache Memory, Including Physical Address Attribute Map (PAAM) and is incorporated herein by reference. The PAAM PC and NR bits do not need to be stored in L 1 D.

FIG. 14B is a block diagram illustrating an implementation of the L 1 D data array 1720 b . Note that the L 1 D Data diagram omits the logic for a write buffer. L 1 D contains a buffer of at least 58 bytes that resides between L 1 D and L 2 , and it is used for both victims and L 1 D write misses. The buffer can hold one 32 byte victim plus two L 1 D doubleword write misses, or it could hold up to ten L 1 D doubleword misses without a victim.

Each D unit of the DSP can access L 1 D 1720 with a load or store of a byte, halfword, word, or doubleword. When the two D units request a simultaneous access, there may or may not be stalls as a result of bank conflicts. Thirty-two bytes are available per cycle (an individual bank holds data for both ways of the cache). A doubleword access can occur at one of four offsets within a line, including 0, 8, 16, and 24 bytes. Assuming random behavior, there is a 25% chance of a conflict between two simultaneous doubleword accesses. Like doubleword requests, byte, halfword, and word requests may or may not have bank conflicts, but they are less likely to happen and easier to avoid from a programming standpoint.

Advantageously, L 1 D cache 1720 b is not affected by a single non-aligned word or double word load or store transaction since a non-aligned transaction is presented to the cache as two aligned transfers. Byte swapping circuitry similar to FIG. 10 and FIG. 11 is included in controller block 1720 a to provide alignment.

FIG. 15 illustrates an exemplary implementation of a digital system that includes DSP 1 packaged in an integrated circuit 40 in a mobile telecommunications device, such as a wireless telephone 15 . Wireless telephone 15 has integrated keyboard 12 and display 14 . As shown in FIG. 15 , DSP 1 is connected to the keyboard 12 , where appropriate via a keyboard adapter (not shown), to the display 14 , where appropriate via a display adapter (not shown) and to radio frequency (RF) circuitry 16 . The RF circuitry 16 is connected to an aerial 18 . Advantageously, by allowing non-aligned accesses into linear regions or circular buffer regions in the memory subsystem of DSP 1 , complex signal processing algorithms can be written in a more efficient manner to satisfy the demand for enhanced wireless telephony functionality. More importantly, non-aligned accesses into linear and circular buffer regions take the same amount of time as aligned access into the same regions, so that real time algorithms operate in a consistent, predictable manner.

Table 15 summarizes instruction operation and execution notations used throughout this document.

