add turingcomplete cpu compiler

Implement fn/return across the full pipeline:
- Lexer: TOK_FN, TOK_RETURN keywords
- Parser: NODE_FN_DEF, NODE_RETURN AST nodes
- Compiler: FunctionEntry table, inline compilation with jump-over and backpatching
- VM: CallFrame stack with variable snapshot for scoped calls and OP_RETURN

docs/todo.md

@@ -0,0 +1,312 @@
+# TODO: Backend mycpu_v2 para j-lang
+
+Roadmap de implementación del generador de código para el CPU v2 de 16-bit.
+El objetivo es que `gencode.h` tome el AST del frontend y produzca:
+- **Texto ensamblador** legible (para debug)
+- **Binario** (array de bytes para cargar en PROM)
+
+---
+
+## Referencia rápida del CPU v2
+
+```
+Instrucción: 8 bytes = [OPCODE:16][PARAM1:16][PARAM2:16][TARGET:16]
+PC cuenta en words de 16-bit → instrucción N está en PC = N × 4
+
+Registros libres:    REG0-REG11 (12 registros, 16 bits cada uno)
+Registros especiales: REG12(RAM_VAL), REG13(RAM_ADDR), REG14(PC), REG15(I/O)
+
+Modos de direccionamiento (codificados en el opcode):
+  base + 0x00 = registro, registro
+  base + 0x40 = inmediato, registro
+  base + 0x80 = registro, inmediato
+  base + 0xC0 = inmediato, inmediato
+
+RAM:
+  Leer:    Escribir dirección en REG13 → REG12 se actualiza automáticamente
+  Escribir: REG12 = valor, REG13 = addr → RSTR (0x18)
+```
+
+### Pseudo-instrucciones útiles
+
+```asm
+MOV #valor, REGn   →  ADD #valor, #0, REGn    (opcode 0xC0)
+MOV REGa, REGb     →  ADD REGa, #0, REGb      (opcode 0x80)
+JMP #addr           →  EQ #0, #0, addr          (opcode 0xD0, siempre true)
+NOP                 →  ADD #0, #0, REG0          (opcode 0xC0)
+```
+
+---
+
+## Convención de registros
+
+```
+REG0-REG3   → Temporales para evaluación de expresiones (expression stack)
+REG4-REG5   → Auxiliares (spill de expresiones profundas)
+REG6-REG11  → Libres / reserva futura (frame pointer, etc.)
+REG12       → RAM VALUE (especial, no tocar directamente)
+REG13       → RAM ADDR (especial, no tocar directamente)
+REG14       → PC (especial)
+REG15       → I/O (especial)
+```
+
+## Almacenamiento de variables
+
+Todas las variables en **RAM**. Tabla nombre→dirección en tiempo de compilación.
+
+```asm
+; Leer variable 'x' (dirección addr) → REGn
+ADD #addr, #0, REG13     ; REG13 = addr
+ADD REG12, #0, REGn      ; REGn = RAM[addr] (lectura automática)
+
+; Escribir variable 'x' (valor en REGn, dirección addr)
+ADD #addr, #0, REG13     ; REG13 = addr
+ADD REGn, #0, REG12      ; REG12 = valor
+RSTR                      ; RAM[addr] = REG12
+```
+
+---
+
+## Fase 0: Infraestructura del emisor
+
+**Objetivo**: Estructuras y funciones base para emitir instrucciones.
+
+- [ ] Struct `Instruction` (opcode, param1, param2, target)
+- [ ] Buffer de instrucciones (array dinámico donde se acumulan)
+- [ ] Función `emit(opcode, p1, p2, target)` — agrega instrucción al buffer
+- [ ] Tabla de variables: mapeo nombre→dirección_RAM (compilación)
+- [ ] Función `lookupOrCreateVar(name)` — busca o asigna dirección RAM
+- [ ] Sistema de labels/backpatching:
+  - [ ] `emitPlaceholder()` → emite instrucción con target=0, retorna índice
+  - [ ] `patchTarget(index, target)` → rellena el target de instrucción emitida
+  - [ ] `currentAddr()` → posición actual (nº de instrucción)
+- [ ] Output ASM: recorrer buffer → texto legible con mnemonicos
+- [ ] Output binario: recorrer buffer → array de bytes (8 bytes/instrucción)
+
+**Criterio**: Emitir instrucciones hardcoded, ver texto ASM y binario generados.
+
+---
+
+## Fase 1: Constantes, asignaciones y print
+
+**Objetivo**: Compilar `x = 42` y `print x`.
+
+- [ ] Compilar `NODE_INT_LIT` → cargar inmediato en REG[depth]
+  ```asm
+  ADD #42, #0, REG0       ; MOV #42, REG0
+  ```
+- [ ] Compilar `NODE_ASSIGN` → evaluar expr → REG0, store en RAM
+  ```asm
+  ; (resultado ya en REG0)
+  ADD #addr, #0, REG13    ; REG13 = dirección de variable
+  ADD REG0, #0, REG12     ; REG12 = valor
+  RSTR                     ; RAM[addr] = valor
+  ```
+- [ ] Compilar `NODE_VAR` → leer de RAM a REG[depth]
+  ```asm
+  ADD #addr, #0, REG13    ; REG13 = dirección
+  ADD REG12, #0, REG0     ; REG0 = RAM[addr]
+  ```
+- [ ] Compilar `NODE_PRINT` → evaluar expr → REG0, copiar a I/O
+  ```asm
+  ADD REG0, #0, REG15     ; OUTPUT = REG0
+  ```
+- [ ] Compilar `NODE_BLOCK` → iterar y compilar cada statement
+
+**Test**: `simple.j` (x = 10). `print 10` → escribe 10 en REG15.
+
+---
+
+## Fase 2: Expresiones aritméticas
+
+**Objetivo**: Compilar `x = 10 + 20 * 3`.
+
+**Estrategia**: Register depth counter. Cada sub-expresión deposita resultado en `REG[depth]`.
+
+- [ ] Variable `int reg_depth = 0` para tracking
+- [ ] Compilar `NODE_BINOP`:
+  ```
+  compilar left  → resultado en REG[depth]
+  depth++
+  compilar right → resultado en REG[depth]
+  depth--
+  emit OP REG[depth], REG[depth+1], REG[depth]
+  ```
+- [ ] Manejar profundidad > 4 → PUSH/POP al stack (spill)
+- [ ] Mapeo de operadores:
+  - `+` → ADD (0x00)
+  - `-` → SUB (0x01)
+  - `*` → MUL (0x02)
+  - `/` → DIV (0x03)
+
+**Test**: `sum.j`, `resta.j`. Verificar que `2 + 3 * 4` da 14.
+
+---
+
+## Fase 3: Comparaciones y control de flujo
+
+**Objetivo**: Compilar `if` y `while`.
+
+### if
+
+- [ ] Compilar `NODE_IF`:
+  ```
+  compilar condición left  → REG0
+  compilar condición right → REG1
+  emit CONDICIONAL_INVERSO REG0, REG1, [placeholder]
+  compilar bloque then
+  patch placeholder → currentAddr() × 4
+  ```
+- [ ] Mapeo de condicionales **inversos** (saltar si la condición es FALSA):
+  - `==` en AST → emit `NEQ` (0x11)
+  - `!=` en AST → emit `EQ`  (0x10)
+  - `<`  en AST → emit `GRE` (0x15) — saltar si >=
+  - `>`  en AST → emit `LSE` (0x13) — saltar si <=
+
+### while
+
+- [ ] Compilar `NODE_WHILE`:
+  ```
+  loop_start = currentAddr()
+  compilar condición left  → REG0
+  compilar condición right → REG1
+  emit CONDICIONAL_INVERSO REG0, REG1, [placeholder_exit]
+  compilar cuerpo
+  emit EQ #0, #0, (loop_start × 4)    ; JMP incondicional
+  patch placeholder_exit → currentAddr() × 4
+  ```
+
+### Recordar
+- **PC = instrucción_index × 4** (cada instrucción = 4 words de 16-bit)
+- El salto incondicional es `EQ #0, #0, target` (0xD0, siempre true)
+
+**Test**: `if.j`, `while.j`. While que cuenta de 0 a 10.
+
+---
+
+## Fase 4: Funciones (CALL/RET)
+
+**Objetivo**: Compilar `fn` definitions y llamadas.
+
+### Convención de llamada
+
+```
+1. Caller pushea argumentos al stack (derecha a izquierda)
+2. Caller ejecuta CALL #dirección (pushea PC+1 al stack, salta)
+3. Callee popea argumentos → variables locales en RAM
+4. Callee ejecuta cuerpo
+5. Callee deja resultado en REG0
+6. Callee ejecuta RET (popea PC del stack, salta)
+7. Caller usa REG0 como valor de retorno
+```
+
+### Tareas
+
+- [ ] Compilar `NODE_FN_DEF`:
+  ```
+  emit JMP [placeholder_skip]           ; saltar sobre el cuerpo
+  fn_addr = currentAddr()
+  para cada param (de derecha a izq):
+    emit POP → REGn
+    store REGn → RAM[param_addr]
+  compilar cuerpo
+  emit RET
+  patch placeholder_skip → currentAddr() × 4
+  registrar fn_name → fn_addr en tabla de funciones
+  ```
+- [ ] Compilar `NODE_CALL`:
+  ```
+  para cada argumento (de izq a der):
+    compilar argumento → REG0
+    emit PUSH REG0
+  emit CALL #(fn_addr × 4)
+  ; resultado queda en REG0
+  ```
+- [ ] Compilar `NODE_RETURN`:
+  ```
+  compilar expresión → REG0
+  emit RET
+  ```
+- [ ] Resolver scope de variables locales:
+  - **Opción simple**: cada función tiene su propio rango de RAM
+  - **Opción avanzada**: frame pointer (registro base + offset para locales)
+
+**Test**: `functions.j`, `custom_fn.j`.
+
+---
+
+## Fase 5: Strings y objetos (avanzado)
+
+**Objetivo**: Soportar strings, clases, campos e instancias.
+
+- [ ] Strings en RAM — caracteres consecutivos, variable apunta a dirección base
+- [ ] Print de strings — loop: leer cada char de RAM → escribir en REG15
+- [ ] Instancias — bloque de RAM con campos, variable apunta a base
+- [ ] Campos — offset fijo desde base de instancia
+- [ ] Métodos — funciones con `self` (dirección de instancia) como primer arg
+- [ ] Constructor — reservar espacio en RAM, llamar a `init`
+
+**Nota**: Requiere un allocator en runtime para reservar memoria dinámica en RAM.
+
+### Estrategia de allocator en runtime
+
+Hay dos opciones, de menor a mayor complejidad:
+
+**Opción A: Bump allocator (recomendado para empezar)**
+
+La más simple. Una dirección de RAM fija (ej: `RAM[0x00FF]`) actúa como "heap pointer" que empieza al final de las variables estáticas. Cada asignación avanza el pointer. No tiene `free`.
+
+```asm
+; alloc(size) — size en REG1, retorna dirección en REG0
+ADD #0x00FF, #0, REG13     ; REG13 = dirección del heap_ptr
+ADD REG12, #0, REG0        ; REG0 = heap_ptr actual (dirección a retornar)
+ADD REG12, REG1, REG12     ; REG12 = heap_ptr + size
+RSTR                        ; guardar nuevo heap_ptr en RAM[0x00FF]
+; REG0 = dirección del bloque asignado
+```
+
+~4 instrucciones. Suficiente para strings literales y concatenaciones simples.
+
+**Opción B: Allocator con metadata (como `allocator.h`, pero en ASM del CPU v2)**
+
+Mismo diseño conceptual que `src/memory/allocator.h` pero implementado como rutina en ensamblador del CPU v2:
+- Cada bloque en RAM: `[size:16][in_use:16][payload...]`
+- Loop que recorre bloques con comparaciones + saltos (first-fit)
+- `free` marca `in_use = 0`
+- ~30-50 instrucciones del CPU v2
+
+Solo necesario si se van a liberar strings (reasignar variables string, concatenaciones temporales).
+
+**Recomendación**: Empezar con bump allocator. Si más adelante se necesita `free`, implementar opción B usando el diseño de `allocator.h` como referencia conceptual.
+
+**Test**: `str.j`, `classes.j`.
+
+---
+
+## Diagrama de dependencias
+
+```
+Fase 0 (infraestructura)
+  │
+  └── Fase 1 (constantes, asignación, print)
+        │
+        └── Fase 2 (aritmética)
+              │
+              └── Fase 3 (if/while)
+                    │
+                    └── Fase 4 (funciones)
+                          │
+                          └── Fase 5 (strings/objetos)
+```
+
+## Verificación por fase
+
+| Fase | Archivos de test |
+|------|-----------------|
+| 1    | `simple.j`      |
+| 2    | `sum.j`, `resta.j` |
+| 3    | `if.j`, `while.j` |
+| 4    | `functions.j`, `custom_fn.j` |
+| 5    | `str.j`, `classes.j` |
+
+Comparar output generado (ASM + binario) con lo que la VM produce para la misma entrada.

mycpu.md

@@ -0,0 +1,103 @@
+# Motivación
+A partir de completar el juego "Turing Complete" y tener un CPU de 8bit con arquitectura LEG funcional, me dispongo a crear un compilador para mi lenguaje "J-LANG".
+
+
+# Estructura del CPU
+
+Arquitectura: 8bit
+PROM:   256 bytes
+RAM:    256 bytes
+STACK:  256 bytes
+
+## Direcciones
+
+0x00 REG0
+0x01 REG1
+0x02 REG2
+0x03 REG3
+0x04 REG4 | RAM VALUE
+0x05 REG5 | RAM ADDR PTR
+0x06 PROGRAM COUNTER
+0x07 INPUT/OUTPUT
+
+### Registros basicos
+
+Desde: 0x00
+Hasta: 0x03
+
+Son registros que almacenan 1 byte.
+
+### Registros conectados a la RAM
+
+Estos registros hacen de "puente" con la ram.
+
+0x05 REG5 es el counter de la ram.
+0x04 REG4 es el valor que contiene RAM[counter(0x05)]
+
+### input/output
+
+Son 1 byte de entrada y otro de salida para interactuar con el juego "Turing Complete"
+
+## OPCODES
+
+Las instrucciones de este cpu tienen que tener un tamaño de 4 bytes cada una.
+Se permiten dos modos de direccionamiento, desde un registro o modo inmediato.
+
+- Desde registro: Se usa el byte para indicar el registro que guarda el valor
+- Inmediato: Se usa el byte como valor directamente
+
+La estructura de una instrucción es:
+[OPCODE] [INPUT0] [INPUT1] [TARGET]
+
+El ultimo byte "target" indica en que registro debe guardarse el resultado de la instruccion.
+
+En los opcodes condicionales, el ultimo byte (TARGET) indica el valor que se escribirá en el PROGRAM_COUNTER si se cumple la condicion.
+
+======== ALU ========
+0x00	ADD r0 r1 t0
+0x01	SUB r0 r1 t0
+0x02	AND r0 r1 t0
+0x03	OR r0 r1 t0
+0x04	NOT r0 r1 t0
+0x05	XOR r0 r1 t0
+---------------------
+0x40	ADD r0 #1 t0        ; # significa inmediato
+0x41	SUB r0 #1 t0
+0x42	AND r0 #1 t0
+0x43	OR r0 #1 t0
+0x44	NOT r0 #1 t0
+0x45	XOR r0 #1 t0
+--------------------
+0x80	ADD #0 r1 t0
+0x81	SUB #0 r1 t0
+0x82	AND #0 r1 t0
+0x83	OR #0 r1 t0
+0x84	NOT #0 r1 t0
+0x85	XOR #0 r1 t0
+--------------------
+0xC0	ADD #0 #1 t0
+0xC1	SUB #0 #1 t0
+0xC2	AND #0 #1 t0
+0xC3	OR #0 #1 t0
+0xC4	NOT #0 #1 t0
+0xC5	XOR #0 #1 t0
+====== CONDITIONAL ======
+0x30	EQ r0 r1 pc	           ; equal
+0x31	NEQ r0 r1 pc           ; not_equal
+0x32	LS r0 r1 pc            ; less
+0x33	LSE r0 r1 pc           ; less_or_equal
+0x34	GR r0 r1 pc            ; greater
+0x35	GRE r0 r1 pc           ; greater_or_equal
+======== RAM ========
+0xE0	RAM_ST ?? ?? ??  ; store value in REG4 in REG5 position RAM[REG5] = REG4
+0xE1	RAM_LD ?? ?? ??  ; no es usa en realidad, la ram siempre está haciendo output en la direccion 0x04
+====== STACK ======
+0x22	PUSH r0 ?? t0           ; ?? no se usa pero debe estar, t0 sobrescribe dicha direccion a 0 
+0x23	POP ?? ?? t0
+0xE2	PUSH #0 ?? t0
+0xE3	POP ?? ?? t0
+====== FUNCTIONS ======
+0x08	CALL r0 ?? ??   ; push de pc+1 en stack y setea pc al valor que contiene r0
+0x09	RET ?? ?? ??    ; pop del stack y escribe el valor en el pc
+0x88	CALL #0
+0x89	RET  ?? ?? ??

