Numpty forgot to add .c files

This commit is contained in:
Ginger Bill
2016-11-23 14:41:20 +00:00
parent 4110324588
commit 7792f009b8
22 changed files with 23799 additions and 0 deletions

View File

@@ -1,3 +1,5 @@
<img src="logo.png" alt="Odin logo" height="74">
# The Odin Programming Language
Odin is fast, concise, readable, pragmatic and open sourced. It is designed with the intent of replacing C with the following goals:

BIN
logo-slim.png Normal file

Binary file not shown.

After

Width:  |  Height:  |  Size: 246 KiB

235
src/array.c Normal file
View File

@@ -0,0 +1,235 @@
#define ARRAY_GROW_FORMULA(x) (2*(x) + 8)
GB_STATIC_ASSERT(ARRAY_GROW_FORMULA(0) > 0);
#define Array(Type_) struct { \
gbAllocator allocator; \
Type_ * e; \
isize count; \
isize capacity; \
}
typedef Array(void) ArrayVoid;
#define array_init_reserve(x_, allocator_, init_capacity_) do { \
GB_ASSERT((x_) != NULL); \
void **e = cast(void **)&((x_)->e); \
(x_)->allocator = (allocator_); \
(x_)->count = 0; \
(x_)->capacity = (init_capacity_); \
*e = gb_alloc((allocator_), gb_size_of(*(x_)->e)*(init_capacity_)); \
} while (0)
#define array_init_count(x_, allocator_, init_count_) do { \
GB_ASSERT((x_) != NULL); \
void **e = cast(void **)&((x_)->e); \
(x_)->allocator = (allocator_); \
(x_)->count = (init_count_); \
(x_)->capacity = (init_count_); \
*e = gb_alloc((allocator_), gb_size_of(*(x_)->e)*(init_count_)); \
} while (0)
#define array_init(x_, allocator_) do { array_init_reserve(x_, allocator_, ARRAY_GROW_FORMULA(0)); } while (0)
#define array_free(x_) do { gb_free((x_)->allocator, (x_)->e); } while (0)
#define array_set_capacity(x_, capacity_) do { array__set_capacity((x_), (capacity_), gb_size_of(*(x_)->e)); } while (0)
#define array_grow(x_, min_capacity_) do { \
isize new_capacity = ARRAY_GROW_FORMULA((x_)->capacity); \
if (new_capacity < (min_capacity_)) { \
new_capacity = (min_capacity_); \
} \
array_set_capacity(x_, new_capacity); \
} while (0)
#define array_add(x_, item_) do { \
if ((x_)->capacity < (x_)->count+1) { \
array_grow(x_, 0); \
} \
(x_)->e[(x_)->count++] = item_; \
} while (0)
#define array_pop(x_) do { GB_ASSERT((x_)->count > 0); (x_)->count--; } while (0)
#define array_clear(x_) do { (x_)->count = 0; } while (0)
#define array_resize(x_, new_count_) do { \
if ((x_)->capacity < (new_count_)) { \
array_grow((x_), (new_count_)); \
} \
(x_)->count = (new_count_); \
} while (0)
#define array_reserve(x_, new_capacity_) do { \
if ((x_)->capacity < (new_capacity_)) { \
array_set_capacity((x_), (new_capacity_)); \
} \
} while (0)
void array__set_capacity(void *ptr, isize capacity, isize element_size) {
GB_ASSERT(ptr != NULL);
ArrayVoid *x = cast(ArrayVoid *)ptr;
GB_ASSERT(element_size > 0);
if (capacity == x->capacity) {
return;
}
if (capacity < x->count) {
if (x->capacity < capacity) {
isize new_capacity = ARRAY_GROW_FORMULA(x->capacity);
if (new_capacity < capacity) {
new_capacity = capacity;
}
array__set_capacity(ptr, new_capacity, element_size);
}
x->count = capacity;
}
{
// TODO(bill): Resize rather than copy and delete
void *new_data = gb_alloc(x->allocator, element_size*capacity);
gb_memmove(new_data, x->e, element_size*x->count);
gb_free(x->allocator, x->e);
x->capacity = capacity;
x->e = new_data;
}
}
#if 0
template <typename T>
struct Array {
gbAllocator allocator;
T * data;
isize count;
isize capacity;
T &operator[](isize index) {
GB_ASSERT_MSG(0 <= index && index < count, "Index out of bounds");
return data[index];
}
T const &operator[](isize index) const {
GB_ASSERT_MSG(0 <= index && index < count, "Index out of bounds");
return data[index];
}
};
template <typename T> void array_init (Array<T> *array, gbAllocator a, isize init_capacity = ARRAY_GROW_FORMULA(0));
template <typename T> void array_init_count (Array<T> *array, gbAllocator a, isize count);
template <typename T> Array<T> array_make (T *data, isize count, isize capacity);
template <typename T> void array_free (Array<T> *array);
template <typename T> void array_add (Array<T> *array, T const &t);
template <typename T> T array_pop (Array<T> *array);
template <typename T> void array_clear (Array<T> *array);
template <typename T> void array_reserve (Array<T> *array, isize capacity);
template <typename T> void array_resize (Array<T> *array, isize count);
template <typename T> void array_set_capacity(Array<T> *array, isize capacity);
template <typename T>
void array_init(Array<T> *array, gbAllocator a, isize init_capacity) {
array->allocator = a;
array->data = gb_alloc_array(a, T, init_capacity);
array->count = 0;
array->capacity = init_capacity;
}
template <typename T>
void array_init_count(Array<T> *array, gbAllocator a, isize count) {
array->allocator = a;
array->data = gb_alloc_array(a, T, count);
array->count = count;
array->capacity = count;
}
template <typename T>
Array<T> array_make(T *data, isize count, isize capacity) {
Array<T> a = {0};
a.data = data;
a.count = count;
a.capacity = capacity;
return a;
}
template <typename T>
void array_free(Array<T> *array) {
if (array->allocator.proc != NULL) {
gb_free(array->allocator, array->data);
}
array->count = 0;
array->capacity = 0;
}
template <typename T>
void array__grow(Array<T> *array, isize min_capacity) {
isize new_capacity = ARRAY_GROW_FORMULA(array->capacity);
if (new_capacity < min_capacity) {
new_capacity = min_capacity;
}
array_set_capacity(array, new_capacity);
}
template <typename T>
void array_add(Array<T> *array, T const &t) {
if (array->capacity < array->count+1) {
array__grow(array, 0);
}
array->data[array->count] = t;
array->count++;
}
template <typename T>
T array_pop(Array<T> *array) {
GB_ASSERT(array->count > 0);
array->count--;
return array->data[array->count];
}
template <typename T>
void array_clear(Array<T> *array) {
array->count = 0;
}
template <typename T>
void array_reserve(Array<T> *array, isize capacity) {
if (array->capacity < capacity) {
array_set_capacity(array, capacity);
}
}
template <typename T>
void array_resize(Array<T> *array, isize count) {
if (array->capacity < count) {
array__grow(array, count);
}
array->count = count;
}
template <typename T>
void array_set_capacity(Array<T> *array, isize capacity) {
if (capacity == array->capacity) {
return;
}
if (capacity < array->count) {
array_resize(array, capacity);
}
T *new_data = NULL;
if (capacity > 0) {
new_data = gb_alloc_array(array->allocator, T, capacity);
gb_memmove(new_data, array->data, gb_size_of(T) * array->capacity);
}
gb_free(array->allocator, array->data);
array->data = new_data;
array->capacity = capacity;
}
#endif

