bloat-buster/src/compiler.bbb

mode_t = typealias u64;
int = typealias s32;
usize = typealias u64;
ssize = typealias s64;
File = typealias s32;

uid_t = typealias u32;
gid_t = typealias u32;
off_t = typealias s64;
ino_t = typealias u64;
dev_t = typealias u64;

timespec = struct
{
    seconds: ssize,
    nanoseconds: ssize,
};

Stat = struct
{
    dev: dev_t,
    ino: ino_t,
    nlink: usize,

    mode: u32,
    uid: uid_t,
    gid: gid_t,
    _: u32,
    rdev: dev_t,
    size: off_t,
    blksize: ssize,
    blocks: s64,

    atim: timespec,
    mtime: timespec,
    ctim: timespec,
    _: [3]ssize,
}

OAccessMode = enum u2
{
    read_only = 0,
    write_only = 1,
    read_write = 2,
}

O = bits u32
{
    access_mode: OAccessMode,
    _: u4,
    creat: u1,
    excl: u1,
    noctty: u1,
    trunc: u1,
    append: u1,
    nonblock: u1,
    dsync: u1,
    async: u1,
    direct: u1,
    _: u1,
    directory: u1,
    nofollow: u1,
    noatime: u1,
    cloexec: u1,
    sync: u1,
    path: u1,
    tmpfile: u1,
    _: u9,
}

[extern] memcmp = fn [cc(c)] (a: &u8, b: &u8, byte_count: usize) int;
[extern] memcpy = fn [cc(c)] (destination: &u8, source: &u8, byte_count: u64) &u8;
[extern] exit = fn [cc(c)] (exit_code: int) noreturn;

[extern] realpath = fn [cc(c)] (source_path: &u8, resolved_path: &u8) &u8;
[extern] mkdir = fn [cc(c)] (path: &u8, mode: mode_t) int;

[extern] open = fn [cc(c)] (path: &u8, o: O, ...) int;
[extern] close = fn [cc(c)] (fd: File) int;
[extern] fstat = fn [cc(c)] (fd: File, s: &Stat) int;
[extern] write = fn [cc(c)] (fd: File, pointer: &u8, byte_count: u64) ssize;
[extern] read = fn [cc(c)] (fd: File, pointer: &u8, byte_count: u64) ssize;

assert = macro (ok: u1) void
{
    if (!ok)
    {
        unreachable;
    }
}

align_forward = fn (value: u64, alignment: u64) u64
{
    assert(alignment != 0);
    >mask = alignment - 1;
    >result = (value + mask) & ~mask;
    return result;
}

string_no_match = #integer_max(u64);

c_string_length = fn (c_string: &u8) u64
{
    >it = c_string;

    while (it.&)
    {
        it = it + 1;
    }

    return #int_from_pointer(it) - #int_from_pointer(c_string);
}

c_string_to_slice = fn (c_string: &u8) []u8
{
    >length = c_string_length(c_string);
    return c_string[0..length];
}

string_equal = fn(a: []u8, b: []u8) u1
{
    >result: #ReturnType = 0;

    if (a.length == b.length)
    {
        result = memcmp(a.pointer, b.pointer, a.length) == 0;
    }

    return result;
}

string_last_character = fn(string: []u8, character: u8) u64
{
    >i = string.length;

    while (i > 0)
    {
        i -= 1;

        if (string[i] == character)
        {
            return i;
        }
    }

    return string_no_match;
}

OS_Linux_PROT = bits u32
{
    read: u1,
    write: u1,
    execute: u1,
    sem: u1,
    _: u28,
}

OS_Linux_MAP_Type = enum u4
{
    shared = 0x1,
    private = 0x2,
    shared_validate = 0x3,
}

OS_Linux_MAP = bits u32
{
    type: OS_Linux_MAP_Type,
    fixed: u1,
    anonymous: u1,
    bit_32: u1,
    _: u1,
    grows_down: u1,
    _: u2,
    deny_write: u1,
    executable: u1,
    locked: u1,
    no_reserve: u1,
    populate: u1,
    non_block: u1,
    stack: u1,
    huge_tlb: u1,
    sync: u1,
    fixed_noreplace: u1,
    _: u5,
    uninitialized: u1,
    _: u5,
}

