@neovand/zilion 0.1.1 → 0.1.2

package/README.md

@@ -145,7 +145,7 @@ Emulator bugs hide in undocumented corners, so Zilion's Z80 core is developed ag
 
 The core implements the full documented instruction set, the **CB**, **ED**, **DD/FD (IX/IY)**, and **DDCB/FDCB** prefix pages (including the undocumented DDCB register-copy side effect and the CPI/CPD undocumented-flag quirk), shadow registers, and `EXX`/`EX AF,AF'`.
 
-> **Scope & honesty:** Zilion is a batch execution core, not a cycle-accurate machine emulator. There are no interrupts and no I/O ports (`IN` reads 0, `OUT` is a no-op), and time is counted in instructions, not T-states — every instruction advances the step counter by one. The `R` refresh register increments per M1 (opcode/prefix) fetch and `HALT` idles correctly (re-executing itself until the step budget runs out, matching a real Z80's PC/R behavior). If you need cycle-accurate single-machine emulation, use a dedicated emulator; if you need to run a zillion Z80s fast, use Zilion.
+> **Scope & honesty:** Zilion is a batch execution core, not a cycle-accurate machine emulator. There are no interrupts and no I/O ports (`IN` reads 0, `OUT` is a no-op), and time is counted in instructions, not T-states — every instruction advances the step counter by one. The `R` refresh register increments per M1 (opcode/prefix) fetch and `HALT` idles correctly (re-executing itself until the step budget runs out, matching a real Z80's PC/R behavior). The one documented simplification: the internal `WZ`/`memptr` register isn't modeled, so the *undocumented* F3/F5 flag bits of `BIT n,(HL)` come from the operand rather than the memory address (this never affects control flow). Everything else — the full documented instruction set across the base, CB, ED, and DD/FD/DDCB/FDCB pages — is differential-tested against a real-Z80 reference to zero divergence in documented behavior. If you need cycle-accurate single-machine emulation, use a dedicated emulator; if you need to run a zillion Z80s fast, use Zilion.
 
 ## Performance
 

package/dist/z80-core.wgsl.d.ts

@@ -1,2 +1,2 @@
-export declare const Z80_CORE_WGSL = "\n// === Z80 CPU State (per invocation) ===\nvar<private> cpu_a: u32;\nvar<private> cpu_f: u32;\nvar<private> cpu_b: u32;\nvar<private> cpu_c: u32;\nvar<private> cpu_d: u32;\nvar<private> cpu_e: u32;\nvar<private> cpu_h: u32;\nvar<private> cpu_l: u32;\nvar<private> cpu_sp: u32;\nvar<private> cpu_pc: u32;\nvar<private> cpu_a2: u32;\nvar<private> cpu_f2: u32;\nvar<private> cpu_b2: u32;\nvar<private> cpu_c2: u32;\nvar<private> cpu_d2: u32;\nvar<private> cpu_e2: u32;\nvar<private> cpu_h2: u32;\nvar<private> cpu_l2: u32;\nvar<private> cpu_halted: u32;\nvar<private> cpu_iff1: u32;\nvar<private> cpu_iff2: u32;\nvar<private> cpu_ix: u32;\nvar<private> cpu_iy: u32;\nvar<private> cpu_i: u32; // interrupt vector register (I)\nvar<private> cpu_r: u32; // memory refresh register (R); bits 0-6 count M1 cycles, bit 7 preserved\n\n// Increment R once per M1 (opcode/prefix) fetch: 7-bit counter, bit 7 preserved.\nfn r_inc() { cpu_r = (cpu_r & 0x80u) | ((cpu_r + 1u) & 0x7fu); }\n// Index-prefix state for the instruction currently executing:\n//   idx_mode: 0 = HL, 1 = IX, 2 = IY\n//   idx_disp: sign-extended displacement for (IX+d)/(IY+d)\n//   idx_uses_mem: 1 when this instruction dereferences (IX+d)/(IY+d), which\n//     means H/L operands are NOT substituted by IXH/IXL (real Z80 rule).\nvar<private> idx_mode: u32;\nvar<private> idx_disp: u32;\nvar<private> idx_uses_mem: u32;\n\n// HOST CONTRACT: the host shader must declare the following BEFORE this core:\n//   fn mem_read(addr: u32) -> u32               // read one byte from memory\n//   fn mem_write(addr: u32, val: u32)           // write one byte to memory\n//   fn on_fetch_opcode(op: u32) -> bool         // return true to skip (NOP) an\n//                                               // opcode after prefix resolution\n// This lets the host choose the memory model (mask, modulo, storage buffer, \u2026)\n// and hook opcode execution (e.g. instruction suppression). See buildComputeShader.\n\n// Z80 flag bits\nconst CF: u32 = 0x01u;\nconst NF: u32 = 0x02u;\nconst PF: u32 = 0x04u;\nconst F3: u32 = 0x08u;\nconst HF: u32 = 0x10u;\nconst F5: u32 = 0x20u;\nconst ZF: u32 = 0x40u;\nconst SFl: u32 = 0x80u;\n\n\nfn z80_fetch() -> u32 {\n    let val = mem_read(cpu_pc);\n    cpu_pc = (cpu_pc + 1u) & 0xffffu;\n    return val;\n}\n\nfn z80_fetch_word() -> u32 {\n    let lo = z80_fetch();\n    let hi = z80_fetch();\n    return (hi << 8u) | lo;\n}\n\nfn z80_push16(val: u32) {\n    cpu_sp = (cpu_sp - 1u) & 0xffffu;\n    mem_write(cpu_sp, (val >> 8u) & 0xffu);\n    cpu_sp = (cpu_sp - 1u) & 0xffffu;\n    mem_write(cpu_sp, val & 0xffu);\n}\n\nfn z80_pop16() -> u32 {\n    let lo = mem_read(cpu_sp);\n    cpu_sp = (cpu_sp + 1u) & 0xffffu;\n    let hi = mem_read(cpu_sp);\n    cpu_sp = (cpu_sp + 1u) & 0xffffu;\n    return (hi << 8u) | lo;\n}\n\nfn signed_byte(b: u32) -> i32 {\n    let sb = i32(b);\n    if (sb > 127) { return sb - 256; }\n    return sb;\n}\n\n// === Register Access ===\nfn get_bc() -> u32 { return (cpu_b << 8u) | cpu_c; }\nfn get_de() -> u32 { return (cpu_d << 8u) | cpu_e; }\nfn get_hl() -> u32 { return (cpu_h << 8u) | cpu_l; }\nfn get_af() -> u32 { return (cpu_a << 8u) | cpu_f; }\n\nfn set_bc(v: u32) { cpu_b = (v >> 8u) & 0xffu; cpu_c = v & 0xffu; }\nfn set_de(v: u32) { cpu_d = (v >> 8u) & 0xffu; cpu_e = v & 0xffu; }\nfn set_hl(v: u32) { cpu_h = (v >> 8u) & 0xffu; cpu_l = v & 0xffu; }\nfn set_af(v: u32) { cpu_a = (v >> 8u) & 0xffu; cpu_f = v & 0xffu; }\n\nfn get_reg(idx: u32) -> u32 {\n    switch(idx) {\n        case 0u: { return cpu_b; }\n        case 1u: { return cpu_c; }\n        case 2u: { return cpu_d; }\n        case 3u: { return cpu_e; }\n        case 4u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { return (idx_reg16() >> 8u) & 0xffu; } // IXH/IYH\n            return cpu_h;\n        }\n        case 5u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { return idx_reg16() & 0xffu; } // IXL/IYL\n            return cpu_l;\n        }\n        case 6u: { return mem_read(idx_addr()); }\n        case 7u: { return cpu_a; }\n        default: { return 0u; }\n    }\n}\n\nfn set_reg(idx: u32, val: u32) {\n    let v = val & 0xffu;\n    switch(idx) {\n        case 0u: { cpu_b = v; }\n        case 1u: { cpu_c = v; }\n        case 2u: { cpu_d = v; }\n        case 3u: { cpu_e = v; }\n        case 4u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { set_idx_reg16((idx_reg16() & 0x00ffu) | (v << 8u)); }\n            else { cpu_h = v; }\n        }\n        case 5u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { set_idx_reg16((idx_reg16() & 0xff00u) | v); }\n            else { cpu_l = v; }\n        }\n        case 6u: { mem_write(idx_addr(), v); }\n        case 7u: { cpu_a = v; }\n        default: {}\n    }\n}\n\nfn get_reg16(idx: u32) -> u32 {\n    switch(idx) {\n        case 0u: { return get_bc(); }\n        case 1u: { return get_de(); }\n        case 2u: { return idx_reg16(); } // HL / IX / IY\n        case 3u: { return cpu_sp; }\n        default: { return 0u; }\n    }\n}\n\nfn set_reg16(idx: u32, val: u32) {\n    let v = val & 0xffffu;\n    switch(idx) {\n        case 0u: { set_bc(v); }\n        case 1u: { set_de(v); }\n        case 2u: { set_idx_reg16(v); } // HL / IX / IY\n        case 3u: { cpu_sp = v; }\n        default: {}\n    }\n}\n\nfn get_reg16_af(idx: u32) -> u32 {\n    if (idx == 3u) { return get_af(); }\n    return get_reg16(idx);\n}\n\nfn set_reg16_af(idx: u32, val: u32) {\n    if (idx == 3u) { set_af(val); } else { set_reg16(idx, val); }\n}\n\n// === Index register (IX/IY) helpers ===\n// The 16-bit register the current prefix maps HL to (HL itself when no prefix).\nfn idx_reg16() -> u32 {\n    if (idx_mode == 1u) { return cpu_ix; }\n    if (idx_mode == 2u) { return cpu_iy; }\n    return get_hl();\n}\nfn set_idx_reg16(v: u32) {\n    if (idx_mode == 1u) { cpu_ix = v & 0xffffu; }\n    else if (idx_mode == 2u) { cpu_iy = v & 0xffffu; }\n    else { set_hl(v); }\n}\n// Address used for (HL) / (IX+d) / (IY+d).\nfn idx_addr() -> u32 {\n    if (idx_mode != 0u) { return (idx_reg16() + idx_disp) & 0xffffu; }\n    return get_hl();\n}\n// Sign-extend a displacement byte to 16 bits (two's complement).