#include <ctype.h>
#include <stdlib.h>
#include <string.h>
#pragma clang diagnostic ignored "-Wparentheses"
#pragma clang diagnostic ignored "-Wreturn-type"
#pragma clang diagnostic ignored "-Wempty-body"

// pu2!
// maybe compacter? idk

#define A 8
// change this for error reporting on unknown instructions!
// examples:
// #define A __builtin_trap()
// ^ does SIGTRAP
// #define A my_custom_cpu_error_trap()
// ^ calls my_custom_cpu_error_trap()

#define __PU2_SPEED_IMPROVEMENT
#define __PU2_EXTENDED_INSTRUCTION_SET

#define S (1 << 25)
#define J while
#define H !--c

#ifndef __PU2_SPEED_IMPROVEMENT

    #ifndef __PU2_EXTENDED_INSTRUCTION_SET

// clang-format off
    m[S],c,v,a,b,u;pu2(){J(c=m[m[7]++],v=m[m[7]++],c){H?a=v:H?a=m[b]:H?u?m[a]=b:0:H?v=a,a=b,b=v:H?a+=b:H?a-=b:H?a&=b:H?a=!a:H?u=v:H?u=a:H?a=u:H?u=b:H?b=u:A;}}
    // clang-format on

    #else

// clang-format off
    m[S],c,v,a,b,u;pu2(){J(c=m[m[7]++],v=m[m[7]++],c){H?a=v:H?a=m[b]:H?u?m[a]=b:0:H?v=a,a=b,b=v:H?a+=b:H?a-=b:H?a&=b:H?a=!a:H?u=v:H?u=a:H?a=u:H?u=b:H?b=u:H?m[0]+=v*2:H?m[v]=m[a]:H?a=m[v]:H?m[a]=v:H?m[a]=m[v]:A;}}
    // clang-format on

    #endif

#else

    #include <stdio.h>

    #undef A
    #define A                                                                                      \
        ({                                                                                         \
            printf("ERROR: Invalid instruction!\n");                                               \
            exit(1);                                                                               \
        })

// pu2 optimized for speed implements a few new instructions for speed,
// and has a "bake" mode which will preprocess the memory and perform the required modifications.
//
// 14: nop
// 15: m[imm] = m[a]
// 16: a = m[imm]
// 17: m[a] = imm
// 18: m[a] = m[imm]

// not implemented yet, was lazy
// bake(int start, int end) {}

__attribute__((aligned(64))) m[ S ], a, b, u, c;
typedef struct {
    int c;
    int v;
} vminst;
pu2() {
    vminst *M = (vminst *) m;

    int s1, s2, s3;

    const __attribute__((aligned(64))) void *PRECOMPUTED_TABLE[]
        = { &&C_ZERO,     &&C_ONE,      &&C_TWO,        &&C_THREE,    &&C_FOUR,
            &&C_FIVE,     &&C_SIX,      &&C_SEVEN,      &&C_EIGHT,    &&C_NINE,
            &&C_TEN,      &&C_ELEVEN,   &&C_TWELVE,     &&C_THIRTEEN, &&CB_FOURTEEN,
            &&CB_FIFTEEN, &&CB_SIXTEEN, &&CB_SEVENTEEN, &&CB_EIGHTEEN };
    int             pc = m[ 7 ] >> 1;
    register vminst inst;
    J(inst = M[ pc++ ], inst.c) {
        if (__builtin_expect(inst.c > 18, 0)) A;

        goto *PRECOMPUTED_TABLE[ inst.c ];

    C_ZERO:
        break;
    C_ONE:
        a = inst.v;
        continue;
    C_TWO:
        a = m[ b ];
        continue;
    C_THREE:
        if (u) {
            if (__builtin_expect(a == 7, 0)) pc = b >> 1;
            m[ a ] = b;
        }

        continue;
    C_FOUR:
        s1 = a;
        a  = b;
        b  = s1;

        continue;
    C_FIVE:
        a += b;
        continue;
    C_SIX:
        a -= b;
        continue;
    C_SEVEN:
        a = !a;
        continue;
    C_EIGHT:
        u = inst.v;
        continue;
    C_NINE:
        u = inst.v;
        continue;
    C_TEN:
        u = a;
        continue;
    C_ELEVEN:
        a = u;
        continue;
    C_TWELVE:
        u = b;
        continue;
    C_THIRTEEN:
        b = u;
        continue;
    CB_FOURTEEN: // nop
        // this instruction allows for essentially embedding data inside a program, as pu2 will skip
        // right over it
        pc += inst.v;
        continue;
    CB_FIFTEEN:
        if (__builtin_expect(inst.v == 7, 0)) pc = m[ a ] >> 1;
        m[ inst.v ] = m[ a ];
        continue;
    CB_SIXTEEN:
        a = m[ inst.v ];
        continue;
    CB_SEVENTEEN:
        if (__builtin_expect(a == 7, 0)) pc = inst.v >> 1;
        m[ a ] = inst.v;
        continue;
    CB_EIGHTEEN:
        if (__builtin_expect(a == 7, 0)) pc = m[ inst.v ] >> 1;
        m[ a ] = m[ inst.v ];
        continue;
    }

    m[ 7 ] = pc << 1;
}

#endif

// there's just memory. no, really
// ok actually there's c_one registers called a and b
// 0  HLT: halt
// 1  IMM: a = imm
// 2  LOD: a = m[b]
// 3  STR: m[a] = b (only happens if acc != 0)
// 4  SWP: swap a and b
// 5  ADD: a += b
// 6  SUB: a -= b
// 7  AND: a &= b
// 8  NOT: a = !a
// 9  ACC: u = imm
// 10 A2U: u = a
// 11 U2A: a = u
// 12 B2U: u = b
// 13 U2B: b = u

// NOTE: Memory indexes less than 1024 should be considered as temporary/compute memory, and
// programs should be written at index 1024 and upwards.
//
// NOTE: Memory index 7 is the program
// counter, in which the current instruction is read from m[m[7]] at each cycle

// -------------------------

// yeehaw baby it's time for as2
// E is the error code loc. zero it before calling as2
// Y is macro storage. extend as needed
// set O for output before running. keep in mind, it gets modified whilst as2 is running.
//    therefore, i recommend you store a copy somewhere!

// clang-format off
char*C="hltimmlodstrswpaddsubandnotacca2uu2ab2uu2bnopcb1cb2cb3",*Y[1<<10],*E,*L,i;*K;as2(char*I){J(c=*I++){if(c==35){J(*++I-10);c=*I++;}L=C;(!isspace(c)?c-39?c-125?c-123?c-64?({i=0,I--;J(i++,*L)strncmp(L,I,3)?L+=3:({I+=3;*K++=i-1;*K++=(i-2&&i-9)?0:strtol(I,&I,10);goto Q;});E=I;return 1;Q:;}):as2(Y[strtol(I,&I,10)]):({Y[strtol(I,&I,10)]=I+1;J(*I++-125);}):({break;}):({*K++=1;*K++=*I++;}):0);}}
// clang-format on

// #undef S
#undef H
#undef J
#undef E