[extern] mmap = fn [cc(c)] (address: u64, size: u64, protection: OS_Linux_PROT, map: OS_Linux_MAP, file_descriptor: s32, offset: s64) &u8;
[extern] mprotect = fn [cc(c)] (address: u64, size: u64, protection: OS_Linux_PROT) s32;

OS_ProtectionFlags = bits
{
    read: u1,
    write: u1,
    execute: u1,
}

OS_MapFlags = bits
{
    private: u1,
    anonymous: u1,
    no_reserve: u1,
    populate: u1,
}

os_linux_protection_flags = fn(map_flags: OS_ProtectionFlags) OS_Linux_PROT
{
    return {
        .read = map_flags.read,
        .write = map_flags.write,
        .execute = map_flags.execute,
        zero,
    };
}

os_linux_map_flags = fn(map_flags: OS_MapFlags) OS_Linux_MAP
{
    return {
        .type = #select(map_flags.private, .private, .shared),
        .anonymous = map_flags.anonymous,
        .no_reserve = map_flags.no_reserve,
        .populate = map_flags.populate,
        zero,
    };
}

os_reserve = fn (base: u64, size: u64, protection: OS_ProtectionFlags, map: OS_MapFlags) &u8
{
    >protection_flags = os_linux_protection_flags(protection);
    >map_flags = os_linux_map_flags(map);
    >address = mmap(base, size, protection_flags, map_flags, -1, 0);
    if (#int_from_pointer(address) == #integer_max(u64))
    {
        unreachable;
    }

    return address;
}

os_commit = fn (address: u64, size: u64, protection: OS_ProtectionFlags) void
{
    >protection_flags = os_linux_protection_flags(protection);
    >result = mprotect(address, size, protection_flags);
    if (result != 0)
    {
        unreachable;
    }
}

os_make_directory = fn (path: &u8) void
{
    >result = mkdir(path, 0o755);
}

OpenFlags = bits
{
    truncate: u1,
    execute: u1,
    write: u1,
    read: u1,
    create: u1,
    directory: u1,
}

OpenPermissions = bits
{
    read: u1,
    write: u1,
    execute: u1,
}

os_file_open = fn (path: &u8, flags: OpenFlags, permissions: OpenPermissions) File
{
    >access_mode: OAccessMode = undefined;
    if (flags.read and flags.write)
    {
        access_mode = .read_write;
    }
    else if (flags.read)
    {
        access_mode = .read_only;
    }
    else if (flags.write)
    {
        access_mode = .write_only;
    }
    else
    {
        unreachable;
    }

    >o: O = {
        .access_mode = access_mode,
        .trunc = flags.truncate,
        .creat = flags.create,
        .directory = flags.directory,
    };

    >mode: mode_t = #select(permissions.execute, 0o755, 0o644);
    >fd = open(path, o, mode);
    return fd;
}

os_file_close = fn (fd: File) void
{
    >result = close(fd);
    assert(result == 0);
}

os_file_is_valid = fn (fd: File) u1
{
    return fd >= 0;
}

os_file_get_size = fn (fd: File) u64
{
    >stat: Stat = undefined;
    >result = fstat(fd, &stat);
    assert(result == 0);
    return #extend(stat.size);
}

os_file_read_partially = fn (fd: File, pointer: &u8, length: u64) u64
{
    >result = read(fd, pointer, length);
    assert(result > 0);
    return #extend(result);
}

os_file_read = fn (fd: File, buffer: []u8, byte_count: u64) void
{
    assert(byte_count <= buffer.length);

    >total_read_byte_count: u64 = 0;

    while (total_read_byte_count < byte_count)
    {
        >read_byte_count = os_file_read_partially(fd, buffer.pointer + total_read_byte_count, byte_count - total_read_byte_count);
        total_read_byte_count += read_byte_count;
    }
}

os_path_absolute_stack = fn (buffer: []u8, relative_file_path: &u8) []u8
{
    >syscall_result = realpath(relative_file_path, buffer.pointer);