1353
src/checker/checker.c Normal file

File diff suppressed because it is too large Load Diff

545
src/checker/decl.c Normal file
View File

@@ -0,0 +1,545 @@
bool check_is_terminating(AstNode *node);
void check_stmt (Checker *c, AstNode *node, u32 flags);
void check_stmt_list (Checker *c, AstNodeArray stmts, u32 flags);
void check_type_decl (Checker *c, Entity *e, AstNode *type_expr, Type *def, CycleChecker *cycle_checker);
void check_const_decl (Checker *c, Entity *e, AstNode *type_expr, AstNode *init_expr);
void check_proc_decl (Checker *c, Entity *e, DeclInfo *d);
void check_var_decl (Checker *c, Entity *e, Entity **entities, isize entity_count, AstNode *type_expr, AstNode *init_expr);
// NOTE(bill): `content_name` is for debugging and error messages
Type *check_init_variable(Checker *c, Entity *e, Operand *operand, String context_name) {
if (operand->mode == Addressing_Invalid ||
operand->type == t_invalid ||
e->type == t_invalid) {
if (operand->mode == Addressing_Builtin) {
gbString expr_str = expr_to_string(operand->expr);
// TODO(bill): is this a good enough error message?
error(ast_node_token(operand->expr),
"Cannot assign builtin procedure `%s` in %.*s",
expr_str,
LIT(context_name));
operand->mode = Addressing_Invalid;
gb_string_free(expr_str);
}
if (e->type == NULL) {
e->type = t_invalid;
}
return NULL;
}
if (e->type == NULL) {
// NOTE(bill): Use the type of the operand
Type *t = operand->type;
if (is_type_untyped(t)) {
if (t == t_invalid || is_type_untyped_nil(t)) {
error(e->token, "Use of untyped nil in %.*s", LIT(context_name));
e->type = t_invalid;
return NULL;
}
t = default_type(t);
}
e->type = t;
}
check_assignment(c, operand, e->type, context_name);
if (operand->mode == Addressing_Invalid) {
return NULL;
}
return e->type;
}
void check_init_variables(Checker *c, Entity **lhs, isize lhs_count, AstNodeArray inits, String context_name) {
if ((lhs == NULL || lhs_count == 0) && inits.count == 0) {
return;
}
gbTempArenaMemory tmp = gb_temp_arena_memory_begin(&c->tmp_arena);
// NOTE(bill): If there is a bad syntax error, rhs > lhs which would mean there would need to be
// an extra allocation
Array(Operand) operands;
array_init_reserve(&operands, c->tmp_allocator, 2*lhs_count);
for_array(i, inits) {
AstNode *rhs = inits.e[i];
Operand o = {0};
check_multi_expr(c, &o, rhs);
if (o.type->kind != Type_Tuple) {
array_add(&operands, o);
} else {
TypeTuple *tuple = &o.type->Tuple;
for (isize j = 0; j < tuple->variable_count; j++) {
o.type = tuple->variables[j]->type;
array_add(&operands, o);
}
}
}
isize rhs_count = operands.count;
for_array(i, operands) {
if (operands.e[i].mode == Addressing_Invalid) {
rhs_count--;
}
}
isize max = gb_min(lhs_count, rhs_count);
for (isize i = 0; i < max; i++) {
check_init_variable(c, lhs[i], &operands.e[i], context_name);
}
if (rhs_count > 0 && lhs_count != rhs_count) {
error(lhs[0]->token, "Assignment count mismatch `%td` := `%td`", lhs_count, rhs_count);
}
gb_temp_arena_memory_end(tmp);
}
void check_entity_decl(Checker *c, Entity *e, DeclInfo *d, Type *named_type, CycleChecker *cycle_checker) {
if (e->type != NULL) {
return;
}
if (d == NULL) {
DeclInfo **found = map_decl_info_get(&c->info.entities, hash_pointer(e));
if (found) {
d = *found;
} else {
e->type = t_invalid;
set_base_type(named_type, t_invalid);
return;
// GB_PANIC("`%.*s` should been declared!", LIT(e->token.string));
}
}
if (e->kind == Entity_Procedure) {
check_proc_decl(c, e, d);
return;
}
CheckerContext prev = c->context;
c->context.scope = d->scope;
c->context.decl = d;
switch (e->kind) {
case Entity_Constant:
check_const_decl(c, e, d->type_expr, d->init_expr);
break;
case Entity_Variable:
check_var_decl(c, e, d->entities, d->entity_count, d->type_expr, d->init_expr);
break;
case Entity_TypeName:
check_type_decl(c, e, d->type_expr, named_type, cycle_checker);
break;
}
c->context = prev;
}
void check_var_decl_node(Checker *c, AstNode *node) {
ast_node(vd, VarDecl, node);
isize entity_count = vd->names.count;
isize entity_index = 0;
Entity **entities = gb_alloc_array(c->allocator, Entity *, entity_count);
for_array(i, vd->names) {
AstNode *name = vd->names.e[i];
Entity *entity = NULL;
if (name->kind == AstNode_Ident) {
Token token = name->Ident;
String str = token.string;
Entity *found = NULL;
// NOTE(bill): Ignore assignments to `_`
if (str_ne(str, str_lit("_"))) {
found = current_scope_lookup_entity(c->context.scope, str);
}
if (found == NULL) {
entity = make_entity_variable(c->allocator, c->context.scope, token, NULL);
add_entity_definition(&c->info, name, entity);
} else {
TokenPos pos = found->token.pos;
error(token,
"Redeclaration of `%.*s` in this scope\n"
"\tat %.*s(%td:%td)",
LIT(str), LIT(pos.file), pos.line, pos.column);
entity = found;
}
} else {
error(ast_node_token(name), "A variable declaration must be an identifier");
}
if (entity == NULL) {
entity = make_entity_dummy_variable(c->allocator, c->global_scope, ast_node_token(name));
}
entities[entity_index++] = entity;
}
Type *init_type = NULL;
if (vd->type) {
init_type = check_type_extra(c, vd->type, NULL, NULL);
if (init_type == NULL)
init_type = t_invalid;
}
for (isize i = 0; i < entity_count; i++) {
Entity *e = entities[i];
GB_ASSERT(e != NULL);
if (e->flags & EntityFlag_Visited) {
e->type = t_invalid;
continue;
}
e->flags |= EntityFlag_Visited;
if (e->type == NULL)
e->type = init_type;
}
check_init_variables(c, entities, entity_count, vd->values, str_lit("variable declaration"));
for_array(i, vd->names) {
if (entities[i] != NULL) {
add_entity(c, c->context.scope, vd->names.e[i], entities[i]);
}
}
}
void check_init_constant(Checker *c, Entity *e, Operand *operand) {
if (operand->mode == Addressing_Invalid ||
operand->type == t_invalid ||
e->type == t_invalid) {
if (e->type == NULL) {
e->type = t_invalid;
}
return;
}
if (operand->mode != Addressing_Constant) {
// TODO(bill): better error
error(ast_node_token(operand->expr),
"`%.*s` is not a constant", LIT(ast_node_token(operand->expr).string));
if (e->type == NULL) {
e->type = t_invalid;
}
return;
}
// if (!is_type_constant_type(operand->type)) {
// gbString type_str = type_to_string(operand->type);
// defer (gb_string_free(type_str));
// error(ast_node_token(operand->expr),
// "Invalid constant type: `%s`", type_str);
// if (e->type == NULL) {
// e->type = t_invalid;
// }
// return;
// }
if (e->type == NULL) { // NOTE(bill): type inference
e->type = operand->type;
}
check_assignment(c, operand, e->type, str_lit("constant declaration"));
if (operand->mode == Addressing_Invalid) {
return;
}
e->Constant.value = operand->value;
}
void check_const_decl(Checker *c, Entity *e, AstNode *type_expr, AstNode *init_expr) {
GB_ASSERT(e->type == NULL);
if (e->flags & EntityFlag_Visited) {
e->type = t_invalid;
return;
}
e->flags |= EntityFlag_Visited;
if (type_expr) {
Type *t = check_type(c, type_expr);
// if (!is_type_constant_type(t)) {
// gbString str = type_to_string(t);
// defer (gb_string_free(str));
// error(ast_node_token(type_expr),
// "Invalid constant type `%s`", str);
// e->type = t_invalid;
// return;
// }
e->type = t;
}
Operand operand = {0};
if (init_expr) {
check_expr(c, &operand, init_expr);
}
check_init_constant(c, e, &operand);
}
void check_type_decl(Checker *c, Entity *e, AstNode *type_expr, Type *def, CycleChecker *cycle_checker) {
GB_ASSERT(e->type == NULL);
Type *named = make_type_named(c->allocator, e->token.string, NULL, e);
named->Named.type_name = e;
if (def != NULL && def->kind == Type_Named) {
def->Named.base = named;
}
e->type = named;
CycleChecker local_cycle_checker = {0};
if (cycle_checker == NULL) {
cycle_checker = &local_cycle_checker;
}
Type *bt = check_type_extra(c, type_expr, named, cycle_checker_add(cycle_checker, e));
named->Named.base = bt;
named->Named.base = base_type(named->Named.base);
if (named->Named.base == t_invalid) {
gb_printf("check_type_decl: %s\n", type_to_string(named));
}
cycle_checker_destroy(&local_cycle_checker);
}
bool are_signatures_similar_enough(Type *a_, Type *b_) {
GB_ASSERT(a_->kind == Type_Proc);
GB_ASSERT(b_->kind == Type_Proc);
TypeProc *a = &a_->Proc;
TypeProc *b = &b_->Proc;
if (a->param_count != b->param_count) {
return false;
}
if (a->result_count != b->result_count) {
return false;
}
for (isize i = 0; i < a->param_count; i++) {
Type *x = base_type(a->params->Tuple.variables[i]->type);
Type *y = base_type(b->params->Tuple.variables[i]->type);
if (is_type_pointer(x) && is_type_pointer(y)) {
continue;
}
if (!are_types_identical(x, y)) {
return false;
}
}
for (isize i = 0; i < a->result_count; i++) {
Type *x = base_type(a->results->Tuple.variables[i]->type);
Type *y = base_type(b->results->Tuple.variables[i]->type);
if (is_type_pointer(x) && is_type_pointer(y)) {
continue;
}
if (!are_types_identical(x, y)) {
return false;
}
}
return true;
}
void check_proc_decl(Checker *c, Entity *e, DeclInfo *d) {
GB_ASSERT(e->type == NULL);
Type *proc_type = make_type_proc(c->allocator, e->scope, NULL, 0, NULL, 0, false);
e->type = proc_type;
ast_node(pd, ProcDecl, d->proc_decl);
check_open_scope(c, pd->type);
check_procedure_type(c, proc_type, pd->type);
bool is_foreign = (pd->tags & ProcTag_foreign) != 0;
bool is_link_name = (pd->tags & ProcTag_link_name) != 0;
bool is_inline = (pd->tags & ProcTag_inline) != 0;
bool is_no_inline = (pd->tags & ProcTag_no_inline) != 0;
if ((d->scope->is_file || d->scope->is_global) &&
str_eq(e->token.string, str_lit("main"))) {
if (proc_type != NULL) {
TypeProc *pt = &proc_type->Proc;
if (pt->param_count != 0 ||
pt->result_count) {
gbString str = type_to_string(proc_type);
error(e->token,
"Procedure type of `main` was expected to be `proc()`, got %s", str);
gb_string_free(str);
}
}
}
if (is_inline && is_no_inline) {
error(ast_node_token(pd->type),
"You cannot apply both `inline` and `no_inline` to a procedure");
}
if (is_foreign && is_link_name) {
error(ast_node_token(pd->type),
"You cannot apply both `foreign` and `link_name` to a procedure");
}
if (pd->body != NULL) {
if (is_foreign) {
error(ast_node_token(pd->body),
"A procedure tagged as `#foreign` cannot have a body");
}
d->scope = c->context.scope;
GB_ASSERT(pd->body->kind == AstNode_BlockStmt);
check_procedure_later(c, c->curr_ast_file, e->token, d, proc_type, pd->body, pd->tags);
}
if (is_foreign) {
MapEntity *fp = &c->info.foreign_procs;
AstNodeProcDecl *proc_decl = &d->proc_decl->ProcDecl;
String name = proc_decl->name->Ident.string;
if (proc_decl->foreign_name.len > 0) {
name = proc_decl->foreign_name;
}
HashKey key = hash_string(name);
Entity **found = map_entity_get(fp, key);
if (found) {
Entity *f = *found;
TokenPos pos = f->token.pos;
Type *this_type = base_type(e->type);
Type *other_type = base_type(f->type);
if (!are_signatures_similar_enough(this_type, other_type)) {
error(ast_node_token(d->proc_decl),
"Redeclaration of #foreign procedure `%.*s` with different type signatures\n"
"\tat %.*s(%td:%td)",
LIT(name), LIT(pos.file), pos.line, pos.column);
}
} else {
map_entity_set(fp, key, e);
}
} else if (is_link_name) {
MapEntity *fp = &c->info.foreign_procs;
AstNodeProcDecl *proc_decl = &d->proc_decl->ProcDecl;
String name = proc_decl->link_name;
HashKey key = hash_string(name);
Entity **found = map_entity_get(fp, key);
if (found) {
Entity *f = *found;
TokenPos pos = f->token.pos;
error(ast_node_token(d->proc_decl),
"Non unique #link_name for procedure `%.*s`\n"
"\tother at %.*s(%td:%td)",
LIT(name), LIT(pos.file), pos.line, pos.column);
} else {
map_entity_set(fp, key, e);
}
}
check_close_scope(c);
}
void check_var_decl(Checker *c, Entity *e, Entity **entities, isize entity_count, AstNode *type_expr, AstNode *init_expr) {
GB_ASSERT(e->type == NULL);
GB_ASSERT(e->kind == Entity_Variable);
if (e->flags & EntityFlag_Visited) {
e->type = t_invalid;
return;
}
e->flags |= EntityFlag_Visited;
if (type_expr != NULL)
e->type = check_type_extra(c, type_expr, NULL, NULL);
if (init_expr == NULL) {
if (type_expr == NULL)
e->type = t_invalid;
return;
}
if (entities == NULL || entity_count == 1) {
GB_ASSERT(entities == NULL || entities[0] == e);
Operand operand = {0};
check_expr(c, &operand, init_expr);
check_init_variable(c, e, &operand, str_lit("variable declaration"));
}
if (type_expr != NULL) {
for (isize i = 0; i < entity_count; i++)
entities[i]->type = e->type;
}
AstNodeArray inits;
array_init_reserve(&inits, c->allocator, 1);
array_add(&inits, init_expr);
check_init_variables(c, entities, entity_count, inits, str_lit("variable declaration"));
}
void check_proc_body(Checker *c, Token token, DeclInfo *decl, Type *type, AstNode *body) {
GB_ASSERT(body->kind == AstNode_BlockStmt);
CheckerContext old_context = c->context;
c->context.scope = decl->scope;
c->context.decl = decl;
GB_ASSERT(type->kind == Type_Proc);
if (type->Proc.param_count > 0) {
TypeTuple *params = &type->Proc.params->Tuple;
for (isize i = 0; i < params->variable_count; i++) {
Entity *e = params->variables[i];
GB_ASSERT(e->kind == Entity_Variable);
if (!(e->flags & EntityFlag_Anonymous)) {
continue;
}
String name = e->token.string;
Type *t = base_type(type_deref(e->type));
if (is_type_struct(t) || is_type_raw_union(t)) {
Scope **found = map_scope_get(&c->info.scopes, hash_pointer(t->Record.node));
GB_ASSERT(found != NULL);
for_array(i, (*found)->elements.entries) {
Entity *f = (*found)->elements.entries.e[i].value;
if (f->kind == Entity_Variable) {
Entity *uvar = make_entity_using_variable(c->allocator, e, f->token, f->type);
Entity *prev = scope_insert_entity(c->context.scope, uvar);
if (prev != NULL) {
error(e->token, "Namespace collision while `using` `%.*s` of: %.*s", LIT(name), LIT(prev->token.string));
break;
}
}
}
} else {
error(e->token, "`using` can only be applied to variables of type struct or raw_union");
break;
}
}
}
push_procedure(c, type);
{
ast_node(bs, BlockStmt, body);
// TODO(bill): Check declarations first (except mutable variable declarations)
check_stmt_list(c, bs->stmts, 0);
if (type->Proc.result_count > 0) {
if (!check_is_terminating(body)) {
error(bs->close, "Missing return statement at the end of the procedure");
}
}
}
pop_procedure(c);
check_scope_usage(c, c->context.scope);
c->context = old_context;
}

179
src/checker/entity.c Normal file
View File

@@ -0,0 +1,179 @@
typedef struct Scope Scope;
typedef struct Checker Checker;
typedef struct Type Type;
typedef enum BuiltinProcId BuiltinProcId;
typedef enum ImplicitValueId ImplicitValueId;
#define ENTITY_KINDS \
ENTITY_KIND(Invalid) \
ENTITY_KIND(Constant) \
ENTITY_KIND(Variable) \
ENTITY_KIND(TypeName) \
ENTITY_KIND(Procedure) \
ENTITY_KIND(Builtin) \
ENTITY_KIND(ImportName) \
ENTITY_KIND(Nil) \
ENTITY_KIND(ImplicitValue) \
ENTITY_KIND(Count)
typedef enum EntityKind {
#define ENTITY_KIND(k) GB_JOIN2(Entity_, k),
ENTITY_KINDS
#undef ENTITY_KIND
} EntityKind;
String const entity_strings[] = {
#define ENTITY_KIND(k) {cast(u8 *)#k, gb_size_of(#k)-1},
ENTITY_KINDS
#undef ENTITY_KIND
};
typedef enum EntityFlag {
EntityFlag_Visited = 1<<0,
EntityFlag_Used = 1<<1,
EntityFlag_Anonymous = 1<<2,
EntityFlag_Field = 1<<3,
EntityFlag_Param = 1<<4,
EntityFlag_VectorElem = 1<<5,
} EntityFlag;
typedef struct Entity Entity;
struct Entity {
EntityKind kind;
u32 flags;
Token token;
Scope * scope;
Type * type;
AstNode * identifier; // Can be NULL
// TODO(bill): Cleanup how `using` works for entities
Entity * using_parent;
AstNode * using_expr;
union {
struct {
ExactValue value;
} Constant;
struct {
i32 field_index;
i32 field_src_index;
} Variable;
i32 TypeName;
i32 Procedure;
struct {
BuiltinProcId id;
} Builtin;
struct {
String path;
String name;
Scope *scope;
bool used;
} ImportName;
i32 Nil;
struct {
// TODO(bill): Should this be a user-level construct rather than compiler-level?
ImplicitValueId id;
Entity * backing;
} ImplicitValue;
};
};
Entity *alloc_entity(gbAllocator a, EntityKind kind, Scope *scope, Token token, Type *type) {
Entity *entity = gb_alloc_item(a, Entity);
entity->kind = kind;
entity->scope = scope;
entity->token = token;
entity->type = type;
return entity;
}
Entity *make_entity_variable(gbAllocator a, Scope *scope, Token token, Type *type) {
Entity *entity = alloc_entity(a, Entity_Variable, scope, token, type);
return entity;
}
Entity *make_entity_using_variable(gbAllocator a, Entity *parent, Token token, Type *type) {
GB_ASSERT(parent != NULL);
Entity *entity = alloc_entity(a, Entity_Variable, parent->scope, token, type);
entity->using_parent = parent;
entity->flags |= EntityFlag_Anonymous;
return entity;
}
Entity *make_entity_constant(gbAllocator a, Scope *scope, Token token, Type *type, ExactValue value) {
Entity *entity = alloc_entity(a, Entity_Constant, scope, token, type);
entity->Constant.value = value;
return entity;
}
Entity *make_entity_type_name(gbAllocator a, Scope *scope, Token token, Type *type) {
Entity *entity = alloc_entity(a, Entity_TypeName, scope, token, type);
return entity;
}
Entity *make_entity_param(gbAllocator a, Scope *scope, Token token, Type *type, bool anonymous) {
Entity *entity = make_entity_variable(a, scope, token, type);
entity->flags |= EntityFlag_Used;
entity->flags |= EntityFlag_Anonymous*(anonymous != 0);
entity->flags |= EntityFlag_Param;
return entity;
}
Entity *make_entity_field(gbAllocator a, Scope *scope, Token token, Type *type, bool anonymous, i32 field_src_index) {
Entity *entity = make_entity_variable(a, scope, token, type);
entity->Variable.field_src_index = field_src_index;
entity->Variable.field_index = field_src_index;
entity->flags |= EntityFlag_Field;
entity->flags |= EntityFlag_Anonymous*(anonymous != 0);
return entity;
}
Entity *make_entity_vector_elem(gbAllocator a, Scope *scope, Token token, Type *type, i32 field_src_index) {
Entity *entity = make_entity_variable(a, scope, token, type);
entity->Variable.field_src_index = field_src_index;
entity->Variable.field_index = field_src_index;
entity->flags |= EntityFlag_Field;
entity->flags |= EntityFlag_VectorElem;
return entity;
}
Entity *make_entity_procedure(gbAllocator a, Scope *scope, Token token, Type *signature_type) {
Entity *entity = alloc_entity(a, Entity_Procedure, scope, token, signature_type);
return entity;
}
Entity *make_entity_builtin(gbAllocator a, Scope *scope, Token token, Type *type, BuiltinProcId id) {
Entity *entity = alloc_entity(a, Entity_Builtin, scope, token, type);
entity->Builtin.id = id;
return entity;
}
Entity *make_entity_import_name(gbAllocator a, Scope *scope, Token token, Type *type,
String path, String name, Scope *import_scope) {
Entity *entity = alloc_entity(a, Entity_ImportName, scope, token, type);
entity->ImportName.path = path;
entity->ImportName.name = name;
entity->ImportName.scope = import_scope;
return entity;
}
Entity *make_entity_nil(gbAllocator a, String name, Type *type) {
Token token = make_token_ident(name);
Entity *entity = alloc_entity(a, Entity_Nil, NULL, token, type);
return entity;
}
Entity *make_entity_implicit_value(gbAllocator a, String name, Type *type, ImplicitValueId id) {
Token token = make_token_ident(name);
Entity *entity = alloc_entity(a, Entity_ImplicitValue, NULL, token, type);
entity->ImplicitValue.id = id;
return entity;
}
Entity *make_entity_dummy_variable(gbAllocator a, Scope *file_scope, Token token) {
token.string = str_lit("_");
return make_entity_variable(a, file_scope, token, NULL);
}