TABLE 15
Instruction Operation and Execution Notations
Symbol                           Meaning
long                             40-bit register value
+a                               Perform twos-complement addition using the addressing
                                 mode defined by the AMR
−a                               Perform twos-complement subtraction using the
                                 addressing mode defined by the AMR
xor                              Bitwise exclusive OR
not                              Bitwise logical complement
b                                                                        y . . . z Selection of bits y through z of bit string b
>>s                              Shift right with sign extension
>>z                              Shift right with a zero fill
x clear b,e                      Clear a field in x, specified by b (beginning bit)
                                 and e (ending bit)
x exts l,r                       Extract and sign-extend a field in x, specified by 1
                                 (shift left value) and r (shift right value)
x extu l,r                       Extract an unsigned field in x, specified by 1
                                 (shift left value) and r (shift right value)
+s                               Perform twos-complement addition and saturate
                                 the result to the result size, if an overflow
                                 or underflow occurs
−s                               Perform twos-complement subtraction and saturate
                                 the result to the result size, if an
                                 overflow or underflow occurs
x set b,e                        Set field in x, to all 1s specified by b
                                 (beginning bit) and e (ending bit)
lmb0(x)                          Leftmost 0 bit search of x
lmb1(x)                          Leftmost 1 bit search of x
norm(x)                          Leftmost nonredundant sign bit of x
abs(x)                           Absolute value of x
and                              Bitwise AND
bi                               Select bit i of source/destination b
bit_count                        Count the number of bits that are 1 in a specified byte
bit_reverse                      Reverse the order of bits in a 32-bit register
byte0                            8-bit value in the least significant byte position in
                                 32-bit register (bits 0-7)
byte1                            8-bit value in the next to least significant byte
                                 position in 32-bit register (bits 8-15)
byte2                            8-bit value in the next to most significant byte
                                 position in 32-bit register (bits 16-23)
byte3                            8-bit value in the most significant byte position
                                 in 32-bit register (bits 24-31)
bv2                              Bit Vector of two flags for s2 or u2 data type
bv4                              Bit Vector of four flags for s4 or u4 data type
cond                             Check for either creg equal to 0 or creg not equal to 0
creg                             3-bit field specifying a conditional register
cstn                             n-bit constant field (for example, cst5)
dst_h or dst_o                   msb32 of dst (placed in odd register of 64-bit
                                 register pair)
dst_l or dst_e                   lsb32 of dst (placed in even register of a 64-bit
                                 register pair)
dws4                             Four packed signed 16-bit integers in a 64-bit
                                 register pair
dwu4                             Four packed unsigned 16-bit integers in a
                                 64-bit register pair
gmpy                             Galois Field Multiply
i2                               Two packed 16-bit integers in a single 32-bit register
i4                               Four packed 8-bit integers in a single 32-bit register
int                              32-bit integer value
lsbn or LSBn                     n least significant bits (for example, lsb16)
msbn or                          n most significant bits (for example, msb16)
MSBn
nop                              No operation
or                               Bitwise OR
R                                Any general-purpose register
rotl                             Rotate left
sat                              Saturate
sbyte0                           Signed 8-bit value in the least significant byte position
                                 in 32-bit register (bits 0-7)
sbyte1                           Signed 8-bit value in the next to least significant byte
                                 position in 32-bit register (bits 8-15)
sbyte2                           Signed 8-bit value in the next to most significant byte
                                 position in 32-bit register (bits 16-23)
sbyte3                           Signed 8-bit value in the most significant byte position
                                 in 32-bit register (bits 24-31)
scstn                            Signed n-bit constant field (for example, scst7)
se                               Sign-extend
sint                             Signed 32-bit integer value
slsb16                           Signed 16-bit integer value in lower half of 32-bit
                                 register
smsb16                           Signed 16-bit integer value in upper half of 32-bit
                                 register
s2                               Two packed signed 16-bit integers in a single 32-bit
                                 register
s4                               Four packed signed 8-bit integers in a single 32-bit
                                 register
sllong                           Signed 64-bit integer value
ubyte0                           Unsigned 8-bit value in the least significant byte
                                 position in 32-bit register (bits 0-7)
ubyte1                           Unsigned 8-bit value in the next to least significant
                                 byte position in 32-bit register (bits 8-15)
ubyte2                           Unsigned 8-bit value in the next to most significant
                                 byte position in 32-bit register (bits 16-23)
ubyte3                           Unsigned 8-bit value in the most significant byte
                                 position in 32-bit register (bits 24-31)
ucstn                            n-bit unsigned constant field (for example, ucst5)
uint                             Unsigned 32-bit integer value
ullong                           Unsigned 64-bit integer value
ulsb16                           Unsigned 16-bit integer value in lower half of
                                 32-bit register
umsb16                           Unsigned 16-bit integer value in upper half of
                                 32-bit register
u2                               Two packed unsigned 16-bit integers in a single
                                 32-bit register
u4                               Four packed unsigned 8-bit integers in a single
                                 32-bit register
xi2                              Two packed 16-bit integers in a single 32-bit
                                 register that can optionally use cross path
xi4                              Four packed 8-bit integers in a single 32-bit
                                 register that can optionally use cross path
xsint                            Signed 32-bit integer value that can optionally
                                 use cross path
xs2                              Two packed signed 16-bit integers in a single
                                 32-bit register that can optionally use cross path
xs4                              Four packed signed 8-bit integers in a single 32-bit
                                 register that can optionally use cross path
xuint                            Unsigned 32-bit integer value that can optionally
                                 use cross path
xu2                              Two packed unsigned 16-bit integers in a single 32-bit
                                 register that can optionally use cross path
xu4                              Four packed unsigned 8-bit integers in a single 32-bit
                                 register that can optionally use cross path
→                                Assignment
+                                Addition
++                               Increment by one
×                                Multiplication
−                                Subtraction
>                                Greater than
<                                Less than
<<                               Shift left
>>                               Shift right
>=                               Greater than or equal to
<=                               Less than or equal to
==                               Equal to
˜                                Logical Inverse
&                                Logical And