    >result: []u8 = zero;

    if (syscall_result)
    {
        result = c_string_to_slice(syscall_result);
        assert(result.length < buffer.length);
    }

    return result;
}

Arena = struct
{
    reserved_size: u64,
    position: u64,
    os_position: u64,
    granularity: u64,
    reserved: [32]u8,
}

minimum_position: u64 = #byte_size(Arena);

ArenaInitialization = struct
{
    reserved_size: u64,
    granularity: u64,
    initial_size: u64,
}

arena_initialize = fn (initialization: ArenaInitialization) &Arena
{
    >protection_flags: OS_ProtectionFlags = {
        .read = 1,
        .write = 1,
        zero,
    };

    >map_flags: OS_MapFlags = {
        .private = 1,
        .anonymous = 1,
        .no_reserve = 1,
        .populate = 0,
    };

    >arena: &Arena = #pointer_cast(os_reserve(0, initialization.reserved_size, protection_flags, map_flags));
    os_commit(#int_from_pointer(arena), initialization.initial_size, {
        .read = 1,
        .write = 1,
        zero,
    });

    arena.& = {
        .reserved_size = initialization.reserved_size,
        .position = minimum_position,
        .os_position = initialization.initial_size,
        .granularity = initialization.granularity,
        zero,
    };

    return arena;
}

arena_initialize_default = fn (initial_size: u64) &Arena
{
    return arena_initialize({
        .reserved_size = 4 * 1024 * 1024 * 1024,
        .granularity = 4 * 1024,
        .initial_size = initial_size,
    });
}

arena_allocate_bytes = fn (arena: &Arena, size: u64, alignment: u64) &u8
{
    >aligned_offset = align_forward(arena.position, alignment);
    >aligned_size_after = aligned_offset + size;

    >arena_byte_pointer: &u8 = #pointer_cast(arena);

    if (aligned_size_after > arena.os_position)
    {
        >target_committed_size = align_forward(aligned_size_after, arena.granularity);
        >size_to_commit = target_committed_size - arena.os_position;
        >commit_pointer = arena_byte_pointer + arena.os_position;
        os_commit(#int_from_pointer(commit_pointer), size_to_commit, {
            .read = 1,
            .write = 1,
            zero,
        });
        arena.os_position = target_committed_size;
    }

    >result = arena_byte_pointer + aligned_offset;
    arena.position = aligned_size_after;
    assert(arena.position <= arena.os_position);

    return result;
}

arena_allocate = macro [T] (arena: &Arena, count: u64) &T
{
    return #pointer_cast(arena_allocate_bytes(arena, #byte_size(T) * count, #alignof(T)));
}

arena_duplicate_string = fn (arena: &Arena, string: []u8) []u8
{
    >result = arena_allocate_bytes(arena, string.length + 1, 1);
    memcpy(result, string.pointer, string.length);
    return result[..string.length];
}

arena_join_string = fn (arena: &Arena, pieces: [][]u8) []u8
{
    >size: u64 = 0;

    for (p: pieces)
    {
        size += p.length;
    }

    >pointer = arena_allocate_bytes(arena, size + 1, 1);
    >i: u64 = 0;

    for (p: pieces)
    {
        memcpy(pointer + i, p.pointer, p.length);
        i += p.length;
    }

    assert(i == size);
    pointer[i] = 0;

    return pointer[..size];
}

file_read = fn (arena: &Arena, path: []u8) []u8
{
    >fd = os_file_open(path.pointer, { .read = 1 }, { .read = 1 });

    >result: []u8 = zero;

    if (os_file_is_valid(fd))
    {
        >file_size = os_file_get_size(fd);
        >file_buffer = arena_allocate_bytes(arena, file_size, 1);
        result = file_buffer[..file_size];
        os_file_read(fd, result, file_size);
        os_file_close(fd);
    }

    return result;
}

path_absolute = fn (arena: &Arena, relative_file_path: &u8) []u8
{
    >buffer: [4096]u8 = undefined;
    >stack_slice = os_path_absolute_stack(buffer[..], relative_file_path);
    >result = arena_duplicate_string(arena, stack_slice);
    return result;
}