4465
src/checker/expr.c Normal file

File diff suppressed because it is too large Load Diff

1130
src/checker/stmt.c Normal file

File diff suppressed because it is too large Load Diff

1487
src/checker/types.c Normal file

File diff suppressed because it is too large Load Diff

195
src/common.c Normal file
View File

@@ -0,0 +1,195 @@
#define GB_NO_DEFER
#define GB_IMPLEMENTATION
#include "gb/gb.h"
gbAllocator heap_allocator(void) {
return gb_heap_allocator();
}
#include "string.c"
#include "array.c"
gb_global String global_module_path = {0};
gb_global bool global_module_path_set = false;
String get_module_dir() {
if (global_module_path_set) {
return global_module_path;
}
Array(wchar_t) path_buf;
array_init_count(&path_buf, heap_allocator(), 300);
isize len = 0;
for (;;) {
len = GetModuleFileNameW(NULL, &path_buf.e[0], path_buf.count);
if (len == 0) {
return make_string(NULL, 0);
}
if (len < path_buf.count) {
break;
}
array_resize(&path_buf, 2*path_buf.count + 300);
}
gbTempArenaMemory tmp = gb_temp_arena_memory_begin(&string_buffer_arena);
wchar_t *text = gb_alloc_array(string_buffer_allocator, wchar_t, len+1);
GetModuleFileNameW(NULL, text, len);
String path = string16_to_string(heap_allocator(), make_string16(text, len));
for (isize i = path.len-1; i >= 0; i--) {
u8 c = path.text[i];
if (c == '/' || c == '\\') {
break;
}
path.len--;
}
global_module_path = path;
global_module_path_set = true;
gb_temp_arena_memory_end(tmp);
array_free(&path_buf);
return path;
}
String path_to_fullpath(gbAllocator a, String s) {
gbTempArenaMemory tmp = gb_temp_arena_memory_begin(&string_buffer_arena);
String16 string16 = string_to_string16(string_buffer_allocator, s);
String result = {0};
DWORD len = GetFullPathNameW(string16.text, 0, NULL, NULL);
if (len != 0) {
wchar_t *text = gb_alloc_array(string_buffer_allocator, wchar_t, len+1);
GetFullPathNameW(string16.text, len, text, NULL);
text[len] = 0;
result = string16_to_string(a, make_string16(text, len));
}
gb_temp_arena_memory_end(tmp);
return result;
}
i64 next_pow2(i64 n) {
if (n <= 0) {
return 0;
}
n--;
n |= n >> 1;
n |= n >> 2;
n |= n >> 4;
n |= n >> 8;
n |= n >> 16;
n |= n >> 32;
n++;
return n;
}
i64 prev_pow2(i64 n) {
if (n <= 0) {
return 0;
}
n |= n >> 1;
n |= n >> 2;
n |= n >> 4;
n |= n >> 8;
n |= n >> 16;
n |= n >> 32;
return n - (n >> 1);
}
i16 f32_to_f16(f32 value) {
union { u32 i; f32 f; } v;
i32 i, s, e, m;
v.f = value;
i = (i32)v.i;
s = (i >> 16) & 0x00008000;
e = ((i >> 23) & 0x000000ff) - (127 - 15);
m = i & 0x007fffff;
if (e <= 0) {
if (e < -10) return cast(i16)s;
m = (m | 0x00800000) >> (1 - e);
if (m & 0x00001000)
m += 0x00002000;
return cast(i16)(s | (m >> 13));
} else if (e == 0xff - (127 - 15)) {
if (m == 0) {
return cast(i16)(s | 0x7c00); /* NOTE(bill): infinity */
} else {
/* NOTE(bill): NAN */
m >>= 13;
return cast(i16)(s | 0x7c00 | m | (m == 0));
}
} else {
if (m & 0x00001000) {
m += 0x00002000;
if (m & 0x00800000) {
m = 0;
e += 1;
}
}
if (e > 30) {
float volatile f = 1e12f;
int j;
for (j = 0; j < 10; j++)
f *= f; /* NOTE(bill): Cause overflow */
return cast(i16)(s | 0x7c00);
}
return cast(i16)(s | (e << 10) | (m >> 13));
}
}
#define for_array(index_, array_) for (isize index_ = 0; index_ < (array_).count; index_++)
// Doubly Linked Lists
#define DLIST_SET(curr_element, next_element) do { \
(curr_element)->next = (next_element); \
(curr_element)->next->prev = (curr_element); \
(curr_element) = (curr_element)->next; \
} while (0)
#define DLIST_APPEND(root_element, curr_element, next_element) do { \
if ((root_element) == NULL) { \
(root_element) = (curr_element) = (next_element); \
} else { \
DLIST_SET(curr_element, next_element); \
} \
} while (0)
////////////////////////////////////////////////////////////////
//
// Generic Data Structures
//
////////////////////////////////////////////////////////////////
#define MAP_TYPE String
#define MAP_PROC map_string_
#define MAP_NAME MapString
#include "map.c"
#define MAP_TYPE bool
#define MAP_PROC map_bool_
#define MAP_NAME MapBool
#include "map.c"
#define MAP_TYPE isize
#define MAP_PROC map_isize_
#define MAP_NAME MapIsize
#include "map.c"

400
src/exact_value.c Normal file
View File

@@ -0,0 +1,400 @@
#include <math.h>
// TODO(bill): Big numbers
// IMPORTANT TODO(bill): This needs to be completely fixed!!!!!!!!
typedef struct AstNode AstNode;
typedef enum ExactValueKind {
ExactValue_Invalid,
ExactValue_Bool,
ExactValue_String,
ExactValue_Integer,
ExactValue_Float,
ExactValue_Pointer,
ExactValue_Compound, // TODO(bill): Is this good enough?
ExactValue_Count,
} ExactValueKind;
typedef struct ExactValue {
ExactValueKind kind;
union {
bool value_bool;
String value_string;
i64 value_integer; // NOTE(bill): This must be an integer and not a pointer
f64 value_float;
i64 value_pointer;
AstNode *value_compound;
};
} ExactValue;
HashKey hash_exact_value(ExactValue v) {
return hashing_proc(&v, gb_size_of(ExactValue));
}
ExactValue make_exact_value_compound(AstNode *node) {
ExactValue result = {ExactValue_Compound};
result.value_compound = node;
return result;
}
ExactValue make_exact_value_bool(bool b) {
ExactValue result = {ExactValue_Bool};
result.value_bool = (b != 0);
return result;
}
ExactValue make_exact_value_string(String string) {
// TODO(bill): Allow for numbers with underscores in them
ExactValue result = {ExactValue_String};
result.value_string = string;
return result;
}
ExactValue make_exact_value_integer_from_string(String string) {
// TODO(bill): Allow for numbers with underscores in them
ExactValue result = {ExactValue_Integer};
i32 base = 10;
if (string.text[0] == '0') {
switch (string.text[1]) {
case 'b': base = 2; break;
case 'o': base = 8; break;
case 'd': base = 10; break;
case 'x': base = 16; break;
}
}
result.value_integer = gb_str_to_i64(cast(char *)string.text, NULL, base);
return result;
}
ExactValue make_exact_value_integer(i64 i) {
ExactValue result = {ExactValue_Integer};
result.value_integer = i;
return result;
}
ExactValue make_exact_value_float_from_string(String string) {
// TODO(bill): Allow for numbers with underscores in them
ExactValue result = {ExactValue_Float};
result.value_float = gb_str_to_f64(cast(char *)string.text, NULL);
return result;
}
ExactValue make_exact_value_float(f64 f) {
ExactValue result = {ExactValue_Float};
result.value_float = f;
return result;
}
ExactValue make_exact_value_pointer(i64 ptr) {
ExactValue result = {ExactValue_Pointer};
result.value_pointer = ptr;
return result;
}
ExactValue make_exact_value_from_basic_literal(Token token) {
switch (token.kind) {
case Token_String: return make_exact_value_string(token.string);
case Token_Integer: return make_exact_value_integer_from_string(token.string);
case Token_Float: return make_exact_value_float_from_string(token.string);
case Token_Rune: {
Rune r = GB_RUNE_INVALID;
gb_utf8_decode(token.string.text, token.string.len, &r);
// gb_printf("%.*s rune: %d\n", LIT(token.string), r);
return make_exact_value_integer(r);
}
default:
GB_PANIC("Invalid token for basic literal");
break;
}
ExactValue result = {ExactValue_Invalid};
return result;
}
ExactValue exact_value_to_integer(ExactValue v) {
switch (v.kind) {
case ExactValue_Integer:
return v;
case ExactValue_Float: {
i64 i = cast(i64)v.value_float;
f64 f = cast(f64)i;
if (f == v.value_float) {
return make_exact_value_integer(i);
}
} break;
case ExactValue_Pointer:
return make_exact_value_integer(cast(i64)cast(intptr)v.value_pointer);
}
ExactValue r = {ExactValue_Invalid};
return r;
}
ExactValue exact_value_to_float(ExactValue v) {
switch (v.kind) {
case ExactValue_Integer:
return make_exact_value_float(cast(i64)v.value_integer);
case ExactValue_Float:
return v;
}
ExactValue r = {ExactValue_Invalid};
return r;
}
ExactValue exact_unary_operator_value(Token op, ExactValue v, i32 precision) {
switch (op.kind) {
case Token_Add: {
switch (v.kind) {
case ExactValue_Invalid:
case ExactValue_Integer:
case ExactValue_Float:
return v;
}
} break;
case Token_Sub: {
switch (v.kind) {
case ExactValue_Invalid:
return v;
case ExactValue_Integer: {
ExactValue i = v;
i.value_integer = -i.value_integer;
return i;
}
case ExactValue_Float: {
ExactValue i = v;
i.value_float = -i.value_float;
return i;
}
}
} break;
case Token_Xor: {
i64 i = 0;
switch (v.kind) {
case ExactValue_Invalid:
return v;
case ExactValue_Integer:
i = v.value_integer;
i = ~i;
break;
default:
goto failure;
}
// NOTE(bill): unsigned integers will be negative and will need to be
// limited to the types precision
if (precision > 0)
i &= ~((~0ll)<<precision);
return make_exact_value_integer(i);
} break;
case Token_Not: {
switch (v.kind) {
case ExactValue_Invalid: return v;
case ExactValue_Bool:
return make_exact_value_bool(!v.value_bool);
}
} break;
}
failure:
GB_PANIC("Invalid unary operation, %.*s", LIT(token_strings[op.kind]));
ExactValue error_value = {0};
return error_value;
}
// NOTE(bill): Make sure things are evaluated in correct order
i32 exact_value_order(ExactValue v) {
switch (v.kind) {
case ExactValue_Invalid:
return 0;
case ExactValue_Bool:
case ExactValue_String:
return 1;
case ExactValue_Integer:
return 2;
case ExactValue_Float:
return 3;
case ExactValue_Pointer:
return 4;
default:
GB_PANIC("How'd you get here? Invalid Value.kind");
return -1;
}
}
void match_exact_values(ExactValue *x, ExactValue *y) {
if (exact_value_order(*y) < exact_value_order(*x)) {
match_exact_values(y, x);
return;
}
switch (x->kind) {
case ExactValue_Invalid:
*y = *x;
return;
case ExactValue_Bool:
case ExactValue_String:
return;
case ExactValue_Integer:
switch (y->kind) {
case ExactValue_Integer:
return;
case ExactValue_Float:
// TODO(bill): Is this good enough?
*x = make_exact_value_float(cast(f64)x->value_integer);
return;
}
break;
case ExactValue_Float:
if (y->kind == ExactValue_Float)
return;
break;
}
compiler_error("How'd you get here? Invalid ExactValueKind");
}
// TODO(bill): Allow for pointer arithmetic? Or are pointer slices good enough?
ExactValue exact_binary_operator_value(Token op, ExactValue x, ExactValue y) {
match_exact_values(&x, &y);
switch (x.kind) {
case ExactValue_Invalid:
return x;
case ExactValue_Bool:
switch (op.kind) {
case Token_CmpAnd: return make_exact_value_bool(x.value_bool && y.value_bool);
case Token_CmpOr: return make_exact_value_bool(x.value_bool || y.value_bool);
case Token_And: return make_exact_value_bool(x.value_bool & y.value_bool);
case Token_Or: return make_exact_value_bool(x.value_bool | y.value_bool);
default: goto error;
}
break;
case ExactValue_Integer: {
i64 a = x.value_integer;
i64 b = y.value_integer;
i64 c = 0;
switch (op.kind) {
case Token_Add: c = a + b; break;
case Token_Sub: c = a - b; break;
case Token_Mul: c = a * b; break;
case Token_Quo: return make_exact_value_float(fmod(cast(f64)a, cast(f64)b));
case Token_QuoEq: c = a / b; break; // NOTE(bill): Integer division
case Token_Mod: c = a % b; break;
case Token_And: c = a & b; break;
case Token_Or: c = a | b; break;
case Token_Xor: c = a ^ b; break;
case Token_AndNot: c = a&(~b); break;
case Token_Shl: c = a << b; break;
case Token_Shr: c = a >> b; break;
default: goto error;
}
return make_exact_value_integer(c);
} break;
case ExactValue_Float: {
f64 a = x.value_float;
f64 b = y.value_float;
switch (op.kind) {
case Token_Add: return make_exact_value_float(a + b);
case Token_Sub: return make_exact_value_float(a - b);
case Token_Mul: return make_exact_value_float(a * b);
case Token_Quo: return make_exact_value_float(a / b);
default: goto error;
}
} break;
}
error:
ExactValue error_value = {0};
// gb_printf_err("Invalid binary operation: %s\n", token_kind_to_string(op.kind));
return error_value;
}
gb_inline ExactValue exact_value_add(ExactValue x, ExactValue y) { Token op = {Token_Add}; return exact_binary_operator_value(op, x, y); }
gb_inline ExactValue exact_value_sub(ExactValue x, ExactValue y) { Token op = {Token_Sub}; return exact_binary_operator_value(op, x, y); }
gb_inline ExactValue exact_value_mul(ExactValue x, ExactValue y) { Token op = {Token_Mul}; return exact_binary_operator_value(op, x, y); }
gb_inline ExactValue exact_value_quo(ExactValue x, ExactValue y) { Token op = {Token_Quo}; return exact_binary_operator_value(op, x, y); }
gb_inline ExactValue exact_value_shift(Token op, ExactValue x, ExactValue y) { return exact_binary_operator_value(op, x, y); }
i32 cmp_f64(f64 a, f64 b) {
return (a > b) - (a < b);
}
bool compare_exact_values(Token op, ExactValue x, ExactValue y) {
match_exact_values(&x, &y);
switch (x.kind) {
case ExactValue_Invalid:
return false;
case ExactValue_Bool:
switch (op.kind) {
case Token_CmpEq: return x.value_bool == y.value_bool;
case Token_NotEq: return x.value_bool != y.value_bool;
}
break;
case ExactValue_Integer: {
i64 a = x.value_integer;
i64 b = y.value_integer;
switch (op.kind) {
case Token_CmpEq: return a == b;
case Token_NotEq: return a != b;
case Token_Lt: return a < b;
case Token_LtEq: return a <= b;
case Token_Gt: return a > b;
case Token_GtEq: return a >= b;
}
} break;
case ExactValue_Float: {
f64 a = x.value_float;
f64 b = y.value_float;
switch (op.kind) {
case Token_CmpEq: return cmp_f64(a, b) == 0;
case Token_NotEq: return cmp_f64(a, b) != 0;
case Token_Lt: return cmp_f64(a, b) < 0;
case Token_LtEq: return cmp_f64(a, b) <= 0;
case Token_Gt: return cmp_f64(a, b) > 0;
case Token_GtEq: return cmp_f64(a, b) >= 0;
}
} break;
case ExactValue_String: {
String a = x.value_string;
String b = y.value_string;
isize len = gb_min(a.len, b.len);
// TODO(bill): gb_memcompare is used because the strings are UTF-8
switch (op.kind) {
case Token_CmpEq: return gb_memcompare(a.text, b.text, len) == 0;
case Token_NotEq: return gb_memcompare(a.text, b.text, len) != 0;
case Token_Lt: return gb_memcompare(a.text, b.text, len) < 0;
case Token_LtEq: return gb_memcompare(a.text, b.text, len) <= 0;
case Token_Gt: return gb_memcompare(a.text, b.text, len) > 0;
case Token_GtEq: return gb_memcompare(a.text, b.text, len) >= 0;
}
} break;
}
GB_PANIC("Invalid comparison");
return false;
}

