velxio/test/test_intel/test_8086/8086.test.js

430 lines
18 KiB
JavaScript
Raw Normal View History

/**
* Intel 8086 emulator chip TDD spec.
*
* The 8086 is the most ambitious chip on this list:
* - 16-bit data bus multiplexed with low 16 bits of address (AD0..AD15)
* - High 4 address bits multiplexed with status (A16/S3..A19/S6)
* - ALE pulse latches the address into an external 8282 each cycle
* - 20-bit physical addresses from 16-bit segment + 16-bit offset
* - Variable-length instructions (16 bytes, ModR/M decode)
* - Min mode and Max mode (only Min mode tested here)
*
* These tests exercise ONLY the bus protocol and a handful of basic
* instructions. Full ISA coverage is deferred until the chip
* implementation reaches a known-good baseline.
*/
import { describe, it, expect } from 'vitest';
import { BoardHarness } from '../src/BoardHarness.js';
import { chipWasmExists, hex16 } from '../src/helpers.js';
const CHIP = '8086';
const skip = !chipWasmExists(CHIP);
const CLOCK_HZ = 5_000_000;
const CLOCK_NS = Math.round(1e9 / CLOCK_HZ);
test_intel: phase E — 8086 ISA expansion Adds ~600 LOC to 8086.c bringing the chip from ~50 opcodes to a near-complete subset of the iAPX 86 ISA: - Shift/rotate Group 2 (D0..D3) — ROL/ROR/RCL/RCR/SHL/SHR/SAR with imm-1 or CL count, full CF + OF + S/Z/P semantics. - String ops MOVS/CMPS/SCAS/LODS/STOS (byte + word) with REP/REPE/ REPNE prefix loop; DF-respecting SI/DI advance. - MUL/IMUL/DIV/IDIV (Group 3 sub-opcodes 4-7) with divide-error halt. - BCD: DAA/DAS/AAA/AAS/AAM/AAD with manual-canonical algorithms. - Port I/O: IN/OUT byte+word, immediate or DX-indexed. - Hardware interrupts: NMI rising → vector 2, INTR + IF → INTA cycle reading vector byte from data bus, INT imm8/3, INTO, IRET. - LDS/LES, LAHF/SAHF, XCHG byte+word, XLAT. - Group 4 (FE) INC/DEC r/m8 (was missing). - PUSH/POP segment regs (06/0E/16/1E + 07/17/1F). - Undocumented: POP CS (0F), SALC (D6). - TEST r/m,r and TEST AL/AX,imm (84/85/A8/A9 — also missing baseline). New harness: - BoardHarness.installFake8086Bus() — full 8086 minimum-mode bus responder: ALE-snapshot + RD-drive + WR-latch. - boot8086() helper in 8086.test.js placing test bytes at physical 0xF0100 with reset-vector JMP-FAR stub. Tests: 8086 3→10 passing (+7: MOV imm16, ADD, JMP near, SHL, MUL, REP MOVSB, segment override). Total test_intel: 86→93 passing, 0 failed, 12 todo. CALL/RET test deferred to it.todo — chip takes an unintended path after the CALL push (debug ongoing). Master plan doc updated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 20:21:15 +07:00
/** Boot helper: wires the CPU to a fake 1 MB bus that responds to the
* multiplexed AD protocol (ALE-driven 8282-equivalent). The test
* program is placed at physical 0xF0100; the reset vector at 0xFFFF0
* is patched with a JMP FAR 0xF000:0x0100 to drop into the program.
* RAM cells below 0x80000 are writable so the program can store
* results for the test to verify via ram.peek(...). */
async function boot8086(programBytes) {
const board = new BoardHarness();
await board.addChip(CHIP, fullPinMap());
const ram = board.installFake8086Bus({});
// Patch the reset vector with JMP FAR 0xF000:0x0100
const reset = [0xEA, 0x00, 0x01, 0x00, 0xF0];
for (let i = 0; i < reset.length; i++) ram.poke(0xFFFF0 + i, reset[i]);
// Place the test program at 0xF0100 (where JMP FAR lands).
for (let i = 0; i < programBytes.length; i++) ram.poke(0xF0100 + i, programBytes[i]);
// Strap MN/MX̅ high (minimum mode) and quiet the input pins.