mycpu_v2.md

@@ -0,0 +1,178 @@
+# Especificaciones
+
+Arquitectura:           16bit
+Tamaño de instruccion:  16bit
+PROM:                   Ilimitado
+RAM:                    1kB - 20kB
+STACK:                  256 - Ilimitado
+
+# Registros
+
+Cada registro puede almacenar 16 bits
+
+
+| ADDR |   NAME   |   NOTES   |
+| ---- | -------- | --------- |
+| 0x00 |   REG0   |           |
+| 0x01 |   REG1   |           |
+| 0x02 |   REG2   |           |
+| 0x03 |   REG3   |           |
+| 0x04 |   REG4   |           |
+| 0x05 |   REG5   |           |
+| 0x06 |   REG6   |           |
+| 0x07 |   REG7   |           |
+| 0x08 |   REG8   |           |
+| 0x09 |   REG9   |           |
+| 0x0A |   REG10  |           |
+| 0x0B |   REG11  |           |
+| 0x0C |   REG12  | RAM VALUE |
+| 0x0D |   REG13  | RAM ADDR  |
+| 0x0E |    PC    |           |
+| 0x0F |  IN/OUT  |           |
+
+# Opcodes
+
+Las instrucciones en este CPU tienen un tamaño total de 8 bytes, es decir, 4 parametros de 16bit cada uno.
+
+[OPCODE] [PARAM1] [PARAM2] [TARGET1]
+
+PARAM1 y PARAM2 soportan 2 modos de direccionamiento:
+- Modo registro
+- Modo inmediato
+
+TARGET1 indica el registro donde se guardará el resultado.
+
+## ALU
+
+| OPCODE | ADDR | PARAM1 | PARAM2 | TARGET1 | DESCRIPTION |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  ADD   | 0x00 |   R0   |   R1   |   T1    |             |
+|  SUB   | 0x01 |   R0   |   R1   |   T1    |             |
+|  MUL   | 0x02 |   R0   |   R1   |   T1    |             |
+|  DIV   | 0x03 |   R0   |   R1   |   T1    |             |
+|  AND   | 0x04 |   R0   |   R1   |   T1    |             |
+|  OR    | 0x05 |   R0   |   R1   |   T1    |             |
+|  NOT   | 0x06 |   R0   |   R1   |   T1    |             |
+|  NAND  | 0x07 |   R0   |   R1   |   T1    |             |
+|  NOR   | 0x08 |   R0   |   R1   |   T1    |             |
+|  XOR   | 0x09 |   R0   |   R1   |   T1    |             |
+|  XNOR  | 0x0A |   R0   |   R1   |   T1    |             |
+|  NEG   | 0x0B |   R0   |   R1   |   T1    |             |
+|   -    | 0x0C |        |        |         |             |
+|   -    | 0x0D |        |        |         |             |
+|   -    | 0x0E |        |        |         |             |
+|   -    | 0x0F |        |        |         |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  ADD   | 0x40 |   #0   |   R1   |   T1    |             |
+|  SUB   | 0x41 |   #0   |   R1   |   T1    |             |
+|  MUL   | 0x42 |   #0   |   R1   |   T1    |             |
+|  DIV   | 0x43 |   #0   |   R1   |   T1    |             |
+|  AND   | 0x44 |   #0   |   R1   |   T1    |             |
+|  OR    | 0x45 |   #0   |   R1   |   T1    |             |
+|  NOT   | 0x46 |   #0   |   R1   |   T1    |             |
+|  NAND  | 0x47 |   #0   |   R1   |   T1    |             |
+|  NOR   | 0x48 |   #0   |   R1   |   T1    |             |
+|  XOR   | 0x49 |   #0   |   R1   |   T1    |             |
+|  XNOR  | 0x4A |   #0   |   R1   |   T1    |             |
+|  NEG   | 0x4B |   #0   |   R1   |   T1    |             |
+|   -    | 0x4C |        |        |         |             |
+|   -    | 0x4D |        |        |         |             |
+|   -    | 0x4E |        |        |         |             |
+|   -    | 0x4F |        |        |         |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  ADD   | 0x80 |   R0   |   #1   |   T1    |             |
+|  SUB   | 0x81 |   R0   |   #1   |   T1    |             |
+|  MUL   | 0x82 |   R0   |   #1   |   T1    |             |
+|  DIV   | 0x83 |   R0   |   #1   |   T1    |             |
+|  AND   | 0x84 |   R0   |   #1   |   T1    |             |
+|  OR    | 0x85 |   R0   |   #1   |   T1    |             |
+|  NOT   | 0x86 |   R0   |   #1   |   T1    |             |
+|  NAND  | 0x87 |   R0   |   #1   |   T1    |             |
+|  NOR   | 0x88 |   R0   |   #1   |   T1    |             |
+|  XOR   | 0x89 |   R0   |   #1   |   T1    |             |
+|  XNOR  | 0x8A |   R0   |   #1   |   T1    |             |
+|  NEG   | 0x8B |   R0   |   #1   |   T1    |             |
+|   -    | 0x8C |        |        |         |             |
+|   -    | 0x8D |        |        |         |             |
+|   -    | 0x8E |        |        |         |             |
+|   -    | 0x8F |        |        |         |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  ADD   | 0xC0 |   #0   |   #1   |   T1    |             |
+|  SUB   | 0xC1 |   #0   |   #1   |   T1    |             |
+|  MUL   | 0xC2 |   #0   |   #1   |   T1    |             |
+|  DIV   | 0xC3 |   #0   |   #1   |   T1    |             |
+|  AND   | 0xC4 |   #0   |   #1   |   T1    |             |
+|  OR    | 0xC5 |   #0   |   #1   |   T1    |             |
+|  NOT   | 0xC6 |   #0   |   #1   |   T1    |             |
+|  NAND  | 0xC7 |   #0   |   #1   |   T1    |             |
+|  NOR   | 0xC8 |   #0   |   #1   |   T1    |             |
+|  XOR   | 0xC9 |   #0   |   #1   |   T1    |             |
+|  XNOR  | 0xCA |   #0   |   #1   |   T1    |             |
+|  NEG   | 0xCB |   #0   |   #1   |   T1    |             |
+|   -    | 0xCC |        |        |         |             |
+|   -    | 0xCD |        |        |         |             |
+|   -    | 0xCE |        |        |         |             |
+|   -    | 0xCF |        |        |         |             |
+
+## CONDITIONALS
+
+En los condicionales TARGET1 representa a la direccion del PC (Program Counter) que se saltará si se cumple la condicion.
+
+| OPCODE | ADDR | PARAM1 | PARAM2 | TARGET1 | DESCRIPTION |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  EQ    | 0x10 |   R0   |   R1   |   T1    |    equal    |
+|  NEQ   | 0x11 |   R0   |   R1   |   T1    |  not equal  |
+|  LS    | 0x12 |   R0   |   R1   |   T1    |     less    |
+|  LSE   | 0x13 |   R0   |   R1   |   T1    | less or eq  |
+|  GR    | 0x14 |   R0   |   R1   |   T1    |   greater   |
+|  GRE   | 0x15 |   R0   |   R1   |   T1    |greater or eq|
+|        | 0x16 |   R0   |   R1   |   T1    |             |
+|        | 0x17 |   R0   |   R1   |   T1    |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  EQ    | 0x50 |   #0   |   R1   |   T1    |    equal    |
+|  NEQ   | 0x51 |   #0   |   R1   |   T1    |  not equal  |
+|  LS    | 0x52 |   #0   |   R1   |   T1    |     less    |
+|  LSE   | 0x53 |   #0   |   R1   |   T1    | less or eq  |
+|  GR    | 0x54 |   #0   |   R1   |   T1    |   greater   |
+|  GRE   | 0x55 |   #0   |   R1   |   T1    |greater or eq|
+|        | 0x56 |   #0   |   R1   |   T1    |             |
+|        | 0x57 |   #0   |   R1   |   T1    |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  EQ    | 0x90 |   R0   |   #1   |   T1    |    equal    |
+|  NEQ   | 0x91 |   R0   |   #1   |   T1    |  not equal  |
+|  LS    | 0x92 |   R0   |   #1   |   T1    |     less    |
+|  LSE   | 0x93 |   R0   |   #1   |   T1    | less or eq  |
+|  GR    | 0x94 |   R0   |   #1   |   T1    |   greater   |
+|  GRE   | 0x95 |   R0   |   #1   |   T1    |greater or eq|
+|        | 0x96 |   R0   |   #1   |   T1    |             |
+|        | 0x97 |   R0   |   #1   |   T1    |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  EQ    | 0xD0 |   #0   |   #1   |   T1    |    equal    |
+|  NEQ   | 0xD1 |   #0   |   #1   |   T1    |  not equal  |
+|  LS    | 0xD2 |   #0   |   #1   |   T1    |     less    |
+|  LSE   | 0xD3 |   #0   |   #1   |   T1    | less or eq  |
+|  GR    | 0xD4 |   #0   |   #1   |   T1    |   greater   |
+|  GRE   | 0xD5 |   #0   |   #1   |   T1    |greater or eq|
+|        | 0xD6 |   #0   |   #1   |   T1    |             |
+|        | 0xD7 |   #0   |   #1   |   T1    |             |
+
+
+## CONTROL UNIT
+
+| OPCODE | ADDR | PARAM1 | PARAM2 | TARGET1 | DESCRIPTION |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  RSTR  | 0x18 |   --   |   --   |    -    |             |
+|  PUSH  | 0x19 |   R1   |   --   |    -    |             |
+|  POP   | 0x1A |   --   |   --   |    T1   |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  PUSH  | 0x59 |   #1   |   --   |    -    |             |
+
+## FUNCTIONS
+
+| OPCODE | ADDR | PARAM1 | PARAM2 | TARGET1 | DESCRIPTION |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  CALL  | 0x20 |   R1   |   --   |    -    |             |
+|  RET   | 0x21 |   --   |   --   |    -    |             |
+|  HALT  | 0x22 |   --   |   --   |    -    |             |
+| ------ | ---- | ------ | ------ | ------- | ----------- |
+|  CALL  | 0x60 |   #1   |   --   |    -    |             |

projects/classes.j

@@ -0,0 +1,11 @@
+class Dog:
+    fn init(self, name):
+        self.name = name
+
+    fn bark(self):
+        println("guau!")
+
+d = Dog("ahi te va")
+x = d.bark()
+println("Hola ", d.name)
+debugHeap()

projects/comment.j

@@ -0,0 +1,5 @@
+// Esto es un comentario
+println("Hello World!")
+
+// Otro comentario mas
+println(40)

projects/custom_fn.j

@@ -0,0 +1,8 @@
+x = "Hello world!"
+
+fn suma(x, y):
+    fn pow(z):
+        return z * z
+    return x + pow(y)
+
+println(suma(2, 2))

projects/functions.j

@@ -0,0 +1,3 @@
+x = "Hello"
+y = 2 * (4 - 2)
+println(y)

projects/if.j

@@ -0,0 +1,4 @@
+x = 20
+if x < 10:
+    print x
+print -300

projects/mycpu/assign.j

@@ -0,0 +1,3 @@
+x = 10
+y = 512
+z = x + y

projects/mycpu/functions.j

@@ -0,0 +1,6 @@
+g = 2
+
+fn suma(x, y):
+    return x + y
+
+x = suma(5, 2) - g + 1

projects/mycpu/ifs.j

@@ -0,0 +1,12 @@
+counter = 0
+
+fn inc():
+    counter = counter + 1
+
+fn main():
+    if counter < 30:
+        inc()
+        main()
+
+
+main()

projects/mycpu/while.j

@@ -0,0 +1,4 @@
+counter = 0
+
+while counter < 65000:
+    counter = counter + 1

projects/str.j

@@ -0,0 +1,12 @@
+x = 0
+while x < 10:
+    x = x + 1
+
+if x > 9:
+    println "fin"
+
+
+x = "a"
+y = x * 1
+z = y + 2
+println "a" + z

projects/test.j

@@ -0,0 +1,3 @@
+x = 5
+z = x > 2
+print x + z

projects/vm_simple.j

@@ -0,0 +1,2 @@
+x = 8 + 2 * 4
+print(x, end="\n")

projects/while.j

@@ -1,4 +1,5 @@
 x = 0
-while x < 100000000:
+while x < 10000000:
     x = x + 1
-print x
+print(x)
+debugHeap()

readme.md

@@ -1,52 +1,240 @@
 # j-lang
 
-La idea de j-lang es crear un "proto-lenguaje" parecido a python pero implementado desde 0 para validar y aprender más sobre la gestión de memoria.
+Un proto-lenguaje con sintaxis inspirada en Python, implementado desde cero en C. El objetivo es aprender sobre gestion de memoria, tokenizacion, parsing y evaluacion de un lenguaje de programacion.
 
-Actualmente en `mem-heap\src\allocator.h` ya hay una implementeción de un Memory Allocator casi funcional.
+## Estado actual
 
-## 🗺️ Hoja de Ruta: Proyecto Proto-Lenguaje
+Las 5 fases del interprete estan implementadas y funcionando:
-Esta ruta va desde lo más bajo (la memoria) hasta lo más alto (ejecutar código).
 
-### Fase 1: El Cimiento (Gestión de Memoria) 🏗️
+```
-Objetivo: Tener un malloc y free propios que gestionen metadatos compactos.
+Codigo fuente (.j)
+       |
+   [LEXER]      src/frontend/lexer.h
+       |
+   Tokens
+       |
+   [PARSER]     src/frontend/parser.h
+       |
+   AST
+       |
+   [EVAL]       src/vm/eval.h
+       |
+   Ejecucion + GC
+```
 
-Estado: ¡Ya estás aquí!
+### Que funciona
 
-Tareas clave:
+- **Variables y asignacion:** `x = 10`
-- [ ] Terminar CMA_malloc con la cabecera compactada (Size + Marked + InUse).
+- **Aritmetica:** `+`, `-`, `*`, `/` con enteros
-- [ ] Implementar una función CMA_free que pueda liberar un bloque específico.
+- **Comparaciones:** `<`, `>`
+- **Strings:** literales, concatenacion con `+`, `len()`
+- **Control de flujo:** `if` y `while` con bloques indentados (estilo Python)
+- **Funciones built-in:** `print()`, `println()`, `len()`
+- **Llamadas a funciones** con multiples argumentos separados por `,`
+- **Expresiones con parentesis:** `2 * (4 - 2)`
+- **Numeros negativos:** `-300`
+- **Comentarios:** `// esto es un comentario`
 
-### Fase 2: El Modelo de Objetos (Object Model) 📦
+### Ejemplo
-Objetivo: Definir cómo se ve un número, una cadena o una lista dentro de tu memoria C.
-Conexión: Cada objeto de tu lenguaje será un struct en C que comienza con tu CMA_metadata.
 
-Tareas clave:
+```
-- [ ] Crear un enum para los tipos (ENTERO, STRING, LISTA).
+x = 0
-- [ ] Definir el struct Object genérico que envuelve tus datos.
+while x < 10:
+    x = x + 1
 
-### Fase 3: El Front-End (Lexer y Parser) 📖
+if x > 9:
-Objetivo: Convertir el texto del código fuente en algo que C entienda.
+    println("fin")
-
+```
-Tareas clave:
-- [ ] Lexer (Tokenizador): Romper el texto x = 10 en fichas: [ID:x], [OP:=], [NUM:10].
-- [ ] Parser: Organizar esas fichas en un Árbol de Sintaxis Abstracta (AST). Por ejemplo, un nodo "Asignación" que tiene un hijo "x" y otro "10".
-
-### Fase 4: El Motor (Evaluador o VM) ⚙️
-Objetivo: Recorrer el árbol y "hacer" lo que dice.
-
-Tareas clave:
-- [ ] Crear una función recursiva eval(nodo) que ejecute la lógica.
-
-Si es un nodo SUMA, suma los hijos. Si es un nodo IMPRIMIR, muestra en pantalla.
-
-### Fase 5: El Recolector de Basura (Garbage Collector) 🧹
-Objetivo: Automatizar la limpieza.
-
-Tareas clave:
-- [ ] Implementar Mark (Marcar): Recorrer todos los objetos accesibles desde tus variables y poner el bit Marked a 1.
-- [ ] Implementar Sweep (Barrer): Recorrer todo el heap linealmente (usando tu función next_block). Si un bloque tiene Marked == 0 y InUse == 1, llamar a CMA_free.
 