GlobalState = struct
{
    arena: &Arena,
}

global_state: GlobalState = undefined;

global_state_initialize = fn () void
{
    global_state = {
        .arena = arena_initialize_default(2 * 1024 * 1024),
    };
}

fail = fn () noreturn
{
    exit(1);
}

CompilerCommand = enum
{
    compile,
    test,
}

BuildMode = enum
{
    debug_none,
    debug_fast,
    debug_size,
    soft_optimize,
    optimize_for_speed,
    optimize_for_size,
    aggressively_optimize_for_speed,
    aggressively_optimize_for_size,
}

CompileFile = struct
{
    relative_file_path: []u8,
    build_mode: BuildMode,
    has_debug_info: u1,
    silent: u1,
}

base_cache_dir = "bb-cache";

CPUArchitecture = enum
{
    x86_64,
}

OperatingSystem = enum
{
    linux,
}

Target = struct
{
    cpu: CPUArchitecture,
    os: OperatingSystem,
}

target_get_native = fn () Target
{
    return { .cpu = .x86_64, .os = .linux };
}

CompileOptions = struct
{
    content: []u8,
    path: []u8,
    executable: []u8,
    name: []u8,
    objects: [][]u8,
    target: Target,
    build_mode: BuildMode,
    has_debug_info: u1,
    silent: u1,
}

TypeId = enum
{
    void,
    noreturn,
    integer,
}

TypeInteger = struct
{
    bit_count: u32,
    signed: u1,
}

TypeContent = union
{
    integer: TypeInteger,
}

Type = struct
{
    content: TypeContent,
    id: TypeId,
    name: []u8,
}

ValueId = enum
{
    infer_or_ignore,
}

Value = struct
{
    type: &Type,
    id: ValueId,
}

i128_offset: u64 = 64 * 2;
void_offset: u64 = i128_offset + 2;

Module = struct
{
    arena: &Arena,
    base_type_allocation: &Type,
    void_value: &Value,
    // Parser data
    content: []u8,
    offset: u64,
    line_offset: u64,
    line_character_offset: u64,
}

module_integer_type = fn (module: &Module, integer: TypeInteger) &Type
{
    assert(integer.bit_count != 0);
    assert(integer.bit_count <= 64);
    >index = #select(integer.bit_count == 128, i128_offset + #extend(integer.signed), #extend(integer.bit_count) - 1 + 64 * #extend(integer.signed));
    return module.base_type_allocation + index;
}

module_void_type = fn (module: &Module) &Type
{
    return module.base_type_allocation + void_offset;
}

module_noreturn_type = fn (module: &Module) &Type
{
    return module_void_type(module) + 1;
}

is_space = fn (ch: u8) u1
{
    return ch == ' ' or ch == '\n' or ch == '\t' or ch == '\r';
}

is_decimal = fn (ch: u8) u1
{
    return ch >= '0' and ch <= '9';
}

is_octal = fn (ch: u8) u1
{
    return ch >= '0' and ch <= '7';
}

is_binary = fn (ch: u8) u1
{
    return ch == '0' or ch == '1';
}

is_hex_alpha_lower = fn (ch: u8) u1
{
    return ch >= 'a' and ch <= 'f';
}

is_hex_alpha_upper = fn (ch: u8) u1
{
    return ch >= 'A' and ch <= 'F';
}

is_hex_alpha = fn (ch: u8) u1
{
    return is_hex_alpha_lower(ch) or is_hex_alpha_upper(ch);
}

is_hex = fn (ch: u8) u1
{
    return is_decimal(ch) or is_hex_alpha(ch);
}

is_identifier_start = fn (ch: u8) u1
{
    return (ch >= 'a' and ch >= 'z') or (ch >= 'A' and ch <= 'Z') or ch == '_';
}

is_identifier = fn (ch: u8) u1
{
    return is_identifier_start(ch) or is_decimal(ch);
}

report_error = fn () noreturn
{
    #trap();
}

get_line = fn (module: &Module) u32
{
    >line = module.line_offset + 1;
    assert(line <= #integer_max(u32));
    return #truncate(line);
}

get_column = fn (module: &Module) u32
{
    >column = module.offset - module.line_character_offset + 1;
    assert(column <= #integer_max(u32));
    return #truncate(column);
}