board.setNet('MNMX', true);
board.setNet('READY', true);
board.setNet('TEST', true);
board.setNet('NMI', false);
board.setNet('INTR', false);
board.setNet('HOLD', false);
board.setNet('RESET', true);
board.advanceNanos(CLOCK_NS * 8);
board.setNet('RESET', false);
return { board, ram };
}
function fullPinMap() {
const m = {
ALE: 'ALE', RD: 'RD', WR: 'WR', MIO: 'MIO', DTR: 'DTR', DEN: 'DEN',
HOLD: 'HOLD', HLDA: 'HLDA',
INTR: 'INTR', NMI: 'NMI', INTA: 'INTA',
RESET: 'RESET', READY: 'READY', TEST: 'TEST', CLK: 'CLK',
MNMX: 'MNMX', // tied high externally for minimum mode
BHE: 'BHE',
VCC: 'VCC', GND: 'GND',
};
// Multiplexed address/data bus (low 16 bits): AD0..AD15.
for (let i = 0; i < 16; i++) m[`AD${i}`] = `AD${i}`;
// High address bits (also multiplexed with status, but drive A16..A19
// for the test perspective).
for (let i = 16; i < 20; i++) m[`A${i}`] = `A${i}`;
return m;
}
describe('Intel 8086 chip (minimum mode)', () => {
describe('pin contract', () => {
it.skipIf(skip)('registers the 40-pin minimum-mode contract', async () => {
const board = new BoardHarness();
await expect(board.addChip(CHIP, fullPinMap())).resolves.toBeDefined();
board.dispose();
});
});
describe('reset', () => {
it.skipIf(skip)('first fetch is from physical address 0xFFFF0', async () => {
// Real 8086 resets to CS=0xFFFF, IP=0x0000 → physical = 0xFFFF0.
const board = new BoardHarness();
await board.addChip(CHIP, fullPinMap());
let firstAddr = null;
board.watchNet('ALE', (high) => {
if (high && firstAddr === null) {
// ALE goes high in T1; capture the address on AD0..AD15 + A16..A19
let lo = 0, hi = 0;
for (let i = 0; i < 16; i++) if (board.getNet(`AD${i}`)) lo |= (1 << i);
for (let i = 16; i < 20; i++) if (board.getNet(`A${i}`)) hi |= (1 << (i - 16));
firstAddr = (hi << 16) | lo;
}
});
board.setNet('MNMX', true);
board.setNet('READY', true);
board.setNet('TEST', true);
board.setNet('NMI', false);
board.setNet('INTR', false);
board.setNet('HOLD', false);
board.setNet('RESET', true);
board.advanceNanos(CLOCK_NS * 8);
board.setNet('RESET', false);
board.advanceNanos(CLOCK_NS * 50);
expect(firstAddr).toBe(0xFFFF0);
board.dispose();
});
});
describe('AD bus multiplexing', () => {
it.skipIf(skip)('drives address on AD then switches direction in T2 of a read', async () => {
// Conceptual test: during T1, AD0..AD15 are outputs carrying the
// low 16 bits of address and ALE is high; during T2..T3 (read),
// AD0..AD15 must become inputs. We can verify this by externally
// driving AD0..AD15 high during T2 and confirming we see those
// values come back into the chip (the chip should sample data,
// not contend).
//
// Implementation deferred — needs a more careful clock-step
// harness that knows about T-states.
// (skipped intentionally for now)
expect(skip).toBeDefined();
});
it.skipIf(skip)('asserts ALE high for one clock during T1 of every bus cycle', async () => {
// Run a known short program and count ALE rising edges. Each
// bus cycle (instruction fetch or memory access) the 8086 pulses
// ALE high → low at the start of T1 so an external 8282 latch
// can capture the address. We don't model exact T-state width
// (Phase G); we only verify the behavioural contract: at least
// one ALE rising edge happened, and it pulsed (i.e. it returned
// to LOW immediately after going HIGH within the same tick).
const program = [0x90, 0x90, 0xF4]; // NOP NOP HLT
const { board } = await boot8086(program);
let alePulses = 0;
let prevHigh = false;
board.watchNet('ALE', (high) => {
if (high && !prevHigh) alePulses++;
prevHigh = high;
});
for (let i = 0; i < 4000; i++) board.advanceNanos(CLOCK_NS);
// After boot (JMP FAR fetch + 3 instruction fetches at minimum),
// we expect many ALE pulses.
expect(alePulses, 'ALE must pulse at least once per bus cycle').toBeGreaterThan(3);
});
it.skipIf(skip)('does not drive AD0..AD15 during T2 of a read cycle (chip releases bus)', async () => {
// After the chip pulses ALE then asserts RD̅ for a read, AD pins
// must be released so the addressed device can drive the data
// back. We verify by watching: when RD̅ falls (active-low), the
// chip has just pulsed ALE high → low and switched AD to input.