\nfn signext(b: u32) -> u32 {\n    if (b >= 0x80u) { return b | 0xff00u; }\n    return b;\n}\n// Does this main opcode dereference (HL)? (Determines displacement fetch and\n// whether H/L operands are IXH/IXL or real H/L under a DD/FD prefix.)\nfn op_uses_hl_mem(op: u32) -> bool {\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    if (x == 1u) { return (y == 6u || z == 6u) && !(y == 6u && z == 6u); } // LD r,(HL)/(HL),r (not HALT)\n    if (x == 2u) { return z == 6u; }                                        // ALU A,(HL)\n    if (x == 0u) {\n        if (z == 4u || z == 5u || z == 6u) { return y == 6u; }              // INC/DEC (HL), LD (HL),n\n        return false;\n    }\n    return false;\n}\n// Write a REAL 8-bit register (no IX/IY substitution) \u2014 used by the\n// undocumented DDCB/FDCB register-copy side effect.\nfn set_reg_raw(idx: u32, val: u32) {\n    let v = val & 0xffu;\n    switch(idx) {\n        case 0u: { cpu_b = v; }\n        case 1u: { cpu_c = v; }\n        case 2u: { cpu_d = v; }\n        case 3u: { cpu_e = v; }\n        case 4u: { cpu_h = v; }\n        case 5u: { cpu_l = v; }\n        case 7u: { cpu_a = v; }\n        default: {}\n    }\n}\n\n// === Flag Helpers ===\nfn sz_flags(val: u32) -> u32 {\n    var f = val & SFl;\n    if (val == 0u) { f |= ZF; }\n    f |= val & (F3 | F5);\n    return f;\n}\n\nfn parity(val: u32) -> bool {\n    var p = val;\n    p ^= p >> 4u;\n    p ^= p >> 2u;\n    p ^= p >> 1u;\n    return (p & 1u) == 0u;\n}\n\nfn check_cc(cc: u32) -> bool {\n    switch(cc) {\n        case 0u: { return (cpu_f & ZF) == 0u; }\n        case 1u: { return (cpu_f & ZF) != 0u; }\n        case 2u: { return (cpu_f & CF) == 0u; }\n        case 3u: { return (cpu_f & CF) != 0u; }\n        case 4u: { return (cpu_f & PF) == 0u; }\n        case 5u: { return (cpu_f & PF) != 0u; }\n        case 6u: { return (cpu_f & SFl) == 0u; }\n        case 7u: { return (cpu_f & SFl) != 0u; }\n        default: { return false; }\n    }\n}\n\n// === ALU ===\nfn z80_alu(op: u32, val: u32) {\n    let a = cpu_a;\n    let c = cpu_f & CF;\n    switch(op) {\n        case 0u: { // ADD\n            let r = a + val;\n            cpu_f = sz_flags(r & 0xffu) | select(0u, CF, r > 0xffu) |\n                    ((a ^ val ^ r) & HF) |\n                    select(0u, PF, ((~(a ^ val)) & (a ^ r) & 0x80u) != 0u);\n            cpu_a = r & 0xffu;\n        }\n        case 1u: { // ADC\n            let r = a + val + c;\n            cpu_f = sz_flags(r & 0xffu) | select(0u, CF, r > 0xffu) |\n                    ((a ^ val ^ r) & HF) |\n                    select(0u, PF, ((~(a ^ val)) & (a ^ r) & 0x80u) != 0u);\n            cpu_a = r & 0xffu;\n        }\n        case 2u: { // SUB\n            let r = i32(a) - i32(val);\n            let ru = u32(r) & 0xffu;\n            cpu_f = sz_flags(ru) | NF | select(0u, CF, r < 0) |\n                    ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n            cpu_a = ru;\n        }\n        case 3u: { // SBC\n            let r = i32(a) - i32(val) - i32(c);\n            let ru = u32(r) & 0xffu;\n            cpu_f = sz_flags(ru) | NF | select(0u, CF, r < 0) |\n                    ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n            cpu_a = ru;\n        }\n        case 4u: { // AND\n            cpu_a = a & val;\n            cpu_f = sz_flags(cpu_a) | HF | select(0u, PF, parity(cpu_a));\n        }\n        case 5u: { // XOR\n            cpu_a = a ^ val;\n            cpu_f = sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        case 6u: { // OR\n            cpu_a = a | val;\n            cpu_f = sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        case 7u: { // CP\n            let r = i32(a) - i32(val);\n            let ru = u32(r) & 0xffu;\n            cpu_f = (ru & SFl) | select(0u, ZF, ru == 0u) | (val & (F3 | F5)) | NF |\n                    select(0u, CF, r < 0) | ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n        }\n        default: {}\n    }\n}\n\nfn z80_inc8(val: u32) -> u32 {\n    let r = (val + 1u) & 0xffu;\n    cpu_f = (cpu_f & CF) | sz_flags(r) |\n            select(0u, PF, val == 0x7fu) |\n            select(0u, HF, (r & 0x0fu) == 0u);\n    return r;\n}\n\nfn z80_dec8(val: u32) -> u32 {\n    let r = (val - 1u) & 0xffu;\n    cpu_f = (cpu_f & CF) | sz_flags(r) | NF |\n            select(0u, PF, val == 0x80u) |\n            select(0u, HF, (val & 0x0fu) == 0u);\n    return r;\n}\n\nfn z80_add_hl(val: u32) {\n    let hl = idx_reg16(); // ADD HL,rp / ADD IX,rp / ADD IY,rp\n    let r = hl + val;\n    cpu_f = (cpu_f & (SFl | ZF | PF)) |\n            select(0u, CF, r > 0xffffu) |\n            select(0u, HF, ((hl ^ val ^ r) & 0x1000u) != 0u) |\n            ((r >> 8u) & (F3 | F5));\n    set_idx_reg16(r & 0xffffu);\n}\n\n// === Rotate/Shift for accumulator ===\nfn z80_rot_accum(y: u32) {\n    let a = cpu_a;\n    let c = cpu_f & CF;\n    let keep = cpu_f & (SFl | ZF | PF);\n    switch(y) {\n        case 0u: { // RLCA\n            cpu_a = ((a << 1u) | (a >> 7u)) & 0xffu;\n            cpu_f = keep | (a >> 7u) | (cpu_a & (F3 | F5));\n        }\n        case 1u: { // RRCA\n            cpu_a = ((a >> 1u) | (a << 7u)) & 0xffu;\n            cpu_f = keep | (a & 1u) | (cpu_a & (F3 | F5));\n        }\n        case 2u: { // RLA\n            cpu_a = ((a << 1u) | c) & 0xffu;\n            cpu_f = keep | (a >> 7u) | (cpu_a & (F3 | F5));\n        }\n        case 3u: { // RRA\n            cpu_a = ((a >> 1u) | (c << 7u)) & 0xffu;\n            cpu_f = keep | (a & 1u) | (cpu_a & (F3 | F5));\n        }\n        case 4u: { // DAA\n            var correction = 0u;\n            var carry = c;\n            if ((cpu_f & HF) != 0u || (a & 0x0fu) > 9u) { correction |= 0x06u; }\n            if (c != 0u || a > 0x99u) { correction |= 0x60u; carry = 1u; }\n            if ((cpu_f & NF) != 0u) { cpu_a = (a - correction) & 0xffu; }\n            else { cpu_a = (a + correction) & 0xffu; }\n            cpu_f = (cpu_f & NF) | sz_flags(cpu_a) | carry |\n                    ((a ^ cpu_a) & HF) | select(0u, PF, parity(cpu_a));\n        }\n        case 5u: { // CPL\n            cpu_a = (~a) & 0xffu;\n            cpu_f = (cpu_f & (SFl | ZF | PF | CF)) | HF | NF | (cpu_a & (F3 | F5));\n        }\n        case 6u: { // SCF\n            cpu_f = (cpu_f & (SFl | ZF | PF)) | CF | (cpu_a & (F3 | F5));\n        }\n        case 7u: { // CCF\n            cpu_f = (cpu_f & (SFl | ZF | PF)) |\n                    select(0u, HF, c != 0u) |\n                    select(CF, 0u, c != 0u) |\n                    (cpu_a & (F3 | F5));\n        }\n        default: {}\n    }\n}\n\n// === CB Prefix (bit ops, rotates, shifts) ===\nfn z80_cb_rot(op: u32, val: u32) -> u32 {\n    let c = cpu_f & CF;\n    var r = 0u;\n    switch(op) {\n        case 0u: { r = ((val << 1u) | (val >> 7u)) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 1u: { r = ((val >> 1u) | (val << 7u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 2u: { r = ((val << 1u) | c) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 3u: { r = ((val >> 1u) | (c << 7u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 4u: { r = (val << 1u) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 5u: { r = ((val >> 1u) | (val & 0x80u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 6u: { r = ((val << 1u) | 1u) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 7u: { r = (val >> 1u) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        default: { r = val; }\n    }\n    return r;\n}\n\nfn z80_exec_cb() {\n    let op = z80_fetch();\n    r_inc(); // M1: CB-page opcode fetch\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    let val = get_reg(z);\n    switch(x) {\n        case 0u: { set_reg(z, z80_cb_rot(y, val)); }\n        case 1u: { // BIT\n            cpu_f = (cpu_f & CF) | HF |\n                    select(0u, ZF | PF, (val & (1u << y)) == 0u) |\n                    select(0u, SFl, y == 7u && (val & 0x80u) != 0u) |\n                    (val & (F3 | F5));\n        }\n        case 2u: { set_reg(z, val & ~(1u << y)); }\n        case 3u: { set_reg(z, val | (1u << y)); }\n        default: {}\n    }\n}\n\n// DDCB / FDCB: operates on (IX+d)/(IY+d). The displacement (idx_disp) has\n// already been fetched. For rot/shift/RES/SET the result is written to memory\n// AND (undocumented) copied to the real register in the low 3 bits unless it is\n// 6. For BIT, the undocumented F3/F5 come from the high byte of the address.