272
src/main.c Normal file
View File

@@ -0,0 +1,272 @@
#if defined(__cplusplus)
extern "C" {
#endif
#define VERSION_STRING "v0.0.3c"
#include "common.c"
#include "timings.c"
#include "unicode.c"
#include "tokenizer.c"
#include "parser.c"
// #include "printer.c"
#include "checker/checker.c"
#include "ssa.c"
#include "ssa_opt.c"
#include "ssa_print.c"
// #include "vm.c"
// NOTE(bill): `name` is used in debugging and profiling modes
i32 win32_exec_command_line_app(char *name, char *fmt, ...) {
STARTUPINFOW start_info = {gb_size_of(STARTUPINFOW)};
PROCESS_INFORMATION pi = {0};
char cmd_line[4096] = {0};
isize cmd_len;
va_list va;
gbTempArenaMemory tmp;
String16 cmd;
i32 exit_code = 0;
start_info.dwFlags = STARTF_USESTDHANDLES | STARTF_USESHOWWINDOW;
start_info.wShowWindow = SW_SHOW;
start_info.hStdInput = GetStdHandle(STD_INPUT_HANDLE);
start_info.hStdOutput = GetStdHandle(STD_OUTPUT_HANDLE);
start_info.hStdError = GetStdHandle(STD_ERROR_HANDLE);
va_start(va, fmt);
cmd_len = gb_snprintf_va(cmd_line, gb_size_of(cmd_line), fmt, va);
va_end(va);
// gb_printf("%.*s\n", cast(int)cmd_len, cmd_line);
tmp = gb_temp_arena_memory_begin(&string_buffer_arena);
cmd = string_to_string16(string_buffer_allocator, make_string(cast(u8 *)cmd_line, cmd_len-1));
if (CreateProcessW(NULL, cmd.text,
NULL, NULL, true, 0, NULL, NULL,
&start_info, &pi)) {
WaitForSingleObject(pi.hProcess, INFINITE);
GetExitCodeProcess(pi.hProcess, cast(DWORD *)&exit_code);
CloseHandle(pi.hProcess);
CloseHandle(pi.hThread);
} else {
// NOTE(bill): failed to create process
gb_printf_err("Failed to execute command:\n\t%s\n", cmd_line);
exit_code = -1;
}
gb_temp_arena_memory_end(tmp);
return exit_code;
}
typedef enum ArchKind {
ArchKind_x64,
ArchKind_x86,
} ArchKind;
typedef struct ArchData {
BaseTypeSizes sizes;
String llc_flags;
String link_flags;
} ArchData;
ArchData make_arch_data(ArchKind kind) {
ArchData data = {0};
switch (kind) {
case ArchKind_x64:
default:
data.sizes.word_size = 8;
data.sizes.max_align = 16;
data.llc_flags = str_lit("-march=x86-64 ");
data.link_flags = str_lit("/machine:x64 ");
break;
case ArchKind_x86:
data.sizes.word_size = 4;
data.sizes.max_align = 8;
data.llc_flags = str_lit("-march=x86 ");
data.link_flags = str_lit("/machine:x86 ");
break;
}
return data;
}
void usage(char *argv0) {
gb_printf_err("%s is a tool for managing Odin source code\n", argv0);
gb_printf_err("Usage:");
gb_printf_err("\n\t%s command [arguments]\n", argv0);
gb_printf_err("Commands:");
gb_printf_err("\n\tbuild compile .odin file");
gb_printf_err("\n\trun compile and run .odin file");
gb_printf_err("\n\tversion print Odin version");
gb_printf_err("\n\n");
}
int main(int argc, char **argv) {
if (argc < 2) {
usage(argv[0]);
return 1;
}
Timings timings = {0};
timings_init(&timings, str_lit("Total Time"), 128);
// defer (timings_destroy(&timings));
#if 1
init_string_buffer_memory();
init_global_error_collector();
String module_dir = get_module_dir();
init_universal_scope();
char *init_filename = NULL;
bool run_output = false;
String arg1 = make_string_c(argv[1]);
if (str_eq(arg1, str_lit("run"))) {
run_output = true;
init_filename = argv[2];
} else if (str_eq(arg1, str_lit("build"))) {
init_filename = argv[2];
} else if (str_eq(arg1, str_lit("version"))) {
gb_printf("%s version %s", argv[0], VERSION_STRING);
return 0;
} else {
usage(argv[0]);
return 1;
}
// TODO(bill): prevent compiling without a linker
timings_start_section(&timings, str_lit("parse files"));
Parser parser = {0};
if (!init_parser(&parser)) {
return 1;
}
// defer (destroy_parser(&parser));
if (parse_files(&parser, init_filename) != ParseFile_None) {
return 1;
}
#if 1
timings_start_section(&timings, str_lit("type check"));
Checker checker = {0};
ArchData arch_data = make_arch_data(ArchKind_x64);
init_checker(&checker, &parser, arch_data.sizes);
// defer (destroy_checker(&checker));
check_parsed_files(&checker);
#endif
#if 1
ssaGen ssa = {0};
if (!ssa_gen_init(&ssa, &checker)) {
return 1;
}
// defer (ssa_gen_destroy(&ssa));
timings_start_section(&timings, str_lit("ssa gen"));
ssa_gen_tree(&ssa);
timings_start_section(&timings, str_lit("ssa opt"));
ssa_opt_tree(&ssa);
timings_start_section(&timings, str_lit("ssa print"));
ssa_print_llvm_ir(&ssa);
// prof_print_all();
#if 1
timings_start_section(&timings, str_lit("llvm-opt"));
char const *output_name = ssa.output_file.filename;
isize base_name_len = gb_path_extension(output_name)-1 - output_name;
String output = make_string(cast(u8 *)output_name, base_name_len);
i32 optimization_level = 0;
optimization_level = gb_clamp(optimization_level, 0, 3);
i32 exit_code = 0;
// For more passes arguments: http://llvm.org/docs/Passes.html
exit_code = win32_exec_command_line_app("llvm-opt",
"%.*sbin/opt %s -o %.*s.bc "
"-mem2reg "
"-memcpyopt "
"-die "
// "-dse "
// "-dce "
// "-S "
"",
LIT(module_dir),
output_name, LIT(output));
if (exit_code != 0) {
return exit_code;
}
#if 1
timings_start_section(&timings, str_lit("llvm-llc"));
// For more arguments: http://llvm.org/docs/CommandGuide/llc.html
exit_code = win32_exec_command_line_app("llvm-llc",
"%.*sbin/llc %.*s.bc -filetype=obj -O%d "
"%.*s "
// "-debug-pass=Arguments "
"",
LIT(module_dir),
LIT(output),
optimization_level,
LIT(arch_data.llc_flags));
if (exit_code != 0) {
return exit_code;
}
timings_start_section(&timings, str_lit("msvc-link"));
gbString lib_str = gb_string_make(heap_allocator(), "Kernel32.lib");
// defer (gb_string_free(lib_str));
char lib_str_buf[1024] = {0};
for_array(i, parser.foreign_libraries) {
String lib = parser.foreign_libraries.e[i];
isize len = gb_snprintf(lib_str_buf, gb_size_of(lib_str_buf),
" %.*s.lib", LIT(lib));
lib_str = gb_string_appendc(lib_str, lib_str_buf);
}
exit_code = win32_exec_command_line_app("msvc-link",
"link %.*s.obj -OUT:%.*s.exe %s "
"/defaultlib:libcmt "
"/nologo /incremental:no /opt:ref /subsystem:console "
" %.*s "
"",
LIT(output), LIT(output),
lib_str, LIT(arch_data.link_flags));
if (exit_code != 0) {
return exit_code;
}
// timings_print_all(&timings);
if (run_output) {
win32_exec_command_line_app("odin run",
"%.*s.exe", cast(int)base_name_len, output_name);
}
#endif
#endif
#endif
#endif
return 0;
}
#if defined(__cplusplus)
}
#endif

1305
src/old_vm.c Normal file

File diff suppressed because it is too large Load Diff

3250
src/parser.c Normal file

File diff suppressed because it is too large Load Diff

221
src/printer.c Normal file
View File

@@ -0,0 +1,221 @@
gb_inline void print_indent(isize indent) {
while (indent --> 0)
gb_printf(" ");
}
void print_ast(AstNode *node, isize indent) {
if (node == NULL)
return;
switch (node->kind) {
case AstNode_BasicLit:
print_indent(indent);
print_token(node->BasicLit);
break;
case AstNode_Ident:
print_indent(indent);
print_token(node->Ident);
break;
case AstNode_ProcLit:
print_indent(indent);
gb_printf("(proc lit)\n");
print_ast(node->ProcLit.type, indent+1);
print_ast(node->ProcLit.body, indent+1);
break;
case AstNode_CompoundLit:
print_indent(indent);
gb_printf("(compound lit)\n");
print_ast(node->CompoundLit.type, indent+1);
for_array(i, node->CompoundLit.elems) {
print_ast(node->CompoundLit.elems[i], indent+1);
}
break;
case AstNode_TagExpr:
print_indent(indent);
gb_printf("(tag)\n");
print_indent(indent+1);
print_token(node->TagExpr.name);
print_ast(node->TagExpr.expr, indent+1);
break;
case AstNode_UnaryExpr:
print_indent(indent);
print_token(node->UnaryExpr.op);
print_ast(node->UnaryExpr.expr, indent+1);
break;
case AstNode_BinaryExpr:
print_indent(indent);
print_token(node->BinaryExpr.op);
print_ast(node->BinaryExpr.left, indent+1);
print_ast(node->BinaryExpr.right, indent+1);
break;
case AstNode_CallExpr:
print_indent(indent);
gb_printf("(call)\n");
print_ast(node->CallExpr.proc, indent+1);
for_array(i, node->CallExpr.args) {
print_ast(node->CallExpr.args[i], indent+1);
}
break;
case AstNode_SelectorExpr:
print_indent(indent);
gb_printf(".\n");
print_ast(node->SelectorExpr.expr, indent+1);
print_ast(node->SelectorExpr.selector, indent+1);
break;
case AstNode_IndexExpr:
print_indent(indent);
gb_printf("([])\n");
print_ast(node->IndexExpr.expr, indent+1);
print_ast(node->IndexExpr.index, indent+1);
break;
case AstNode_DerefExpr:
print_indent(indent);
gb_printf("(deref)\n");
print_ast(node->DerefExpr.expr, indent+1);
break;
case AstNode_ExprStmt:
print_ast(node->ExprStmt.expr, indent);
break;
case AstNode_IncDecStmt:
print_indent(indent);
print_token(node->IncDecStmt.op);
print_ast(node->IncDecStmt.expr, indent+1);
break;
case AstNode_AssignStmt:
print_indent(indent);
print_token(node->AssignStmt.op);
for_array(i, node->AssignStmt.lhs) {
print_ast(node->AssignStmt.lhs[i], indent+1);
}
for_array(i, node->AssignStmt.rhs) {
print_ast(node->AssignStmt.rhs[i], indent+1);
}
break;
case AstNode_BlockStmt:
print_indent(indent);
gb_printf("(block)\n");
for_array(i, node->BlockStmt.stmts) {
print_ast(node->BlockStmt.stmts[i], indent+1);
}
break;
case AstNode_IfStmt:
print_indent(indent);
gb_printf("(if)\n");
print_ast(node->IfStmt.cond, indent+1);
print_ast(node->IfStmt.body, indent+1);
if (node->IfStmt.else_stmt) {
print_indent(indent);
gb_printf("(else)\n");
print_ast(node->IfStmt.else_stmt, indent+1);
}
break;
case AstNode_ReturnStmt:
print_indent(indent);
gb_printf("(return)\n");
for_array(i, node->ReturnStmt.results) {
print_ast(node->ReturnStmt.results[i], indent+1);
}
break;
case AstNode_ForStmt:
print_indent(indent);
gb_printf("(for)\n");
print_ast(node->ForStmt.init, indent+1);
print_ast(node->ForStmt.cond, indent+1);
print_ast(node->ForStmt.post, indent+1);
print_ast(node->ForStmt.body, indent+1);
break;
case AstNode_DeferStmt:
print_indent(indent);
gb_printf("(defer)\n");
print_ast(node->DeferStmt.stmt, indent+1);
break;
case AstNode_VarDecl:
print_indent(indent);
gb_printf("(decl:var)\n");
for_array(i, node->VarDecl.names) {
print_ast(node->VarDecl.names[i], indent+1);
}
print_ast(node->VarDecl.type, indent+1);
for_array(i, node->VarDecl.values) {
print_ast(node->VarDecl.values[i], indent+1);
}
break;
case AstNode_ConstDecl:
print_indent(indent);
gb_printf("(decl:const)\n");
for_array(i, node->VarDecl.names) {
print_ast(node->VarDecl.names[i], indent+1);
}
print_ast(node->VarDecl.type, indent+1);
for_array(i, node->VarDecl.values) {
print_ast(node->VarDecl.values[i], indent+1);
}
break;
case AstNode_ProcDecl:
print_indent(indent);
gb_printf("(decl:proc)\n");
print_ast(node->ProcDecl.type, indent+1);
print_ast(node->ProcDecl.body, indent+1);
break;
case AstNode_TypeDecl:
print_indent(indent);
gb_printf("(type)\n");
print_ast(node->TypeDecl.name, indent+1);
print_ast(node->TypeDecl.type, indent+1);
break;
case AstNode_ProcType:
print_indent(indent);
gb_printf("(type:proc)(%td -> %td)\n", node->ProcType.params.count, node->ProcType.results.count);
for_array(i, node->ProcType.params) {
print_ast(node->ProcType.params[i], indent+1);
}
if (node->ProcType.results.count > 0) {
print_indent(indent+1);
gb_printf("->\n");
for_array(i, node->ProcType.results) {
print_ast(node->ProcType.results[i], indent+1);
}
}
break;
case AstNode_Parameter:
for_array(i, node->Parameter.names) {
print_ast(node->Parameter.names[i], indent+1);
}
print_ast(node->Parameter.type, indent);
break;
case AstNode_PointerType:
print_indent(indent);
print_token(node->PointerType.token);
print_ast(node->PointerType.type, indent+1);
break;
case AstNode_ArrayType:
print_indent(indent);
gb_printf("[]\n");
print_ast(node->ArrayType.count, indent+1);
print_ast(node->ArrayType.elem, indent+1);
break;
case AstNode_StructType:
print_indent(indent);
gb_printf("(struct)\n");
for_array(i, node->StructType.decls) {
print_ast(node->StructType.decls[i], indent+1);
}
break;
}
// if (node->next)
// print_ast(node->next, indent);
}