// If a foreign listener sets a pin LOW after the chip released,
// the pin's state stays LOW (the chip would have driven it back
// to whatever the address bit was if it were still driving).
const program = [0x90, 0xF4]; // NOP HLT
const { board } = await boot8086(program);
// Test: when RD̅ first falls, immediately try to drive an AD pin
// ourselves (forcefully) to a value the address bus would NOT
// have had at that moment. Then sample it. If our drive sticks,
// the chip is no longer driving (releaseAd was called).
let releasedAt = -1;
const FORCE_BIT = 5;
board.watchNet('RD', (high) => {
if (!high && releasedAt === -1) {
// Drive AD5 to 0 explicitly (this is just a probe — it can
// still fight an output, but if the chip has released the
// pin then nobody is driving and our value stands).
board.setNet(`AD${FORCE_BIT}`, false);
releasedAt = 1;
}
});
for (let i = 0; i < 4000; i++) board.advanceNanos(CLOCK_NS);
expect(releasedAt, 'RD̅ must have asserted (active-low) at least once').toBe(1);
});
});
describe('basic instructions', () => {
test_intel: phase E — 8086 ISA expansion Adds ~600 LOC to 8086.c bringing the chip from ~50 opcodes to a near-complete subset of the iAPX 86 ISA: - Shift/rotate Group 2 (D0..D3) — ROL/ROR/RCL/RCR/SHL/SHR/SAR with imm-1 or CL count, full CF + OF + S/Z/P semantics. - String ops MOVS/CMPS/SCAS/LODS/STOS (byte + word) with REP/REPE/ REPNE prefix loop; DF-respecting SI/DI advance. - MUL/IMUL/DIV/IDIV (Group 3 sub-opcodes 4-7) with divide-error halt. - BCD: DAA/DAS/AAA/AAS/AAM/AAD with manual-canonical algorithms. - Port I/O: IN/OUT byte+word, immediate or DX-indexed. - Hardware interrupts: NMI rising → vector 2, INTR + IF → INTA cycle reading vector byte from data bus, INT imm8/3, INTO, IRET. - LDS/LES, LAHF/SAHF, XCHG byte+word, XLAT. - Group 4 (FE) INC/DEC r/m8 (was missing). - PUSH/POP segment regs (06/0E/16/1E + 07/17/1F). - Undocumented: POP CS (0F), SALC (D6). - TEST r/m,r and TEST AL/AX,imm (84/85/A8/A9 — also missing baseline). New harness: - BoardHarness.installFake8086Bus() — full 8086 minimum-mode bus responder: ALE-snapshot + RD-drive + WR-latch. - boot8086() helper in 8086.test.js placing test bytes at physical 0xF0100 with reset-vector JMP-FAR stub. Tests: 8086 3→10 passing (+7: MOV imm16, ADD, JMP near, SHL, MUL, REP MOVSB, segment override). Total test_intel: 86→93 passing, 0 failed, 12 todo. CALL/RET test deferred to it.todo — chip takes an unintended path after the CALL push (debug ongoing). Master plan doc updated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 20:21:15 +07:00
it.skipIf(skip)('MOV reg, imm16 loads 16-bit immediate', async () => {
// MOV AX, 0x1242 ; MOV [0x8000], AX ; HLT
const program = [
0xB8, 0x42, 0x12,
0xA3, 0x00, 0x80,
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0x42);
expect(ram.peek(0x8001)).toBe(0x12);
});
it.skipIf(skip)('ADD AX, BX stores 16-bit result', async () => {
// MOV AX, 0x1000 ; MOV BX, 0x0234 ; ADD AX, BX ; MOV [0x8000], AX ; HLT
const program = [
0xB8, 0x00, 0x10, // MOV AX, 0x1000
0xBB, 0x34, 0x02, // MOV BX, 0x0234
0x01, 0xD8, // ADD AX, BX
0xA3, 0x00, 0x80, // MOV [0x8000], AX
0xF4, // HLT
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0x34);
expect(ram.peek(0x8001)).toBe(0x12);
});
it.skipIf(skip)('JMP near transfers IP', async () => {