skip_space = fn (module: &Module) void
{
    while (1)
    {
        >iteration_offset = module.offset;

        while (module.offset < module.content.length and? is_space(module.content[module.offset]))
        {
            module.line_offset += #extend(module.content[module.offset] == '\n');
            module.line_character_offset = #select(module.content[module.offset] == '\n', module.offset, module.line_character_offset);
            module.offset += 1;
        }

        if (module.offset + 1 < module.content.length)
        {
            >i = module.offset;
            >is_comment = module.content[i] == '/' and module.content[i + 1] == '/';

            if (is_comment)
            {
                while (module.offset < module.content.length and? module.content[module.offset] != '\n')
                {
                    module.offset += 1;
                }

                if (module.offset < module.content.length)
                {
                    module.line_offset += 1;
                    module.line_character_offset = module.offset;
                    module.offset += 1;
                }
            }
        }

        if (module.offset - iteration_offset == 0)
        {
            break;
        }
    }
}

consume_character_if_match = fn (module: &Module, expected_character: u8) u1
{
    >is_ch: u1 = 0;

    if (module.offset < module.content.length)
    {
        >ch = module.content[module.offset];
        is_ch = expected_character == ch;
        module.offset += #extend(is_ch);
    }

    return is_ch;
}

expect_character = fn (module: &Module, expected_character: u8) void
{
    if (!consume_character_if_match(module, expected_character))
    {
        report_error();
    }
}

parse_identifier = fn (module: &Module) []u8
{
    >start = module.offset;

    if (is_identifier_start(module.content[start]))
    {
        module.offset += 1;

        while (module.offset < module.content.length)
        {
            if (is_identifier(module.content[module.offset]))
            {
                module.offset += 1;
            }
            else
            {
                break;
            }
        }
    }

    if (module.offset - start == 0)
    {
        report_error();
    }

    >result = module.content[start..module.offset];
    return result;
}

parse = fn (module: &Module) void
{
}

emit = fn (module: &Module) void
{
}

compile = fn (arena: &Arena, options: CompileOptions) void
{
    >signs: [2]u1 = [0, 1];

    >base_type_allocation = arena_allocate[Type](arena,
        i128_offset // Basic integer types
        + 2 // u128, s128
        + 2 // void, noreturn
    );

    >type_it = base_type_allocation;

    for (sign: signs)
    {
        for (b: 0..64)
        {
            >bit_count = b + 1;
            >first_digit: u8 = #truncate(#select(bit_count < 10, bit_count % 10 + '0', bit_count / 10 + '0'));
            >second_digit: u8 = #truncate(#select(bit_count > 9, bit_count % 10 + '0', 0));
            >name_buffer: [3]u8 = [ #select(sign, 's', 'u'), first_digit, second_digit ];
            >name_length: u64 = 2 + #extend(bit_count > 9);

            >name = arena_duplicate_string(arena, name_buffer[..name_length]);

            type_it.& = {
                .content = {
                    .integer = {
                        .bit_count = #truncate(bit_count),
                        .signed = sign,
                    },
                },
                .id = .integer,
                .name = name,
            };
            type_it += 1;
        }
    }

    for (sign: signs)
    {
        >name = #select(sign, "s128", "u128");
        type_it.& = {
            .content = {
                .integer = {
                    .bit_count = 128,
                    .signed = sign,
                },
            },
            .id = .integer,
            .name = name,
        };
        type_it += 1;
    }

    >void_type = type_it;
    type_it += 1;
    >noreturn_type = type_it;
    type_it += 1;

    void_type.& = {
        .id = .void,
        .name = "void",
    };

    noreturn_type.& = {
        .id = .noreturn,
        .name = "noreturn",
    };

    >void_value = arena_allocate[Value](arena, 1);
    void_value.& = {
        .type = void_type,
        .id = .infer_or_ignore,
    };

    >module: Module = {
        .arena = arena,
        .base_type_allocation = base_type_allocation,
        .void_value = void_value,
        .content = options.content,
        .offset = 0,
        .line_offset = 0,
        .line_character_offset = 0,
    };

    parse(&module);
    emit(&module);
}

compile_file = fn (arena: &Arena, compile_options: CompileFile) void
{
    >relative_file_path = compile_options.relative_file_path;
    if (relative_file_path.length < 5)
    {
        fail();
    }