 ## Estructura del proyecto
 
-- vm: maquina virtual de j-lang
+```
-- projects: carpeta con scripts en j-lang
+src/
+  frontend/
+    lexer.h        Tokenizador: texto -> tokens
+    parser.h       Parser: tokens -> AST
+  memory/
+    allocator.h    Memory allocator custom (heap simulado)
+    gc.h           Garbage collector (mark-and-sweep)
+  objects/
+    object.h       Modelo de objetos (int, float, string, list)
+  vm/
+    eval.h         Evaluador: recorre el AST y ejecuta
+  main.c           Punto de entrada
+projects/            Scripts de ejemplo en .j
+docs/
+  roadmap.md       Roadmap detallado de implementacion
+```
+
+### Memory allocator
+
+Heap simulado sobre un array de bytes con metadatos por bloque (`size`, `in_use`, `marked`). Soporta asignacion, liberacion, reutilizacion de bloques libres (first-fit) y crecimiento automatico cuando se queda sin espacio.
+
+### Garbage collector
+
+Mark-and-sweep: marca los objetos alcanzables desde las variables del environment, barre los no marcados y fusiona bloques libres contiguos.
+
+### Modelo de objetos
+
+Los valores del lenguaje se representan como `Object` con tagged union. Tipos soportados: `OBJ_INT`, `OBJ_FLOAT`, `OBJ_STRING`, `OBJ_LIST`, `OBJ_NONE`. Los objetos viven en el heap custom y se referencian por offset (no punteros absolutos).
+
+## Compilar y ejecutar
+
+```bash
+gcc src/main.c -o run
+./run projects/sum.j
+```
+
+---
+
+## Roadmap: que falta para hacer un juego 2D con JLang
+
+Para poder escribir un juego 2D tipo "mover un personaje por pantalla, disparar, colisiones" con JLang, harian falta estos bloques:
+
+### 1. Funciones de usuario
+
+Lo mas urgente. Sin funciones no se puede organizar nada.
+
+```
+fn update(dt):
+    player_x = player_x + speed * dt
+
+fn draw():
+    draw_rect(player_x, player_y, 32, 32)
+```
+
+Implica: nuevo token `fn`, nodo `NODE_FUNC_DEF` en el AST, almacenar el cuerpo de la funcion en el environment, y un mecanismo de scopes (variables locales vs globales).
+
+### 2. Return
+
+Las funciones necesitan devolver valores.
+
+```
+fn distance(x1, y1, x2, y2):
+    dx = x1 - x2
+    dy = y1 - y2
+    return sqrt(dx * dx + dy * dy)
+```
+
+### 3. Structs o clases
+
+Para representar entidades del juego (jugador, enemigos, balas...).
+
+```
+class Entity:
+    x = 0
+    y = 0
+    w = 32
+    h = 32
+
+player = Entity()
+player.x = 100
+player.y = 200
+```
+
+Implica: acceso a campos con `.`, constructor, almacenar la definicion de la clase como un objeto mas en el heap.
+
+### 4. Listas funcionales
+
+Las listas ya existen como tipo (`OBJ_LIST`) pero no hay sintaxis para usarlas. Se necesitan para manejar colecciones de entidades.
+
+```
+enemies = [Enemy(), Enemy(), Enemy()]
+append(enemies, Enemy())
+i = 0
+while i < len(enemies):
+    update(enemies[i])
+    i = i + 1
+```
+
+Implica: sintaxis `[...]`, acceso por indice `lista[i]`, `append()`, `len()` para listas.
+
+### 5. Else / elif
+
+Imprescindible para logica de juego.
+
+```
+if key == "left":
+    player_x = player_x - speed
+elif key == "right":
+    player_x = player_x + speed
+else:
+    speed = 0
+```
+
+### 6. For loops
+
+Iterar de forma mas limpia que con `while`.
+
+```
+for enemy in enemies:
+    draw_rect(enemy.x, enemy.y, enemy.w, enemy.h)
+```
+
+### 7. Operadores que faltan
+
+- `%` (modulo) - util para animaciones ciclicas, wrapping
+- `==`, `!=` (ya tokenizados pero no evaluados completamente)
+- `<=`, `>=`
+- `and`, `or`, `not` - operadores logicos
+- `+=`, `-=` - azucar sintactico
+
+### 8. Floats funcionales
+
+El tipo `OBJ_FLOAT` existe pero no se puede usar desde el lenguaje. Para un juego se necesita aritmetica de punto flotante para posiciones, velocidades, delta time, etc.
+
+```
+player_x = 100.0
+speed = 2.5
+player_x = player_x + speed * dt
+```
+
+### 9. Libreria grafica (FFI a C)
+
+El punto critico. JLang necesita poder llamar a una libreria grafica en C como SDL2 o raylib. Hay dos caminos:
+
+**Opcion A: Built-in functions (mas facil)**
+Registrar funciones C directamente en el evaluador, como ya se hace con `print`:
+
+```c
+// En el eval, junto a print/println:
+if (strcmp(name, "draw_rect") == 0) { SDL_RenderFillRect(...); }
+if (strcmp(name, "key_pressed") == 0) { ... }
+```
+
+**Opcion B: FFI generico (mas ambicioso)**
+Un sistema para enlazar funciones C arbitrarias desde JLang.
+
+Las funciones minimas para un juego serian:
+
+| Funcion | Descripcion |
+|---|---|
+| `create_window(w, h, title)` | Crear ventana |
+| `clear()` | Limpiar pantalla |
+| `draw_rect(x, y, w, h, r, g, b)` | Dibujar rectangulo |
+| `draw_image(path, x, y)` | Dibujar imagen/sprite |
+| `present()` | Mostrar frame |
+| `key_pressed(key)` | Consultar tecla |
+| `get_dt()` | Delta time entre frames |
+| `random(min, max)` | Numero aleatorio |
+
+### 10. Funciones matematicas
+
+`sqrt()`, `sin()`, `cos()`, `abs()`, `random()`. Todas se pueden registrar como built-ins que llamen a `math.h`.
+
+### Orden sugerido de implementacion
+
+```
+ 1. Funciones de usuario + return    (sin esto no se puede hacer nada)
+ 2. Else / elif
+ 3. Floats funcionales
+ 4. Operadores que faltan (%, <=, >=, and, or)
+ 5. Listas con sintaxis ([], indexado, append)
+ 6. For loops
+ 7. Structs o clases
+ 8. Built-ins graficos (SDL2/raylib)
+ 9. Funciones matematicas
+10. Juego 2D funcional
+```
+
+Los pasos 1-7 son trabajo puro de lenguaje (lexer/parser/eval). El paso 8 es donde JLang toca el mundo real: linkear con SDL2 o raylib a la hora de compilar y exponer las funciones como built-ins en el evaluador.

src/backend/bytecode/compiler.h

@@ -0,0 +1,470 @@
+#ifndef JLANG_COMPILER_H
+#define JLANG_COMPILER_H
+
+#include "../../frontend/parser.h"
+#include "opcodes.h"
+#include <string.h>
+
+typedef struct {
+  char *method_name;
+  int entry_point;
+  int param_count;
+  char **param_names;
+} MethodEntry;
+
+typedef struct {
+  char *name;
+  MethodEntry methods[16];
+  int method_count;
+} ClassEntry;
+
+typedef struct {
+  char *name;
+  int entry_point; // indice de la primera instruccion
+  int param_count;
+  char **param_names; // nombres de parametros
+} FunctionEntry;
+
+typedef struct {
+  Instruction code[4096]; // bytecodes
+  int code_count;
+  char *constants[256]; // pool de strings literales
+  int const_count;
+  char *names[256]; // tabla de nombres (variables + funciones)
+  int name_count;
+  FunctionEntry functions[64];
+  int func_count;
+  ClassEntry classes[16];
+  int class_count;
+} Chunk;
+
+int emit(Chunk *chunk, Instruction instr) {
+  chunk->code[chunk->code_count++] = instr;
+  return chunk->code_count - 1;
+}
+
+int add_constant(Chunk *chunk, char *str) {
+  for (int i = 0; i < chunk->const_count; i++) {
+    if (strcmp(chunk->constants[i], str) == 0) {
+      return i;
+    }
+  }
+
+  chunk->constants[chunk->const_count++] = str;
+  return chunk->const_count - 1;
+}
+
+Instruction make_instruction(OpCode op) {
+  Instruction instr;
+  instr.op = op;
+  return instr;
+}
+
+int add_name(Chunk *chunk, char *name) {
+  for (int i = 0; i < chunk->name_count; i++) {
+    if (strcmp(chunk->names[i], name) == 0) {
+      return i;
+    }
+  }
+
+  chunk->names[chunk->name_count++] = name;
+  return chunk->name_count - 1;
+}
+
+int compile_node(Chunk *chunk, ASTNode *node) {
+  switch (node->type) {
+  case NODE_INT_LIT: {
+    Instruction instr = make_instruction(OP_CONST_INT);
+    instr.operand.int_val = node->data.int_val;
+    return emit(chunk, instr);
+  }
+  case NODE_STRING_LIT: {
+    Instruction instr = make_instruction(OP_CONST_STRING);
+    instr.operand.str_index = add_constant(chunk, node->data.string_val);
+    return emit(chunk, instr);
+  }
+  case NODE_VAR: {
+    Instruction instr = make_instruction(OP_LOAD_VAR);
+    instr.operand.var_index = add_name(chunk, node->data.string_val);
+    return emit(chunk, instr);
+  }
+  case NODE_ASSIGN: {
+    compile_node(chunk, node->data.assign.value);
+    Instruction instr = make_instruction(OP_STORE_VAR);
+    instr.operand.var_index = add_name(chunk, node->data.assign.name);
+    return emit(chunk, instr);
+  }
+  case NODE_CALL: {
+    // Compilar cada argumento y pushear al stack
+    for (int i = 0; i < node->data.call.arg_count; i++) {
+      compile_node(chunk, node->data.call.args[i]);
+    }
+
+    // Verificar si es constructor de una clase
+    for (int i = 0; i < chunk->class_count; i++) {
+      if (strcmp(chunk->classes[i].name, node->data.call.name) == 0) {
+        // Buscar init
+        for (int m = 0; m < chunk->classes[i].method_count; m++) {
+          if (strcmp(chunk->classes[i].methods[m].method_name, "init") == 0) {
+            int expected =
+                chunk->classes[i].methods[m].param_count - 1; // -1 por self
+            if (node->data.call.arg_count != expected) {
+              printf("error: %s() espera %d args, pero recibio %d\n",
+                     node->data.call.name, expected, node->data.call.arg_count);
+              exit(1);
+            }
+            break;
+          }
+        }
+        break;
+      }
+    }
+
+    // Registrar el nombre de la funcion
+    Instruction instr = make_instruction(OP_CALL);
+    instr.operand.call.arg_count = node->data.call.arg_count;
+    instr.operand.call.name_index = add_name(chunk, node->data.call.name);
+    return emit(chunk, instr);
+  }
+  case NODE_BLOCK: {
+    int n = node->data.block.count;
+
+    // NOP for gc
+    emit(chunk, make_instruction(OP_NOP));
+
+    for (int i = 0; i < n; i++) {
+      compile_node(chunk, node->data.block.stmts[i]);
+    }
+
+    return 0;
+  }
+  case NODE_BINOP: {
+    int leftOffset = compile_node(chunk, node->data.binop.left);
+    int rightOffset = compile_node(chunk, node->data.binop.right);
+
+    OpCode opCode;
+    switch (node->data.binop.op) {
+    case '+':
+      opCode = OP_ADD;
+      break;
+    case '-':
+      opCode = OP_SUB;
+      break;
+    case '*':
+      opCode = OP_MUL;
+      break;
+    case '/':
+      opCode = OP_DIV;
+      break;
+    case '>':
+      opCode = OP_CMP_GT;
+      break;
+    case '<':
+      opCode = OP_CMP_LT;
+      break;
+    default:
+      break;
+    }
+    emit(chunk, make_instruction(opCode));
+    return 0;
+  }
+  case NODE_WHILE: {
+    int loop_start = chunk->code_count;
+    compile_node(chunk, node->data.while_loop.cond);
+    // jump if zero, zero = false
+    Instruction instr = make_instruction(OP_JUMP_IF_ZERO);
+    instr.operand.jump_target = -1;
+    int jump_offset = emit(chunk, instr);
+
+    // compile body
+    compile_node(chunk, node->data.while_loop.body);
+
+    instr = make_instruction(OP_JUMP);
+    instr.operand.jump_target = loop_start;
+    emit(chunk, instr);
+
+    // Bachpatching
+    chunk->code[jump_offset].operand.jump_target = chunk->code_count;
+    break;
+  }
+
+  case NODE_IF: {
+    // compile condition
+    compile_node(chunk, node->data.if_statement.cond);
+
+    // add jump if zero
+    Instruction instr = make_instruction(OP_JUMP_IF_ZERO);
+    instr.operand.jump_target = -1;
+    int jump_offset = emit(chunk, instr);
+
+    // compile body
+    compile_node(chunk, node->data.if_statement.body);
+
+    chunk->code[jump_offset].operand.jump_target = chunk->code_count;
+    break;
+  }
+
+  case NODE_RETURN: {
+    if (node->data.ret.value) {
+      compile_node(chunk, node->data.ret.value);
+    }
+    emit(chunk, make_instruction(OP_RETURN));
+    return 0;
+  }
+  case NODE_FN_DEF: {
+    // emitir jmp para ignorar la funcion por defecto
+    Instruction jump = make_instruction(OP_JUMP);
+    jump.operand.jump_target = -1; // backpatch despues
+    int jump_idx = emit(chunk, jump);
+
+    // registrar entrypoint de la funcion
+    int entry = chunk->code_count;
+    FunctionEntry *fn = &chunk->functions[chunk->func_count++];
+    fn->name = node->data.fn_def.name;
+    fn->entry_point = entry;
+    fn->param_count = node->data.fn_def.param_count;
+    fn->param_names = node->data.fn_def.params;
+
+    // emitir store_var para cada parametro (orden inverso al stack)
+    for (int i = node->data.fn_def.param_count - 1; i >= 0; i--) {
+      Instruction store = make_instruction(OP_STORE_VAR);
+      store.operand.var_index = add_name(chunk, node->data.fn_def.params[i]);
+      emit(chunk, store);
+    }
+
+    // compilar el cuerpo
+    compile_node(chunk, node->data.fn_def.body);
+
+    // emitir el return implicito (por si no hay return explicito)
+    emit(chunk, make_instruction(OP_RETURN));
+
+    // backpatch jump
+    chunk->code[jump_idx].operand.jump_target = chunk->code_count;
+    break;
+  }
+
+  case NODE_CLASS_DEF: {
+    // Registrar ClassEntry
+    ClassEntry *cls = &chunk->classes[chunk->class_count++];
+    cls->name = node->data.class_def.name;
+    cls->method_count = 0;
+
+    // Pre-registrar self en la tabla de nombres
+    add_name(chunk, "self");
+
+    int totalMethods = node->data.class_def.method_count;
+    for (int i = 0; i < totalMethods; i++) {
+      ASTNode *method = node->data.class_def.methods[i];
+
+      // jump over
+      Instruction jump = make_instruction(OP_JUMP);
+      jump.operand.jump_target = -1;
+      int jump_idx = emit(chunk, jump);
+
+      // Registrar method entry
+      int entry = chunk->code_count;
+      MethodEntry *me = &cls->methods[cls->method_count++];
+      me->method_name = method->data.fn_def.name;
+      me->entry_point = entry;
+      me->param_count = method->data.fn_def.param_count;
+      me->param_names = method->data.fn_def.params;
+
+      // store_var para cada parametro (orden inverso)
+      for (int p = method->data.fn_def.param_count - 1; p >= 0; p--) {
+        Instruction store = make_instruction(OP_STORE_VAR);
+        store.operand.var_index =
+            add_name(chunk, method->data.fn_def.params[p]);
+        emit(chunk, store);
+      }
+
+      // Compilar body
+      compile_node(chunk, method->data.fn_def.body);
+
+      // return implicito
+      emit(chunk, make_instruction(OP_RETURN));
+
+      // Backpatch
+      chunk->code[jump_idx].operand.jump_target = chunk->code_count;
+    }
+    break;
+  }
+  case NODE_DOT_ACCESS: {
+    compile_node(chunk, node->data.dot_access.object);
+    Instruction instr = make_instruction(OP_GET_FIELD);
+    instr.operand.var_index = add_name(chunk, node->data.dot_access.field);
+    emit(chunk, instr);
+    break;
+  }
+  case NODE_DOT_ASSIGN: {
+    compile_node(chunk, node->data.dot_assign.value);  // push valor
+    compile_node(chunk, node->data.dot_assign.object); // push instancia
+    Instruction instr = make_instruction(OP_SET_FIELD);
+    instr.operand.var_index = add_name(chunk, node->data.dot_assign.field);
+    emit(chunk, instr);
+    break;
+  }
+  case NODE_METHOD_CALL: {
+    compile_node(chunk, node->data.method_call.object); // push instancia
+    for (int i = 0; i < node->data.method_call.arg_count; i++) {
+      compile_node(chunk, node->data.method_call.args[i]);
+    }
+    Instruction instr = make_instruction(OP_CALL_METHOD);
+    instr.operand.call.name_index =
+        add_name(chunk, node->data.method_call.method);
+    instr.operand.call.arg_count = node->data.method_call.arg_count;
+    emit(chunk, instr);
+    break;
+  }
+
+  default:
+    break;
+  }
+
+  return 0;
+}
+
+Chunk *compile(ASTNode *root) {
+  // Create chunk
+  Chunk *chunk = (Chunk *)malloc(sizeof(Chunk));
+
+  // Set arrays to 0
+  memset(chunk, 0, sizeof(Chunk));
+
+  compile_node(chunk, root);
+
+  Instruction instr;
+  instr.op = OP_HALT;
+  emit(chunk, instr);
+
+  return chunk;
+}
+
+void print_chunk(Chunk *chunk) {
+  printf("=== Names (%d) ===\n", chunk->name_count);
+  for (int i = 0; i < chunk->name_count; i++) {
+    printf("  [%d] %s\n", i, chunk->names[i]);
+  }
+
+  printf("=== Constants (%d) ===\n", chunk->const_count);
+  for (int i = 0; i < chunk->const_count; i++) {
+    printf("  [%d] \"%s\"\n", i, chunk->constants[i]);
+  }
+
+  printf("=== Bytecode (%d instructions) ===\n", chunk->code_count);
+  for (int i = 0; i < chunk->code_count; i++) {
+    Instruction instr = chunk->code[i];
+    printf("%04d  ", i);
+    switch (instr.op) {
+    case OP_CONST_INT:
+      printf("CONST_INT %d", instr.operand.int_val);
+      break;
+    case OP_CONST_STRING:
+      printf("CONST_STRING [%d] \"%s\"", instr.operand.str_index,
+             chunk->constants[instr.operand.str_index]);
+      break;
+    case OP_POP:
+      printf("POP");
+      break;
+    case OP_ADD:
+      printf("ADD");
+      break;
+    case OP_SUB:
+      printf("SUB");
+      break;
+    case OP_MUL:
+      printf("MUL");
+      break;
+    case OP_DIV:
+      printf("DIV");
+      break;
+    case OP_NEG:
+      printf("NEG");
+      break;
+    case OP_CMP_LT:
+      printf("CMP_LT");
+      break;
+    case OP_CMP_GT:
+      printf("CMP_GT");
+      break;
+    case OP_LOAD_VAR:
+      printf("LOAD_VAR [%d] %s", instr.operand.var_index,
+             chunk->names[instr.operand.var_index]);
+      break;
+    case OP_STORE_VAR:
+      printf("STORE_VAR [%d] %s", instr.operand.var_index,
+             chunk->names[instr.operand.var_index]);
+      break;
+    case OP_JUMP:
+      printf("JUMP -> %04d", instr.operand.jump_target);
+      break;
+    case OP_JUMP_IF_ZERO:
+      printf("JUMP_IF_ZERO -> %04d", instr.operand.jump_target);
+      break;
+    case OP_CALL:
+      printf("CALL %s(%d args)", chunk->names[instr.operand.call.name_index],
+             instr.operand.call.arg_count);
+      break;
+    case OP_RETURN:
+      printf("RETURN");
+      break;
+    case OP_NOP:
+      printf("NOP");
+      break;
+    case OP_HALT:
+      printf("HALT");
+      break;
+    case OP_GET_FIELD:
+      printf("GET_FIELD [%d] %s", instr.operand.var_index,
+             chunk->names[instr.operand.var_index]);
+      break;
+    case OP_SET_FIELD:
+      printf("SET_FIELD [%d] %s", instr.operand.var_index,
+             chunk->names[instr.operand.var_index]);
+      break;
+    case OP_CALL_METHOD:
+      printf("CALL_METHOD %s(%d args)",
+             chunk->names[instr.operand.call.name_index],
+             instr.operand.call.arg_count);
+      break;
+
+    default:
+      printf("UNKNOWN op=%d", instr.op);
+      break;
+    }
+    printf("\n");
+  }
+  printf("=== User Functions ===\n");
+  for (int i = 0; i < chunk->func_count; i++) {
+    FunctionEntry *fn = &chunk->functions[i];
+    printf("[%.4d] %s(", fn->entry_point, fn->name);
+
+    for (int p = 0; p < fn->param_count; p++) {
+      printf("%s", fn->param_names[p]);
+
+      if (p < fn->param_count - 1) {
+        printf(", ");
+      }
+    }
+    printf(")\n");
+  }
+
+  printf("=== Classes ===\n");
+  for (int i = 0; i < chunk->class_count; i++) {
+    ClassEntry *cls = &chunk->classes[i];
+    printf("class %s (%d methods)\n", cls->name, cls->method_count);
+    for (int m = 0; m < cls->method_count; m++) {
+      MethodEntry *me = &cls->methods[m];
+      printf("  [%.4d] %s(", me->entry_point, me->method_name);
+      for (int p = 0; p < me->param_count; p++) {
+        printf("%s", me->param_names[p]);
+        if (p < me->param_count - 1)
+          printf(", ");
+      }
+      printf(")\n");
+    }
+  }
+
+  printf("=== End ===\n");
+}
+
+#endif