\nfn z80_exec_idxcb(cbop: u32) {\n    let addr = (idx_reg16() + idx_disp) & 0xffffu;\n    let cx = (cbop >> 6u) & 3u;\n    let cy = (cbop >> 3u) & 7u;\n    let cz = cbop & 7u;\n    let val = mem_read(addr);\n    switch(cx) {\n        case 0u: {\n            let r = z80_cb_rot(cy, val);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        case 1u: { // BIT n,(IX+d)\n            cpu_f = (cpu_f & CF) | HF |\n                    select(0u, ZF | PF, (val & (1u << cy)) == 0u) |\n                    select(0u, SFl, cy == 7u && (val & 0x80u) != 0u) |\n                    ((addr >> 8u) & (F3 | F5));\n        }\n        case 2u: {\n            let r = val & ~(1u << cy);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        case 3u: {\n            let r = val | (1u << cy);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        default: {}\n    }\n}\n\n// === Block Transfer (ED prefix) ===\nfn z80_ldi() {\n    let val = mem_read(get_hl());\n    mem_write(get_de(), val);\n    set_hl((get_hl() + 1u) & 0xffffu);\n    set_de((get_de() + 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    let n = (val + cpu_a) & 0xffu;\n    cpu_f = (cpu_f & (SFl | ZF | CF)) |\n            select(0u, PF, get_bc() != 0u) |\n            (n & F3) | select(0u, F5, (n & 0x02u) != 0u);\n}\n\nfn z80_ldd() {\n    let val = mem_read(get_hl());\n    mem_write(get_de(), val);\n    set_hl((get_hl() - 1u) & 0xffffu);\n    set_de((get_de() - 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    let n = (val + cpu_a) & 0xffu;\n    cpu_f = (cpu_f & (SFl | ZF | CF)) |\n            select(0u, PF, get_bc() != 0u) |\n            (n & F3) | select(0u, F5, (n & 0x02u) != 0u);\n}\n\nfn z80_cpi() {\n    let val = mem_read(get_hl());\n    let r = (cpu_a - val) & 0xffu;\n    let hf = (cpu_a ^ val ^ r) & HF;\n    // Undocumented F3/F5 come from n = A-(HL)-HF (bit 3 -> F3, bit 1 -> F5).\n    let n = (r - select(0u, 1u, hf != 0u)) & 0xffu;\n    set_hl((get_hl() + 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    cpu_f = (cpu_f & CF) | (r & SFl) | select(0u, ZF, r == 0u) | NF |\n            hf | select(0u, PF, get_bc() != 0u) |\n            (n & F3) | ((n & 0x02u) << 4u);\n}\n\nfn z80_cpd() {\n    let val = mem_read(get_hl());\n    let r = (cpu_a - val) & 0xffu;\n    let hf = (cpu_a ^ val ^ r) & HF;\n    let n = (r - select(0u, 1u, hf != 0u)) & 0xffu;\n    set_hl((get_hl() - 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    cpu_f = (cpu_f & CF) | (r & SFl) | select(0u, ZF, r == 0u) | NF |\n            hf | select(0u, PF, get_bc() != 0u) |\n            (n & F3) | ((n & 0x02u) << 4u);\n}\n\n// === ED Prefix ===\nfn z80_exec_ed() {\n    let op = z80_fetch();\n    r_inc(); // M1: ED-page opcode fetch\n    switch(op) {\n        case 0xa0u: { z80_ldi(); }\n        case 0xa8u: { z80_ldd(); }\n        case 0xb0u: { z80_ldi(); if (get_bc() != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } } // LDIR\n        case 0xb8u: { z80_ldd(); if (get_bc() != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } } // LDDR\n        case 0xa1u: { z80_cpi(); }\n        case 0xa9u: { z80_cpd(); }\n        case 0xb1u: { z80_cpi(); if (get_bc() != 0u && (cpu_f & ZF) == 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } }\n        case 0xb9u: { z80_cpd(); if (get_bc() != 0u && (cpu_f & ZF) == 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } }\n        // NEG\n        case 0x44u, 0x4cu, 0x54u, 0x5cu, 0x64u, 0x6cu, 0x74u, 0x7cu: {\n            let a = cpu_a; cpu_a = 0u; z80_alu(2u, a);\n        }\n        // RETN/RETI\n        case 0x45u, 0x4du, 0x55u, 0x5du, 0x65u, 0x6du, 0x75u, 0x7du: {\n            cpu_iff1 = cpu_iff2; cpu_pc = z80_pop16();\n        }\n        // LD I,A / LD R,A / LD A,I / LD A,R\n        case 0x47u: { cpu_i = cpu_a; } // LD I,A\n        case 0x4fu: { cpu_r = cpu_a; } // LD R,A\n        // LD A,I / LD A,R: load I or R into A. Flags: S/Z (+F3/F5) from the loaded\n        // value, PF = IFF2, N/H reset, C preserved.\n        case 0x57u: { cpu_a = cpu_i; cpu_f = (cpu_f & CF) | sz_flags(cpu_i) | select(0u, PF, cpu_iff2 != 0u); }\n        case 0x5fu: { cpu_a = cpu_r; cpu_f = (cpu_f & CF) | sz_flags(cpu_r) | select(0u, PF, cpu_iff2 != 0u); }\n        // LD (nn), rr\n        case 0x43u, 0x53u, 0x63u, 0x73u: {\n            let nn = z80_fetch_word();\n            let rp = (op >> 4u) & 3u;\n            let val = get_reg16(rp);\n            mem_write(nn, val & 0xffu);\n            mem_write((nn + 1u) & 0xffffu, (val >> 8u) & 0xffu);\n        }\n        // LD rr, (nn)\n        case 0x4bu, 0x5bu, 0x6bu, 0x7bu: {\n            let nn = z80_fetch_word();\n            let rp = (op >> 4u) & 3u;\n            let lo = mem_read(nn);\n            let hi = mem_read((nn + 1u) & 0xffffu);\n            set_reg16(rp, (hi << 8u) | lo);\n        }\n        // ADC HL, rr\n        case 0x4au, 0x5au, 0x6au, 0x7au: {\n            let rp = (op >> 4u) & 3u;\n            let hl = get_hl();\n            let val = get_reg16(rp);\n            let c = cpu_f & CF;\n            let r = hl + val + c;\n            cpu_f = ((r >> 8u) & SFl) | select(0u, ZF, (r & 0xffffu) == 0u) |\n                    select(0u, HF, ((hl ^ val ^ r) & 0x1000u) != 0u) |\n                    select(0u, PF, ((~(hl ^ val)) & (hl ^ r) & 0x8000u) != 0u) |\n                    select(0u, CF, r > 0xffffu) | ((r >> 8u) & (F3 | F5));\n            set_hl(r & 0xffffu);\n        }\n        // SBC HL, rr\n        case 0x42u, 0x52u, 0x62u, 0x72u: {\n            let rp = (op >> 4u) & 3u;\n            let hl = get_hl();\n            let val = get_reg16(rp);\n            let c = cpu_f & CF;\n            let r = i32(hl) - i32(val) - i32(c);\n            let ru = u32(r) & 0xffffu;\n            cpu_f = ((ru >> 8u) & SFl) | select(0u, ZF, ru == 0u) | NF |\n                    select(0u, HF, ((hl ^ val ^ u32(r)) & 0x1000u) != 0u) |\n                    select(0u, PF, (((hl ^ val) & (hl ^ u32(r))) & 0x8000u) != 0u) |\n                    select(0u, CF, r < 0) | ((ru >> 8u) & (F3 | F5));\n            set_hl(ru);\n        }\n        // RRD\n        case 0x67u: {\n            let m = mem_read(get_hl());\n            mem_write(get_hl(), ((cpu_a << 4u) | (m >> 4u)) & 0xffu);\n            cpu_a = (cpu_a & 0xf0u) | (m & 0x0fu);\n            cpu_f = (cpu_f & CF) | sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        // RLD\n        case 0x6fu: {\n            let m = mem_read(get_hl());\n            mem_write(get_hl(), ((m << 4u) | (cpu_a & 0x0fu)) & 0xffu);\n            cpu_a = (cpu_a & 0xf0u) | (m >> 4u);\n            cpu_f = (cpu_f & CF) | sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        // IN r,(C) - simplified, just set to 0\n        case 0x40u, 0x48u, 0x50u, 0x58u, 0x60u, 0x68u, 0x70u, 0x78u: {\n            set_reg((op >> 3u) & 7u, 0u);\n        }\n        default: {} // unknown ED ops = NOP\n    }\n}\n\n// === Main Opcode Execution ===\nfn z80_exec_x0(y: u32, z: u32, p: u32, q: u32) {\n    switch(z) {\n        case 0u: {\n            switch(y) {\n                case 0u: {} // NOP\n                case 1u: { // EX AF,AF'\n                    var t = cpu_a; cpu_a = cpu_a2; cpu_a2 = t;\n                    t = cpu_f; cpu_f = cpu_f2; cpu_f2 = t;\n                }\n                case 2u: { // DJNZ\n                    let d = signed_byte(z80_fetch());\n                    cpu_b = (cpu_b - 1u) & 0xffu;\n                    if (cpu_b != 0u) { cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu; }\n                }\n                case 3u: { // JR\n                    let d = signed_byte(z80_fetch());\n                    cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu;\n                }\n                default: { // JR cc (y-4)\n                    let d = signed_byte(z80_fetch());\n                    if (check_cc(y - 4u)) { cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu; }\n                }\n            }\n        }\n        case 1u: {\n            if (q == 0u) { set_reg16(p, z80_fetch_word()); }\n            else { z80_add_hl(get_reg16(p)); }\n        }\n        case 2u: {\n            if (q == 0u) {\n                switch(p) {\n                    case 0u: { mem_write(get_bc(), cpu_a); }\n                    case 1u: { mem_write(get_de(), cpu_a); }\n                    case 2u: { let nn = z80_fetch_word(); let hl = idx_reg16(); mem_write(nn, hl & 