Fabrication of digital system 1 and 1300 involves multiple steps of implanting various amounts of impurities into a semiconductor substrate and diffusing the impurities to selected depths within the substrate to form transistor devices. Masks are formed to control the placement of the impurities. Multiple layers of conductive material and insulative material are deposited and etched to interconnect the various devices. These steps are performed in a clean room environment.

A significant portion of the cost of producing the data processing device involves testing. While in wafer form, individual devices are biased to an operational state and probe tested for basic operational functionality. The wafer is then separated into individual dice which may be sold as bare die or packaged. After packaging, finished parts are biased into an operational state and tested for operational functionality.

Thus, a digital system is provided with a processor having an improved instruction set architecture. The .D units can access words and double words on any byte boundary by using non-aligned load and store instructions, maintain the same instruction execution timing for aligned and non-aligned memory accesses. Advantageously, a significant amount of additional hardware is not required to perform the non-aligned accesses because a single non-aligned access can be performed using the resources of two separate aligned access ports.

As used herein, the terms “applied,” “connected,” and “connection” mean electrically connected, including where additional elements may be in the electrical connection path. “Associated” means a controlling relationship, such as a memory resource that is controlled by an associated port. The terms assert, assertion, de-assert, de-assertion, negate and negation are used to avoid confusion when dealing with a mixture of active high and active low signals. Assert and assertion are used to indicate that a signal is rendered active, or logically true. De-assert, de-assertion, negate, and negation are used to indicate that a signal is rendered inactive, or logically false.

While the invention has been described with reference to illustrative embodiments, this description is not intended to be construed in a limiting sense. Various other embodiments of the invention will be apparent to persons skilled in the art upon reference to this description. For example, more than two target memory ports may be provided. Different data widths may be provided, such as 128-bit data items, for example. As long as the size of a non-aligned data item is less than or equal to the size of each aligned access port, then two access ports can be shared to provide a single non-aligned access without adding significant additional resources.

Scaling circuitry may be included or not included within the address generation circuitry. Scaling/non-scaling can be selectively included in instructions for data sizes other than double words.

Circular buffer address circuitry may be included or not included within the address generation circuitry.

It is therefore contemplated that the appended claims will cover any such modifications of the embodiments as fall within the true scope and spirit of the invention.

Claims

1. A digital system, comprising;a microprocessor having at least a first load/store unit and a second load/store unit; a memory subsystem having at least first memory port connected to the first load/store unit and a second memory port connected to the second load/store unit; address generation circuitry in the first load/store unit having a first address output connected to the first memory port and a second address output selectively connected to the second memory port, the address generation circuitry operable to provide a first address on the first address output and a second address on the second address output; and an extraction circuit connected to the first memory port, wherein the extraction circuit is operable to provide a first non-aligned data item to the first load/store unit extracted from a first data item accessed in response to the first address and from a second data item accessed in response to the second address.

2. The digital system of claim 1, further comprising insertion circuitry connected to the first memory port, wherein the insertion circuitry is operable to receive a second non-aligned data item from the first load/store unit and to store a first portion of the second non-aligned data item in a first data item in the memory subsystem responsive to the first address and to store a second portion of the second non-aligned data item in a second data item in the memory subsystem responsive to the second address.

3. The digital system of claim 2, wherein the address generation circuitry is operable to provide the first address and second address to the memory subsystem to access the memory subsystem in response to a first instruction type and to provide only the first address to the memory subsystem to access the memory subsystem in response to a second instruction type.