src/backend/bytecode/opcodes.h

@@ -0,0 +1,43 @@
+#ifndef JLANG_OPCODES_H
+#define JLANG_OPCODES_H
+
+typedef enum {
+  OP_CONST_INT,    // push entero inmediato
+  OP_CONST_STRING, // push string desde pool de constantes (alloc en heap)
+  OP_POP,          // descarta top del stack
+  OP_ADD,
+  OP_SUB,
+  OP_MUL,
+  OP_DIV, // aritmetica
+  OP_NEG, // negacion unaria
+  OP_CMP_LT,
+  OP_CMP_GT,       // comparacion -> push 0 o 1
+  OP_LOAD_VAR,     // push variable por indice
+  OP_STORE_VAR,    // pop -> guardar en variable por indice
+  OP_JUMP,         // salto incondicional
+  OP_JUMP_IF_ZERO, // pop -> si false, saltar
+  OP_CALL,         // llamar built-in por indice de nombre
+  OP_RETURN,       // retornar de funcion (pop call frame)
+  OP_NOP,
+  OP_HALT,
+  OP_GET_FIELD,   // TOS=instance, operand=name_idx → push field value
+  OP_SET_FIELD,   // TOS=instance, TOS-1=value, operand=name_idx → set field
+  OP_CALL_METHOD, // TOS-N-1=instance + N args, operand={name_idx, arg_count}
+
+} OpCode;
+
+typedef struct {
+  OpCode op;
+  union {
+    int int_val;     // OP_CONST_INT
+    int str_index;   // OP_CONST_STRING: indice a pool de constantes
+    int var_index;   // OP_LOAD_VAR, OP_STORE_VAR
+    int jump_target; // OP_JUMP, OP_JUMP_IF_ZERO
+    struct {
+      int name_index;
+      int arg_count;
+    } call; // OP_CALL
+  } operand;
+} Instruction;
+
+#endif

src/backend/bytecode/value.h

@@ -0,0 +1,26 @@
+#ifndef JLANG_VALUE_H
+#define JLANG_VALUE_H
+
+#include "opcodes.h"
+#include <stdlib.h>
+
+typedef enum
+{
+    VAL_INT,
+    VAL_FLOAT,
+    VAL_OBJ,
+    VAL_NONE,
+} ValueType;
+
+typedef struct
+{
+    ValueType type;
+    union
+    {
+        int int_val;
+        double float_val;
+        size_t heap_offset; // para strings, listas
+    } as;
+} Value;
+
+#endif

src/backend/bytecode/vm.h

@@ -0,0 +1,518 @@
+#ifndef JLANG_VM_H
+#define JLANG_VM_H
+
+#include "../../memory/gc.h"
+#include "compiler.h"
+#include "value.h"
+
+typedef struct {
+  int return_ip;         // a donde volver
+  int saved_sp;          // base del stack
+  Value saved_vars[256]; // variables del caller (snapshot)
+  int saved_var_set[256];
+  int is_constructor;
+  Value constructor_instance;
+} CallFrame;
+
+typedef struct {
+  Chunk *chunk;
+  int ip; // instruction pointer
+  Value stack[1024];
+  int sp;           // stack pointer
+  Value vars[256];  // variables por indice
+  int var_set[256]; // 0=no definida, 1=definida
+
+  CallFrame frames[64];
+  int frame_count;
+
+  JLANG_memory_allocator *allocator;
+} VM;
+
+void run_vm(VM *vm) {
+  while (1) {
+    Instruction instr = vm->chunk->code[vm->ip];
+
+    switch (instr.op) {
+    case OP_HALT:
+      // Stop vm
+      return;
+    case OP_JUMP: {
+      // Go to instruction
+      vm->ip = instr.operand.jump_target;
+      continue;
+    }
+
+    case OP_JUMP_IF_ZERO: {
+      // pop from stack
+      Value var1 = vm->stack[--vm->sp];
+      if (var1.as.int_val == 0) {
+        vm->ip = instr.operand.jump_target;
+        continue;
+      }
+      break;
+    }
+
+    case OP_CONST_INT: {
+      // push value to stack
+      Value v = {0};
+      v.type = VAL_INT;
+      v.as.int_val = instr.operand.int_val;
+      vm->stack[vm->sp++] = v;
+      break;
+    }
+
+    case OP_CONST_STRING: {
+      // Create obj
+      size_t strOffsetHeap = obj_new_string(
+          vm->allocator, vm->chunk->constants[instr.operand.str_index]);
+
+      // Push to stack
+      Value v = {0};
+      v.type = VAL_OBJ;
+      v.as.heap_offset = strOffsetHeap;
+
+      vm->stack[vm->sp++] = v;
+      break;
+    }
+
+    case OP_STORE_VAR: {
+      // pop del stack
+      Value v = vm->stack[--vm->sp];
+      int idx = instr.operand.var_index;
+
+      // store vm->vars and mark vm->var_set
+      vm->vars[idx] = v;
+      vm->var_set[idx] = 1;
+      break;
+    }
+
+    case OP_LOAD_VAR: {
+      // get from vm->var
+      int idx = instr.operand.var_index;
+      if (!vm->var_set[idx]) {
+        printf("error: variable '%s' no definida\n", vm->chunk->names[idx]);
+        return;
+      }
+      Value v = vm->vars[idx];
+      // push to stack
+      vm->stack[vm->sp++] = v;
+      break;
+    }
+
+    case OP_CALL: {
+      int nameIdx = instr.operand.call.name_index;
+      char *name = vm->chunk->names[nameIdx];
+
+      // check if is an user function
+      FunctionEntry *fn = NULL;
+      for (int i = 0; i < vm->chunk->func_count; i++) {
+        if (strcmp(vm->chunk->functions[i].name, name) == 0) {
+          fn = &vm->chunk->functions[i];
+          break;
+        }
+      }
+
+      if (fn != NULL) {
+        // Guardar estado actual en call frame
+        CallFrame *frame = &vm->frames[vm->frame_count++];
+        frame->return_ip = vm->ip + 1; // volver a la siguiente instruccion
+        frame->saved_sp = vm->sp - fn->param_count;
+        frame->is_constructor = 0;
+        memcpy(frame->saved_vars, vm->vars, sizeof(vm->vars));
+        memcpy(frame->saved_var_set, vm->var_set, sizeof(vm->var_set));
+
+        // Saltar al entrypoint
+        vm->ip = fn->entry_point;
+        continue; // no hacer ip++
+      }
+
+      // Buscar en classes[]
+      ClassEntry *cls = NULL;
+      int class_idx = -1;
+      for (int i = 0; i < vm->chunk->class_count; i++) {
+        if (strcmp(vm->chunk->classes[i].name, name) == 0) {
+          cls = &vm->chunk->classes[i];
+          class_idx = i;
+          break;
+        }
+      }
+
+      if (cls != NULL) {
+        // alloc instancia
+        size_t instOffset =
+            obj_new_instance(vm->allocator, class_idx, 8, sizeof(Value));
+        Value instVal = {0};
+        instVal.type = VAL_OBJ;
+        instVal.as.heap_offset = instOffset;
+
+        // buscar init
+        MethodEntry *init = NULL;
+        for (int m = 0; m < cls->method_count; m++) {
+          if (strcmp(cls->methods[m].method_name, "init") == 0) {
+            init = &cls->methods[m];
+            break;
+          }
+        }
+
+        if (init != NULL) {
+          // insertar instancia bajo de los args para self
+          int nArgs = instr.operand.call.arg_count;
+          for (int a = vm->sp - 1; a >= vm->sp - nArgs; a--) {
+            vm->stack[a + 1] = vm->stack[a];
+          }
+          vm->stack[vm->sp - nArgs] = instVal;
+          vm->sp++;
+
+          // Save frame
+          CallFrame *frame = &vm->frames[vm->frame_count++];
+          frame->return_ip = vm->ip + 1;
+          frame->saved_sp = vm->sp - nArgs - 1;
+          frame->is_constructor = 1;
+          frame->constructor_instance = instVal;
+          memcpy(frame->saved_vars, vm->vars, sizeof(vm->vars));
+          ;
+          memcpy(frame->saved_var_set, vm->var_set, sizeof(vm->var_set));
+
+          vm->ip = init->entry_point;
+          continue;
+        } else {
+          // No hay init, descartar args y devolver instancia
+          vm->sp -= instr.operand.call.arg_count;
+          vm->stack[vm->sp++] = instVal;
+        }
+        break;
+      }
+
+      if (strcmp(name, "print") == 0 || strcmp(name, "println") == 0) {
+
+        int nParams = instr.operand.call.arg_count;
+        for (int i = 0; i < nParams; i++) {
+          Value v = vm->stack[vm->sp - nParams + i];
+          switch (v.type) {
+          case VAL_INT:
+            printf("%d", v.as.int_val);
+            break;
+          case VAL_OBJ: {
+            // Get object from heap
+            obj_print(vm->allocator, v.as.heap_offset, "", "");
+            break;
+          }
+          default:
+            break;
+          }
+        }
+        vm->sp -= nParams;
+
+        if (strcmp(name, "println") == 0) {
+          printf("\n");
+        }
+
+      } else if (strcmp(name, "debugHeap") == 0) {
+        printf("\n");
+        JLANG_visualize(vm->allocator);
+        break;
+      } else if (strcmp(name, "len") == 0) {
+        // pop value from stack
+        Value var1 = vm->stack[--vm->sp];
+        if (var1.type == VAL_OBJ) {
+          // Resolve obj
+          Object *obj = JLANG_RESOLVE(vm->allocator, var1.as.heap_offset);
+
+          Value result = {0};
+          result.type = VAL_INT;
+          result.as.int_val = obj->data.string_val.length;
+
+          // push to stack
+          vm->stack[vm->sp++] = result;
+        }
+        break;
+      } else {
+        printf("error: function '%s' not found!\n", name);
+        return;
+      }
+
+      break;
+    }
+
+    case OP_RETURN: {
+      // Captrurar valor de retorno si hay alguno en el stack
+      Value return_val = {0};
+      int has_return = 0;
+      if (vm->sp > vm->frames[vm->frame_count - 1].saved_sp) {
+        return_val = vm->stack[--vm->sp];
+        has_return = 1;
+      }
+
+      // Restaurar call frame
+      CallFrame *frame = &vm->frames[--vm->frame_count];
+      vm->ip = frame->return_ip;
+      vm->sp = frame->saved_sp;
+      memcpy(vm->vars, frame->saved_vars, sizeof(vm->vars));
+      memcpy(vm->var_set, frame->saved_var_set, sizeof(vm->var_set));
+
+      // Push return value
+      if (frame->is_constructor) {
+        vm->stack[vm->sp++] = frame->constructor_instance;
+      } else if (has_return) {
+        vm->stack[vm->sp++] = return_val;
+      } else {
+        Value nil = {0};
+        vm->stack[vm->sp++] = nil;
+      }
+
+      continue;
+    }
+
+    case OP_ADD: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      if (var1.type != var2.type) {
+        printf("panic: var types mismatch on OP_ADD\n");
+        return;
+      }
+
+      Value result = {0};
+
+      if (var1.type == VAL_INT) {
+        result.type = VAL_INT;
+        result.as.int_val = var1.as.int_val + var2.as.int_val;
+      } else if (var1.type == VAL_OBJ) {
+        // resolve obj
+        Object *obj1 = JLANG_RESOLVE(vm->allocator, var1.as.heap_offset);
+        Object *obj2 = JLANG_RESOLVE(vm->allocator, var2.as.heap_offset);
+
+        // get chars
+        char *str1 = JLANG_RESOLVE(vm->allocator, obj1->data.string_val.chars);
+        char *str2 = JLANG_RESOLVE(vm->allocator, obj2->data.string_val.chars);
+
+        // tmp char buffer
+        size_t total =
+            obj1->data.string_val.length + obj2->data.string_val.length;
+        char *tmpBuffer = (char *)malloc(total + 1);
+        memcpy(tmpBuffer, str1, obj1->data.string_val.length);
+        memcpy(tmpBuffer + obj1->data.string_val.length, str2,
+               obj2->data.string_val.length);
+        tmpBuffer[total] = '\0';
+
+        // Create new str
+        size_t strHeapIndex = obj_new_string(vm->allocator, tmpBuffer);
+        free(tmpBuffer);
+        // set value
+        result.type = VAL_OBJ;
+        result.as.heap_offset = strHeapIndex;
+      }
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_SUB: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      Value result = {0};
+      result.type = VAL_INT;
+      result.as.int_val = var1.as.int_val - var2.as.int_val;
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_MUL: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      Value result = {0};
+      result.type = VAL_INT;
+      result.as.int_val = var1.as.int_val * var2.as.int_val;
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_DIV: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      Value result = {0};
+      result.type = VAL_INT;
+      result.as.int_val = var1.as.int_val / var2.as.int_val;
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_CMP_GT: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      Value result = {0};
+      result.type = VAL_INT;
+      result.as.int_val = var1.as.int_val > var2.as.int_val;
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_CMP_LT: {
+      // Pop from stack
+      Value var2 = vm->stack[--vm->sp];
+      Value var1 = vm->stack[--vm->sp];
+
+      Value result = {0};
+      result.type = VAL_INT;
+      result.as.int_val = var1.as.int_val < var2.as.int_val;
+
+      // Push to stack
+      vm->stack[vm->sp++] = result;
+      break;
+    }
+
+    case OP_NOP: {
+      // Pass gc
+      size_t roots[512];
+      int root_count = 0;
+
+      for (int i = 0; i < 256; i++) {
+        if (vm->var_set[i] && vm->vars[i].type == VAL_OBJ) {
+          roots[root_count++] = vm->vars[i].as.heap_offset;
+
+          // si es instancia agregar fields
+          Object *obj =
+              JLANG_RESOLVE(vm->allocator, vm->vars[i].as.heap_offset);
+          if (obj->type == OBJ_INSTANCE) {
+            Value *values = (Value *)JLANG_RESOLVE(
+                vm->allocator, obj->data.instance_val.field_values);
+            for (int f = 0; f < obj->data.instance_val.field_count; f++) {
+              if (values[f].type == VAL_OBJ) {
+                roots[root_count++] = values[f].as.heap_offset;
+              }
+            }
+          }
+        }
+      }
+
+      for (int i = 0; i < vm->sp; i++) {
+        if (vm->stack[i].type == VAL_OBJ) {
+          roots[root_count++] = vm->stack[i].as.heap_offset;
+        }
+      }
+
+      gc_collect(vm->allocator, roots, root_count);
+
+      break;
+    }
+
+    case OP_GET_FIELD: {
+      Value instance = vm->stack[--vm->sp];
+      Object *obj = JLANG_RESOLVE(vm->allocator, instance.as.heap_offset);
+      int name_idx = instr.operand.var_index;
+
+      int *names = (int *)JLANG_RESOLVE(vm->allocator,
+                                        obj->data.instance_val.field_names);
+      Value *values = (Value *)JLANG_RESOLVE(
+          vm->allocator, obj->data.instance_val.field_values);
+
+      int found = 0;
+      for (int i = 0; i < obj->data.instance_val.field_count; i++) {
+        if (names[i] == name_idx) {
+          vm->stack[vm->sp++] = values[i];
+          found = 1;
+          break;
+        }
+      }
+
+      if (!found) {
+        printf("error: field '%s' not found\n", vm->chunk->names[name_idx]);
+        return;
+      }
+      break;
+    }
+
+    case OP_SET_FIELD: {
+      Value instance = vm->stack[--vm->sp];
+      Value value = vm->stack[--vm->sp];
+      int name_idx = instr.operand.var_index;
+
+      Object *obj = JLANG_RESOLVE(vm->allocator, instance.as.heap_offset);
+      int *names = (int *)JLANG_RESOLVE(vm->allocator,
+                                        obj->data.instance_val.field_names);
+      Value *values = (Value *)JLANG_RESOLVE(
+          vm->allocator, obj->data.instance_val.field_values);
+
+      int found = 0;
+      for (int i = 0; i < obj->data.instance_val.field_count; i++) {
+        if (names[i] == name_idx) {
+          values[i] = value;
+          found = 1;
+          break;
+        }
+      }
+
+      if (!found) {
+        // Agregar campo nuevo
+        int idx = obj->data.instance_val.field_count++;
+        names[idx] = name_idx;
+        values[idx] = value;
+      }
+      break;
+    }
+
+    case OP_CALL_METHOD: {
+      int method_name_idx = instr.operand.call.name_index;
+      int arg_count = instr.operand.call.arg_count;
+      char *method_name = vm->chunk->names[method_name_idx];
+
+      // la instancia está bajo de los args
+      Value instance = vm->stack[vm->sp - arg_count - 1];
+      Object *obj = JLANG_RESOLVE(vm->allocator, instance.as.heap_offset);
+      ClassEntry *cls = &vm->chunk->classes[obj->data.instance_val.class_index];
+
+      // Buscar metodo
+      MethodEntry *method = NULL;
+      for (int i = 0; i < cls->method_count; i++) {
+        if (strcmp(cls->methods[i].method_name, method_name) == 0) {
+          method = &cls->methods[i];
+          break;
+        }
+      }
+
+      if (method == NULL) {
+        printf("error: method '%s' not found in class '%s'\n", method_name,
+               cls->name);
+        return;
+      }
+
+      // Save frame
+      CallFrame *frame = &vm->frames[vm->frame_count++];
+      frame->return_ip = vm->ip + 1;
+      frame->saved_sp = vm->sp - arg_count - 1;
+      frame->is_constructor = 0;
+      memcpy(frame->saved_vars, vm->vars, sizeof(vm->vars));
+      memcpy(frame->saved_var_set, vm->var_set, sizeof(vm->var_set));
+
+      vm->ip = method->entry_point;
+      continue;
+    }
+
+    default:
+      break;
+    }
+
+    // go to next instruction
+    vm->ip++;
+  }
+}
+
+#endif