    >extension_start = string_last_character(relative_file_path, '.');
    if (extension_start == string_no_match)
    {
        fail();
    }

    if (!string_equal(relative_file_path[extension_start..], ".bbb"))
    {
        fail();
    }

    >separator_index = string_last_character(relative_file_path, '/');
    if (separator_index == string_no_match)
    {
        separator_index = 0;
    }
    
    >base_start = separator_index + #extend(separator_index != 0 or relative_file_path[separator_index] == '/');
    >base_name = relative_file_path[base_start..extension_start];

    >is_compiler = string_equal(relative_file_path, "src/compiler.bbb");
    >outputh_path_dir = arena_join_string(arena, [
        base_cache_dir,
        #select(is_compiler, "/compiler/", "/"),
        #enum_name(compile_options.build_mode),
        "_",
        #select(compile_options.has_debug_info, "di", "nodi"),
    ][..]);

    os_make_directory(base_cache_dir.pointer);
    if (is_compiler)
    {
        >compiler_dir = arena_join_string(arena, [ base_cache_dir, "/compiler" ][..]);
        os_make_directory(compiler_dir.pointer);
    }
    os_make_directory(outputh_path_dir.pointer);

    >outputh_path_base = arena_join_string(arena, [ outputh_path_dir, "/", base_name ][..]);

    >output_object_path = arena_join_string(arena, [ outputh_path_base, ".o" ][..]);
    >output_executable_path = outputh_path_base;

    >file_content = file_read(arena, relative_file_path);
    >file_path = path_absolute(arena, relative_file_path.pointer);
    >c_abi_object_path = ""; // TODO

    >objects: [][]u8 = undefined;
    if (string_equal(base_name, "c_abi"))
    {
        objects = [ output_object_path, c_abi_object_path ][..];
    }
    else
    {
        objects = [ output_object_path ][..];
    }

    >options: CompileOptions = {
        .executable = output_executable_path,
        .objects = objects,
        .name = base_name,
        .build_mode = compile_options.build_mode,
        .content = file_content,
        .path = file_path,
        .has_debug_info = compile_options.has_debug_info,
        .target = target_get_native(),
        .silent = compile_options.silent,
    };

    compile(arena, options);
}

[export] main = fn [cc(c)] (argument_count: u32, argv: &&u8) s32
{
    global_state_initialize();
    >arena = global_state.arena;

    if (argument_count < 2)
    {
        return 1;
    }

    >command_string = c_string_to_slice(argv[1]);

    >command_string_to_enum = #string_to_enum(CompilerCommand, command_string);
    if (!command_string_to_enum.is_valid)
    {
        return 1;
    }

    >command = command_string_to_enum.enum_value;
    switch (command)
    {
        .compile =>
        {
            if (argument_count < 3)
            {
                return 1;
            }

            >build_mode: BuildMode = .debug_none;
            >has_debug_info: u1 = 1;

            if (argument_count >= 4)
            {
                >build_mode_string_to_enum = #string_to_enum(BuildMode, c_string_to_slice(argv[3]));
                if (!build_mode_string_to_enum.is_valid)
                {
                    return 1;
                }
                build_mode = build_mode_string_to_enum.enum_value;
            }

            if (argument_count >= 5)
            {
                >has_debug_info_string = c_string_to_slice(argv[4]);
                if (string_equal(has_debug_info_string, "true"))
                {
                    has_debug_info = 1;
                }
                else if (string_equal(has_debug_info_string, "false"))
                {
                    has_debug_info = 0;
                }
                else
                {
                    return 1;
                }
            }

            >relative_file_path_pointer = argv[2];
            if (!relative_file_path_pointer)
            {
                return 1;
            }

            >relative_file_path = c_string_to_slice(relative_file_path_pointer);

            compile_file(arena, {
                .relative_file_path = relative_file_path,
                .build_mode = build_mode,
                .has_debug_info = has_debug_info,
                .silent = 0,
            });
        },
        .test =>
        {
            // TODO
            #trap();
        },
    }

    return 0;
}