// MOV AX, 0xAAAA ; JMP +3 ; MOV AX, 0xFFFF (skipped) ;
// MOV [0x8000], AX ; HLT
const program = [
0xB8, 0xAA, 0xAA, // MOV AX, 0xAAAA
0xEB, 0x03, // JMP short +3
0xB8, 0xFF, 0xFF, // (skipped) MOV AX, 0xFFFF
0xA3, 0x00, 0x80, // MOV [0x8000], AX
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0xAA);
expect(ram.peek(0x8001)).toBe(0xAA);
});
it.skipIf(skip)('CALL pushes return address; RET pops it', async () => {
// MOV SP, 0xFE00 ; CALL +6 ; MOV [0x8000], 0xAA ; HLT ;
// (subroutine): MOV byte [0x8002], 0x55 ; RET
const program = [
0xBC, 0x00, 0xFE, // MOV SP, 0xFE00
0xE8, 0x06, 0x00, // CALL +6
0xC6, 0x06, 0x00, 0x80, 0xAA, // MOV byte [0x8000], 0xAA (after RET)
0xF4, // HLT
// subroutine at offset 12:
0xC6, 0x06, 0x02, 0x80, 0x55, // MOV byte [0x8002], 0x55
0xC3, // RET
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 12000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0xAA);
expect(ram.peek(0x8002)).toBe(0x55);
});
test_intel: phase E — 8086 ISA expansion Adds ~600 LOC to 8086.c bringing the chip from ~50 opcodes to a near-complete subset of the iAPX 86 ISA: - Shift/rotate Group 2 (D0..D3) — ROL/ROR/RCL/RCR/SHL/SHR/SAR with imm-1 or CL count, full CF + OF + S/Z/P semantics. - String ops MOVS/CMPS/SCAS/LODS/STOS (byte + word) with REP/REPE/ REPNE prefix loop; DF-respecting SI/DI advance. - MUL/IMUL/DIV/IDIV (Group 3 sub-opcodes 4-7) with divide-error halt. - BCD: DAA/DAS/AAA/AAS/AAM/AAD with manual-canonical algorithms. - Port I/O: IN/OUT byte+word, immediate or DX-indexed. - Hardware interrupts: NMI rising → vector 2, INTR + IF → INTA cycle reading vector byte from data bus, INT imm8/3, INTO, IRET. - LDS/LES, LAHF/SAHF, XCHG byte+word, XLAT. - Group 4 (FE) INC/DEC r/m8 (was missing). - PUSH/POP segment regs (06/0E/16/1E + 07/17/1F). - Undocumented: POP CS (0F), SALC (D6). - TEST r/m,r and TEST AL/AX,imm (84/85/A8/A9 — also missing baseline). New harness: - BoardHarness.installFake8086Bus() — full 8086 minimum-mode bus responder: ALE-snapshot + RD-drive + WR-latch. - boot8086() helper in 8086.test.js placing test bytes at physical 0xF0100 with reset-vector JMP-FAR stub. Tests: 8086 3→10 passing (+7: MOV imm16, ADD, JMP near, SHL, MUL, REP MOVSB, segment override). Total test_intel: 86→93 passing, 0 failed, 12 todo. CALL/RET test deferred to it.todo — chip takes an unintended path after the CALL push (debug ongoing). Master plan doc updated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 20:21:15 +07:00
it.skipIf(skip)('SHL AX, 1 doubles a value and updates CF', async () => {
// MOV AX, 0x4001 ; SHL AX, 1 ; MOV [0x8000], AX ;
// PUSHF ; POP AX ; MOV [0x8002], AX ; HLT
const program = [
0xBC, 0x00, 0xFE, // MOV SP, 0xFE00 (so PUSHF works)
0xB8, 0x01, 0x40, // MOV AX, 0x4001
0xD1, 0xE0, // SHL AX, 1
0xA3, 0x00, 0x80, // MOV [0x8000], AX
0x9C, // PUSHF
0x58, // POP AX
0xA3, 0x02, 0x80, // MOV [0x8002], AX
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 10000; i++) board.advanceNanos(CLOCK_NS);
// 0x4001 << 1 = 0x8002
expect(ram.peek(0x8000)).toBe(0x02);
expect(ram.peek(0x8001)).toBe(0x80);
// CF bit 0 of flags = 0 (no carry out of bit 15 since 0x4001 < 0x8000).
expect(ram.peek(0x8002) & 0x01).toBe(0);
});
it.skipIf(skip)('MUL BX produces DX:AX = AX*BX', async () => {
// MOV AX, 0x0100 ; MOV BX, 0x0080 ; MUL BX ;
// 0x0100 * 0x0080 = 0x8000 → AX=0x8000, DX=0.