src/backend/eval/eval.h

@@ -1,8 +1,8 @@
 #ifndef JLANG_EVAL_H
 #define JLANG_EVAL_H
 
-#include "../frontend/parser.h"
+#include "../../frontend/parser.h"
-#include "../memory/gc.h"
+#include "../../memory/gc.h"
 
 typedef struct {
   char *name;
@@ -41,16 +41,8 @@ void env_set(Environment *env, const char *name, size_t value) {
 
 int step = 0;
 
-size_t eval(ASTNode *node, Environment *env, void *allocator, int debug, int gc) {
+size_t eval(ASTNode *node, Environment *env, void *allocator, int debug,
-
+            int gc) {
-  // Run GC
-  if (gc) {
-    size_t roots[256];
-    for (int i = 0; i < env->count; i++) {
-      roots[i] = env->vars[i].value;
-    }
-    gc_collect(allocator, roots, env->count);
-  }
 
   if (debug > 0) {
     step++;
@@ -62,6 +54,8 @@ size_t eval(ASTNode *node, Environment *env, void *allocator, int debug, int gc)
   }
 
   switch (node->type) {
+  case NODE_STRING_LIT:
+    return obj_new_string(allocator, node->data.string_val);
   case NODE_INT_LIT:
     return obj_new_int(allocator, node->data.int_val);
   case NODE_VAR:
@@ -81,6 +75,23 @@ size_t eval(ASTNode *node, Environment *env, void *allocator, int debug, int gc)
 
     // Operar (ints por ahora)
     if (node->data.binop.op == '+') {
+      if (l->type == OBJ_STRING) {
+        int n = l->data.string_val.length + r->data.string_val.length;
+        char *tempBuff = (char *)malloc(n + 1);
+        // Copy left text
+        memcpy(tempBuff, JLANG_RESOLVE(allocator, l->data.string_val.chars),
+               l->data.string_val.length);
+
+        // Copy right text
+        memcpy(tempBuff + l->data.string_val.length,
+               JLANG_RESOLVE(allocator, r->data.string_val.chars),
+               r->data.string_val.length);
+        tempBuff[n] = '\0';
+
+        size_t newObj = obj_new_string(allocator, tempBuff);
+        free(tempBuff);
+        return newObj;
+      }
       return obj_new_int(allocator, l->data.int_val + r->data.int_val);
     } else if (node->data.binop.op == '-') {
       return obj_new_int(allocator, l->data.int_val - r->data.int_val);
@@ -94,13 +105,16 @@ size_t eval(ASTNode *node, Environment *env, void *allocator, int debug, int gc)
       return obj_new_int(allocator, l->data.int_val > r->data.int_val);
     }
   }
-  case NODE_PRINT: {
-    size_t val = eval(node->data.print.expr, env, allocator, debug, gc);
-    obj_print(allocator, val, "");
-    printf("\n");
-    return val;
-  }
   case NODE_BLOCK:
+    // Run GC
+    if (gc) {
+      size_t roots[256];
+      for (int i = 0; i < env->count; i++) {
+        roots[i] = env->vars[i].value;
+      }
+      gc_collect(allocator, roots, env->count);
+    }
+
     for (int i = 0; i < node->data.block.count; i++)
       eval(node->data.block.stmts[i], env, allocator, debug, gc);
     return 0;
@@ -112,6 +126,51 @@ size_t eval(ASTNode *node, Environment *env, void *allocator, int debug, int gc)
         break;
       eval(node->data.while_loop.body, env, allocator, debug, gc);
     }
+    return 0;
+  case NODE_IF: {
+    size_t cond = eval(node->data.while_loop.cond, env, allocator, debug, gc);
+    Object *obj = (Object *)JLANG_RESOLVE(allocator, cond);
+    if (obj->data.int_val > 0) {
+      eval(node->data.while_loop.body, env, allocator, debug, gc);
+    }
+    break;
+  }
+  case NODE_CALL: {
+    if (strcmp(node->data.call.name, "print") == 0) {
+      if (node->data.call.arg_count > 0) {
+        size_t val = eval(node->data.call.args[0], env, allocator, debug, gc);
+        obj_print(allocator, val, "", "");
+        return val;
+      }
+      
+      printf("");
+      return 0;
+    }
+    if (strcmp(node->data.call.name, "println") == 0) {
+      if (node->data.call.arg_count > 0) {
+        size_t val = eval(node->data.call.args[0], env, allocator, debug, gc);
+        obj_print(allocator, val, "", "");
+        printf("\n");
+        return val;
+      }
+
+      printf("\n");
+      return 0;
+    }
+    if (strcmp(node->data.call.name, "len") == 0) {
+      if (node->data.call.arg_count == 1) {
+        size_t val = eval(node->data.call.args[0], env, allocator, debug, gc);
+        Object *obj = (Object *) JLANG_RESOLVE(allocator, val);
+        
+        if (obj->type == OBJ_STRING) {
+          return obj_new_int(allocator, obj->data.string_val.length);
+        }
+      }
+      return 0;
+    }
+    printf("ERROR: funcion '%s' no definida\n", node->data.call.name);
+    exit(1);
+  }
   default:
     break;
   }

