#ifndef AMALGAM_BYTECODE_H #define AMALGAM_BYTECODE_H #include "../std/defs.h" #include "../std/misc.h" #include "../std/buffer.h" #include "../defs.h" #include /*doc(Instructions) Variable length instructions. Instruction size ranges from 1 to 5 bytes. # Instruction formats Instructions can be in 7 different formats: 1. 1 byte: Opcode(u8) 2. 2 bytes: Opcode(u8) + register(i8) 3. 3 bytes: Opcode(u8) + register(i8) + register(i8) 4. 3 bytes:\ 4.1 Opcode(u8) + intermediate(u16)\ 4.2 Opcode(u8) + data(u16)\ 4.3 Opcode(u8) + label(i16)\ 4.4 Opcode(u8) + register(i8) + num_args(u8) 5. 4 bytes: Opcode(u8) + register(i8) + register(i8) + register(i8) 6. 4 bytes:\ 6.1 Opcode(u8) + register(i8) + label(i16)\ 6.2 Opcode(u8) + register(i8) + intermediate(u16)\ 6.3 Opcode(u8) + register(i8) + data(u16)\ 6.4 Opcode(u8) + flags(u8) + num_local_var_reg(u16)\ 6.5 Opcode(u8) + stack_offset(i24) 7. 5 bytes: Opcode(u8) + index(u8) + index(u16) + num_args(u8) # Registers Registers have a range of 128. Local variables start from register 0 and increment while parameters start from -1 and decrement. Registers have the scope of functions and reset after instructions reach a new function (AMAL_OP_FUNC_START). If import index for call and calle is 0, then that means the function resides in the same file the function call is being called from. Which means that import index 1 is actually import index 0 into the import list. */ /* Important: The number of fields in this enum can't exceed 255 */ typedef enum { AMAL_OP_NOP, /* No operation (do nothing). This can be used for patching code */ AMAL_OP_SETZ, /* setz reg - Set register value to 0 */ AMAL_OP_MOV, /* mov dst, src - Move src register to dst register */ AMAL_OP_MOVI, /* movi dst, src - Move src intermediate to dst register */ AMAL_OP_MOVD, /* movd dst, src - Move src data to dst register */ AMAL_OP_ADD, /* add dst, reg1, reg2 - Add reg1 and reg2 and put the result in dst */ AMAL_OP_SUB, /* sub dst, reg1, reg2 - Substract reg2 from reg1 and put the result in dst */ AMAL_OP_IMUL, /* imul dst, reg1, reg2 - Signed multiplication */ AMAL_OP_MUL, /* mul dst, reg1, reg2 - Unsigned multiplication */ AMAL_OP_IDIV, /* idiv dst, reg1, reg2 - Signed division */ AMAL_OP_DIV, /* div dst, reg1, reg2 - Unsigned division */ AMAL_OP_PUSH, /* push reg - Push register onto stack */ AMAL_OP_PUSHI, /* pushi int - Push intermediate onto stack */ AMAL_OP_PUSHD, /* pushd data - Push data onto stack */ AMAL_OP_PUSH_RET, /* push_ret reg - Push register onto stack as a return value of the next function call */ AMAL_OP_CALL, /* call ii, fi, num_args - Call a function in imported file (ii, import index) using function index (fi) and num_args arguments. ii is u8, fi is u16 and num_args is u8 */ AMAL_OP_CALLR, /* callr reg, num_args - Call a function using a register. Used for function pointers. num_args is u8 */ AMAL_OP_CALLE, /* calle ii, efi, num_args - Call an extern function in imported file (ii, import index) using extern function index (efi) and num_args arguments. ii is u8, efi is u16 and num_args is u8 */ AMAL_OP_CMP, /* cmp dst, reg1, reg2 - Set dst to 1 if reg1 equals reg2, otherwise set it to 0 */ AMAL_OP_JZ, /* jz reg, label - Jump to label in the current function if reg is zero. label is u16 */ AMAL_OP_JMP, /* jmp label - Unconditional jump to label in the current function. label is u16 */ AMAL_OP_RET, /* ret reg - Return from the function with reg result */ AMAL_OP_FUNC_START, /* func_start flags, num_local_var_reg - Start of a function which has @num_local_var_reg local variable registers allocated and has the flag @flag. @flag is u8 and @num_local_var_reg is u16 */ AMAL_OP_FUNC_END, /* func_end - End of a function. Implementation should do a ret here */ AMAL_OP_LABEL /* label - Label. This is the target of a jump instruction. Jump instructions only jump to labels in the same function scope */ } AmalOpcode; #define AMAL_BYTECODE_MAGIC_NUMBER (u32)0xdec05eba #define AMAL_BYTECODE_MAJOR_VERSION 1 #define AMAL_BYTECODE_MINOR_VERSION 0 #define AMAL_BYTECODE_PATCH_VERSION 0 #define AMAL_BYTECODE_SECTION_MAGIC_NUMBER (u32)0x004005e4 /* "section\0" in ascii */ #define AMAL_BYTECODE_NUM_REGISTERS 256 typedef enum { FUNC_FLAG_NONE = 0, FUNC_FLAG_EXPORTED = 1 << 0 } amal_func_flag; typedef u8 AmalOpcodeType; /* TODO: Make sure this pragma pack works on all platforms */ #pragma pack(push, 1) typedef struct { u32 magic_number; /* AMAL_BYTECODE_MAGIC_NUMBER */ u8 major_version; u8 minor_version; u8 patch_version; u8 endian; /* 0 = little endian, 1 = big endian */ } BytecodeHeader; #pragma pack(pop) /* TODO: Make sure this pragma pack works on all platforms */ #pragma pack(push, 1) typedef struct { u32 func_offset; u8 num_params; u32 params_num_pointers; u32 params_fixed_size; u8 num_return_types; u32 return_types_num_pointers; u32 return_types_fixed_size; } BytecodeHeaderFunction; #pragma pack(pop) /* TODO: Make sure this pragma pack works on all platforms */ #pragma pack(push, 1) typedef struct { u8 num_params; u32 params_num_pointers; u32 params_fixed_size; u8 num_return_types; u32 return_types_num_pointers; u32 return_types_fixed_size; } BytecodeHeaderExternFunction; #pragma pack(pop) /* TODO: Make sure this pragma pack works on all platforms */ #pragma pack(push, 1) typedef struct { u32 function_index; #define parser_index function_index u32 extern_function_index; } BytecodeHeaderImport; #pragma pack(pop) struct Bytecode { Buffer/**/ data; usize import_index; /* Reference inside @data where imports start */ u32 funcs_index; /* Reference inside @data where funcs start */ u32 extern_funcs_index; /* Reference inside @data where extern funcs start */ u32 offset; /* Offset that this bytecode starts from in the final program (all bytecodes combined) */ }; typedef struct { jmp_buf env; Bytecode *bytecode; Parser *parser; /* borrowed */ } BytecodeCompilerContext; CHECK_RESULT int bytecode_init(Bytecode *self, ArenaAllocator *allocator); CHECK_RESULT int buffer_append_header(Buffer *program_data); /* longjump to self->env on failure */ void generate_bytecode_from_ssa(BytecodeCompilerContext *self); #endif