4. The digital system of claim 3, wherein the first instruction type is a non-aligned access type, and wherein the second instruction type is an aligned access type.

5. The digital system of claim 4, wherein the second load/store unit comprises address generation circuitry with a first address output selectively connected to the second memory port, such that the second load/store unit is operable to transfer a data item to the second memory port in parallel with the first load/store unit transferring a data item to the first memory port.

6. The digital system of claim 5, wherein the address generation circuitry of the second load/store unit is operable to provide the first address on a second address output selectively connected to the first memory port and the second address on the first address output for accessing the memory subsystem for an instruction of the first type.

7. The digital system of claim 5, wherein the second load/store unit is operable to execute a non-memory access instruction in parallel with the first load/store unit accessing the memory subsystem for an instruction of the first type.

8. The digital system of claim 1, wherein the memory subsystem is a cache memory.

9. The digital system of claim 1, wherein the microprocessor is a digital signal processor.

10. The digital system of claim 1 being a cellular telephone, further comprising:an integrated keyboard connected to the CPU via a keyboard adapter; a display, connected to the CPU via a display adapter; radio frequency (RF) circuitry connected to the CPU; and an aerial connected to the RF circuitry.

11. The digital system of claim 1, wherein the address circuitry comprises:combination circuitry connected to receive a base address value and an offset value, operable to combine the base address value and the offset value to form a base-offset address, wherein the base-offset address is selectively coupled to the first address output; and adder circuitry connected to receive the base-offset address and a line size value, operable to add the line size value to the base-offset address to form an augmented address, wherein the augmented address is selectively coupled the second address output.

12. The digital system of claim 1, wherein the memory subsystem comprises:a plurality of memory banks connected to the extraction circuitry; decode circuitry connected to the first memory port and to the second memory port; and a plurality of address multiplexers connected respectively to the plurality of memory banks with first input of each of the plurality of multiplexers connected a receive an address from the first memory port and a second input connected to receive an address from the second memory port, each of the plurality of address multiplexers having a select control separately connected to the decode circuitry, such that the decode circuitry is operable to individually control each of the plurality of address multiplexers.

13. A method of operating a microprocessor, comprising the steps of:fetching a first instruction for execution, wherein the first instruction is a non-aligned access type instruction and wherein the first instruction references a non-aligned data item in a memory subsystem region; decoding the instruction to form a plurality of fields; forming a first address and accessing a first data item via a first port of the memory subsystem; forming a second address and accessing a second data item via a second port of the memory subsystem, such that the first address and second address are formed in a simultaneous manner and such that the first data item and the second data item are accessed in a simultaneous manner: and extracting the non-aligned data item from the first data item and the second data item.

14. The method of claim 13, wherein, wherein the step of forming a first address comprises the step of combining a base address value and an offset value in accordance with a one of the plurality of fields of the instruction.

15. The method of claim 14, wherein the step of forming a second address comprises the step of adding a line size value in accordance with another one of the fields of the instruction.

16. The method of claim 13 wherein the step of extracting loads a data value from the non-aligned data item into the microprocessor.

17. The method of claim 13, wherein the step of extracting stores a data value in the non-aligned data item in the memory subsystem by storing a first portion of the non-aligned data item in a first aligned data item and storing a second portion of the non-aligned data item in a second aligned data item.

18. The method of claim 13, wherein the first data item is an aligned data item and wherein the second data item is an aligned data item.

19. The method of claim 13, further comprising the steps of:fetching a second instruction and a third instruction for parallel execution, wherein the second instruction and the third instruction are both aligned access type instructions; forming a third address in accordance with the second instruction and accessing a third data item via the first port of the memory subsystem; forming a fourth address in accordance with the third instruction and accessing a fourth data item via the second port of the memory subsystem, such that the third address and the fourth address are formed in a simultaneous manner and such that the third data item and the fourth data item are accessed in a simultaneous manner.