0xffu); mem_write((nn+1u) & 0xffffu, (hl >> 8u) & 0xffu); }\n                    case 3u: { mem_write(z80_fetch_word(), cpu_a); }\n                    default: {}\n                }\n            } else {\n                switch(p) {\n                    case 0u: { cpu_a = mem_read(get_bc()); }\n                    case 1u: { cpu_a = mem_read(get_de()); }\n                    case 2u: { let nn = z80_fetch_word(); let lo = mem_read(nn); let hi = mem_read((nn+1u) & 0xffffu); set_idx_reg16((hi << 8u) | lo); }\n                    case 3u: { cpu_a = mem_read(z80_fetch_word()); }\n                    default: {}\n                }\n            }\n        }\n        case 3u: {\n            if (q == 0u) { set_reg16(p, (get_reg16(p) + 1u) & 0xffffu); }\n            else { set_reg16(p, (get_reg16(p) - 1u) & 0xffffu); }\n        }\n        case 4u: { set_reg(y, z80_inc8(get_reg(y))); }\n        case 5u: { set_reg(y, z80_dec8(get_reg(y))); }\n        case 6u: { set_reg(y, z80_fetch()); }\n        case 7u: { z80_rot_accum(y); }\n        default: {}\n    }\n}\n\nfn z80_exec_x3(y: u32, z: u32, p: u32, q: u32) {\n    switch(z) {\n        case 0u: { if (check_cc(y)) { cpu_pc = z80_pop16(); } }\n        case 1u: {\n            if (q == 0u) { set_reg16_af(p, z80_pop16()); }\n            else {\n                switch(p) {\n                    case 0u: { cpu_pc = z80_pop16(); } // RET\n                    case 1u: { // EXX\n                        var t = cpu_b; cpu_b = cpu_b2; cpu_b2 = t;\n                        t = cpu_c; cpu_c = cpu_c2; cpu_c2 = t;\n                        t = cpu_d; cpu_d = cpu_d2; cpu_d2 = t;\n                        t = cpu_e; cpu_e = cpu_e2; cpu_e2 = t;\n                        t = cpu_h; cpu_h = cpu_h2; cpu_h2 = t;\n                        t = cpu_l; cpu_l = cpu_l2; cpu_l2 = t;\n                    }\n                    case 2u: { cpu_pc = idx_reg16(); } // JP (HL)/(IX)/(IY)\n                    case 3u: { cpu_sp = idx_reg16(); } // LD SP,HL/IX/IY\n                    default: {}\n                }\n            }\n        }\n        case 2u: { let nn = z80_fetch_word(); if (check_cc(y)) { cpu_pc = nn; } }\n        case 3u: {\n            switch(y) {\n                case 0u: { cpu_pc = z80_fetch_word(); } // JP nn\n                case 1u: { z80_exec_cb(); }\n                case 2u: { z80_fetch(); } // OUT (n),A - no I/O device, discard (matches zff outPort no-op)\n                case 3u: { z80_fetch(); cpu_a = 0u; } // IN A,(n) - no I/O device, reads 0 (matches zff inPort\u21920)\n                case 4u: { // EX (SP),HL / EX (SP),IX / EX (SP),IY\n                    let lo = mem_read(cpu_sp);\n                    let hi = mem_read((cpu_sp + 1u) & 0xffffu);\n                    let hl = idx_reg16();\n                    mem_write(cpu_sp, hl & 0xffu);\n                    mem_write((cpu_sp + 1u) & 0xffffu, (hl >> 8u) & 0xffu);\n                    set_idx_reg16((hi << 8u) | lo);\n                }\n                case 5u: { // EX DE,HL\n                    let td = cpu_d; let te = cpu_e;\n                    cpu_d = cpu_h; cpu_e = cpu_l;\n                    cpu_h = td; cpu_l = te;\n                }\n                case 6u: { cpu_iff1 = 0u; cpu_iff2 = 0u; }\n                case 7u: { cpu_iff1 = 1u; cpu_iff2 = 1u; }\n                default: {}\n            }\n        }\n        case 4u: { let nn = z80_fetch_word(); if (check_cc(y)) { z80_push16(cpu_pc); cpu_pc = nn; } }\n        case 5u: {\n            if (q == 0u) { z80_push16(get_reg16_af(p)); }\n            else {\n                switch(p) {\n                    case 0u: { let nn = z80_fetch_word(); z80_push16(cpu_pc); cpu_pc = nn; } // CALL nn\n                    case 1u, 3u: {} // DD/FD prefix handled in z80_step\n                    case 2u: { z80_exec_ed(); }\n                    default: {}\n                }\n            }\n        }\n        case 6u: { z80_alu(y, z80_fetch()); }\n        case 7u: { z80_push16(cpu_pc); cpu_pc = y * 8u; } // RST\n        default: {}\n    }\n}\n\nfn z80_execute(op: u32) {\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    let p = (y >> 1u) & 3u;\n    let q = y & 1u;\n    switch(x) {\n        case 0u: { z80_exec_x0(y, z, p, q); }\n        case 1u: {\n            if (y == 6u && z == 6u) {\n                // HALT: mark halted and back PC up onto the HALT opcode so the\n                // CPU re-executes it every step until an interrupt.\n                cpu_halted = 1u;\n                cpu_pc = (cpu_pc - 1u) & 0xffffu;\n            }\n            else { set_reg(y, get_reg(z)); }\n        }\n        case 2u: { z80_alu(y, get_reg(z)); }\n        case 3u: { z80_exec_x3(y, z, p, q); }\n        default: {}\n    }\n}\n\nfn z80_step() {\n    // While halted, the CPU keeps executing HALT: each step is an M1 fetch of\n    // the same opcode (R increments), PC stays put. It resumes only on interrupt.\n    if (cpu_halted != 0u) { r_inc(); return; }\n    // Reset per-instruction index-prefix state.\n    idx_mode = 0u;\n    idx_uses_mem = 0u;\n    idx_disp = 0u;\n\n    var op = z80_fetch();\n    r_inc(); // M1: opcode (or prefix) fetch\n    // A DD/FD prefix selects IX/IY for the following opcode.\n    if (op == 0xddu || op == 0xfdu) {\n        idx_mode = select(2u, 1u, op == 0xddu);\n        let next = z80_fetch();\n        r_inc(); // M1: opcode after the prefix\n        // A prefix immediately followed by another prefix or ED is a wasted M1:\n        // this step consumes just the prefix; back up so the next step restarts\n        // at the following byte (matches real Z80 timing and avoids an\n        // unbounded fetch loop on all-prefix programs).\n        if (next == 0xddu || next == 0xfdu || next == 0xedu) {\n            cpu_pc = (cpu_pc - 1u) & 0xffffu;\n            return;\n        }\n        op = next;\n    }\n\n    // Host hook: skip execution (treat as NOP) if requested.\n    if (on_fetch_opcode(op)) { return; }\n\n    if (idx_mode != 0u) {\n        if (op == 0xcbu) {\n            // DDCB/FDCB: displacement precedes the CB opcode.\n            idx_disp = signext(z80_fetch());\n            let cbop = z80_fetch();\n            z80_exec_idxcb(cbop);\n            return;\n        }\n        if (op_uses_hl_mem(op)) {\n            idx_uses_mem = 1u;\n            idx_disp = signext(z80_fetch());\n        }\n    }\n    z80_execute(op);\n}\n";
+export declare const Z80_CORE_WGSL = "\n// === Z80 CPU State (per invocation) ===\nvar<private> cpu_a: u32;\nvar<private> cpu_f: u32;\nvar<private> cpu_b: u32;\nvar<private> cpu_c: u32;\nvar<private> cpu_d: u32;\nvar<private> cpu_e: u32;\nvar<private> cpu_h: u32;\nvar<private> cpu_l: u32;\nvar<private> cpu_sp: u32;\nvar<private> cpu_pc: u32;\nvar<private> cpu_a2: u32;\nvar<private> cpu_f2: u32;\nvar<private> cpu_b2: u32;\nvar<private> cpu_c2: u32;\nvar<private> cpu_d2: u32;\nvar<private> cpu_e2: u32;\nvar<private> cpu_h2: u32;\nvar<private> cpu_l2: u32;\nvar<private> cpu_halted: u32;\nvar<private> cpu_iff1: u32;\nvar<private> cpu_iff2: u32;\nvar<private> cpu_ix: u32;\nvar<private> cpu_iy: u32;\nvar<private> cpu_i: u32; // interrupt vector register (I)\nvar<private> cpu_r: u32; // memory refresh register (R); bits 0-6 count M1 cycles, bit 7 preserved\n\n// Increment R once per M1 (opcode/prefix) fetch: 7-bit counter, bit 7 preserved.\nfn r_inc() { cpu_r = (cpu_r & 0x80u) | ((cpu_r + 1u) & 0x7fu); }\n// Index-prefix state for the instruction currently executing:\n//   idx_mode: 0 = HL, 1 = IX, 2 = IY\n//   idx_disp: sign-extended displacement for (IX+d)/(IY+d)\n//   idx_uses_mem: 1 when this instruction dereferences (IX+d)/(IY+d), which\n//     means H/L operands are NOT substituted by IXH/IXL (real Z80 rule).\nvar<private> idx_mode: u32;\nvar<private> idx_disp: u32;\nvar<private> idx_uses_mem: u32;\n\n// HOST CONTRACT: the host shader must declare the following BEFORE this core:\n//   fn mem_read(addr: u32) -> u32               // read one byte from memory\n//   fn mem_write(addr: u32, val: u32)           // write one byte to memory\n//   fn on_fetch_opcode(op: u32) -> bool         // return true to skip (NOP) an\n//                                               // opcode after prefix resolution\n// This lets the host choose the memory model (mask, modulo, storage buffer, \u2026)\n// and hook opcode execution (e.g. instruction suppression). See buildComputeShader.