INTEL 80386DX

INTEL 80386DX

The 80386 processor was introduced in 1987 as the first 32 bit member of the family. It has 32 bit registers and both 32 bit data and address buses. It is software compatible with the previous generations through the preservation of the older register set within the 80386’s newer extended register model and through a special 8086 emulation mode where the 80386 behaves like a very fast 8086. The processor has an on-chip paging memory manage-ment unit which can be used to support multitasking and demand paging virtual memory schemes if required.

Architecture

The 80386 has eight general-purpose 32 bit registers EAX, EBX, ECX, EDX, ESI, EDI, EBP and ESP. These general-purpose registers are used for storing either data or addresses. To ensure compatibility with the earlier 8086 processor, the lower half of each register can be accessed as a 16-bit register (AX, BX, CX, DX, SI, DI, BP and SP). The AX, BX, CX and DX registers can be also accessed as 8 bit registers by changing the X suffix for either H or L thus creating the 8088 registers AH, AL, BH, BL and so on.

To generate a 32 bit physical address, six segment registers (CS, SS, DS, ES, FS, GS) are used with addresses from the general registers or instruction pointer. The code segment (CS) is used with the instruction pointer to create the addresses used for instruction fetches and any stack access uses the SS register. The remaining segment registers are used for data addresses.

Each segment register has an associated descriptor register which is used to program and control the on-chip memory man-agement unit. These descriptor registers — controlled by the operating system and not normally accessible to the application programmer — hold the base address, segment limit and various attribute bits that describe the segment‘s properties.

The 80386 can run in three different modes: the real mode, where the size of each segment is limited to 64 kbytes, just like the 8088 and 8086; a protected mode, where the largest segment size is increased to 4 Gbytes; and a special version of the protected mode that creates multiple virtual 8086 processor environments.

The 32 bit flag register contains the normal carry zero, auxiliary carry, parity, sign and overflow flags. The resume flag is used with the trap 1 flag during debug operations to stop and start the processor. The remaining flags are used for system control to select virtual mode, nested task operation and input/output privilege level.

For external input and output, a separate peripheral ad-dress facility is available similar to that found on the 8086. As an alternative, memory mapping is also supported (like the M68000 family) where the peripheral is located within the main memory map.

Interrupt facilities

The 80386 has two external interrupt signals which can be used to allow external devices to interrupt the processor. The INTR input generates a maskable interrupt while the NMI gener-ates a non-maskable interrupt and naturally has the higher prior-ity of the two.

During an interrupt cycle, the processor carries out two interrupt acknowledge bus cycles and reads an 8 bit vector number on D0–D7 during the second cycle. This vector number is then used to locate, within the vector table, the address of the corre-sponding interrupt service routine. The NMI interrupt is auto-matically assigned the vector number of 2.

Software interrupts can be generated by executing the INT n instruction where n is the vector number for the interrupt. The vector table consists of 4 byte entries for each vector and starts at memory location 0 when the processor is running in the real mode. In the protected mode, each vector is 8 bytes long. The vector table is very similar to that of the 80286.

Vector         Function

0        Divide error

1        Debug exception

2        Non-masked interrupt NMI

3        One byte interrupt INT

4        Interrupt on overflow INTO

S       Array bounds check BOUND

6        Invalid opcode

7        Device not available

8        Double fault

9        Coprocessor segment overrun

10      Invalid TSS

11      Segment not present

12      Stack fault

13      General protection fault

14      Page fault

15      Reserved

16      Coprocessor error

17-32 Reserved

33-255        INT n trap instructions

Instruction set

The 80386 instruction set is essentially a superset of the 8086 instruction set. The format follows the dyadic approach and uses two operands as sources with one of them also duplicating as a destination. Arithmetic and other similar operations thus follow the A+B=B type of format (like the M68000). When the processor is operating in the real mode — like an 8086 processor — its instruction set, data types and register model is essentially re-stricted to a that of the 8086. In its protected mode, the full 80386 instruction set, data types and register model becomes available. Supported data types include bits, bit fields, bytes, words (16 bits), long words (32 bits) and quad words (64 bits). Data can be signed or unsigned binary, packed or unpacked BCD, character bytes and strings. In addition, there is a further group of instructions that can be used when the CPU is running in protected mode only. They provide access to the memory management and control registers. Typically, they are not available to the user programmer and are left to the operating system to use.

LSL   Load segment limit

LTR  Load task register

SGDT         Store global descriptor table

SIDT Store interrupt descriptor table

STR  Store task register

SLDT          Store local descriptor table

SMSW        Store machine status word

VERR         Verify segment for reading

VERW        Verify segment for writing

Addressing modes provided are:

Register direct      (Register contains operand)

Immediate   (Instruction contains data)

Displacement       (8/16 bits)

Base address        (Uses BX or BP register)

Index (Uses DI or SI register)

80387 floating point coprocessor

The 80386 can also be used with the 80387 floating point coprocessor to provide acceleration for floating point calculations. If the device is not present, it is possible to emulate the floating point operations in software, but at a far lower performance.

Feature comparison

There is a derivative of the 80386DX called the 80386SX which provides a lower cost device while retaining the same architecture. To reduce the cost, it uses an external 16 bit data bus and a 24 bit memory bus. The SX device is not pin compatible with the DX device. These slight differences can cause quite different levels of performance which can mean the difference between performing a function of not.

In addition, Intel have produced an 80386SL device for portable PCs which incorporates a power control module that provides support for efficient power conservation.

Although Intel designed the 80386 series, the processor has been successfully cloned by other manufacturers (both technically and legally) such as AMD, SGS Thomson, and Cyrix. Their ver-sions are available at far higher clock speeds than the Intel origi-nals and many PCs are now using them. These are now available with all the peripherals needed to create a PC AT clone integrated on the chip and these are extensively used to create embedded systems based on the PC architecture.