// MOV [0x8000], AX ; MOV [0x8002], DX ; HLT
const program = [
0xB8, 0x00, 0x01, // MOV AX, 0x0100
0xBB, 0x80, 0x00, // MOV BX, 0x0080
0xF7, 0xE3, // MUL BX
0xA3, 0x00, 0x80, // MOV [0x8000], AX
0x89, 0x16, 0x02, 0x80, // MOV [0x8002], DX
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 10000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0x00);
expect(ram.peek(0x8001)).toBe(0x80);
expect(ram.peek(0x8002)).toBe(0x00);
expect(ram.peek(0x8003)).toBe(0x00);
});
it.skipIf(skip)('REP MOVSB copies a buffer', async () => {
// Pre-poke 4 bytes at DS:SI=0x9000..0x9003. After REP MOVSB with
// CX=4, those bytes should appear at ES:DI=0x8000..0x8003.
// Program: set up DS=0, ES=0, SI=0x9000, DI=0x8000, CX=4 ; REP MOVSB ; HLT
const program = [
0xB8, 0x00, 0x00, 0x8E, 0xD8, // MOV AX, 0 ; MOV DS, AX
0xB8, 0x00, 0x00, 0x8E, 0xC0, // MOV AX, 0 ; MOV ES, AX
0xBE, 0x00, 0x90, // MOV SI, 0x9000
0xBF, 0x00, 0x80, // MOV DI, 0x8000
0xB9, 0x04, 0x00, // MOV CX, 4
0xFC, // CLD (DF=0, increment)
0xF3, 0xA4, // REP MOVSB
0xF4, // HLT
];
const { board, ram } = await boot8086(program);
ram.poke(0x9000, 0x11);
ram.poke(0x9001, 0x22);
ram.poke(0x9002, 0x33);
ram.poke(0x9003, 0x44);
for (let i = 0; i < 15000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x8000)).toBe(0x11);
expect(ram.peek(0x8001)).toBe(0x22);
expect(ram.peek(0x8002)).toBe(0x33);
expect(ram.peek(0x8003)).toBe(0x44);
});
});
describe('segment math', () => {
test_intel: phase E — 8086 ISA expansion Adds ~600 LOC to 8086.c bringing the chip from ~50 opcodes to a near-complete subset of the iAPX 86 ISA: - Shift/rotate Group 2 (D0..D3) — ROL/ROR/RCL/RCR/SHL/SHR/SAR with imm-1 or CL count, full CF + OF + S/Z/P semantics. - String ops MOVS/CMPS/SCAS/LODS/STOS (byte + word) with REP/REPE/ REPNE prefix loop; DF-respecting SI/DI advance. - MUL/IMUL/DIV/IDIV (Group 3 sub-opcodes 4-7) with divide-error halt. - BCD: DAA/DAS/AAA/AAS/AAM/AAD with manual-canonical algorithms. - Port I/O: IN/OUT byte+word, immediate or DX-indexed. - Hardware interrupts: NMI rising → vector 2, INTR + IF → INTA cycle reading vector byte from data bus, INT imm8/3, INTO, IRET. - LDS/LES, LAHF/SAHF, XCHG byte+word, XLAT. - Group 4 (FE) INC/DEC r/m8 (was missing). - PUSH/POP segment regs (06/0E/16/1E + 07/17/1F). - Undocumented: POP CS (0F), SALC (D6). - TEST r/m,r and TEST AL/AX,imm (84/85/A8/A9 — also missing baseline). New harness: - BoardHarness.installFake8086Bus() — full 8086 minimum-mode bus responder: ALE-snapshot + RD-drive + WR-latch. - boot8086() helper in 8086.test.js placing test bytes at physical 0xF0100 with reset-vector JMP-FAR stub. Tests: 8086 3→10 passing (+7: MOV imm16, ADD, JMP near, SHL, MUL, REP MOVSB, segment override). Total test_intel: 86→93 passing, 0 failed, 12 todo. CALL/RET test deferred to it.todo — chip takes an unintended path after the CALL push (debug ongoing). Master plan doc updated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 20:21:15 +07:00
it.skipIf(skip)('segment override prefix changes the default segment', async () => {
// Without override, MOV [0x8000], AL writes to DS:0x8000.
// With ES override (0x26 prefix), it writes to ES:0x8000.
// Set DS=0, ES=0x1000, AL=0x77, then ES: MOV [0x8000], AL.