\n\n// Z80 flag bits\nconst CF: u32 = 0x01u;\nconst NF: u32 = 0x02u;\nconst PF: u32 = 0x04u;\nconst F3: u32 = 0x08u;\nconst HF: u32 = 0x10u;\nconst F5: u32 = 0x20u;\nconst ZF: u32 = 0x40u;\nconst SFl: u32 = 0x80u;\n\n\nfn z80_fetch() -> u32 {\n    let val = mem_read(cpu_pc);\n    cpu_pc = (cpu_pc + 1u) & 0xffffu;\n    return val;\n}\n\nfn z80_fetch_word() -> u32 {\n    let lo = z80_fetch();\n    let hi = z80_fetch();\n    return (hi << 8u) | lo;\n}\n\nfn z80_push16(val: u32) {\n    cpu_sp = (cpu_sp - 1u) & 0xffffu;\n    mem_write(cpu_sp, (val >> 8u) & 0xffu);\n    cpu_sp = (cpu_sp - 1u) & 0xffffu;\n    mem_write(cpu_sp, val & 0xffu);\n}\n\nfn z80_pop16() -> u32 {\n    let lo = mem_read(cpu_sp);\n    cpu_sp = (cpu_sp + 1u) & 0xffffu;\n    let hi = mem_read(cpu_sp);\n    cpu_sp = (cpu_sp + 1u) & 0xffffu;\n    return (hi << 8u) | lo;\n}\n\nfn signed_byte(b: u32) -> i32 {\n    let sb = i32(b);\n    if (sb > 127) { return sb - 256; }\n    return sb;\n}\n\n// === Register Access ===\nfn get_bc() -> u32 { return (cpu_b << 8u) | cpu_c; }\nfn get_de() -> u32 { return (cpu_d << 8u) | cpu_e; }\nfn get_hl() -> u32 { return (cpu_h << 8u) | cpu_l; }\nfn get_af() -> u32 { return (cpu_a << 8u) | cpu_f; }\n\nfn set_bc(v: u32) { cpu_b = (v >> 8u) & 0xffu; cpu_c = v & 0xffu; }\nfn set_de(v: u32) { cpu_d = (v >> 8u) & 0xffu; cpu_e = v & 0xffu; }\nfn set_hl(v: u32) { cpu_h = (v >> 8u) & 0xffu; cpu_l = v & 0xffu; }\nfn set_af(v: u32) { cpu_a = (v >> 8u) & 0xffu; cpu_f = v & 0xffu; }\n\nfn get_reg(idx: u32) -> u32 {\n    switch(idx) {\n        case 0u: { return cpu_b; }\n        case 1u: { return cpu_c; }\n        case 2u: { return cpu_d; }\n        case 3u: { return cpu_e; }\n        case 4u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { return (idx_reg16() >> 8u) & 0xffu; } // IXH/IYH\n            return cpu_h;\n        }\n        case 5u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { return idx_reg16() & 0xffu; } // IXL/IYL\n            return cpu_l;\n        }\n        case 6u: { return mem_read(idx_addr()); }\n        case 7u: { return cpu_a; }\n        default: { return 0u; }\n    }\n}\n\nfn set_reg(idx: u32, val: u32) {\n    let v = val & 0xffu;\n    switch(idx) {\n        case 0u: { cpu_b = v; }\n        case 1u: { cpu_c = v; }\n        case 2u: { cpu_d = v; }\n        case 3u: { cpu_e = v; }\n        case 4u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { set_idx_reg16((idx_reg16() & 0x00ffu) | (v << 8u)); }\n            else { cpu_h = v; }\n        }\n        case 5u: {\n            if (idx_mode != 0u && idx_uses_mem == 0u) { set_idx_reg16((idx_reg16() & 0xff00u) | v); }\n            else { cpu_l = v; }\n        }\n        case 6u: { mem_write(idx_addr(), v); }\n        case 7u: { cpu_a = v; }\n        default: {}\n    }\n}\n\nfn get_reg16(idx: u32) -> u32 {\n    switch(idx) {\n        case 0u: { return get_bc(); }\n        case 1u: { return get_de(); }\n        case 2u: { return idx_reg16(); } // HL / IX / IY\n        case 3u: { return cpu_sp; }\n        default: { return 0u; }\n    }\n}\n\nfn set_reg16(idx: u32, val: u32) {\n    let v = val & 0xffffu;\n    switch(idx) {\n        case 0u: { set_bc(v); }\n        case 1u: { set_de(v); }\n        case 2u: { set_idx_reg16(v); } // HL / IX / IY\n        case 3u: { cpu_sp = v; }\n        default: {}\n    }\n}\n\nfn get_reg16_af(idx: u32) -> u32 {\n    if (idx == 3u) { return get_af(); }\n    return get_reg16(idx);\n}\n\nfn set_reg16_af(idx: u32, val: u32) {\n    if (idx == 3u) { set_af(val); } else { set_reg16(idx, val); }\n}\n\n// === Index register (IX/IY) helpers ===\n// The 16-bit register the current prefix maps HL to (HL itself when no prefix).\nfn idx_reg16() -> u32 {\n    if (idx_mode == 1u) { return cpu_ix; }\n    if (idx_mode == 2u) { return cpu_iy; }\n    return get_hl();\n}\nfn set_idx_reg16(v: u32) {\n    if (idx_mode == 1u) { cpu_ix = v & 0xffffu; }\n    else if (idx_mode == 2u) { cpu_iy = v & 0xffffu; }\n    else { set_hl(v); }\n}\n// Address used for (HL) / (IX+d) / (IY+d).\nfn idx_addr() -> u32 {\n    if (idx_mode != 0u) { return (idx_reg16() + idx_disp) & 0xffffu; }\n    return get_hl();\n}\n// Sign-extend a displacement byte to 16 bits (two's complement).\nfn signext(b: u32) -> u32 {\n    if (b >= 0x80u) { return b | 0xff00u; }\n    return b;\n}\n// Does this main opcode dereference (HL)? (Determines displacement fetch and\n// whether H/L operands are IXH/IXL or real H/L under a DD/FD prefix.)\nfn op_uses_hl_mem(op: u32) -> bool {\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    if (x == 1u) { return (y == 6u || z == 6u) && !(y == 6u && z == 6u); } // LD r,(HL)/(HL),r (not HALT)\n    if (x == 2u) { return z == 6u; }                                        // ALU A,(HL)\n    if (x == 0u) {\n        if (z == 4u || z == 5u || z == 6u) { return y == 6u; }              // INC/DEC (HL), LD (HL),n\n        return false;\n    }\n    return false;\n}\n// Write a REAL 8-bit register (no IX/IY substitution) \u2014 used by the\n// undocumented DDCB/FDCB register-copy side effect.\nfn set_reg_raw(idx: u32, val: u32) {\n    let v = val & 0xffu;\n    switch(idx) {\n        case 0u: { cpu_b = v; }\n        case 1u: { cpu_c = v; }\n        case 2u: { cpu_d = v; }\n        case 3u: { cpu_e = v; }\n        case 4u: { cpu_h = v; }\n        case 5u: { cpu_l = v; }\n        case 7u: { cpu_a = v; }\n        default: {}\n    }\n}\n\n// === Flag Helpers ===\nfn sz_flags(val: u32) -> u32 {\n    var f = val & SFl;\n    if (val == 0u) { f |= ZF; }\n    f |= val & (F3 | F5);\n    return f;\n}\n\nfn parity(val: u32) -> bool {\n    var p = val;\n    p ^= p >> 4u;\n    p ^= p >> 2u;\n    p ^= p >> 1u;\n    return (p & 1u) == 0u;\n}\n\nfn check_cc(cc: u32) -> bool {\n    switch(cc) {\n        case 0u: { return (cpu_f & ZF) == 0u; }\n        case 1u: { return (cpu_f & ZF) != 0u; }\n        case 2u: { return (cpu_f & CF) == 0u; }\n        case 3u: { return (cpu_f & CF) != 0u; }\n        case 4u: { return (cpu_f & PF) == 0u; }\n        case 5u: { return (cpu_f & PF) != 0u; }\n        case 6u: { return (cpu_f & SFl) == 0u; }\n        case 7u: { return (cpu_f & SFl) != 0u; }\n        default: { return false; }\n    }\n}\n\n// === ALU ===\nfn z80_alu(op: u32, val: u32) {\n    let a = cpu_a;\n    let c = cpu_f & CF;\n    switch(op) {\n        case 0u: { // ADD\n            let r = a + val;\n            cpu_f = sz_flags(r & 0xffu) | select(0u, CF, r > 0xffu) |\n                    ((a ^ val ^ r) & HF) |\n                    select(0u, PF, ((~(a ^ val)) & (a ^ r) & 0x80u) != 0u);\n            cpu_a = r & 0xffu;\n        }\n        case 1u: { // ADC\n            let r = a + val + c;\n            cpu_f = sz_flags(r & 0xffu) | select(0u, CF, r > 0xffu) |\n                    ((a ^ val ^ r) & HF) |\n                    select(0u, PF, ((~(a ^ val)) & (a ^ r) & 0x80u) != 0u);\n            cpu_a = r & 0xffu;\n        }\n        case 2u: { // SUB\n            let r = i32(a) - i32(val);\n            let ru = u32(r) & 0xffu;\n            cpu_f = sz_flags(ru) | NF | select(0u, CF, r < 0) |\n                    ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n            cpu_a = ru;\n        }\n        case 3u: { // SBC\n            let r = i32(a) - i32(val) - i32(c);\n            let ru = u32(r) & 0xffu;\n            cpu_f = sz_flags(ru) | NF | select(0u, CF, r < 0) |\n                    ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n            cpu_a = ru;\n        }\n        case 4u: { // AND\n            cpu_a = a & val;\n            cpu_f = sz_flags(cpu_a) | HF | select(0u, PF, parity(cpu_a));\n        }\n        case 5u: { // XOR\n            cpu_a = a ^ val;\n            cpu_f = sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        case 6u: { // OR\n            cpu_a = a | val;\n            cpu_f = sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        case 7u: { // CP\n            let r = i32(a) - i32(val);\n            let ru = u32(r) & 0xffu;\n            cpu_f = (ru & SFl) | select(0u, ZF, ru == 0u) | (val & (F3 | F5)) | NF |\n                    select(0u, CF, r < 0) | ((a ^ val ^ u32(r)) & HF) |\n                    select(0u, PF, (((a ^ val) & (a ^ u32(r))) & 0x80u) != 0u);\n        }\n        default: {}\n    }\n}\n\nfn z80_inc8(val: u32) -> u32 {\n    let r = (val + 1u) & 0xffu;\n    cpu_f = (cpu_f & CF) | sz_flags(r) |\n            select(0u, PF, val == 0x7fu) |\n            select(0u, HF, (r & 0x0fu) == 0u);\n    return r;\n}\n\nfn z80_dec8(val: u32) -> u32 {\n    let r = (val - 1u) & 0xffu;\n    cpu_f = (cpu_f & CF) | sz_flags(r) | NF |\n            select(0u, PF, val == 0x80u) |\n            select(0u, HF, (val & 0x0fu) == 0u);\n    return r;\n}\n\nfn z80_add_hl(val: u32) {\n    let hl = idx_reg16(); // ADD HL,rp / ADD IX,rp / ADD IY,rp\n    let r = hl + val;\n    cpu_f = (cpu_f & (SFl | ZF | PF)) |\n            select(0u, CF, r > 0xffffu) |\n            select(0u, HF, ((hl ^ val ^ r) & 0x1000u) != 0u) |\n            ((r >> 8u) & (F3 | F5));\n    set_idx_reg16(r & 0xffffu);\n}\n\n// === Rotate/Shift for accumulator ===\nfn z80_rot_accum(y: u32) {\n    let a = cpu_a;\n    let c = cpu_f & CF;\n    let keep = cpu_f & (SFl | ZF | PF);\n    switch(y) {\n        case 0u: { // RLCA\n            cpu_a = ((a << 1u) | (a >> 7u)) & 0xffu;\n            cpu_f = keep | (a >> 7u) | (cpu_a & (F3 | F5));\n        }\n        case 1u: { // RRCA\n            cpu_a = ((a >> 1u) | (a << 7u)) & 0xffu;\n            cpu_f = keep | (a & 1u) | (cpu_a & (F3 | F5));\n        }\n        case 2u: { // RLA\n            cpu_a = ((a << 1u) | c) & 0xffu;\n            cpu_f = keep | (a >> 7u) | (cpu_a & (F3 | F5));\n        }\n        case 3u: { // RRA\n            cpu_a = ((a >> 1u) | (c << 7u)) & 0xffu;\n            cpu_f = keep | (a & 1u) | (cpu_a & (F3 | F5));\n        }\n        case 4u: { // DAA\n            var correction = 0u;\n            var carry = c;\n            if ((cpu_f & HF) != 0u || (a & 0x0fu) > 9u) { correction |= 0x06u; }\n            if (c != 0u || a > 0x99u) { correction |= 0x60u; carry = 1u; }\n            if ((cpu_f & NF) != 0u) { cpu_a = (a - correction) & 0xffu; }\n            else { cpu_a = (a + correction) & 0xffu; }\n            cpu_f = (cpu_f & NF) | sz_flags(cpu_a) | carry |\n                    ((a ^ cpu_a) & HF) | select(0u, PF, parity(cpu_a));\n        }\n        case 5u: { // CPL\n            cpu_a = (~a) & 0xffu;\n            cpu_f = (cpu_f & (SFl | ZF | PF | CF)) | HF | NF | (cpu_a & (F3 | F5));\n        }\n        case 6u: { // SCF\n            cpu_f = (cpu_f & (SFl | ZF | PF)) | CF | (cpu_a & (F3 | F5));\n        }\n        case 7u: { // CCF\n            cpu_f = (cpu_f & (SFl | ZF | PF)) |\n                    select(0u, HF, c != 0u) |\n                    select(CF, 0u, c != 0u) |\n                    (cpu_a & (F3 | F5));\n        }\n        default: {}\n    }\n}\n\n// === CB Prefix (bit ops, rotates, shifts) ===\nfn z80_cb_rot(op: u32, val: u32) -> u32 {\n    let c = cpu_f & CF;\n    var r = 0u;\n    switch(op) {\n        case 0u: { r = ((val << 1u) | (val >> 7u)) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 1u: { r = ((val >> 1u) | (val << 7u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 2u: { r = ((val << 1u) | c) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 3u: { r = ((val >> 1u) | (c << 7u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 4u: { r = (val << 1u) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 5u: { r = ((val >> 1u) | (val & 0x80u)) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        case 6u: { r = ((val << 1u) | 1u) & 0xffu; cpu_f = sz_flags(r) | (val >> 7u) | select(0u, PF, parity(r)); }\n        case 7u: { r = (val >> 1u) & 0xffu; cpu_f = sz_flags(r) | (val & 1u) | select(0u, PF, parity(r)); }\n        default: { r = val; }\n    }\n    return r;\n}\n\nfn z80_exec_cb() {\n    let op = z80_fetch();\n    r_inc(); // M1: CB-page opcode fetch\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    let val = get_reg(z);\n    switch(x) {\n        case 0u: { set_reg(z, z80_cb_rot(y, val)); }\n        case 1u: { // BIT\n            cpu_f = (cpu_f & CF) | HF |\n                    select(0u, ZF | PF, (val & (1u << y)) == 0u) |\n                    select(0u, SFl, y == 7u && (val & 0x80u) != 0u) |\n                    (val & (F3 | F5));\n        }\n        case 2u: { set_reg(z, val & ~(1u << y)); }\n        case 3u: { set_reg(z, val | (1u << y)); }\n        default: {}\n    }\n}\n\n// DDCB / FDCB: operates on (IX+d)/(IY+d). The displacement (idx_disp) has\n// already been fetched. For rot/shift/RES/SET the result is written to memory\n// AND (undocumented) copied to the real register in the low 3 bits unless it is\n// 6. For BIT, the undocumented F3/F5 come from the high byte of the address.\nfn z80_exec_idxcb(cbop: u32) {\n    let addr = (idx_reg16() + idx_disp) & 0xffffu;\n    let cx = (cbop >> 6u) & 3u;\n    let cy = (cbop >> 3u) & 7u;\n    let cz = cbop & 7u;\n    let val = mem_read(addr);\n    switch(cx) {\n        case 0u: {\n            let r = z80_cb_rot(cy, val);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        case 1u: { // BIT n,(IX+d)\n            cpu_f = (cpu_f & CF) | HF |\n                    select(0u, ZF | PF, (val & (1u << cy)) == 0u) |\n                    select(0u, SFl, cy == 7u && (val & 0x80u) != 0u) |\n                    ((addr >> 8u) & (F3 | F5));\n        }\n        case 2u: {\n            let r = val & ~(1u << cy);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        case 3u: {\n            let r = val | (1u << cy);\n            mem_write(addr, r);\n            if (cz != 6u) { set_reg_raw(cz, r); }\n        }\n        default: {}\n    }\n}\n\n// === Block Transfer (ED prefix) ===\nfn z80_ldi() {\n    let val = mem_read(get_hl());\n    mem_write(get_de(), val);\n    set_hl((get_hl() + 1u) & 0xffffu);\n    set_de((get_de() + 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    let n = (val + cpu_a) & 0xffu;\n    cpu_f = (cpu_f & (SFl | ZF | CF)) |\n            select(0u, PF, get_bc() != 0u) |\n            (n & F3) | select(0u, F5, (n & 0x02u) != 0u);\n}\n\nfn z80_ldd() {\n    let val = mem_read(get_hl());\n    mem_write(get_de(), val);\n    set_hl((get_hl() - 1u) & 0xffffu);\n    set_de((get_de() - 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    let n = (val + cpu_a) & 0xffu;\n    cpu_f = (cpu_f & (SFl | ZF | CF)) |\n            select(0u, PF, get_bc() != 0u) |\n            (n & F3) | select(0u, F5, (n & 0x02u) != 0u);\n}\n\nfn z80_cpi() {\n    let val = mem_read(get_hl());\n    let r = (cpu_a - val) & 0xffu;\n    let hf = (cpu_a ^ val ^ r) & HF;\n    // Undocumented F3/F5 come from n = A-(HL)-HF (bit 3 -> F3, bit 1 -> F5).\n    let n = (r - select(0u, 1u, hf != 0u)) & 0xffu;\n    set_hl((get_hl() + 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    cpu_f = (cpu_f & CF) | (r & SFl) | select(0u, ZF, r == 0u) | NF |\n            hf | select(0u, PF, get_bc() != 0u) |\n            (n & F3) | ((n & 0x02u) << 4u);\n}\n\nfn z80_cpd() {\n    let val = mem_read(get_hl());\n    let r = (cpu_a - val) & 0xffu;\n    let hf = (cpu_a ^ val ^ r) & HF;\n    let n = (r - select(0u, 1u, hf != 0u)) & 0xffu;\n    set_hl((get_hl() - 1u) & 0xffffu);\n    set_bc((get_bc() - 1u) & 0xffffu);\n    cpu_f = (cpu_f & CF) | (r & SFl) | select(0u, ZF, r == 0u) | NF |\n            hf | select(0u, PF, get_bc() != 0u) |\n            (n & F3) | ((n & 0x02u) << 4u);\n}\n\n// === ED Prefix ===\nfn z80_exec_ed() {\n    let op = z80_fetch();\n    r_inc(); // M1: ED-page opcode fetch\n    switch(op) {\n        case 0xa0u: { z80_ldi(); }\n        case 0xa8u: { z80_ldd(); }\n        case 0xb0u: { z80_ldi(); if (get_bc() != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } } // LDIR\n        case 0xb8u: { z80_ldd(); if (get_bc() != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } } // LDDR\n        case 0xa1u: { z80_cpi(); }\n        case 0xa9u: { z80_cpd(); }\n        case 0xb1u: { z80_cpi(); if (get_bc() != 0u && (cpu_f & ZF) == 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } }\n        case 0xb9u: { z80_cpd(); if (get_bc() != 0u && (cpu_f & ZF) == 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; } }\n        // NEG\n        case 0x44u, 0x4cu, 0x54u, 0x5cu, 0x64u, 0x6cu, 0x74u, 0x7cu: {\n            let a = cpu_a; cpu_a = 0u; z80_alu(2u, a);\n        }\n        // RETN/RETI\n        case 0x45u, 0x4du, 0x55u, 0x5du, 0x65u, 0x6du, 0x75u, 0x7du: {\n            cpu_iff1 = cpu_iff2; cpu_pc = z80_pop16();\n        }\n        // LD I,A / LD R,A / LD A,I / LD A,R\n        case 0x47u: { cpu_i = cpu_a; } // LD I,A\n        case 0x4fu: { cpu_r = cpu_a; } // LD R,A\n        // LD A,I / LD A,R: load I or R into A. Flags: S/Z (+F3/F5) from the loaded\n        // value, PF = IFF2, N/H reset, C preserved.