src/backend/mycpu/gencode.h

@@ -0,0 +1,504 @@
+#ifndef JLANG_MYCPU_H
+#define JLANG_MYCPU_H
+#include "../../frontend/parser.h"
+#include "opcodes.h"
+
+#define CPU_NOT_FOUND 0xFFFF
+
+typedef struct {
+  char *name;
+  short entry_point;
+  int param_count;
+  char **param_names;
+} CPUFunctionEntry;
+
+typedef struct {
+  CPUInstruction code[4096];
+  int code_count;
+  char *names[256];
+  unsigned short name_addr[256];
+  int name_count;
+  CPUFunctionEntry functions[64];
+  int func_count;
+  unsigned short ram_offset;
+  char *current_fn; // NULL si estamos en global
+  char **comments[4096];
+  int comments_per_code[4096];
+} CPUChunk;
+
+unsigned short cpu_malloc(CPUChunk *chunk, size_t size) {
+  // Save current offset
+  short current_addr = chunk->ram_offset;
+
+  // Increase offset
+  chunk->ram_offset += size;
+
+  return current_addr;
+}
+
+short cpu_emit(CPUChunk *chunk, CPUInstruction instr) {
+  chunk->code[chunk->code_count++] = instr;
+  return chunk->code_count - 1;
+}
+
+CPUInstruction cpu_make_instruction(CPUOpCode op) {
+  CPUInstruction instr = {0};
+  instr.op = op;
+  return instr;
+}
+
+unsigned short cpu_find_name(CPUChunk *chunk, char *name) {
+  for (int i = 0; i < chunk->name_count; i++) {
+    if (strcmp(chunk->names[i], name) == 0) {
+      return chunk->name_addr[i];
+    }
+  }
+
+  return CPU_NOT_FOUND;
+}
+
+unsigned short cpu_add_name(CPUChunk *chunk, char *name) {
+  for (int i = 0; i < chunk->name_count; i++) {
+    if (strcmp(chunk->names[i], name) == 0) {
+      return chunk->name_addr[i];
+    }
+  }
+
+  chunk->names[chunk->name_count++] = name;
+
+  // Asignar hueco en la ram
+  chunk->name_addr[chunk->name_count - 1] = cpu_malloc(chunk, 1);
+
+  return chunk->name_addr[chunk->name_count - 1];
+}
+
+void cpu_comment(CPUChunk *chunk, char *comment) {
+  unsigned short pc = chunk->code_count;
+  // Get comment count for pc
+  int total_comments = chunk->comments_per_code[pc];
+  chunk->comments_per_code[pc] += 1;
+
+  if (total_comments == 0) {
+    // Allocate space for comments
+    chunk->comments[pc] = malloc(sizeof(char *) * 64);
+  }
+
+  chunk->comments[pc][total_comments] = comment;
+}
+
+short cpu_compile_node(CPUChunk *chunk, ASTNode *node, int depth) {
+  if (depth > 12) {
+    printf("error: max register depth reached\n");
+    exit(1);
+  }
+
+  switch (node->type) {
+  case NODE_INT_LIT: {
+    // Mostrar comentario
+    char comment[64];
+    snprintf(comment, sizeof(comment), "NODE_INT_LIT %d", node->data.int_val);
+    cpu_comment(chunk, strdup(comment));
+
+    // Cargar inmediato en REG[depth]
+    CPUInstruction instr = cpu_make_instruction(II_ADD);
+    instr.param1 = node->data.int_val;
+    instr.param2 = 0;
+    instr.target = depth; // REG[depth]
+    cpu_emit(chunk, instr);
+    break;
+  }
+  case NODE_VAR: {
+    // Leer de ram a REG[depth]
+
+    // Obtener addr de la variable de la tabla o crearla
+    unsigned short var_addr;
+    if (chunk->current_fn) {
+      char mangled[128];
+      snprintf(mangled, sizeof(mangled), "%s.%s", chunk->current_fn,
+               node->data.string_val);
+
+      // Primero buscar variable local de la funcion
+      var_addr = cpu_find_name(chunk, strdup(mangled));
+
+      if (var_addr == CPU_NOT_FOUND) {
+        // Si no existe buscamos la variable global
+        var_addr = cpu_add_name(chunk, node->data.string_val);
+      }
+    } else {
+      var_addr = cpu_add_name(chunk, node->data.string_val);
+    }
+
+    // Add comment
+    cpu_comment(chunk, "REG13 = var_addr");
+
+    // REG13 = var_addr
+    CPUInstruction set_addr = cpu_make_instruction(II_ADD);
+    set_addr.param1 = var_addr;
+    set_addr.param2 = 0;
+    set_addr.target = 0x0D;
+    cpu_emit(chunk, set_addr);
+
+    // REG12 -> REG[DEPTH]
+    CPUInstruction mov_val = cpu_make_instruction(RI_ADD);
+    mov_val.param1 = 0x0C;
+    mov_val.param2 = 0;
+    mov_val.target = depth;
+    cpu_emit(chunk, mov_val);
+
+    break;
+  }
+  case NODE_ASSIGN: {
+
+    char comment[64];
+    snprintf(comment, sizeof(comment), "ASSIGN %s", node->data.assign.name);
+    cpu_comment(chunk, strdup(comment));
+
+    // 1. Compilar la expresion, el resultado va a REG[depth]
+    cpu_compile_node(chunk, node->data.assign.value, depth);
+
+    // 2. Obtener direccion ram de la variable
+    unsigned short addr;
+    if (chunk->current_fn) {
+      char mangled[128];
+      snprintf(mangled, sizeof(mangled), "%s.%s", chunk->current_fn,
+               node->data.assign.name);
+      addr = cpu_find_name(chunk, strdup(mangled));
+      if (addr == CPU_NOT_FOUND) {
+        addr = cpu_add_name(chunk, node->data.assign.name);
+      }
+    } else {
+      addr = cpu_add_name(chunk, node->data.assign.name);
+    }
+
+    // 3. REG13 = addr
+    CPUInstruction set_addr = cpu_make_instruction(II_ADD);
+    set_addr.param1 = addr;
+    set_addr.param2 = 0;
+    set_addr.target = 0x0D;
+    cpu_emit(chunk, set_addr);
+
+    // 4. REG12 = REG[depth]
+    CPUInstruction set_value = cpu_make_instruction(RI_ADD);
+    set_value.param1 = depth; // REG[depth]
+    set_value.param2 = 0;
+    set_value.target = 0x0C;
+    cpu_emit(chunk, set_value);
+
+    // 5. RSTR -> RAM[REG13] = REG12 -> STORE RAM
+    cpu_emit(chunk, cpu_make_instruction(RR_STR));
+    break;
+  }
+  case NODE_CALL: {
+    char comment[64];
+    snprintf(comment, sizeof(comment), "NODE_CALL %s", node->data.call.name);
+    cpu_comment(chunk, strdup(comment));
+    
+    // Pushear argumentos al stack
+    for (int i = 0; i < node->data.call.arg_count; i++) {
+      cpu_compile_node(chunk, node->data.call.args[i], depth);
+
+      // push reg[depth]
+      cpu_comment(chunk, "RI_PUSH");
+      CPUInstruction push_arg = cpu_make_instruction(RI_PUSH);
+      push_arg.param1 = depth;
+      push_arg.target =
+          0xff; // si se deja en 0, el push sobrescribe ese registro a 0
+      cpu_emit(chunk, push_arg);
+    }
+
+    // Buscar funcion en la tabla de funciones
+    unsigned short fn_addr = CPU_NOT_FOUND;
+    for (int i = 0; i < chunk->func_count; i++) {
+      if (strcmp(chunk->functions[i].name, node->data.call.name) == 0) {
+        fn_addr = chunk->functions[i].entry_point;
+      }
+    }
+
+    if (fn_addr == CPU_NOT_FOUND) {
+      printf("error: funcion '%s' no definida\n", node->data.call.name);
+      exit(1);
+    }
+
+    cpu_comment(chunk, "IR_CALL");
+    CPUInstruction call_fn = cpu_make_instruction(IR_CALL);
+    call_fn.param1 = fn_addr * 4;
+    cpu_emit(chunk, call_fn);
+
+    if (depth != 0) {
+      CPUInstruction mov = cpu_make_instruction(RI_ADD);
+      mov.param1 = 0;
+      mov.param2 = 0;
+      mov.target = depth;
+      cpu_emit(chunk, mov);
+    }
+    break;
+  }
+  case NODE_BLOCK: {
+    // Iterar statments y compilar
+    for (int i = 0; i < node->data.block.count; i++) {
+      cpu_compile_node(chunk, node->data.block.stmts[i], depth);
+    }
+    break;
+  }
+  case NODE_RETURN: {
+    if (node->data.ret.value) {
+      // Por convencion el resultado va a REG0
+      cpu_compile_node(chunk, node->data.ret.value, 0);
+    }
+
+    CPUInstruction ret = cpu_make_instruction(RR_RET);
+    ret.target = 0xff;
+    cpu_emit(chunk, ret);
+    break;
+  }
+  case NODE_BINOP: {
+    cpu_compile_node(chunk, node->data.binop.left, depth);
+    cpu_compile_node(chunk, node->data.binop.right, depth + 1);
+
+    CPUOpCode op;
+    switch (node->data.binop.op) {
+    case '+': {
+      op = RR_ADD;
+      break;
+    }
+    case '-': {
+      op = RR_SUB;
+      break;
+    }
+    case '*': {
+      op = RR_MUL;
+      break;
+    }
+    case '/': {
+      op = RR_DIV;
+      break;
+    }
+    default: {
+      op = RR_ADD;
+      break;
+    }
+    }
+
+    CPUInstruction binop = cpu_make_instruction(op);
+    binop.param1 = depth;
+    binop.param2 = depth + 1;
+    binop.target = depth;
+    cpu_emit(chunk, binop);
+    break;
+  }
+
+  case NODE_FN_DEF: {
+    char comment[64];
+    snprintf(comment, sizeof(comment), "NODE_FN_DEF %s", node->data.fn_def.name);
+    cpu_comment(chunk, strdup(comment));
+
+    // emitir jmp para ignorar la funcion por defecto
+    CPUInstruction jump = cpu_make_instruction(II_ADD);
+    jump.param1 = 0x00;
+    jump.param2 = 0x00;
+    jump.target = 0x0E;
+    short jump_idx = cpu_emit(chunk, jump);
+
+    // Registrar entrypoint
+    short entry_point = chunk->code_count;
+    CPUFunctionEntry *fn = &chunk->functions[chunk->func_count++];
+    fn->name = node->data.fn_def.name;
+    fn->entry_point = entry_point;
+    fn->param_count = node->data.fn_def.param_count;
+    fn->param_names = node->data.fn_def.params;
+
+    CPUInstruction pop_ret = cpu_make_instruction(RR_POP);
+    pop_ret.target = 0x0B; // REG11 como temp
+    cpu_emit(chunk, pop_ret);
+
+    // Pop de los params del stack a RAM
+    for (int i = node->data.fn_def.param_count - 1; i >= 0; i--) {
+      // POP -> reg0
+      CPUInstruction pop = cpu_make_instruction(RR_POP);
+      pop.target = 0x00;
+      cpu_emit(chunk, pop);
+
+      // guardar en ram como variable local
+      char *param_name = node->data.fn_def.params[i];
+      char mangled[128];
+      snprintf(mangled, sizeof(mangled), "%s.%s", node->data.fn_def.name,
+               param_name);
+      short addr = cpu_add_name(chunk, strdup(mangled));
+      CPUInstruction set_addr = cpu_make_instruction(II_ADD);
+      set_addr.param1 = addr;
+      set_addr.target = 0x0D;
+      cpu_emit(chunk, set_addr);
+
+      CPUInstruction set_val = cpu_make_instruction(RI_ADD);
+      set_val.param1 = 0x00;
+      set_val.target = 0x0C;
+      cpu_emit(chunk, set_val);
+
+      cpu_emit(chunk, cpu_make_instruction(RR_STR));
+    }
+
+    CPUInstruction push_ret = cpu_make_instruction(RR_PUSH);
+    push_ret.param1 = 0x0B;
+    cpu_emit(chunk, push_ret);
+
+    chunk->current_fn = node->data.fn_def.name;
+    // Compilar body
+    cpu_compile_node(chunk, node->data.fn_def.body, 0);
+
+    chunk->current_fn = NULL;
+
+    // RET implicito
+    cpu_emit(chunk, cpu_make_instruction(RR_RET));
+
+    // Backpatch del jump
+    chunk->code[jump_idx].param1 = chunk->code_count * 4;
+    break;
+  }
+
+  case NODE_IF: {
+    cpu_comment(chunk, "NODE_IF");
+    ASTNode *cond = node->data.if_statement.cond;
+
+    // Compilar miembros de la expresion y mover a depth y depth+1
+    cpu_compile_node(chunk, cond->data.binop.left, depth);
+    cpu_compile_node(chunk, cond->data.binop.right, depth + 1);
+
+    // Opcode inverso (saltar si la conficion es falsa)
+    CPUOpCode skip_op;
+    switch (cond->data.binop.op) {
+    case '>': {
+      skip_op = RR_LSE;
+      break;
+    }
+    case '<': {
+      skip_op = RR_GRE;
+      break;
+    }
+    case 'e': {
+      skip_op = RR_NEQ;
+      break;
+    }
+    case 'n': {
+      skip_op = RR_EQ;
+      break;
+    }
+
+    default: {
+      printf("error: binop op not supported '%c'\n", cond->data.binop.op);
+      exit(1);
+    }
+    }
+
+    // Emitir salto con placeholder
+    CPUInstruction jmp = cpu_make_instruction(skip_op);
+    jmp.param1 = depth;
+    jmp.param2 = depth + 1;
+    jmp.target = 0;
+    short jmp_idx = cpu_emit(chunk, jmp);
+
+    // Compilar body
+    cpu_compile_node(chunk, node->data.if_statement.body, depth);
+
+    // Backpatch
+    chunk->code[jmp_idx].target = chunk->code_count * 4;
+    break;
+  }
+  case NODE_WHILE: {
+
+    cpu_comment(chunk, "NODE_WHILE");
+    // Guardar entrada al bucle
+    short entry = chunk->code_count * 4;
+
+    // chequear condicional, si no se cumple, salir del bucle
+    ASTNode *cond = node->data.while_loop.cond;
+
+    // Compilar miembros de la expresion y mover a depth y depth+1
+    cpu_compile_node(chunk, cond->data.binop.left, depth);
+    cpu_compile_node(chunk, cond->data.binop.right, depth + 1);
+
+    // Opcode inverso (saltar si la conficion es falsa)
+    CPUOpCode skip_op;
+    switch (cond->data.binop.op) {
+    case '>': {
+      skip_op = RR_LSE;
+      break;
+    }
+    case '<': {
+      skip_op = RR_GRE;
+      break;
+    }
+    case 'e': {
+      skip_op = RR_NEQ;
+      break;
+    }
+    case 'n': {
+      skip_op = RR_EQ;
+      break;
+    }
+
+    default: {
+      printf("error: binop op not supported '%c'\n", cond->data.binop.op);
+      exit(1);
+    }
+    }
+
+    // Emitir salto con placeholder
+    CPUInstruction jmp = cpu_make_instruction(skip_op);
+    jmp.param1 = depth;
+    jmp.param2 = depth + 1;
+    jmp.target = 0;
+    short jmp_idx = cpu_emit(chunk, jmp);
+
+    // Compilar body
+    cpu_compile_node(chunk, node->data.while_loop.body, depth);
+
+    // Saltar al principio
+    CPUInstruction entry_jmp = cpu_make_instruction(II_ADD);
+    entry_jmp.param1 = entry;
+    entry_jmp.target = 0x0E;
+    cpu_emit(chunk, entry_jmp);
+
+    // Backpatch
+    chunk->code[jmp_idx].target = chunk->code_count * 4;
+
+    break;
+  }
+  default:
+    break;
+  }
+
+  return -1;
+}
+
+CPUChunk *compileAST(ASTNode *root) {
+  CPUChunk *chunk = (CPUChunk *)malloc(sizeof(CPUChunk));
+  memset(chunk, 0, sizeof(CPUChunk));
+
+  cpu_compile_node(chunk, root, 0);
+
+  // Añadir HALT
+  cpu_comment(chunk, "HALT");
+  CPUInstruction halt = cpu_make_instruction(RR_HALT);
+  cpu_emit(chunk, halt);
+
+  return chunk;
+}
+
+void cpu_print_code(CPUChunk *chunk, int show_comments) {
+  printf("========= BINARY =========\n");
+  for (int i = 0; i < chunk->code_count; i++) {
+
+    if (show_comments)  {
+        // Mostrar comentarios
+        for (int c = 0; c < chunk->comments_per_code[i]; c++) {
+            printf("# %s\n", chunk->comments[i][c]);
+        }
+    }
+
+    CPUInstruction instr = chunk->code[i];
+    printf("0x%04X 0x%04X 0x%04X 0x%04X\n", instr.op, instr.param1,
+           instr.param2, instr.target);
+  }
+}
+
+#endif

src/backend/mycpu/opcodes.h

@@ -0,0 +1,129 @@
+#ifndef JLANG_MYCPU_OOCODES_H
+#define JLANG_MYCPU_OOCODES_H
+
+typedef enum {
+    // Register, Register
+    RR_ADD,
+    RR_SUB,
+    RR_MUL,
+    RR_DIV,
+    RR_AND,
+    RR_OR,
+    RR_NOT,
+    RR_NAND,
+    RR_NOR,
+    RR_XOR,
+    RR_XNOR,
+    RR_NEG,
+
+    RR_EQ = 0x10,
+    RR_NEQ,
+    RR_LS,
+    RR_LSE,
+    RR_GR,
+    RR_GRE,
+
+    RR_STR = 0x18,
+    RR_PUSH,
+    RR_POP,
+
+    RR_CALL = 0x20,
+    RR_RET,
+    RR_HALT,
+
+    // Inmediate, Register
+    IR_ADD = 0x40,
+    IR_SUB,
+    IR_MUL,
+    IR_DIV,
+    IR_AND,
+    IR_OR,
+    IR_NOT,
+    IR_NAND,
+    IR_NOR,
+    IR_XOR,
+    IR_XNOR,
+    IR_NEG,
+
+    IR_EQ = 0x50,
+    IR_NEQ,
+    IR_LS,
+    IR_LSE,
+    IR_GR,
+    IR_GRE,
+
+    IR_STR = 0x58,
+    IR_PUSH,
+    IR_POP,
+
+    IR_CALL = 0x60,
+    IR_RET,
+    IR_HALT,
+
+    // Register, Inmediate
+    RI_ADD = 0x80,
+    RI_SUB,
+    RI_MUL,
+    RI_DIV,
+    RI_AND,
+    RI_OR,
+    RI_NOT,
+    RI_NAND,
+    RI_NOR,
+    RI_XOR,
+    RI_XNOR,
+    RI_NEG,
+
+    RI_EQ = 0x90,
+    RI_NEQ,
+    RI_LS,
+    RI_LSE,
+    RI_GR,
+    RI_GRE,
+
+    RI_STR = 0x98,
+    RI_PUSH,
+    RI_POP,
+
+    RI_CALL = 0xA0,
+    RI_RET,
+    RI_HALT,
+
+    // Inmediate, Inmediate
+    II_ADD = 0xC0,
+    II_SUB,
+    II_MUL,
+    II_DIV,
+    II_AND,
+    II_OR,
+    II_NOT,
+    II_NAND,
+    II_NOR,
+    II_XOR,
+    II_XNOR,
+    II_NEG,
+
+    II_EQ = 0xD0,
+    II_NEQ,
+    II_LS,
+    II_LSE,
+    II_GR,
+    II_GRE,
+
+    II_STR = 0xD8,
+    II_PUSH,
+    II_POP,
+
+    II_CALL = 0xE0,
+    II_RET,
+    II_HALT,
+} CPUOpCode;
+
+typedef struct {
+    CPUOpCode op;
+    unsigned short param1;
+    unsigned short param2;
+    unsigned short target;
+} CPUInstruction;
+
+#endif

src/frontend/lexer.h

@@ -14,10 +14,12 @@ typedef enum {
   TOK_STRING, // "hola"
 
   // Identificadores y keywords
-  TOK_ID,    // x, foo, mi_var
+  TOK_ID,     // x, foo, mi_var
-  TOK_PRINT, // print
+  TOK_IF,     // if
-  TOK_IF,    // if
+  TOK_WHILE,  // while
-  TOK_WHILE, // while
+  TOK_FN,     // fn
+  TOK_RETURN, // return
+  TOK_CLASS,  // class
 