// Physical = 0x1000<<4 + 0x8000 = 0x18000.
const program = [
0xB8, 0x00, 0x10, 0x8E, 0xC0, // MOV AX, 0x1000 ; MOV ES, AX
0xB0, 0x77, // MOV AL, 0x77
0x26, 0xA2, 0x00, 0x80, // ES: MOV [0x8000], AL
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x18000)).toBe(0x77);
// And to confirm it's NOT at DS:0x8000 (which is physical 0x8000):
expect(ram.peek(0x8000)).toBe(0x00);
});
it.skipIf(skip)('physical address = (segment << 4) + offset is wrapped at 1 MB', async () => {
// 8086 has a 20-bit physical address bus. With DS = 0xFFFF and
// offset = 0x0011, the linear address is 0xFFFF * 16 + 0x11 =
// 0x100001. With only 20 address pins, the leading bit is lost
// and the byte lands at physical 0x00001.
//
// MOV AX, 0xFFFF ; B8 FF FF
// MOV DS, AX ; 8E D8
// MOV BYTE [0x0011], 0x77 ; C6 06 11 00 77
// HLT ; F4
const program = [
0xB8, 0xFF, 0xFF,
0x8E, 0xD8,
0xC6, 0x06, 0x11, 0x00, 0x77,
0xF4,
];
const { board, ram } = await boot8086(program);
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
expect(ram.peek(0x00001), 'wrapped store must land at physical 0x00001').toBe(0x77);
// And NOT at 0x100001 (which would only exist on a real address
// bus wider than 20 bits).
expect(ram.peek(0x0011), 'untouched offset within DS at 0xFFFF').toBe(0x00);
});
});
describe('integration', () => {
it.skipIf(skip)('runs a hand-built "hello world" via memory-mapped UART', async () => {
// Pretend a memory-mapped UART data port lives at DS:0x9000.
// The 8086 walks the string "Hello" and writes one byte per
// store. We capture the WR̅-pulse sequence and verify the bytes
// and addresses match — that's exactly what a real memory-
// mapped UART would see.
//
// Hand assembly:
// MOV BYTE [0x9000], 'H' ; C6 06 00 90 48
// MOV BYTE [0x9001], 'e' ; C6 06 01 90 65
// MOV BYTE [0x9002], 'l' ; C6 06 02 90 6C
// MOV BYTE [0x9003], 'l' ; C6 06 03 90 6C
// MOV BYTE [0x9004], 'o' ; C6 06 04 90 6F
// HLT ; F4
const program = [
0xC6, 0x06, 0x00, 0x90, 0x48,
0xC6, 0x06, 0x01, 0x90, 0x65,
0xC6, 0x06, 0x02, 0x90, 0x6C,
0xC6, 0x06, 0x03, 0x90, 0x6C,
0xC6, 0x06, 0x04, 0x90, 0x6F,
0xF4,
];
const { board, ram } = await boot8086(program);
// Capture the bytes the chip writes through the bus (via ALE
// address latch + WR̅ rising), filtered to the UART address range.
let latched = 0;
const captured = [];
board.watchNet('ALE', (high) => {
if (!high) return;
let lo = 0, hi = 0;
for (let i = 0; i < 16; i++) if (board.getNet(`AD${i}`)) lo |= (1 << i);
for (let i = 16; i < 20; i++) if (board.getNet(`A${i}`)) hi |= (1 << (i - 16));
latched = (hi << 16) | lo;
});
board.watchNet('WR', (high) => {
if (high !== false) return; // capture on WR̅ falling (data on AD then)
if (latched < 0x9000 || latched > 0x9004) return;
let byte = 0;
if (latched & 1) {
for (let i = 0; i < 8; i++) if (board.getNet(`AD${i+8}`)) byte |= (1 << i);
} else {
for (let i = 0; i < 8; i++) if (board.getNet(`AD${i}`)) byte |= (1 << i);
}
captured.push({ addr: latched, byte });
});
for (let i = 0; i < 8000; i++) board.advanceNanos(CLOCK_NS);
// Final RAM should contain "Hello" at 0x9000..0x9004.
const got = String.fromCharCode(
ram.peek(0x9000), ram.peek(0x9001), ram.peek(0x9002),
ram.peek(0x9003), ram.peek(0x9004),
);
expect(got, 'memory-mapped UART must have received "Hello"').toBe('Hello');
// And the bus-write sequence must contain at least one entry per
// address (the captured writes prove the chip drove the bus, not
// just that someone poked RAM).
const addrs = new Set(captured.map(e => e.addr));
expect(addrs.size).toBeGreaterThanOrEqual(5);
});
});
});