\n        case 0x57u: { cpu_a = cpu_i; cpu_f = (cpu_f & CF) | sz_flags(cpu_i) | select(0u, PF, cpu_iff2 != 0u); }\n        case 0x5fu: { cpu_a = cpu_r; cpu_f = (cpu_f & CF) | sz_flags(cpu_r) | select(0u, PF, cpu_iff2 != 0u); }\n        // LD (nn), rr\n        case 0x43u, 0x53u, 0x63u, 0x73u: {\n            let nn = z80_fetch_word();\n            let rp = (op >> 4u) & 3u;\n            let val = get_reg16(rp);\n            mem_write(nn, val & 0xffu);\n            mem_write((nn + 1u) & 0xffffu, (val >> 8u) & 0xffu);\n        }\n        // LD rr, (nn)\n        case 0x4bu, 0x5bu, 0x6bu, 0x7bu: {\n            let nn = z80_fetch_word();\n            let rp = (op >> 4u) & 3u;\n            let lo = mem_read(nn);\n            let hi = mem_read((nn + 1u) & 0xffffu);\n            set_reg16(rp, (hi << 8u) | lo);\n        }\n        // ADC HL, rr\n        case 0x4au, 0x5au, 0x6au, 0x7au: {\n            let rp = (op >> 4u) & 3u;\n            let hl = get_hl();\n            let val = get_reg16(rp);\n            let c = cpu_f & CF;\n            let r = hl + val + c;\n            cpu_f = ((r >> 8u) & SFl) | select(0u, ZF, (r & 0xffffu) == 0u) |\n                    select(0u, HF, ((hl ^ val ^ r) & 0x1000u) != 0u) |\n                    select(0u, PF, ((~(hl ^ val)) & (hl ^ r) & 0x8000u) != 0u) |\n                    select(0u, CF, r > 0xffffu) | ((r >> 8u) & (F3 | F5));\n            set_hl(r & 0xffffu);\n        }\n        // SBC HL, rr\n        case 0x42u, 0x52u, 0x62u, 0x72u: {\n            let rp = (op >> 4u) & 3u;\n            let hl = get_hl();\n            let val = get_reg16(rp);\n            let c = cpu_f & CF;\n            let r = i32(hl) - i32(val) - i32(c);\n            let ru = u32(r) & 0xffffu;\n            cpu_f = ((ru >> 8u) & SFl) | select(0u, ZF, ru == 0u) | NF |\n                    select(0u, HF, ((hl ^ val ^ u32(r)) & 0x1000u) != 0u) |\n                    select(0u, PF, (((hl ^ val) & (hl ^ u32(r))) & 0x8000u) != 0u) |\n                    select(0u, CF, r < 0) | ((ru >> 8u) & (F3 | F5));\n            set_hl(ru);\n        }\n        // RRD\n        case 0x67u: {\n            let m = mem_read(get_hl());\n            mem_write(get_hl(), ((cpu_a << 4u) | (m >> 4u)) & 0xffu);\n            cpu_a = (cpu_a & 0xf0u) | (m & 0x0fu);\n            cpu_f = (cpu_f & CF) | sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        // RLD\n        case 0x6fu: {\n            let m = mem_read(get_hl());\n            mem_write(get_hl(), ((m << 4u) | (cpu_a & 0x0fu)) & 0xffu);\n            cpu_a = (cpu_a & 0xf0u) | (m >> 4u);\n            cpu_f = (cpu_f & CF) | sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));\n        }\n        // IN r,(C): no I/O device, so the port reads 0. Store it (except for the\n        // reg-6 form IN (C), which only sets flags) and set S/Z/F3/F5/P flags.\n        case 0x40u, 0x48u, 0x50u, 0x58u, 0x60u, 0x68u, 0x70u, 0x78u: {\n            let val = 0u;\n            let reg = (op >> 3u) & 7u;\n            if (reg != 6u) { set_reg(reg, val); }\n            cpu_f = (cpu_f & CF) | sz_flags(val) | select(0u, PF, parity(val));\n        }\n        // Block input: INI (A2) / IND (AA) / INIR (B2) / INDR (BA). No I/O\n        // device, so the port reads 0. Writes it to (HL), decrements B, moves HL,\n        // sets the block-I/O flags, and repeats (R-forms) while B != 0.\n        case 0xa2u, 0xaau, 0xb2u, 0xbau: {\n            let val = 0u;\n            mem_write(get_hl(), val);\n            cpu_b = (cpu_b - 1u) & 0xffu;\n            let dec = (op & 0x08u) != 0u;\n            let s = (val + cpu_c) & 0xffu;\n            let other = select((s + 1u) & 0xffu, (s - 1u) & 0xffu, dec);\n            cpu_f = select(0u, NF, (val & 0x80u) != 0u) |\n                    select(0u, HF | CF, other < val) |\n                    select(0u, PF, parity((other & 7u) ^ cpu_b)) |\n                    sz_flags(cpu_b);\n            if (dec) { set_hl((get_hl() - 1u) & 0xffffu); } else { set_hl((get_hl() + 1u) & 0xffffu); }\n            if ((op & 0x10u) != 0u && cpu_b != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; }\n        }\n        // Block output: OUTI (A3) / OUTD (AB) / OTIR (B3) / OTDR (BB). Reads (HL),\n        // decrements B, moves HL, sends to the (absent) port, sets flags, repeats.\n        case 0xa3u, 0xabu, 0xb3u, 0xbbu: {\n            let val = mem_read(get_hl());\n            cpu_b = (cpu_b - 1u) & 0xffu;\n            let dec = (op & 0x08u) != 0u;\n            if (dec) { set_hl((get_hl() - 1u) & 0xffffu); } else { set_hl((get_hl() + 1u) & 0xffffu); }\n            let other = (val + cpu_l) & 0xffu;\n            cpu_f = select(0u, NF, (val & 0x80u) != 0u) |\n                    select(0u, HF | CF, other < val) |\n                    select(0u, PF, parity((other & 7u) ^ cpu_b)) |\n                    sz_flags(cpu_b);\n            if ((op & 0x10u) != 0u && cpu_b != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; }\n        }\n        default: {} // unknown ED ops = NOP\n    }\n}\n\n// === Main Opcode Execution ===\nfn z80_exec_x0(y: u32, z: u32, p: u32, q: u32) {\n    switch(z) {\n        case 0u: {\n            switch(y) {\n                case 0u: {} // NOP\n                case 1u: { // EX AF,AF'\n                    var t = cpu_a; cpu_a = cpu_a2; cpu_a2 = t;\n                    t = cpu_f; cpu_f = cpu_f2; cpu_f2 = t;\n                }\n                case 2u: { // DJNZ\n                    let d = signed_byte(z80_fetch());\n                    cpu_b = (cpu_b - 1u) & 0xffu;\n                    if (cpu_b != 0u) { cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu; }\n                }\n                case 3u: { // JR\n                    let d = signed_byte(z80_fetch());\n                    cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu;\n                }\n                default: { // JR cc (y-4)\n                    let d = signed_byte(z80_fetch());\n                    if (check_cc(y - 4u)) { cpu_pc = u32(i32(cpu_pc) + d) & 0xffffu; }\n                }\n            }\n        }\n        case 1u: {\n            if (q == 0u) { set_reg16(p, z80_fetch_word()); }\n            else { z80_add_hl(get_reg16(p)); }\n        }\n        case 2u: {\n            if (q == 0u) {\n                switch(p) {\n                    case 0u: { mem_write(get_bc(), cpu_a); }\n                    case 1u: { mem_write(get_de(), cpu_a); }\n                    case 2u: { let nn = z80_fetch_word(); let hl = idx_reg16(); mem_write(nn, hl & 0xffu); mem_write((nn+1u) & 0xffffu, (hl >> 8u) & 0xffu); }\n                    case 3u: { mem_write(z80_fetch_word(), cpu_a); }\n                    default: {}\n                }\n            } else {\n                switch(p) {\n                    case 0u: { cpu_a = mem_read(get_bc()); }\n                    case 1u: { cpu_a = mem_read(get_de()); }\n                    case 2u: { let nn = z80_fetch_word(); let lo = mem_read(nn); let hi = mem_read((nn+1u) & 0xffffu); set_idx_reg16((hi << 8u) | lo); }\n                    case 3u: { cpu_a = mem_read(z80_fetch_word()); }\n                    default: {}\n                }\n            }\n        }\n        case 3u: {\n            if (q == 0u) { set_reg16(p, (get_reg16(p) + 1u) & 0xffffu); }\n            else { set_reg16(p, (get_reg16(p) - 1u) & 0xffffu); }\n        }\n        case 4u: { set_reg(y, z80_inc8(get_reg(y))); }\n        case 5u: { set_reg(y, z80_dec8(get_reg(y))); }\n        case 6u: { set_reg(y, z80_fetch()); }\n        case 7u: { z80_rot_accum(y); }\n        default: {}\n    }\n}\n\nfn z80_exec_x3(y: u32, z: u32, p: u32, q: u32) {\n    switch(z) {\n        case 0u: { if (check_cc(y)) { cpu_pc = z80_pop16(); } }\n        case 1u: {\n            if (q == 0u) { set_reg16_af(p, z80_pop16()); }\n            else {\n                switch(p) {\n                    case 0u: { cpu_pc = z80_pop16(); } // RET\n                    case 1u: { // EXX\n                        var t = cpu_b; cpu_b = cpu_b2; cpu_b2 = t;\n                        t = cpu_c; cpu_c = cpu_c2; cpu_c2 = t;\n                        t = cpu_d; cpu_d = cpu_d2; cpu_d2 = t;\n                        t = cpu_e; cpu_e = cpu_e2; cpu_e2 = t;\n                        t = cpu_h; cpu_h = cpu_h2; cpu_h2 = t;\n                        t = cpu_l; cpu_l = cpu_l2; cpu_l2 = t;\n                    }\n                    case 2u: { cpu_pc = idx_reg16(); } // JP (HL)/(IX)/(IY)\n                    case 3u: { cpu_sp = idx_reg16(); } // LD SP,HL/IX/IY\n                    default: {}\n                }\n            }\n        }\n        case 2u: { let nn = z80_fetch_word(); if (check_cc(y)) { cpu_pc = nn; } }\n        case 3u: {\n            switch(y) {\n                case 0u: { cpu_pc = z80_fetch_word(); } // JP nn\n                case 1u: { z80_exec_cb(); }\n                case 2u: { z80_fetch(); } // OUT (n),A - no I/O device, discard (matches zff outPort no-op)\n                case 3u: { z80_fetch(); cpu_a = 0u; } // IN A,(n) - no I/O device, reads 0 (matches zff inPort\u21920)\n                case 4u: { // EX (SP),HL / EX (SP),IX / EX (SP),IY\n                    let lo = mem_read(cpu_sp);\n                    let hi = mem_read((cpu_sp + 1u) & 0xffffu);\n                    let hl = idx_reg16();\n                    mem_write(cpu_sp, hl & 0xffu);\n                    mem_write((cpu_sp + 1u) & 0xffffu, (hl >> 8u) & 0xffu);\n                    set_idx_reg16((hi << 8u) | lo);\n                }\n                case 5u: { // EX DE,HL\n                    let td = cpu_d; let te = cpu_e;\n                    cpu_d = cpu_h; cpu_e = cpu_l;\n                    cpu_h = td; cpu_l = te;\n                }\n                case 6u: { cpu_iff1 = 0u; cpu_iff2 = 0u; }\n                case 7u: { cpu_iff1 = 1u; cpu_iff2 = 1u; }\n                default: {}\n            }\n        }\n        case 4u: { let nn = z80_fetch_word(); if (check_cc(y)) { z80_push16(cpu_pc); cpu_pc = nn; } }\n        case 5u: {\n            if (q == 0u) { z80_push16(get_reg16_af(p)); }\n            else {\n                switch(p) {\n                    case 0u: { let nn = z80_fetch_word(); z80_push16(cpu_pc); cpu_pc = nn; } // CALL nn\n                    case 1u, 3u: {} // DD/FD prefix handled in z80_step\n                    case 2u: { z80_exec_ed(); }\n                    default: {}\n                }\n            }\n        }\n        case 6u: { z80_alu(y, z80_fetch()); }\n        case 7u: { z80_push16(cpu_pc); cpu_pc = y * 8u; } // RST\n        default: {}\n    }\n}\n\nfn z80_execute(op: u32) {\n    let x = (op >> 6u) & 3u;\n    let y = (op >> 3u) & 7u;\n    let z = op & 7u;\n    let p = (y >> 1u) & 3u;\n    let q = y & 1u;\n    switch(x) {\n        case 0u: { z80_exec_x0(y, z, p, q); }\n        case 1u: {\n            if (y == 6u && z == 6u) {\n                // HALT: mark halted and back PC up onto the HALT opcode so the\n                // CPU re-executes it every step until an interrupt.\n                cpu_halted = 1u;\n                cpu_pc = (cpu_pc - 1u) & 0xffffu;\n            }\n            else { set_reg(y, get_reg(z)); }\n        }\n        case 2u: { z80_alu(y, get_reg(z)); }\n        case 3u: { z80_exec_x3(y, z, p, q); }\n        default: {}\n    }\n}\n\nfn z80_step() {\n    // While halted, the CPU keeps executing HALT: each step is an M1 fetch of\n    // the same opcode (R increments), PC stays put. It resumes only on interrupt.\n    if (cpu_halted != 0u) { r_inc(); return; }\n    // Reset per-instruction index-prefix state.\n    idx_mode = 0u;\n    idx_uses_mem = 0u;\n    idx_disp = 0u;\n\n    var op = z80_fetch();\n    r_inc(); // M1: opcode (or prefix) fetch\n    // A DD/FD prefix selects IX/IY for the following opcode.\n    if (op == 0xddu || op == 0xfdu) {\n        idx_mode = select(2u, 1u, op == 0xddu);\n        let next = z80_fetch();\n        r_inc(); // M1: opcode after the prefix\n        // A prefix immediately followed by another prefix or ED is a wasted M1:\n        // this step consumes just the prefix; back up so the next step restarts\n        // at the following byte (matches real Z80 timing and avoids an\n        // unbounded fetch loop on all-prefix programs).\n        if (next == 0xddu || next == 0xfdu || next == 0xedu) {\n            cpu_pc = (cpu_pc - 1u) & 0xffffu;\n            return;\n        }\n        op = next;\n    }\n\n    // Host hook: skip execution (treat as NOP) if requested.\n    if (on_fetch_opcode(op)) { return; }\n\n    if (idx_mode != 0u) {\n        if (op == 0xcbu) {\n            // DDCB/FDCB: displacement precedes the CB opcode.\n            idx_disp = signext(z80_fetch());\n            let cbop = z80_fetch();\n            z80_exec_idxcb(cbop);\n            return;\n        }\n        if (op_uses_hl_mem(op)) {\n            idx_uses_mem = 1u;\n            idx_disp = signext(z80_fetch());\n        }\n    }\n    z80_execute(op);\n}\n";
 //# sourceMappingURL=z80-core.wgsl.d.ts.map

package/dist/z80-core.wgsl.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"z80-core.wgsl.d.ts","sourceRoot":"","sources":["../src/z80-core.wgsl.ts"],"names":[],"mappings":"AAMA,eAAO,MAAM,aAAa,gv6BA+xBzB,CAAC"}
+{"version":3,"file":"z80-core.wgsl.d.ts","sourceRoot":"","sources":["../src/z80-core.wgsl.ts"],"names":[],"mappings":"AAMA,eAAO,MAAM,aAAa,iw+BAk0BzB,CAAC"}

package/dist/z80-core.wgsl.js

@@ -601,9 +601,44 @@ fn z80_exec_ed() {
             cpu_a = (cpu_a & 0xf0u) | (m >> 4u);
             cpu_f = (cpu_f & CF) | sz_flags(cpu_a) | select(0u, PF, parity(cpu_a));
         }
-        // IN r,(C) - simplified, just set to 0
+        // IN r,(C): no I/O device, so the port reads 0. Store it (except for the
+        // reg-6 form IN (C), which only sets flags) and set S/Z/F3/F5/P flags.
         case 0x40u, 0x48u, 0x50u, 0x58u, 0x60u, 0x68u, 0x70u, 0x78u: {
-            set_reg((op >> 3u) & 7u, 0u);
+            let val = 0u;
+            let reg = (op >> 3u) & 7u;
+            if (reg != 6u) { set_reg(reg, val); }
+            cpu_f = (cpu_f & CF) | sz_flags(val) | select(0u, PF, parity(val));
+        }
+        // Block input: INI (A2) / IND (AA) / INIR (B2) / INDR (BA). No I/O
+        // device, so the port reads 0. Writes it to (HL), decrements B, moves HL,
+        // sets the block-I/O flags, and repeats (R-forms) while B != 0.
+        case 0xa2u, 0xaau, 0xb2u, 0xbau: {
+            let val = 0u;
+            mem_write(get_hl(), val);
+            cpu_b = (cpu_b - 1u) & 0xffu;
+            let dec = (op & 0x08u) != 0u;
+            let s = (val + cpu_c) & 0xffu;
+            let other = select((s + 1u) & 0xffu, (s - 1u) & 0xffu, dec);
+            cpu_f = select(0u, NF, (val & 0x80u) != 0u) |
+                    select(0u, HF | CF, other < val) |
+                    select(0u, PF, parity((other & 7u) ^ cpu_b)) |
+                    sz_flags(cpu_b);
+            if (dec) { set_hl((get_hl() - 1u) & 0xffffu); } else { set_hl((get_hl() + 1u) & 0xffffu); }
+            if ((op & 0x10u) != 0u && cpu_b != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; }
+        }
+        // Block output: OUTI (A3) / OUTD (AB) / OTIR (B3) / OTDR (BB). Reads (HL),
+        // decrements B, moves HL, sends to the (absent) port, sets flags, repeats.
+        case 0xa3u, 0xabu, 0xb3u, 0xbbu: {
+            let val = mem_read(get_hl());
+            cpu_b = (cpu_b - 1u) & 0xffu;
+            let dec = (op & 0x08u) != 0u;
+            if (dec) { set_hl((get_hl() - 1u) & 0xffffu); } else { set_hl((get_hl() + 1u) & 0xffffu); }
+            let other = (val + cpu_l) & 0xffu;
+            cpu_f = select(0u, NF, (val & 0x80u) != 0u) |
+                    select(0u, HF | CF, other < val) |
+                    select(0u, PF, parity((other & 7u) ^ cpu_b)) |
+                    sz_flags(cpu_b);
+            if ((op & 0x10u) != 0u && cpu_b != 0u) { cpu_pc = (cpu_pc - 2u) & 0xffffu; }
         }
         default: {} // unknown ED ops = NOP
     }

package/dist/z80-core.wgsl.js.map

@@ -1 +1 @@
-{"version":3,"file":"z80-core.wgsl.js","sourceRoot":"","sources":["../src/z80-core.wgsl.ts"],"names":[],"mappings":"AAAA,0EAA0E;AAC1E,+DAA+D;AAC/D,EAAE;AACF,+EAA+E;AAC/E,kEAAkE;AAElE,MAAM,CAAC,MAAM,aAAa,GAAG,UAAU,CAAC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CA+xBvC,CAAC"}
+{"version":3,"file":"z80-core.wgsl.js","sourceRoot":"","sources":["../src/z80-core.wgsl.ts"],"names":[],"mappings":"AAAA,0EAA0E;AAC1E,+DAA+D;AAC/D,EAAE;AACF,+EAA+E;AAC/E,kEAAkE;AAElE,MAAM,CAAC,MAAM,aAAa,GAAG,UAAU,CAAC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;CAk0BvC,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@neovand/zilion",
-	"version": "0.1.1",
+	"version": "0.1.2",
 	"publishConfig": {
 		"access": "public"
 	},