   // Operadores
   TOK_ASSIGN, // =
@@ -34,6 +36,8 @@ typedef enum {
   TOK_LPAREN,  // (
   TOK_RPAREN,  // )
   TOK_COLON,   // :
+  TOK_COMMA,   // ,
+  TOK_DOT,     // .
   TOK_NEWLINE, // \n (significativo, como en Python)
   TOK_INDENT,  // aumento de indentacion
   TOK_DEDENT,  // reduccion de indentacion
@@ -87,12 +91,17 @@ Token *tokenize(const char *source, int *token_count) {
         pos++;
       }
       int new_level = spaces / 4; // 4 espacios = 1 nivel
-      if (new_level > indent_level) {
+      if (source[pos] != '\n' && source[pos] != '\0') {
-        tokens[count++] = make_token(TOK_INDENT, "INDENT");
+        if (new_level > indent_level) {
-      } else if (new_level < indent_level) {
+          for (int l = indent_level; l < new_level; l++)
-        tokens[count++] = make_token(TOK_DEDENT, "DEDENT");
+            tokens[count++] = make_token(TOK_INDENT, "INDENT");
+        } else if (new_level < indent_level) {
+          for (int l = indent_level; l > new_level; l--)
+            tokens[count++] = make_token(TOK_DEDENT, "DEDENT");
+        }
+        indent_level = new_level;
       }
-      indent_level = new_level;
+
     } else if (c == '+') {
       tokens[count++] = make_token(TOK_PLUS, "+");
       pos++;
@@ -106,8 +115,21 @@ Token *tokenize(const char *source, int *token_count) {
       tokens[count++] = make_token(TOK_SLASH, "/");
       pos++;
     } else if (c == '=') {
-      tokens[count++] = make_token(TOK_ASSIGN, "=");
+      if (source[pos + 1] == '=') {
-      pos++;
+        tokens[count++] = make_token(TOK_EQ, "==");
+        pos += 2;
+      } else {
+        tokens[count++] = make_token(TOK_ASSIGN, "=");
+        pos++;
+      }
+    } else if (c == '!') {
+      if (source[pos + 1] == '=') {
+        tokens[count++] = make_token(TOK_NEQ, "!=");
+        pos += 2;
+      } else {
+        printf("WARN: caracter no reconocido '!' en pos %d\n", pos);
+        exit(1);
+      }
     } else if (c == '<') {
       tokens[count++] = make_token(TOK_LT, "<");
       pos++;
@@ -117,24 +139,51 @@ Token *tokenize(const char *source, int *token_count) {
     } else if (c == ':') {
       tokens[count++] = make_token(TOK_COLON, ":");
       pos++;
+    } else if (c == ',') {
+      tokens[count++] = make_token(TOK_COMMA, ",");
+      pos++;
+    } else if (c == '.') {
+      tokens[count++] = make_token(TOK_DOT, ".");
+      pos++;
+    } else if (c == '(') {
+      tokens[count++] = make_token(TOK_LPAREN, "(");
+      pos++;
+    } else if (c == ')') {
+      tokens[count++] = make_token(TOK_RPAREN, ")");
+      pos++;
+    } else if (c == '"') {
+      // Leer todo hasta el proximo '"'
+      pos++; // consumir '"'
+      int start = pos;
+      while (source[pos] != '"')
+        pos++;
+      tokens[count++] = make_token(TOK_STRING, substr(source, start, pos));
+      pos++; // consumir '"'
     } else if (c >= '0' && c <= '9') {
       // Leer todos los digitos consecutivos
       int start = pos;
       while (source[pos] >= '0' && source[pos] <= '9')
         pos++;
       tokens[count++] = make_token(TOK_INT, substr(source, start, pos));
-    } else if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) {
+    } else if ((c >= 'a' && c <= 'z') || (c == '_') || (c >= 'A' && c <= 'Z')) {
       // Leer todos los caracteres consecutivos
       int start = pos;
-      while (isalnum(source[pos]))
+      while (isalnum(source[pos]) || source[pos] == '_')
         pos++;
       char *word = substr(source, start, pos);
+      // tokens[count++] = make_token(TOK_ID, word);
 
       // Comprobar si es una keyword reservada
-      if (strcmp(word, "print") == 0) {
+      if (strcmp(word, "if") == 0) {
-        tokens[count++] = make_token(TOK_PRINT, word);
+        tokens[count++] = make_token(TOK_IF, word);
       } else if (strcmp(word, "while") == 0) {
         tokens[count++] = make_token(TOK_WHILE, word);
+      } else if (strcmp(word, "fn") == 0) {
+        tokens[count++] = make_token(TOK_FN, word);
+      } else if (strcmp(word, "return") == 0) {
+        tokens[count++] = make_token(TOK_RETURN, word);
+      } else if (strcmp(word, "class") == 0) {
+        tokens[count++] = make_token(TOK_CLASS, word);
       } else {
         tokens[count++] = make_token(TOK_ID, word);
       }

src/frontend/parser.h

@@ -12,10 +12,17 @@ typedef enum {
   NODE_VAR,        // referencia a variable
   NODE_ASSIGN,     // asignacion: x = expr
   NODE_BINOP,      // operacion binaria: a + b
-  NODE_PRINT,      // print(expr)
+  NODE_NOP,        // noop
   NODE_IF,         // if cond: bloque
   NODE_WHILE,      // while cond: bloque
   NODE_BLOCK,      // secuencia de statements
+  NODE_CALL,
+  NODE_FN_DEF,    // definicion de funcion
+  NODE_RETURN,    // return expr
+  NODE_CLASS_DEF, // definicion de clase
+  NODE_DOT_ACCESS,
+  NODE_DOT_ASSIGN,
+  NODE_METHOD_CALL,
 } NodeType;
 
 typedef struct ASTNode {
@@ -43,6 +50,44 @@ typedef struct ASTNode {
       struct ASTNode *cond;
       struct ASTNode *body;
     } while_loop; // NODE_WHILE
+    struct {
+      struct ASTNode *cond;
+      struct ASTNode *body;
+    } if_statement; // NODE_IF
+    struct {
+      char *name;
+      struct ASTNode **args;
+      int arg_count;
+    } call;
+    struct {
+      char *name;
+      char **params;
+      int param_count;
+      struct ASTNode *body;
+    } fn_def; // NODE_FN_DEF
+    struct {
+      struct ASTNode *value; // expresion de retorno
+    } ret;                   // NODE_RETURN
+    struct {
+      char *name;
+      struct ASTNode **methods;
+      int method_count;
+    } class_def;
+    struct {
+      struct ASTNode *object;
+      char *field;
+    } dot_access;
+    struct {
+      struct ASTNode *object;
+      char *field;
+      struct ASTNode *value;
+    } dot_assign;
+    struct {
+      struct ASTNode *object;
+      char *method;
+      struct ASTNode **args;
+      int arg_count;
+    } method_call;
   } data;
 } ASTNode;
 
@@ -54,19 +99,106 @@ ASTNode *make_node(NodeType type) {
 
 int pos = 0;
 
-ASTNode *parse_term(Token *tokens) {
+ASTNode *parse_expr(Token *tokens);
+ASTNode *parse_term(Token *tokens);
+ASTNode *parse_factor(Token *tokens);
+
+ASTNode *parse_factor(Token *tokens) {
   if (tokens[pos].type == TOK_INT) {
     ASTNode *node = make_node(NODE_INT_LIT);
     node->data.int_val = atoi(tokens[pos].value);
     pos++;
     return node;
-  } else if (tokens[pos].type == TOK_ID) {
+  } else if (tokens[pos].type == TOK_STRING) {
-    ASTNode *node = make_node(NODE_VAR);
+    ASTNode *node = make_node(NODE_STRING_LIT);
     node->data.string_val = tokens[pos].value;
     pos++;
     return node;
+  } else if (tokens[pos].type == TOK_ID) {
+    ASTNode *node;
+    if (tokens[pos + 1].type == TOK_LPAREN) {
+      // Function call
+      char *name = tokens[pos].value;
+      pos++; // consumir ID
+      pos++; // consumir (
+
+      // Parsear argumentos
+      ASTNode **args =
+          (ASTNode **)malloc(sizeof(ASTNode *) * 16); // Max 16 parametros
+      int arg_count = 0;
+
+      if (tokens[pos].type != TOK_RPAREN) {
+        args[arg_count++] = parse_expr(tokens);
+        while (tokens[pos].type == TOK_COMMA) {
+          pos++; // Consumir ","
+          args[arg_count++] = parse_expr(tokens);
+        }
+      }
+      pos++; // consumir ")"
+      node = make_node(NODE_CALL);
+      node->data.call.name = name;
+      node->data.call.args = args;
+      node->data.call.arg_count = arg_count;
+    } else {
+      node = make_node(NODE_VAR);
+      node->data.string_val = tokens[pos].value;
+      pos++;
+    }
+
+    // Loop post-fix para dot access
+    while (tokens[pos].type == TOK_DOT) {
+      pos++; // consumir .
+      char *field = tokens[pos].value;
+      pos++; // consumir field name
+
+      if (tokens[pos].type == TOK_LPAREN) {
+        // obj.method(args...)
+        pos++; // consumir '('
+        ASTNode **args = (ASTNode **)malloc(sizeof(ASTNode *) * 16);
+        int arg_count = 0;
+        if (tokens[pos].type != TOK_RPAREN) {
+          args[arg_count++] = parse_expr(tokens);
+          while (tokens[pos].type == TOK_COMMA) {
+            pos++;
+            args[arg_count++] = parse_expr(tokens);
+          }
+        }
+        pos++; // consumir ')'
+        ASTNode *mc = make_node(NODE_METHOD_CALL);
+        mc->data.method_call.object = node;
+        mc->data.method_call.method = field;
+        mc->data.method_call.args = args;
+        mc->data.method_call.arg_count = arg_count;
+        node = mc;
+      } else {
+        // obj.field
+        ASTNode *da = make_node(NODE_DOT_ACCESS);
+        da->data.dot_access.object = node;
+        da->data.dot_access.field = field;
+        node = da;
+      }
+    }
+
+    return node;
+
+  } else if (tokens[pos].type == TOK_LPAREN) {
+    pos++; // consumir (
+    ASTNode *expr = parse_expr(tokens);
+    pos++; // consumir )
+    return expr;
+  } else if (tokens[pos].type == TOK_MINUS) {
+    pos++; // consumir '-'
+    ASTNode *term = parse_term(tokens);
+    ASTNode *neg = make_node(NODE_BINOP);
+    neg->data.binop.op = '-';
+    neg->data.binop.left = make_node(NODE_INT_LIT);
+    neg->data.binop.left->data.int_val = 0;
+    neg->data.binop.right = term;
+    return neg;
   }
-  printf("ERROR: esperaba INT o ID, encontré tipo %d\n", tokens[pos].type);
+
+  printf("ERROR: esperaba INT o ID, encontre tipo %d value: %s\n",
+         tokens[pos].type, tokens[pos].value);
   exit(1);
 }
 
@@ -74,9 +206,16 @@ ASTNode *parse_expr(Token *tokens) {
   ASTNode *left = parse_term(tokens);
 
   while (tokens[pos].type == TOK_PLUS || tokens[pos].type == TOK_MINUS ||
-         tokens[pos].type == TOK_STAR || tokens[pos].type == TOK_SLASH ||
+         tokens[pos].type == TOK_LT || tokens[pos].type == TOK_GT ||
-         tokens[pos].type == TOK_LT || tokens[pos].type == TOK_GT) {
+         tokens[pos].type == TOK_EQ || tokens[pos].type == TOK_NEQ) {
-    char op = tokens[pos].value[0]; // +,-,*,/
+    char op;
+    if (tokens[pos].type == TOK_EQ) {
+      op = 'e';
+    } else if (tokens[pos].type == TOK_NEQ) {
+      op = 'n';
+    } else {
+      op = tokens[pos].value[0]; // +,-,*,/
+    }
     pos++;
     ASTNode *right = parse_term(tokens);
 
@@ -89,8 +228,69 @@ ASTNode *parse_expr(Token *tokens) {
   return left;
 }
 
+ASTNode *parse_term(Token *tokens) {
+  ASTNode *left = parse_factor(tokens);
+
+  while (tokens[pos].type == TOK_STAR || tokens[pos].type == TOK_SLASH) {
+    char op = tokens[pos].value[0];
+    pos++;
+    ASTNode *right = parse_factor(tokens);
+
+    ASTNode *binop = make_node(NODE_BINOP);
+    binop->data.binop.op = op;
+    binop->data.binop.left = left;
+    binop->data.binop.right = right;
+    left = binop;
+  }
+  return left;
+}
+
 ASTNode *parse_statement(Token *tokens) {
   if (tokens[pos].type == TOK_ID) {
+    if (tokens[pos + 1].type == TOK_DOT) {
+      ASTNode *expr = parse_expr(tokens);
+
+      if (tokens[pos].type == TOK_ASSIGN) {
+        // dot_assign: self.name = expr
+        pos++; // consumir '='
+        ASTNode *value = parse_expr(tokens);
+        ASTNode *node = make_node(NODE_DOT_ASSIGN);
+        node->data.dot_assign.object = expr->data.dot_access.object;
+        node->data.dot_assign.field = expr->data.dot_access.field;
+        node->data.dot_assign.value = value;
+        return node;
+      }
+
+      // si no hay '=', es un method call como (d.speak())
+      return expr;
+    }
+
+    if (tokens[pos + 1].type == TOK_LPAREN) {
+      // Es una funcion
+      char *name = tokens[pos].value;
+      pos++; // consumir ID
+      pos++; // consumir "("
+
+      // Parsear argumentos
+      ASTNode **args =
+          (ASTNode **)malloc(sizeof(ASTNode *) * 16); // Max 16 parametros
+      int arg_count = 0;
+
+      if (tokens[pos].type != TOK_RPAREN) {
+        args[arg_count++] = parse_expr(tokens);
+        while (tokens[pos].type == TOK_COMMA) {
+          pos++; // Consumir ","
+          args[arg_count++] = parse_expr(tokens);
+        }
+      }
+      pos++; // consumir ")"
+      ASTNode *node = make_node(NODE_CALL);
+      node->data.call.name = name;
+      node->data.call.args = args;
+      node->data.call.arg_count = arg_count;
+      return node;
+    }
+
     char *name = tokens[pos].value;
     pos++; // consumir ID
     pos++; // consumir "="
@@ -101,13 +301,21 @@ ASTNode *parse_statement(Token *tokens) {
     node->data.assign.value = value;
     return node;
   }
-  if (tokens[pos].type == TOK_PRINT) {
-    pos++; // consumir "print"
-    ASTNode *expr = parse_expr(tokens);
 
-    ASTNode *node = make_node(NODE_PRINT);
+  // Parse comments
-    node->data.print.expr = expr;
+  if (tokens[pos].type == TOK_SLASH) {
-    return node;
+    if (tokens[pos + 1].type == TOK_SLASH) {
+      pos++; // consumir /
+      pos++; // consumir /
+
+      // Consumir hasta NewLine
+      while (tokens[pos].type != TOK_NEWLINE)
+        pos++;
+
+      pos++; // consumir newline
+
+      return make_node(NODE_NOP);
+    }
   }
 
   if (tokens[pos].type == TOK_WHILE) {
@@ -124,7 +332,7 @@ ASTNode *parse_statement(Token *tokens) {
     while (tokens[pos].type != TOK_DEDENT) {
       body->data.block.stmts[body->data.block.count++] =
           parse_statement(tokens);
-      if (tokens[pos].type == TOK_NEWLINE) {
+      while (tokens[pos].type == TOK_NEWLINE) {
         pos++;
       }
     }
@@ -135,6 +343,109 @@ ASTNode *parse_statement(Token *tokens) {
     node->data.while_loop.body = body;
     return node;
   }
+
+  if (tokens[pos].type == TOK_IF) {
+    pos++; // consumir if
+    ASTNode *cond = parse_expr(tokens);
+    pos++; // consumir :
+    pos++; // consumir NEWLINE
+    pos++; // consumir INDENT
+
+    // Parsear bloque de statements hasta DEDENT
+    ASTNode *body = make_node(NODE_BLOCK);
+    body->data.block.stmts = (ASTNode **)malloc(sizeof(ASTNode *) * 256);
+    body->data.block.count = 0;
+    while (tokens[pos].type != TOK_DEDENT) {
+      body->data.block.stmts[body->data.block.count++] =
+          parse_statement(tokens);
+      while (tokens[pos].type == TOK_NEWLINE) {
+        pos++;
+      }
+    }
+    pos++; // Consumir DEDENT
+
+    ASTNode *node = make_node(NODE_IF);
+    node->data.while_loop.cond = cond;
+    node->data.while_loop.body = body;
+    return node;
+  }
+
+  if (tokens[pos].type == TOK_CLASS) {
+    pos++; // consumir 'class'
+    char *name = tokens[pos].value;
+    pos++; // consumir nombre
+    pos++; // consumir :
+    pos++; // consumir NEWLINE
+    pos++; // consumir INDENT
+
+    ASTNode **methods = (ASTNode **)malloc(sizeof(ASTNode *) * 16);
+    int method_count = 0;
+    while (tokens[pos].type != TOK_DEDENT) {
+      methods[method_count++] = parse_statement(tokens);
+      while (tokens[pos].type == TOK_NEWLINE) {
+        pos++;
+      }
+    }
+    pos++; // consumir DEDENT
+
+    ASTNode *node = make_node(NODE_CLASS_DEF);
+    node->data.class_def.name = name;
+    node->data.class_def.methods = methods;
+    node->data.class_def.method_count = method_count;
+    return node;
+  }
+
+  if (tokens[pos].type == TOK_FN) {
+    pos++; // consumir "fn"
+    char *name = tokens[pos].value;
+    pos++; // consumir name
+    pos++; // consumir "("
+
+    // Parsear parametros (max 16)
+    char **params = malloc(sizeof(char *) * 16);
+    int param_count = 0;
+    if (tokens[pos].type != TOK_RPAREN) {
+      params[param_count++] = tokens[pos].value;
+      pos++;
+      while (tokens[pos].type == TOK_COMMA) {
+        pos++; // consumir ","
+        params[param_count++] = tokens[pos].value;
+        pos++;
+      }
+    }
+    pos++; // consumir ")"
+    pos++; // consumir ":"
+    pos++; // consumir NEWLINE
+    pos++; // consumir INDENT
+
+    // Parsear bloque de statements hasta DEDENT
+    ASTNode *body = make_node(NODE_BLOCK);
+    body->data.block.stmts = (ASTNode **)malloc(sizeof(ASTNode *) * 256);
+    body->data.block.count = 0;
+    while (tokens[pos].type != TOK_DEDENT) {
+      body->data.block.stmts[body->data.block.count++] =
+          parse_statement(tokens);
+      while (tokens[pos].type == TOK_NEWLINE) {
+        pos++;
+      }
+    }
+    pos++; // Consumir DEDENT
+
+    ASTNode *node = make_node(NODE_FN_DEF);
+    node->data.fn_def.name = name;
+    node->data.fn_def.params = params;
+    node->data.fn_def.param_count = param_count;
+    node->data.fn_def.body = body;
+    return node;
+  }
+
+  if (tokens[pos].type == TOK_RETURN) {
+    pos++;
+    ASTNode *node = make_node(NODE_RETURN);
+    node->data.ret.value = parse_expr(tokens);
+    return node;
+  }
+
   printf("ERROR: statement inesperado\n");
   exit(1);
 }
@@ -145,7 +456,7 @@ ASTNode *parse(Token *tokens, int token_count) {
   block->data.block.count = 0;
 
   while (pos < token_count) {
-    if (tokens[pos].type == TOK_NEWLINE) {
+    while (tokens[pos].type == TOK_NEWLINE) {
       pos++; // Saltar newlines sueltos
       continue;
     }
@@ -202,11 +513,6 @@ void ast_print(ASTNode *node, const char *prefix, int is_last) {
     ast_print(node->data.binop.right, new_prefix, 1);
     break;
 