5419
src/ssa.c Normal file

File diff suppressed because it is too large Load Diff

493
src/ssa_opt.c Normal file
View File

@@ -0,0 +1,493 @@
// Optimizations for the SSA code
void ssa_opt_add_operands(ssaValueArray *ops, ssaInstr *i) {
switch (i->kind) {
case ssaInstr_Comment:
break;
case ssaInstr_Local:
break;
case ssaInstr_ZeroInit:
array_add(ops, i->ZeroInit.address);
break;
case ssaInstr_Store:
array_add(ops, i->Store.address);
array_add(ops, i->Store.value);
break;
case ssaInstr_Load:
array_add(ops, i->Load.address);
break;
case ssaInstr_ArrayElementPtr:
array_add(ops, i->ArrayElementPtr.address);
array_add(ops, i->ArrayElementPtr.elem_index);
break;
case ssaInstr_StructElementPtr:
array_add(ops, i->StructElementPtr.address);
break;
case ssaInstr_PtrOffset:
array_add(ops, i->PtrOffset.address);
array_add(ops, i->PtrOffset.offset);
break;
case ssaInstr_ArrayExtractValue:
array_add(ops, i->ArrayExtractValue.address);
break;
case ssaInstr_StructExtractValue:
array_add(ops, i->StructExtractValue.address);
break;
case ssaInstr_Conv:
array_add(ops, i->Conv.value);
break;
case ssaInstr_Jump:
break;
case ssaInstr_If:
array_add(ops, i->If.cond);
break;
case ssaInstr_Return:
if (i->Return.value != NULL) {
array_add(ops, i->Return.value);
}
break;
case ssaInstr_Select:
array_add(ops, i->Select.cond);
break;
case ssaInstr_Phi:
for_array(j, i->Phi.edges) {
array_add(ops, i->Phi.edges.e[j]);
}
break;
case ssaInstr_Unreachable: break;
case ssaInstr_BinaryOp:
array_add(ops, i->BinaryOp.left);
array_add(ops, i->BinaryOp.right);
break;
case ssaInstr_Call:
array_add(ops, i->Call.value);
for (isize j = 0; j < i->Call.arg_count; j++) {
array_add(ops, i->Call.args[j]);
}
break;
case ssaInstr_VectorExtractElement:
array_add(ops, i->VectorExtractElement.vector);
array_add(ops, i->VectorExtractElement.index);
break;
case ssaInstr_VectorInsertElement:
array_add(ops, i->VectorInsertElement.vector);
array_add(ops, i->VectorInsertElement.elem);
array_add(ops, i->VectorInsertElement.index);
break;
case ssaInstr_VectorShuffle:
array_add(ops, i->VectorShuffle.vector);
break;
case ssaInstr_StartupRuntime:
break;
case ssaInstr_BoundsCheck:
array_add(ops, i->BoundsCheck.index);
array_add(ops, i->BoundsCheck.len);
break;
case ssaInstr_SliceBoundsCheck:
array_add(ops, i->SliceBoundsCheck.low);
array_add(ops, i->SliceBoundsCheck.high);
array_add(ops, i->SliceBoundsCheck.max);
break;
}
}
void ssa_opt_block_replace_pred(ssaBlock *b, ssaBlock *from, ssaBlock *to) {
for_array(i, b->preds) {
ssaBlock *pred = b->preds.e[i];
if (pred == from) {
b->preds.e[i] = to;
}
}
}
void ssa_opt_block_replace_succ(ssaBlock *b, ssaBlock *from, ssaBlock *to) {
for_array(i, b->succs) {
ssaBlock *succ = b->succs.e[i];
if (succ == from) {
b->succs.e[i] = to;
}
}
}
bool ssa_opt_block_has_phi(ssaBlock *b) {
return b->instrs.e[0]->Instr.kind == ssaInstr_Phi;
}
ssaValueArray ssa_get_block_phi_nodes(ssaBlock *b) {
ssaValueArray phis = {0};
for_array(i, b->instrs) {
ssaInstr *instr = &b->instrs.e[i]->Instr;
if (instr->kind != ssaInstr_Phi) {
phis = b->instrs;
phis.count = i;
return phis;
}
}
return phis;
}
void ssa_remove_pred(ssaBlock *b, ssaBlock *p) {
ssaValueArray phis = ssa_get_block_phi_nodes(b);
isize i = 0;
for_array(j, b->preds) {
ssaBlock *pred = b->preds.e[j];
if (pred != p) {
b->preds.e[i] = b->preds.e[j];
for_array(k, phis) {
ssaInstrPhi *phi = &phis.e[k]->Instr.Phi;
phi->edges.e[i] = phi->edges.e[j];
}
i++;
}
}
b->preds.count = i;
for_array(k, phis) {
ssaInstrPhi *phi = &phis.e[k]->Instr.Phi;
phi->edges.count = i;
}
}
void ssa_remove_dead_blocks(ssaProcedure *proc) {
isize j = 0;
for_array(i, proc->blocks) {
ssaBlock *b = proc->blocks.e[i];
if (b == NULL) {
continue;
}
// NOTE(bill): Swap order
b->index = j;
proc->blocks.e[j++] = b;
}
proc->blocks.count = j;
}
void ssa_mark_reachable(ssaBlock *b) {
isize const WHITE = 0;
isize const BLACK = -1;
b->index = BLACK;
for_array(i, b->succs) {
ssaBlock *succ = b->succs.e[i];
if (succ->index == WHITE) {
ssa_mark_reachable(succ);
}
}
}
void ssa_remove_unreachable_blocks(ssaProcedure *proc) {
isize const WHITE = 0;
isize const BLACK = -1;
for_array(i, proc->blocks) {
proc->blocks.e[i]->index = WHITE;
}
ssa_mark_reachable(proc->blocks.e[0]);
for_array(i, proc->blocks) {
ssaBlock *b = proc->blocks.e[i];
if (b->index == WHITE) {
for_array(j, b->succs) {
ssaBlock *c = b->succs.e[j];
if (c->index == BLACK) {
ssa_remove_pred(c, b);
}
}
// NOTE(bill): Mark as empty but don't actually free it
// As it's been allocated with an arena
proc->blocks.e[i] = NULL;
}
}
ssa_remove_dead_blocks(proc);
}
bool ssa_opt_block_fusion(ssaProcedure *proc, ssaBlock *a) {
if (a->succs.count != 1) {
return false;
}
ssaBlock *b = a->succs.e[0];
if (b->preds.count != 1) {
return false;
}
if (ssa_opt_block_has_phi(b)) {
return false;
}
array_pop(&a->instrs); // Remove branch at end
for_array(i, b->instrs) {
array_add(&a->instrs, b->instrs.e[i]);
ssa_set_instr_parent(b->instrs.e[i], a);
}
array_clear(&a->succs);
for_array(i, b->succs) {
array_add(&a->succs, b->succs.e[i]);
}
// Fix preds links
for_array(i, b->succs) {
ssa_opt_block_replace_pred(b->succs.e[i], b, a);
}
proc->blocks.e[b->index] = NULL;
return true;
}
void ssa_opt_blocks(ssaProcedure *proc) {
ssa_remove_unreachable_blocks(proc);
#if 1
bool changed = true;
while (changed) {
changed = false;
for_array(i, proc->blocks) {
ssaBlock *b = proc->blocks.e[i];
if (b == NULL) {
continue;
}
GB_ASSERT(b->index == i);
if (ssa_opt_block_fusion(proc, b)) {
changed = true;
}
// TODO(bill): other simple block optimizations
}
}
#endif
ssa_remove_dead_blocks(proc);
}
void ssa_opt_build_referrers(ssaProcedure *proc) {
gbTempArenaMemory tmp = gb_temp_arena_memory_begin(&proc->module->tmp_arena);
ssaValueArray ops = {0}; // NOTE(bill): Act as a buffer
array_init_reserve(&ops, proc->module->tmp_allocator, 64); // HACK(bill): This _could_ overflow the temp arena
for_array(i, proc->blocks) {
ssaBlock *b = proc->blocks.e[i];
for_array(j, b->instrs) {
ssaValue *instr = b->instrs.e[j];
array_clear(&ops);
ssa_opt_add_operands(&ops, &instr->Instr);
for_array(k, ops) {
ssaValue *op = ops.e[k];
if (op == NULL) {
continue;
}
ssaValueArray *refs = ssa_value_referrers(op);
if (refs != NULL) {
array_add(refs, instr);
}
}
}
}
gb_temp_arena_memory_end(tmp);
}
// State of Lengauer-Tarjan algorithm
// Based on this paper: http://jgaa.info/accepted/2006/GeorgiadisTarjanWerneck2006.10.1.pdf
typedef struct ssaLTState {
isize count;
// NOTE(bill): These are arrays
ssaBlock **sdom; // Semidominator
ssaBlock **parent; // Parent in DFS traversal of CFG
ssaBlock **ancestor;
} ssaLTState;
// §2.2 - bottom of page
void ssa_lt_link(ssaLTState *lt, ssaBlock *p, ssaBlock *q) {
lt->ancestor[q->index] = p;
}
i32 ssa_lt_depth_first_search(ssaLTState *lt, ssaBlock *p, i32 i, ssaBlock **preorder) {
preorder[i] = p;
p->dom.pre = i++;
lt->sdom[p->index] = p;
ssa_lt_link(lt, NULL, p);
for_array(index, p->succs) {
ssaBlock *q = p->succs.e[index];
if (lt->sdom[q->index] == NULL) {
lt->parent[q->index] = p;
i = ssa_lt_depth_first_search(lt, q, i, preorder);
}
}
return i;
}
ssaBlock *ssa_lt_eval(ssaLTState *lt, ssaBlock *v) {
ssaBlock *u = v;
for (;
lt->ancestor[v->index] != NULL;
v = lt->ancestor[v->index]) {
if (lt->sdom[v->index]->dom.pre < lt->sdom[u->index]->dom.pre) {
u = v;
}
}
return u;
}
typedef struct ssaDomPrePost {
i32 pre, post;
} ssaDomPrePost;
ssaDomPrePost ssa_opt_number_dom_tree(ssaBlock *v, i32 pre, i32 post) {
ssaDomPrePost result = {pre, post};
v->dom.pre = pre++;
for_array(i, v->dom.children) {
result = ssa_opt_number_dom_tree(v->dom.children.e[i], result.pre, result.post);
}
v->dom.post = post++;
result.pre = pre;
result.post = post;
return result;
}
// NOTE(bill): Requires `ssa_opt_blocks` to be called before this
void ssa_opt_build_dom_tree(ssaProcedure *proc) {
// Based on this paper: http://jgaa.info/accepted/2006/GeorgiadisTarjanWerneck2006.10.1.pdf
gbTempArenaMemory tmp = gb_temp_arena_memory_begin(&proc->module->tmp_arena);
isize n = proc->blocks.count;
ssaBlock **buf = gb_alloc_array(proc->module->tmp_allocator, ssaBlock *, 5*n);
ssaLTState lt = {0};
lt.count = n;
lt.sdom = &buf[0*n];
lt.parent = &buf[1*n];
lt.ancestor = &buf[2*n];
ssaBlock **preorder = &buf[3*n];
ssaBlock **buckets = &buf[4*n];
ssaBlock *root = proc->blocks.e[0];
// Step 1 - number vertices
i32 pre_num = ssa_lt_depth_first_search(&lt, root, 0, preorder);
gb_memmove(buckets, preorder, n*gb_size_of(preorder[0]));
for (i32 i = n-1; i > 0; i--) {
ssaBlock *w = preorder[i];
// Step 3 - Implicitly define idom for nodes
for (ssaBlock *v = buckets[i]; v != w; v = buckets[v->dom.pre]) {
ssaBlock *u = ssa_lt_eval(&lt, v);
if (lt.sdom[u->index]->dom.pre < i) {
v->dom.idom = u;
} else {
v->dom.idom = w;
}
}
// Step 2 - Compute all sdoms
lt.sdom[w->index] = lt.parent[w->index];
for_array(pred_index, w->preds) {
ssaBlock *v = w->preds.e[pred_index];
ssaBlock *u = ssa_lt_eval(&lt, v);
if (lt.sdom[u->index]->dom.pre < lt.sdom[w->index]->dom.pre) {
lt.sdom[w->index] = lt.sdom[u->index];
}
}
ssa_lt_link(&lt, lt.parent[w->index], w);
if (lt.parent[w->index] == lt.sdom[w->index]) {
w->dom.idom = lt.parent[w->index];
} else {
buckets[i] = buckets[lt.sdom[w->index]->dom.pre];
buckets[lt.sdom[w->index]->dom.pre] = w;
}
}
// The rest of Step 3
for (ssaBlock *v = buckets[0]; v != root; v = buckets[v->dom.pre]) {
v->dom.idom = root;
}
// Step 4 - Explicitly define idom for nodes (in preorder)
for (isize i = 1; i < n; i++) {
ssaBlock *w = preorder[i];
if (w == root) {
w->dom.idom = NULL;
} else {
// Weird tree relationships here!
if (w->dom.idom != lt.sdom[w->index]) {
w->dom.idom = w->dom.idom->dom.idom;
}
// Calculate children relation as inverse of idom
if (w->dom.idom->dom.children.e == NULL) {
// TODO(bill): Is this good enough for memory allocations?
array_init(&w->dom.idom->dom.children, heap_allocator());
}
array_add(&w->dom.idom->dom.children, w);
}
}
ssa_opt_number_dom_tree(root, 0, 0);
gb_temp_arena_memory_end(tmp);
}
void ssa_opt_mem2reg(ssaProcedure *proc) {
// TODO(bill): ssa_opt_mem2reg
}
void ssa_opt_tree(ssaGen *s) {
s->opt_called = true;
for_array(member_index, s->module.procs) {
ssaProcedure *proc = s->module.procs.e[member_index];
if (proc->blocks.count == 0) { // Prototype/external procedure
continue;
}
ssa_opt_blocks(proc);
#if 1
ssa_opt_build_referrers(proc);
ssa_opt_build_dom_tree(proc);
// TODO(bill): ssa optimization
// [ ] cse (common-subexpression) elim
// [ ] copy elim
// [ ] dead code elim
// [ ] dead store/load elim
// [ ] phi elim
// [ ] short circuit elim
// [ ] bounds check elim
// [ ] lift/mem2reg
// [ ] lift/mem2reg
ssa_opt_mem2reg(proc);
#endif
GB_ASSERT(proc->blocks.count > 0);
ssa_number_proc_registers(proc);
}
}

1439
src/ssa_print.c Normal file

File diff suppressed because it is too large Load Diff

422
src/string.c Normal file
View File

@@ -0,0 +1,422 @@
gb_global gbArena string_buffer_arena = {0};
gb_global gbAllocator string_buffer_allocator = {0};
void init_string_buffer_memory(void) {
// NOTE(bill): This should be enough memory for file systems
gb_arena_init_from_allocator(&string_buffer_arena, heap_allocator(), gb_megabytes(1));
string_buffer_allocator = gb_arena_allocator(&string_buffer_arena);
}
// NOTE(bill): Used for UTF-8 strings
typedef struct String {
u8 * text;
isize len;
} String;
// NOTE(bill): used for printf style arguments
#define LIT(x) ((int)(x).len), (x).text
typedef struct String16 {
wchar_t *text;
isize len;
} String16;
gb_inline String make_string(u8 *text, isize len) {
String s;
s.text = text;
if (len < 0) {
len = gb_strlen(cast(char *)text);
}
s.len = len;
return s;
}
gb_inline String16 make_string16(wchar_t *text, isize len) {
String16 s;
s.text = text;
s.len = len;
return s;
}
gb_inline String make_string_c(char *text) {
return make_string(cast(u8 *)cast(void *)text, gb_strlen(text));
}
#define str_lit(c_str) make_string(cast(u8 *)c_str, gb_size_of(c_str)-1)
gb_inline bool are_strings_equal(String a, String b) {
if (a.len == b.len) {
return gb_memcompare(a.text, b.text, a.len) == 0;
}
return false;
}
gb_inline bool str_eq_ignore_case(String a, String b) {
if (a.len == b.len) {
for (isize i = 0; i < a.len; i++) {
char x = cast(char)a.text[i];
char y = cast(char)b.text[i];
if (gb_char_to_lower(x) != gb_char_to_lower(y))
return false;
}
return true;
}
return false;
}
int string_compare(String x, String y) {
if (x.len == y.len &&
x.text == y.text) {
return 0;
}
isize n = gb_min(x.len, y.len);
isize fast = n/gb_size_of(isize) + 1;
isize offset = (fast-1)*gb_size_of(isize);
isize curr_block = 0;
if (n <= gb_size_of(isize)) {
fast = 0;
}
isize *la = cast(isize *)x.text;
isize *lb = cast(isize *)y.text;
for (; curr_block < fast; curr_block++) {
if (la[curr_block] ^ lb[curr_block]) {
for (isize pos = curr_block*gb_size_of(isize); pos < n; pos++) {
if (x.text[pos] ^ y.text[pos]) {
return cast(int)x.text[pos] - cast(int)y.text[pos];
}
}
}
}
for (; offset < n; offset++) {
if (x.text[offset] ^ y.text[offset]) {
return cast(int)x.text[offset] - cast(int)y.text[offset];
}
}
return 0;
}
GB_COMPARE_PROC(string_cmp_proc) {
String x = *(String *)a;
String y = *(String *)b;
return string_compare(x, y);
}
// gb_inline bool operator ==(String a, String b) { return are_strings_equal(a, b) != 0; }
// gb_inline bool operator !=(String a, String b) { return !operator==(a, b); }
// gb_inline bool operator < (String a, String b) { return string_compare(a, b) < 0; }
// gb_inline bool operator > (String a, String b) { return string_compare(a, b) > 0; }
// gb_inline bool operator <=(String a, String b) { return string_compare(a, b) <= 0; }
// gb_inline bool operator >=(String a, String b) { return string_compare(a, b) >= 0; }
// template <size_t N> gb_inline bool operator ==(String a, char const (&b)[N]) { return a == make_string(cast(u8 *)b, N-1); }
// template <size_t N> gb_inline bool operator !=(String a, char const (&b)[N]) { return a != make_string(cast(u8 *)b, N-1); }
// template <size_t N> gb_inline bool operator ==(char const (&a)[N], String b) { return make_string(cast(u8 *)a, N-1) == b; }
// template <size_t N> gb_inline bool operator !=(char const (&a)[N], String b) { return make_string(cast(u8 *)a, N-1) != b; }
gb_inline bool str_eq(String a, String b) { return are_strings_equal(a, b) != 0; }
gb_inline bool str_ne(String a, String b) { return !str_eq(a, b); }
gb_inline bool str_lt(String a, String b) { return string_compare(a, b) < 0; }
gb_inline bool str_gt(String a, String b) { return string_compare(a, b) > 0; }
gb_inline bool str_le(String a, String b) { return string_compare(a, b) <= 0; }
gb_inline bool str_ge(String a, String b) { return string_compare(a, b) >= 0; }
gb_inline isize string_extension_position(String str) {
isize dot_pos = -1;
isize i = str.len;
bool seen_dot = false;
while (i --> 0) {
if (str.text[i] == GB_PATH_SEPARATOR)
break;
if (str.text[i] == '.') {
dot_pos = i;
break;
}
}
return dot_pos;
}
gb_inline bool string_has_extension(String str, String ext) {
if (str.len > ext.len+1) {
u8 *s = str.text+str.len - ext.len-1;
if (s[0] == '.') {
s++;
return gb_memcompare(s, ext.text, ext.len) == 0;
}
return false;
}
return false;
}
bool string_contains_char(String s, u8 c) {
for (isize i = 0; i < s.len; i++) {
if (s.text[i] == c)
return true;
}
return false;
}
// TODO(bill): Make this non-windows specific
String16 string_to_string16(gbAllocator a, String s) {
if (s.len < 1) {
return make_string16(NULL, 0);
}
int len = MultiByteToWideChar(CP_UTF8, MB_ERR_INVALID_CHARS,
cast(char *)s.text, s.len, NULL, 0);
if (len == 0) {
return make_string16(NULL, 0);
}
wchar_t *text = gb_alloc_array(a, wchar_t, len+1);
int len1 = MultiByteToWideChar(CP_UTF8, MB_ERR_INVALID_CHARS,
cast(char *)s.text, s.len, text, len);
if (len1 == 0) {
gb_free(a, text);
return make_string16(NULL, 0);
}
text[len] = 0;
return make_string16(text, len-1);
}
String string16_to_string(gbAllocator a, String16 s) {
if (s.len < 1) {
return make_string(NULL, 0);
}
int len = WideCharToMultiByte(CP_UTF8, WC_ERR_INVALID_CHARS,
s.text, s.len, NULL, 0,
NULL, NULL);
if (len == 0) {
return make_string(NULL, 0);
}
u8 *text = gb_alloc_array(a, u8, len+1);
int len1 = WideCharToMultiByte(CP_UTF8, WC_ERR_INVALID_CHARS,
s.text, s.len, cast(char *)text, len,
NULL, NULL);
if (len1 == 0) {
gb_free(a, text);
return make_string(NULL, 0);
}
text[len] = 0;
return make_string(text, len-1);
}
bool unquote_char(String s, u8 quote, Rune *rune, bool *multiple_bytes, String *tail_string) {
if (s.text[0] == quote &&
(quote == '$' || quote == '"')) {
return false;
} else if (s.text[0] >= 0x80) {
Rune r = -1;
isize size = gb_utf8_decode(s.text, s.len, &r);
*rune = r;
*multiple_bytes = true;
*tail_string = make_string(s.text+size, s.len-size);
return true;
} else if (s.text[0] != '\\') {
*rune = s.text[0];
*tail_string = make_string(s.text+1, s.len-1);
return true;
}
if (s.len <= 1) {
return false;
}
u8 c = s.text[1];
s = make_string(s.text+2, s.len-2);
switch (c) {
default: return false;
case 'a': *rune = '\a'; break;
case 'b': *rune = '\b'; break;
case 'f': *rune = '\f'; break;
case 'n': *rune = '\n'; break;
case 'r': *rune = '\r'; break;
case 't': *rune = '\t'; break;
case 'v': *rune = '\v'; break;
case '\\': *rune = '\\'; break;
case '$':
case '"':
if (c != quote) {
return false;
}
*rune = c;
break;
case '0':
case '1':
case '2':
case '3':
case '4':
case '5':
case '6':
case '7': {
i32 r = gb_digit_to_int(c);
if (s.len < 2) {
return false;
}
for (isize i = 0; i < 2; i++) {
i32 d = gb_digit_to_int(s.text[i]);
if (d < 0 || d > 7) {
return false;
}
r = (r<<3) | d;
}
s = make_string(s.text+2, s.len-2);
if (r > 0xff) {
return false;
}
*rune = r;
} break;
case 'x':
case 'u':
case 'U': {
isize count = 0;
switch (c) {
case 'x': count = 2; break;
case 'u': count = 4; break;
case 'U': count = 8; break;
}
Rune r = 0;
if (s.len < count) {
return false;
}
for (isize i = 0; i < count; i++) {
i32 d = gb_hex_digit_to_int(s.text[i]);
if (d < 0) {
return false;
}
r = (r<<4) | d;
}
s = make_string(s.text+count, s.len-count);
if (c == 'x') {
*rune = r;
break;
}
if (r > GB_RUNE_MAX) {
return false;
}
*rune = r;
*multiple_bytes = true;
} break;
}
*tail_string = s;
return true;
}
// 0 == failure
// 1 == original memory
// 2 == new allocation
i32 unquote_string(gbAllocator a, String *s_) {
GB_ASSERT(s_ != NULL);
String s = *s_;
isize n = s.len;
if (n < 2)
return 0;
u8 quote = s.text[0];
if (quote != s.text[n-1])
return 0;
s.text += 1;
s.len -= 2;
if (quote == '`') {
if (string_contains_char(s, '`')) {
return 0;
}
*s_ = s;
return 1;
}
if (quote != '"' && quote != '$')
return 0;
if (string_contains_char(s, '\n'))
return 0;
if (!string_contains_char(s, '\\') && !string_contains_char(s, quote)) {
if (quote == '"') {
*s_ = s;
return 1;
} else if (quote == '$') {
Rune r = GB_RUNE_INVALID;
isize size = gb_utf8_decode(s.text, s.len, &r);
if ((size == s.len) && (r != -1 || size != 1)) {
*s_ = s;
return 1;
}
}
}
u8 rune_temp[4] = {0};
isize buf_len = 3*s.len / 2;
u8 *buf = gb_alloc_array(a, u8, buf_len);
isize offset = 0;
while (s.len > 0) {
String tail_string = {0};
Rune r = 0;
bool multiple_bytes = false;
bool success = unquote_char(s, quote, &r, &multiple_bytes, &tail_string);
if (!success) {
gb_free(a, buf);
return 0;
}
s = tail_string;
if (r < 0x80 || !multiple_bytes) {
buf[offset++] = cast(u8)r;
} else {
isize size = gb_utf8_encode_rune(rune_temp, r);
gb_memmove(buf+offset, rune_temp, size);
offset += size;
}
if (quote == '$' && s.len != 0) {
gb_free(a, buf);
return 0;
}
}
*s_ = make_string(buf, offset);
return 2;
}

105
src/timings.c Normal file
View File

@@ -0,0 +1,105 @@
typedef struct TimeStamp {
u64 start;
u64 finish;
String label;
} TimeStamp;
typedef struct Timings {
TimeStamp total;
Array(TimeStamp) sections;
u64 freq;
} Timings;
u64 win32_time_stamp_time_now(void) {
LARGE_INTEGER counter;
QueryPerformanceCounter(&counter);
return counter.QuadPart;
}
u64 win32_time_stamp__freq(void) {
gb_local_persist LARGE_INTEGER win32_perf_count_freq = {0};
if (!win32_perf_count_freq.QuadPart) {
QueryPerformanceFrequency(&win32_perf_count_freq);
GB_ASSERT(win32_perf_count_freq.QuadPart != 0);
}
return win32_perf_count_freq.QuadPart;
}
u64 time_stamp_time_now(void) {
#if defined(GB_SYSTEM_WINDOWS)
return win32_time_stamp_time_now();
#else
#error time_stamp_time_now
#endif
}
u64 time_stamp__freq(void) {
#if defined(GB_SYSTEM_WINDOWS)
return win32_time_stamp__freq();
#else
#error time_stamp__freq
#endif
}
TimeStamp make_time_stamp(String label) {
TimeStamp ts = {0};
ts.start = time_stamp_time_now();
ts.label = label;
return ts;
}
void timings_init(Timings *t, String label, isize buffer_size) {
array_init_reserve(&t->sections, heap_allocator(), buffer_size);
t->total = make_time_stamp(label);
t->freq = time_stamp__freq();
}
void timings_destroy(Timings *t) {
array_free(&t->sections);
}
void timings__stop_current_section(Timings *t) {
if (t->sections.count > 0) {
t->sections.e[t->sections.count-1].finish = time_stamp_time_now();
}
}
void timings_start_section(Timings *t, String label) {
timings__stop_current_section(t);
array_add(&t->sections, make_time_stamp(label));
}
f64 time_stamp_as_ms(TimeStamp ts, u64 freq) {
GB_ASSERT_MSG(ts.finish >= ts.start, "time_stamp_as_ms - %.*s", LIT(ts.label));
return 1000.0 * cast(f64)(ts.finish - ts.start) / cast(f64)freq;
}
void timings_print_all(Timings *t) {
timings__stop_current_section(t);
t->total.finish = time_stamp_time_now();
char const SPACES[] = " ";
isize max_len = t->total.label.len;
for_array(i, t->sections) {
TimeStamp ts = t->sections.e[i];
max_len = gb_max(max_len, ts.label.len);
}
GB_ASSERT(max_len <= gb_size_of(SPACES)-1);
gb_printf("%.*s%.*s - %.3f ms\n",
LIT(t->total.label),
cast(int)(max_len-t->total.label.len), SPACES,
time_stamp_as_ms(t->total, t->freq));
for_array(i, t->sections) {
TimeStamp ts = t->sections.e[i];
gb_printf("%.*s%.*s - %.3f ms\n",
LIT(ts.label),
cast(int)(max_len-ts.label.len), SPACES,
time_stamp_as_ms(ts, t->freq));
}
}

816
src/tokenizer.c Normal file
View File

@@ -0,0 +1,816 @@
#define TOKEN_KINDS \
TOKEN_KIND(Token_Invalid, "Invalid"), \
TOKEN_KIND(Token_EOF, "EOF"), \
TOKEN_KIND(Token_Comment, "Comment"), \
\
TOKEN_KIND(Token__LiteralBegin, "_LiteralBegin"), \
TOKEN_KIND(Token_Identifier, "Identifier"), \
TOKEN_KIND(Token_Integer, "Integer"), \
TOKEN_KIND(Token_Float, "Float"), \
TOKEN_KIND(Token_Rune, "Rune"), \
TOKEN_KIND(Token_String, "String"), \
TOKEN_KIND(Token__LiteralEnd, "_LiteralEnd"), \
\
TOKEN_KIND(Token__OperatorBegin, "_OperatorBegin"), \
TOKEN_KIND(Token_Eq, "="), \
TOKEN_KIND(Token_Not, "!"), \
TOKEN_KIND(Token_Hash, "#"), \
TOKEN_KIND(Token_At, "@"), \
TOKEN_KIND(Token_Pointer, "^"), \
TOKEN_KIND(Token_Maybe, "?"), \
TOKEN_KIND(Token_Add, "+"), \
TOKEN_KIND(Token_Sub, "-"), \
TOKEN_KIND(Token_Mul, "*"), \
TOKEN_KIND(Token_Quo, "/"), \
TOKEN_KIND(Token_Mod, "%"), \
TOKEN_KIND(Token_And, "&"), \
TOKEN_KIND(Token_Or, "|"), \
TOKEN_KIND(Token_Xor, "~"), \
TOKEN_KIND(Token_AndNot, "&~"), \
TOKEN_KIND(Token_Shl, "<<"), \
TOKEN_KIND(Token_Shr, ">>"), \
\
TOKEN_KIND(Token_as, "as"), \
TOKEN_KIND(Token_transmute, "transmute"), \
TOKEN_KIND(Token_down_cast, "down_cast"), \
TOKEN_KIND(Token_union_cast, "union_cast"), \
\
TOKEN_KIND(Token_Prime, "'"), \
TOKEN_KIND(Token_DoublePrime, "''"), \
\
TOKEN_KIND(Token_CmpAnd, "&&"), \
TOKEN_KIND(Token_CmpOr, "||"), \
\
TOKEN_KIND(Token__AssignOpBegin, "_AssignOpBegin"), \
TOKEN_KIND(Token_AddEq, "+="), \
TOKEN_KIND(Token_SubEq, "-="), \
TOKEN_KIND(Token_MulEq, "*="), \
TOKEN_KIND(Token_QuoEq, "/="), \
TOKEN_KIND(Token_ModEq, "%="), \
TOKEN_KIND(Token_AndEq, "&="), \
TOKEN_KIND(Token_OrEq, "|="), \
TOKEN_KIND(Token_XorEq, "~="), \
TOKEN_KIND(Token_AndNotEq, "&~="), \
TOKEN_KIND(Token_ShlEq, "<<="), \
TOKEN_KIND(Token_ShrEq, ">>="), \
TOKEN_KIND(Token_CmpAndEq, "&&="), \
TOKEN_KIND(Token_CmpOrEq, "||="), \
TOKEN_KIND(Token__AssignOpEnd, "_AssignOpEnd"), \
TOKEN_KIND(Token_Increment, "++"), \
TOKEN_KIND(Token_Decrement, "--"), \
TOKEN_KIND(Token_ArrowRight, "->"), \
TOKEN_KIND(Token_ArrowLeft, "<-"), \
\
TOKEN_KIND(Token__ComparisonBegin, "_ComparisonBegin"), \
TOKEN_KIND(Token_CmpEq, "=="), \
TOKEN_KIND(Token_NotEq, "!="), \
TOKEN_KIND(Token_Lt, "<"), \
TOKEN_KIND(Token_Gt, ">"), \
TOKEN_KIND(Token_LtEq, "<="), \
TOKEN_KIND(Token_GtEq, ">="), \
TOKEN_KIND(Token__ComparisonEnd, "_ComparisonEnd"), \
\
TOKEN_KIND(Token_OpenParen, "("), \
TOKEN_KIND(Token_CloseParen, ")"), \
TOKEN_KIND(Token_OpenBracket, "["), \
TOKEN_KIND(Token_CloseBracket, "]"), \
TOKEN_KIND(Token_OpenBrace, "{"), \
TOKEN_KIND(Token_CloseBrace, "}"), \
TOKEN_KIND(Token_Colon, ":"), \
TOKEN_KIND(Token_Semicolon, ";"), \
TOKEN_KIND(Token_Period, "."), \
TOKEN_KIND(Token_Comma, ","), \
TOKEN_KIND(Token_Ellipsis, ".."), \
TOKEN_KIND(Token_RangeExclusive, "..<"), \
TOKEN_KIND(Token__OperatorEnd, "_OperatorEnd"), \
\
TOKEN_KIND(Token__KeywordBegin, "_KeywordBegin"), \
TOKEN_KIND(Token_type, "type"), \
TOKEN_KIND(Token_proc, "proc"), \
TOKEN_KIND(Token_match, "match"), \
TOKEN_KIND(Token_break, "break"), \
TOKEN_KIND(Token_continue, "continue"), \
TOKEN_KIND(Token_fallthrough, "fallthrough"), \
TOKEN_KIND(Token_case, "case"), \
TOKEN_KIND(Token_default, "default"), \
TOKEN_KIND(Token_then, "then"), \
TOKEN_KIND(Token_if, "if"), \
TOKEN_KIND(Token_else, "else"), \
TOKEN_KIND(Token_for, "for"), \
TOKEN_KIND(Token_range, "range"), \
TOKEN_KIND(Token_defer, "defer"), \
TOKEN_KIND(Token_return, "return"), \
TOKEN_KIND(Token_struct, "struct"), \
TOKEN_KIND(Token_union, "union"), \
TOKEN_KIND(Token_raw_union, "raw_union"), \
TOKEN_KIND(Token_enum, "enum"), \
TOKEN_KIND(Token_using, "using"), \
TOKEN_KIND(Token_asm, "asm"), \
TOKEN_KIND(Token_volatile, "volatile"), \
TOKEN_KIND(Token_atomic, "atomic"), \
TOKEN_KIND(Token_push_allocator, "push_allocator"), \
TOKEN_KIND(Token_push_context, "push_context"), \
TOKEN_KIND(Token__KeywordEnd, "_KeywordEnd"), \
TOKEN_KIND(Token_Count, "")
typedef enum TokenKind {
#define TOKEN_KIND(e, s) e
TOKEN_KINDS
#undef TOKEN_KIND
} TokenKind;
String const token_strings[] = {
#define TOKEN_KIND(e, s) {cast(u8 *)s, gb_size_of(s)-1}
TOKEN_KINDS
#undef TOKEN_KIND
};
typedef struct TokenPos {
String file;
isize line;
isize column;
} TokenPos;
i32 token_pos_cmp(TokenPos a, TokenPos b) {
if (a.line == b.line) {
if (a.column == b.column) {
isize min_len = gb_min(a.file.len, b.file.len);
return gb_memcompare(a.file.text, b.file.text, min_len);
}
return (a.column < b.column) ? -1 : +1;
}
return (a.line < b.line) ? -1 : +1;
}
bool token_pos_are_equal(TokenPos a, TokenPos b) {
return token_pos_cmp(a, b) == 0;
}
// NOTE(bill): Text is UTF-8, thus why u8 and not char
typedef struct Token {
TokenKind kind;
String string;
TokenPos pos;
} Token;
Token empty_token = {Token_Invalid};
Token blank_token = {Token_Identifier, {cast(u8 *)"_", 1}};
Token make_token_ident(String s) {
Token t = {Token_Identifier, s};
return t;
}
typedef struct ErrorCollector {
TokenPos prev;
i64 count;
i64 warning_count;
gbMutex mutex;
} ErrorCollector;
gb_global ErrorCollector global_error_collector;
void init_global_error_collector(void) {
gb_mutex_init(&global_error_collector.mutex);
}
void warning(Token token, char *fmt, ...) {
gb_mutex_lock(&global_error_collector.mutex);
global_error_collector.warning_count++;
// NOTE(bill): Duplicate error, skip it
if (!token_pos_are_equal(global_error_collector.prev, token.pos)) {
va_list va;
global_error_collector.prev = token.pos;
va_start(va, fmt);
gb_printf_err("%.*s(%td:%td) Warning: %s\n",
LIT(token.pos.file), token.pos.line, token.pos.column,
gb_bprintf_va(fmt, va));
va_end(va);
}
gb_mutex_unlock(&global_error_collector.mutex);
}
void error(Token token, char *fmt, ...) {
gb_mutex_lock(&global_error_collector.mutex);
global_error_collector.count++;
// NOTE(bill): Duplicate error, skip it
if (!token_pos_are_equal(global_error_collector.prev, token.pos)) {
va_list va;
global_error_collector.prev = token.pos;
va_start(va, fmt);
gb_printf_err("%.*s(%td:%td) %s\n",
LIT(token.pos.file), token.pos.line, token.pos.column,
gb_bprintf_va(fmt, va));
va_end(va);
}
gb_mutex_unlock(&global_error_collector.mutex);
}
void syntax_error(Token token, char *fmt, ...) {
gb_mutex_lock(&global_error_collector.mutex);
global_error_collector.count++;
// NOTE(bill): Duplicate error, skip it
if (!token_pos_are_equal(global_error_collector.prev, token.pos)) {
va_list va;
global_error_collector.prev = token.pos;
va_start(va, fmt);
gb_printf_err("%.*s(%td:%td) Syntax Error: %s\n",
LIT(token.pos.file), token.pos.line, token.pos.column,
gb_bprintf_va(fmt, va));
va_end(va);
}
gb_mutex_unlock(&global_error_collector.mutex);
}
void compiler_error(char *fmt, ...) {
va_list va;
va_start(va, fmt);
gb_printf_err("Internal Compiler Error: %s\n",
gb_bprintf_va(fmt, va));
va_end(va);
gb_exit(1);
}
gb_inline bool token_is_literal(Token t) {
return gb_is_between(t.kind, Token__LiteralBegin+1, Token__LiteralEnd-1);
}
gb_inline bool token_is_operator(Token t) {
return gb_is_between(t.kind, Token__OperatorBegin+1, Token__OperatorEnd-1);
}
gb_inline bool token_is_keyword(Token t) {
return gb_is_between(t.kind, Token__KeywordBegin+1, Token__KeywordEnd-1);
}
gb_inline bool token_is_comparison(Token t) {
return gb_is_between(t.kind, Token__ComparisonBegin+1, Token__ComparisonEnd-1);
}
gb_inline bool token_is_shift(Token t) {
return t.kind == Token_Shl || t.kind == Token_Shr;
}
gb_inline void print_token(Token t) { gb_printf("%.*s\n", LIT(t.string)); }
typedef enum TokenizerInitError {
TokenizerInit_None,
TokenizerInit_Invalid,
TokenizerInit_NotExists,
TokenizerInit_Permission,
TokenizerInit_Empty,
TokenizerInit_Count,
} TokenizerInitError;
typedef struct Tokenizer {
String fullpath;
u8 *start;
u8 *end;
Rune curr_rune; // current character
u8 * curr; // character pos
u8 * read_curr; // pos from start
u8 * line; // current line pos
isize line_count;
isize error_count;
Array(String) allocated_strings;
} Tokenizer;
void tokenizer_err(Tokenizer *t, char *msg, ...) {
va_list va;
isize column = t->read_curr - t->line+1;
if (column < 1)
column = 1;
gb_printf_err("%.*s(%td:%td) Syntax error: ", LIT(t->fullpath), t->line_count, column);
va_start(va, msg);
gb_printf_err_va(msg, va);
va_end(va);
gb_printf_err("\n");
t->error_count++;
}
void advance_to_next_rune(Tokenizer *t) {
if (t->read_curr < t->end) {
Rune rune;
isize width = 1;
t->curr = t->read_curr;
if (t->curr_rune == '\n') {
t->line = t->curr;
t->line_count++;
}
rune = *t->read_curr;
if (rune == 0) {
tokenizer_err(t, "Illegal character NUL");
} else if (rune >= 0x80) { // not ASCII
width = gb_utf8_decode(t->read_curr, t->end-t->read_curr, &rune);
if (rune == GB_RUNE_INVALID && width == 1)
tokenizer_err(t, "Illegal UTF-8 encoding");
else if (rune == GB_RUNE_BOM && t->curr-t->start > 0)
tokenizer_err(t, "Illegal byte order mark");
}
t->read_curr += width;
t->curr_rune = rune;
} else {
t->curr = t->end;
if (t->curr_rune == '\n') {
t->line = t->curr;
t->line_count++;
}
t->curr_rune = GB_RUNE_EOF;
}
}
TokenizerInitError init_tokenizer(Tokenizer *t, String fullpath) {
TokenizerInitError err = TokenizerInit_None;
char *c_str = gb_alloc_array(heap_allocator(), char, fullpath.len+1);
memcpy(c_str, fullpath.text, fullpath.len);
c_str[fullpath.len] = '\0';
// TODO(bill): Memory map rather than copy contents
gbFileContents fc = gb_file_read_contents(heap_allocator(), true, c_str);
gb_zero_item(t);
if (fc.data != NULL) {
t->start = cast(u8 *)fc.data;
t->line = t->read_curr = t->curr = t->start;
t->end = t->start + fc.size;
t->fullpath = fullpath;
t->line_count = 1;
advance_to_next_rune(t);
if (t->curr_rune == GB_RUNE_BOM) {
advance_to_next_rune(t); // Ignore BOM at file beginning
}
array_init(&t->allocated_strings, heap_allocator());
} else {
gbFile f = {0};
gbFileError file_err = gb_file_open(&f, c_str);
switch (file_err) {
case gbFileError_Invalid: err = TokenizerInit_Invalid; break;
case gbFileError_NotExists: err = TokenizerInit_NotExists; break;
case gbFileError_Permission: err = TokenizerInit_Permission; break;
}
if (err == TokenizerInit_None && gb_file_size(&f) == 0) {
err = TokenizerInit_Empty;
}
gb_file_close(&f);
}
gb_free(heap_allocator(), c_str);
return err;
}
gb_inline void destroy_tokenizer(Tokenizer *t) {
if (t->start != NULL) {
gb_free(heap_allocator(), t->start);
}
for_array(i, t->allocated_strings) {
gb_free(heap_allocator(), t->allocated_strings.e[i].text);
}
array_free(&t->allocated_strings);
}
void tokenizer_skip_whitespace(Tokenizer *t) {
while (rune_is_whitespace(t->curr_rune)) {
advance_to_next_rune(t);
}
}
gb_inline i32 digit_value(Rune r) {
if (gb_char_is_digit(cast(char)r)) {
return r - '0';
} else if (gb_is_between(cast(char)r, 'a', 'f')) {
return r - 'a' + 10;
} else if (gb_is_between(cast(char)r, 'A', 'F')) {
return r - 'A' + 10;
}
return 16; // NOTE(bill): Larger than highest possible
}
gb_inline void scan_mantissa(Tokenizer *t, i32 base) {
// TODO(bill): Allow for underscores in numbers as a number separator
// TODO(bill): Is this a good idea?
// while (digit_value(t->curr_rune) < base || t->curr_rune == '_')
while (digit_value(t->curr_rune) < base) {
advance_to_next_rune(t);
}
}
Token scan_number_to_token(Tokenizer *t, bool seen_decimal_point) {
Token token = {0};
token.kind = Token_Integer;
token.string = make_string(t->curr, 1);
token.pos.file = t->fullpath;
token.pos.line = t->line_count;
token.pos.column = t->curr-t->line+1;
if (seen_decimal_point) {
token.kind = Token_Float;
scan_mantissa(t, 10);
goto exponent;
}
if (t->curr_rune == '0') {
u8 *prev = t->curr;
advance_to_next_rune(t);
if (t->curr_rune == 'b') { // Binary
advance_to_next_rune(t);
scan_mantissa(t, 2);
if (t->curr - prev <= 2)
token.kind = Token_Invalid;
} else if (t->curr_rune == 'o') { // Octal
advance_to_next_rune(t);
scan_mantissa(t, 8);
if (t->curr - prev <= 2)
token.kind = Token_Invalid;
} else if (t->curr_rune == 'd') { // Decimal
advance_to_next_rune(t);
scan_mantissa(t, 10);
if (t->curr - prev <= 2)
token.kind = Token_Invalid;
} else if (t->curr_rune == 'x') { // Hexadecimal
advance_to_next_rune(t);
scan_mantissa(t, 16);
if (t->curr - prev <= 2)
token.kind = Token_Invalid;
} else {
seen_decimal_point = false;
scan_mantissa(t, 10);
if (t->curr_rune == '.' || t->curr_rune == 'e' || t->curr_rune == 'E') {
seen_decimal_point = true;
goto fraction;
}
}
token.string.len = t->curr - token.string.text;
return token;
}
scan_mantissa(t, 10);
fraction:
if (t->curr_rune == '.') {
token.kind = Token_Float;
advance_to_next_rune(t);
scan_mantissa(t, 10);
}
exponent:
if (t->curr_rune == 'e' || t->curr_rune == 'E') {
token.kind = Token_Float;
advance_to_next_rune(t);
if (t->curr_rune == '-' || t->curr_rune == '+') {
advance_to_next_rune(t);
}
scan_mantissa(t, 10);
}
token.string.len = t->curr - token.string.text;
return token;
}
// Quote == " for string
bool scan_escape(Tokenizer *t, Rune quote) {
isize len = 0;
u32 base = 0, max = 0, x = 0;
Rune r = t->curr_rune;
if (r == 'a' ||
r == 'b' ||
r == 'f' ||
r == 'n' ||
r == 'r' ||
r == 't' ||
r == 'v' ||
r == '\\' ||
r == quote) {
advance_to_next_rune(t);
return true;
} else if (gb_is_between(r, '0', '7')) {
len = 3; base = 8; max = 255;
} else if (r == 'x') {
advance_to_next_rune(t);
len = 2; base = 16; max = 255;
} else if (r == 'u') {
advance_to_next_rune(t);
len = 4; base = 16; max = GB_RUNE_MAX;
} else if (r == 'U') {
advance_to_next_rune(t);
len = 8; base = 16; max = GB_RUNE_MAX;
} else {
if (t->curr_rune < 0)
tokenizer_err(t, "Escape sequence was not terminated");
else
tokenizer_err(t, "Unknown escape sequence");
return false;
}
while (len --> 0) {
u32 d = cast(u32)digit_value(t->curr_rune);
if (d >= base) {
if (t->curr_rune < 0)
tokenizer_err(t, "Escape sequence was not terminated");
else
tokenizer_err(t, "Illegal character %d in escape sequence", t->curr_rune);
return false;
}
x = x*base + d;
advance_to_next_rune(t);
}
return true;
}
gb_inline TokenKind token_kind_variant2(Tokenizer *t, TokenKind a, TokenKind b) {
if (t->curr_rune == '=') {
advance_to_next_rune(t);
return b;
}
return a;
}
gb_inline TokenKind token_kind_variant3(Tokenizer *t, TokenKind a, TokenKind b, Rune ch_c, TokenKind c) {
if (t->curr_rune == '=') {
advance_to_next_rune(t);
return b;
}
if (t->curr_rune == ch_c) {
advance_to_next_rune(t);
return c;
}
return a;
}
gb_inline TokenKind token_kind_variant4(Tokenizer *t, TokenKind a, TokenKind b, Rune ch_c, TokenKind c, Rune ch_d, TokenKind d) {
if (t->curr_rune == '=') {
advance_to_next_rune(t);
return b;
} else if (t->curr_rune == ch_c) {
advance_to_next_rune(t);
return c;
} else if (t->curr_rune == ch_d) {
advance_to_next_rune(t);
return d;
}
return a;
}
gb_inline TokenKind token_kind_dub_eq(Tokenizer *t, Rune sing_rune, TokenKind sing, TokenKind sing_eq, TokenKind dub, TokenKind dub_eq) {
if (t->curr_rune == '=') {
advance_to_next_rune(t);
return sing_eq;
} else if (t->curr_rune == sing_rune) {
advance_to_next_rune(t);
if (t->curr_rune == '=') {
advance_to_next_rune(t);
return dub_eq;
}
return dub;
}
return sing;
}
Token tokenizer_get_token(Tokenizer *t) {
Token token = {0};
Rune curr_rune;
tokenizer_skip_whitespace(t);
token.string = make_string(t->curr, 1);
token.pos.file = t->fullpath;
token.pos.line = t->line_count;
token.pos.column = t->curr - t->line + 1;
curr_rune = t->curr_rune;
if (rune_is_letter(curr_rune)) {
token.kind = Token_Identifier;
while (rune_is_letter(t->curr_rune) || rune_is_digit(t->curr_rune)) {
advance_to_next_rune(t);
}
token.string.len = t->curr - token.string.text;
// NOTE(bill): All keywords are > 1
if (token.string.len > 1) {
if (str_eq(token.string, token_strings[Token_as])) {
token.kind = Token_as;
} else if (str_eq(token.string, token_strings[Token_transmute])) {
token.kind = Token_transmute;
} else if (str_eq(token.string, token_strings[Token_down_cast])) {
token.kind = Token_down_cast;
} else if (str_eq(token.string, token_strings[Token_union_cast])) {
token.kind = Token_union_cast;
} else {
for (i32 k = Token__KeywordBegin+1; k < Token__KeywordEnd; k++) {
if (str_eq(token.string, token_strings[k])) {
token.kind = cast(TokenKind)k;
break;
}
}
}
}
} else if (gb_is_between(curr_rune, '0', '9')) {
token = scan_number_to_token(t, false);
} else {
advance_to_next_rune(t);
switch (curr_rune) {
case GB_RUNE_EOF:
token.kind = Token_EOF;
break;
case '\'':
token.kind = Token_Prime;
if (t->curr_rune == '\'') {
advance_to_next_rune(t);
token.kind = Token_DoublePrime;
}
break;
case '`': // Raw String Literal
case '"': // String Literal
{
Rune quote = curr_rune;
token.kind = Token_String;
if (curr_rune == '"') {
for (;;) {
Rune r = t->curr_rune;
if (r == '\n' || r < 0) {
tokenizer_err(t, "String literal not terminated");
break;
}
advance_to_next_rune(t);
if (r == quote)
break;
if (r == '\\')
scan_escape(t, '"');
}
} else {
for (;;) {
Rune r = t->curr_rune;
if (r < 0) {
tokenizer_err(t, "String literal not terminated");
break;
}
advance_to_next_rune(t);
if (r == quote)
break;
}
}
token.string.len = t->curr - token.string.text;
i32 success = unquote_string(heap_allocator(), &token.string);
if (success > 0) {
if (success == 2) {
array_add(&t->allocated_strings, token.string);
}
return token;
} else {
tokenizer_err(t, "Invalid string literal");
}
} break;
case '.':
token.kind = Token_Period; // Default
if (gb_is_between(t->curr_rune, '0', '9')) { // Might be a number
token = scan_number_to_token(t, true);
} else if (t->curr_rune == '.') { // Could be an ellipsis
advance_to_next_rune(t);
token.kind = Token_Ellipsis;
if (t->curr_rune == '<') {
advance_to_next_rune(t);
token.kind = Token_RangeExclusive;
}
}
break;
case '#': token.kind = Token_Hash; break;
case '@': token.kind = Token_At; break;
case '^': token.kind = Token_Pointer; break;
case '?': token.kind = Token_Maybe; break;
case ';': token.kind = Token_Semicolon; break;
case ',': token.kind = Token_Comma; break;
case '(': token.kind = Token_OpenParen; break;
case ')': token.kind = Token_CloseParen; break;
case '[': token.kind = Token_OpenBracket; break;
case ']': token.kind = Token_CloseBracket; break;
case '{': token.kind = Token_OpenBrace; break;
case '}': token.kind = Token_CloseBrace; break;
case ':': token.kind = Token_Colon; break;
case '*': token.kind = token_kind_variant2(t, Token_Mul, Token_MulEq); break;
case '%': token.kind = token_kind_variant2(t, Token_Mod, Token_ModEq); break;
case '=': token.kind = token_kind_variant2(t, Token_Eq, Token_CmpEq); break;
case '~': token.kind = token_kind_variant2(t, Token_Xor, Token_XorEq); break;
case '!': token.kind = token_kind_variant2(t, Token_Not, Token_NotEq); break;
case '+': token.kind = token_kind_variant3(t, Token_Add, Token_AddEq, '+', Token_Increment); break;
case '-': token.kind = token_kind_variant4(t, Token_Sub, Token_SubEq, '-', Token_Decrement, '>', Token_ArrowRight); break;
case '/': {
if (t->curr_rune == '/') {
while (t->curr_rune != '\n') {
advance_to_next_rune(t);
}
token.kind = Token_Comment;
} else if (t->curr_rune == '*') {
isize comment_scope = 1;
advance_to_next_rune(t);
while (comment_scope > 0) {
if (t->curr_rune == '/') {
advance_to_next_rune(t);
if (t->curr_rune == '*') {
advance_to_next_rune(t);
comment_scope++;
}
} else if (t->curr_rune == '*') {
advance_to_next_rune(t);
if (t->curr_rune == '/') {
advance_to_next_rune(t);
comment_scope--;
}
} else {
advance_to_next_rune(t);
}
}
token.kind = Token_Comment;
} else {
token.kind = token_kind_variant2(t, Token_Quo, Token_QuoEq);
}
} break;
case '<':
if (t->curr_rune == '-') {
token.kind = Token_ArrowLeft;
} else {
token.kind = token_kind_dub_eq(t, '<', Token_Lt, Token_LtEq, Token_Shl, Token_ShlEq);
}
break;
case '>':
token.kind = token_kind_dub_eq(t, '>', Token_Gt, Token_GtEq, Token_Shr, Token_ShrEq);
break;
case '&':
token.kind = Token_And;
if (t->curr_rune == '~') {
token.kind = Token_AndNot;
advance_to_next_rune(t);
if (t->curr_rune == '=') {
token.kind = Token_AndNotEq;
advance_to_next_rune(t);
}
} else {
token.kind = token_kind_dub_eq(t, '&', Token_And, Token_AndEq, Token_CmpAnd, Token_CmpAndEq);
}
break;
case '|': token.kind = token_kind_dub_eq(t, '|', Token_Or, Token_OrEq, Token_CmpOr, Token_CmpOrEq); break;
default:
if (curr_rune != GB_RUNE_BOM) {
u8 str[4] = {0};
int len = cast(int)gb_utf8_encode_rune(str, curr_rune);
tokenizer_err(t, "Illegal character: %.*s (%d) ", len, str, curr_rune);
}
token.kind = Token_Invalid;
break;
}
}
token.string.len = t->curr - token.string.text;
return token;
}

66
src/unicode.c Normal file
View File

@@ -0,0 +1,66 @@
#pragma warning(push)
#pragma warning(disable: 4245)
// #include "utf8proc/utf8proc.h"
#include "utf8proc/utf8proc.c"
#pragma warning(pop)
bool rune_is_letter(Rune r) {
if ((r < 0x80 && gb_char_is_alpha(cast(char)r)) ||
r == '_') {
return true;
}
switch (utf8proc_category(r)) {
case UTF8PROC_CATEGORY_LU:
case UTF8PROC_CATEGORY_LL:
case UTF8PROC_CATEGORY_LT:
case UTF8PROC_CATEGORY_LM:
case UTF8PROC_CATEGORY_LO:
return true;
}
return false;
}
bool rune_is_digit(Rune r) {
if (r < 0x80 && gb_is_between(r, '0', '9')) {
return true;
}
return utf8proc_category(r) == UTF8PROC_CATEGORY_ND;
}
bool rune_is_whitespace(Rune r) {
switch (r) {
case ' ':
case '\t':
case '\n':
case '\r':
return true;
}
return false;
}
bool is_string_an_identifier(String s) {
if (s.len < 1) {
return false;
}
isize offset = 0;
while (offset < s.len) {
bool ok = false;
Rune r = -1;
isize size = gb_utf8_decode(s.text+offset, s.len-offset, &r);
if (offset == 0) {
ok = rune_is_letter(r);
} else {
ok = rune_is_letter(r) || rune_is_digit(r);
}
if (!ok) {
return false;
}
offset += size;
}
return offset == s.len;
}