-  case NODE_PRINT:
-    printf("NODE_PRINT\n");
-    ast_print(node->data.print.expr, new_prefix, 1);
-    break;
-
   case NODE_BLOCK:
     printf("NODE_BLOCK\n");
     for (int i = 0; i < node->data.block.count; i++) {
@@ -215,6 +521,69 @@ void ast_print(ASTNode *node, const char *prefix, int is_last) {
     }
     break;
 
+  case NODE_IF:
+    printf("NODE_IF\n");
+    ast_print(node->data.while_loop.cond, new_prefix, 0);
+    ast_print(node->data.while_loop.body, new_prefix, 1);
+    break;
+  case NODE_NOP:
+    printf("NODE_NOOP\n");
+    break;
+
+  case NODE_CALL:
+    printf("NODE_CALL(\"%s\")\n", node->data.call.name);
+    for (int i = 0; i < node->data.call.arg_count; i++) {
+      ast_print(node->data.call.args[i], new_prefix,
+                i == node->data.call.arg_count - 1);
+    }
+    break;
+
+  case NODE_FN_DEF:
+    printf("NODE_FN_DEF(\"%s\"", node->data.fn_def.name);
+    for (int i = 0; i < node->data.fn_def.param_count; i++) {
+      printf(", %s", node->data.fn_def.params[i]);
+    }
+    printf(")\n");
+    ast_print(node->data.fn_def.body, new_prefix, 1);
+    break;
+
+  case NODE_RETURN:
+    printf("NODE_RETURN\n");
+    if (node->data.ret.value) {
+      ast_print(node->data.ret.value, new_prefix, 1);
+    }
+    break;
+
+  case NODE_CLASS_DEF:
+    printf("NODE_CLASS_DEF(\"%s\")\n", node->data.class_def.name);
+    for (int i = 0; i < node->data.class_def.method_count; i++) {
+      ast_print(node->data.class_def.methods[i], new_prefix,
+                i == node->data.class_def.method_count - 1);
+    }
+    break;
+
+  case NODE_DOT_ACCESS:
+    printf("NODE_DOT_ACCESS(.%s)\n", node->data.dot_access.field);
+    ast_print(node->data.dot_access.object, new_prefix, 1);
+    break;
+
+  case NODE_DOT_ASSIGN:
+    printf("NODE_DOT_ASSIGN(.%s)\n", node->data.dot_assign.field);
+    ast_print(node->data.dot_assign.object, new_prefix, 0);
+    ast_print(node->data.dot_assign.value, new_prefix, 1);
+    break;
+
+  case NODE_METHOD_CALL:
+    printf("NODE_METHOD_CALL(.%s, %d args)\n", node->data.method_call.method,
+           node->data.method_call.arg_count);
+    ast_print(node->data.method_call.object, new_prefix,
+              node->data.method_call.arg_count == 0);
+    for (int i = 0; i < node->data.method_call.arg_count; i++) {
+      ast_print(node->data.method_call.args[i], new_prefix,
+                i == node->data.method_call.arg_count - 1);
+    }
+    break;
+
   default:
     printf("UNKNOWN\n");
     break;

src/main.c

@@ -1,15 +1,18 @@
-#include "vm/eval.h"
+#include "backend/eval/eval.h"
+#include "backend/bytecode/compiler.h"
+#include "backend/bytecode/vm.h"
+#include "backend/mycpu/gencode.h"
 
 int main(int argc, char **argv) {
-  if (argc != 2) {
+  if (argc != 3) {
-    printf("usage: %s <path to .j file>\n", argv[0]);
+    printf("usage: %s eval|vm|asm|mycpu <path to .j file>\n", argv[0]);
     exit(1);
   }
   // Creamos un allocator
   JLANG_memory_allocator *allocPtr = JLANG_CreateAllocator();
 
   // Read file from argv
-  FILE *fptr = fopen(argv[1], "r");
+  FILE *fptr = fopen(argv[2], "r");
   if (fptr == NULL) {
     printf("error leyendo: %s\n", argv[1]);
     exit(1);
@@ -24,20 +27,40 @@ int main(int argc, char **argv) {
 
   fclose(fptr);
 
+  printf("=== CODE ===\n");
   printf("%s\n", buff);
 
   // Lexer test
   int totalTokens = 0;
   Token *tokens = tokenize(buff, &totalTokens);
-  printf("totalTokens=%d\n", totalTokens);
+  printf("=== INFO ===\n");
   ASTNode *block = parse(tokens, totalTokens);
   ast_debug(block);
-  
+ 
-  Environment env = {0};
+  if (strcmp(argv[1], "eval") == 0) {
-  eval(block, &env, allocPtr, 0, 1);
+    Environment env = {0};
+    eval(block, &env, allocPtr, 0, 1);
+
+    // printf("\nheapSize=%zu\n", allocPtr->size);
+    // JLANG_visualize(allocPtr);
+  } else if (strcmp(argv[1], "vm") == 0){
+    Chunk* chunk = compile(block);
+    VM vm = {0};
+    vm.chunk = chunk;
+    vm.allocator = allocPtr;
+    print_chunk(chunk);
+    run_vm(&vm);
+
+    // printf("\n");
+    // JLANG_visualize(allocPtr);
+  } else if (strcmp(argv[1], "mycpu") == 0){
+    CPUChunk* chunk = compileAST(block);
+    cpu_print_code(chunk, 1);
+    
+  } else {
+    printf("panic: WIP\n");
+  }
 
-  printf("heapSize=%zu\n", allocPtr->size);
-  // JLANG_visualize(allocPtr);
 
   return 0;
 }

src/memory/allocator.h

@@ -49,9 +49,7 @@ void *JLANG_CreateAllocator() {
   allocator->size = 1 * 1024;
 
   // ensure all memory is zero
-  for (int i = 0; i < 1 * 1024; i++) {
+  memset(allocator->memory, 0, 1024);
-    allocator->memory[i] = 0;
-  }
 
   return allocator;
 }

src/memory/gc.h

@@ -4,6 +4,7 @@
 #include "../objects/object.h"
 #include "allocator.h"
 
+
 void gc_collect(JLANG_memory_allocator *allocPtr, size_t *roots,
                 int root_count) {
   // Stage 1. Mark blocks
@@ -29,67 +30,79 @@ void gc_collect(JLANG_memory_allocator *allocPtr, size_t *roots,
                              objPtr->data.string_val.chars -
                              sizeof(JLANG_metadata));
       itemsHeader->marked = 1;
-    }
+    } else if (objPtr->type == OBJ_INSTANCE) {
-  }
+      JLANG_metadata *namesHeader =
+          (JLANG_metadata *)((char *)allocPtr->memory +
+                             objPtr->data.instance_val.field_names -
+                             sizeof(JLANG_metadata));
+      namesHeader->marked = 1;
 
-  // Stage 2. Sweep memory
+      JLANG_metadata *valuesHeader =
-  JLANG_metadata *currentHead = (JLANG_metadata *)allocPtr->memory;
+          (JLANG_metadata *)((char *)allocPtr->memory +
-  while (currentHead->size != 0) {
+                             objPtr->data.instance_val.field_values -
-    size_t blockOffset = (char *)currentHead - (char *)allocPtr->memory;
+                             sizeof(JLANG_metadata));
-
+      valuesHeader->marked = 1;
-    if (currentHead->marked == 0 && currentHead->in_use) {
+      }
-      // Free block
-      JLANG_free(allocPtr, blockOffset + sizeof(JLANG_metadata));
     }
 
-    // Reset mark
+    // Stage 2. Sweep memory
-    currentHead->marked = 0;
+    JLANG_metadata *currentHead = (JLANG_metadata *)allocPtr->memory;
+    while (currentHead->size != 0) {
+      size_t blockOffset = (char *)currentHead - (char *)allocPtr->memory;
 
-    currentHead = (JLANG_metadata *)((char *)currentHead + currentHead->size +
+      if (currentHead->marked == 0 && currentHead->in_use) {
-                                     sizeof(JLANG_metadata));
+        // Free block
-  }
+        JLANG_free(allocPtr, blockOffset + sizeof(JLANG_metadata));
+      }
 
-  // Stage 3. Join free blocks
+      // Reset mark
-  currentHead = (JLANG_metadata *)allocPtr->memory;
+      currentHead->marked = 0;
-  while (currentHead->size != 0) {
 
-    if (currentHead->in_use == 0) {
+      currentHead = (JLANG_metadata *)((char *)currentHead + currentHead->size +
-      // iterate until last next free block
+                                       sizeof(JLANG_metadata));
-      JLANG_metadata *nextHeader =
+    }
-          (JLANG_metadata *)((char *)currentHead + sizeof(JLANG_metadata) +
+
-                             currentHead->size);
+    // Stage 3. Join free blocks
-      JLANG_metadata *lastHeader = currentHead;
+    currentHead = (JLANG_metadata *)allocPtr->memory;
-      while (1) {
+    while (currentHead->size != 0) {
-        if (nextHeader->in_use == 1 || nextHeader->size == 0) {
+
-          // Volvemos atrás
+      if (currentHead->in_use == 0) {
-          break;
+        // iterate until last next free block
+        JLANG_metadata *nextHeader =
+            (JLANG_metadata *)((char *)currentHead + sizeof(JLANG_metadata) +
+                               currentHead->size);
+        JLANG_metadata *lastHeader = currentHead;
+        while (1) {
+          if (nextHeader->in_use == 1 || nextHeader->size == 0) {
+            // Volvemos atrás
+            break;
+          }
+
+          lastHeader = nextHeader;
+          nextHeader =
+              (JLANG_metadata *)((char *)nextHeader + sizeof(JLANG_metadata) +
+                                 nextHeader->size);
         }
 
-        lastHeader = nextHeader;
+        // Make from currentHead to nextHeader + nextHeader(size) one block
-        nextHeader =
+        size_t startIndex = (char *)currentHead - (char *)allocPtr->memory;
-            (JLANG_metadata *)((char *)nextHeader + sizeof(JLANG_metadata) +
+        size_t endIndex =
-                               nextHeader->size);
+            ((char *)lastHeader + sizeof(JLANG_metadata) + lastHeader->size) -
+            (char *)allocPtr->memory;
+
+        // Set to 0
+        for (int i = 0; i < (endIndex - startIndex); i++) {
+          allocPtr->memory[i + startIndex] = 0;
+        }
+
+        // Create valid header
+        currentHead = (JLANG_metadata *)((char *)allocPtr->memory + startIndex);
+        currentHead->size = (endIndex - startIndex) - sizeof(JLANG_metadata);
       }
 
-      // Make from currentHead to nextHeader + nextHeader(size) one block
+      currentHead = (JLANG_metadata *)((char *)currentHead + currentHead->size +
-      size_t startIndex = (char *)currentHead - (char *)allocPtr->memory;
+                                       sizeof(JLANG_metadata));
-      size_t endIndex =
-          ((char *)lastHeader + sizeof(JLANG_metadata) + lastHeader->size) -
-          (char *)allocPtr->memory;
-
-      // Set to 0
-      for (int i = 0; i < (endIndex - startIndex); i++) {
-        allocPtr->memory[i + startIndex] = 0;
-      }
-
-      // Create valid header
-      currentHead = (JLANG_metadata *) ((char *)allocPtr->memory + startIndex);
-      currentHead->size = (endIndex - startIndex) - sizeof(JLANG_metadata);
     }
-
-    currentHead = (JLANG_metadata *)((char *)currentHead + currentHead->size +
-                                     sizeof(JLANG_metadata));
   }
-}
 
 #endif

src/objects/object.h

@@ -3,11 +3,10 @@
 
 #include "../memory/allocator.h"
 
-
 #define JLANG_RESOLVE(alloc, offset)                                           \
   ((void *)(((JLANG_memory_allocator *)(alloc))->memory + (offset)))
 
-typedef enum { OBJ_INT, OBJ_FLOAT, OBJ_STRING, OBJ_LIST, OBJ_NONE } ObjectType;
+typedef enum { OBJ_INT, OBJ_FLOAT, OBJ_STRING, OBJ_LIST, OBJ_NONE, OBJ_INSTANCE } ObjectType;
 
 typedef struct Object {
   ObjectType type;
@@ -23,9 +22,38 @@ typedef struct Object {
       int count;
       int capacity;
     } list_val;
+    struct {
+      int class_index;     // indice en Chunk.classes[]
+      size_t field_names;  // heap offset -> array de int
+      size_t field_values; // heap offset -> array de Value
+      int field_count;
+      int field_capacity;
+    } instance_val;
   } data;
 } Object;
 
+size_t obj_new_instance(void *allocator, int class_index, int capacity, size_t value_size) {
+  size_t offset = JLANG_malloc(allocator, sizeof(Object));
+  Object *objPtr = (Object *)JLANG_RESOLVE(allocator, offset);
+  objPtr->type = OBJ_INSTANCE;
+  objPtr->data.instance_val.class_index = class_index;
+  objPtr->data.instance_val.field_count = 0;
+  objPtr->data.instance_val.field_capacity = capacity;
+
+  // alloc array de nombres
+  size_t namesOffset = JLANG_malloc(allocator, capacity * sizeof(int));
+  objPtr = (Object *) JLANG_RESOLVE(allocator, offset); // re-resolve
+
+  // alloc array de valores
+  size_t valuesOffset = JLANG_malloc(allocator, capacity * value_size);
+  objPtr = (Object *) JLANG_RESOLVE(allocator, offset); // re-resolve
+
+  objPtr->data.instance_val.field_names = namesOffset;
+  objPtr->data.instance_val.field_values = valuesOffset;
+
+  return offset;
+}
+
 size_t obj_new_int(void *allocator, int value) {
   // Object *objPtr = (Object *)JLANG_malloc(allocator, sizeof(Object));
   size_t offset = JLANG_malloc(allocator, sizeof(Object));
@@ -91,10 +119,16 @@ void obj_free(void *allocator, size_t offset) {
     JLANG_free(allocator, obj->data.list_val.items);
   }
 
+  if (obj->type == OBJ_INSTANCE) {
+    JLANG_free(allocator, obj->data.instance_val.field_names);
+    JLANG_free(allocator, obj->data.instance_val.field_values);
+  }
+
   JLANG_free(allocator, offset);
 }
 
-void obj_print(void *allocator, size_t offset, const char *preffix) {
+void obj_print(void *allocator, size_t offset, const char *preffix,
+               const char *suffix) {
   Object *obj = (Object *)JLANG_RESOLVE(allocator, offset);
 
   switch (obj->type) {
@@ -117,7 +151,7 @@ void obj_print(void *allocator, size_t offset, const char *preffix) {
       if (items[i] == offset) {
         printf("<self:0x%zu>", offset);
       } else {
-        obj_print(allocator, items[i], "\"");
+        obj_print(allocator, items[i], "\"", "\"");
       }
 
       if (i < obj->data.list_val.capacity - 1) {
@@ -133,8 +167,8 @@ void obj_print(void *allocator, size_t offset, const char *preffix) {
     }
     printf("%s", (char *)JLANG_RESOLVE(allocator, obj->data.string_val.chars));
 
-    if (strcmp(preffix, "") != 0) {
+    if (strcmp(suffix, "") != 0) {
-      printf("%s", preffix);
+      printf("%s", suffix);
     }
     break;
   default: