78 Commits

Author SHA1 Message Date
e7d56dd4bd Clean up some styles. 2020-07-15 13:56:03 -07:00
a4fedb276d Adjust margin spacing. 2020-07-15 13:18:34 -07:00
277c0a2ce6 Rework sidenote spacing and TOC. 2020-07-15 13:13:47 -07:00
ef3c61e9e6 Make table of contents dark. 2020-06-30 22:15:22 -07:00
1908126607 Add border to code. 2020-06-30 21:31:16 -07:00
2d77f8489f Move hiding code into margin SCSS. 2020-06-30 21:22:19 -07:00
0371651fdd Fix headings on Starbound post. 2020-06-24 23:01:35 -07:00
01734d24f7 Get started on tables of contents. 2020-06-24 22:46:22 -07:00
71fc0546e0 Move move code into common 'margin node' mixin. 2020-06-24 22:06:08 -07:00
871a745702 Extract margin variables and mixins into separate file. 2020-06-24 14:21:56 -07:00
3f0df8ae0d Add links for 12th part of compiler series. 2020-06-21 22:21:43 -07:00
1746011c16 Publish 12th part of compiler series. 2020-06-21 00:51:04 -07:00
7c4cfbf3d4 Fix typechecking of mutually recursive functions. 2020-06-21 00:47:26 -07:00
8524e098a8 Make proofreading-based fixes. 2020-06-20 23:50:26 -07:00
971f58da9b Finish draft of part 12 of compiler series. 2020-06-20 22:03:57 -07:00
c496be1031 Finish implementation description in part 12. 2020-06-20 20:46:54 -07:00
21851e3a9c Add more content to part 12. 2020-06-19 02:22:08 -07:00
600d5b91ea Remove unneeded parent class. 2020-06-18 23:06:13 -07:00
09b90c3bbc Add line numbers to codelines shortode. 2020-06-18 22:30:01 -07:00
f6ca13d6dc Add more implementation content to part 12. 2020-06-18 22:29:38 -07:00
9c4d7c514f Add more content to post 12 draft. 2020-06-16 23:32:09 -07:00
ad1946e9fb Add first draft of lambdas. 2020-06-14 02:00:20 -07:00
68910458e8 Properly handle null types in pattern typechecking. 2020-06-14 00:43:39 -07:00
240e87eca4 Use mangled names in variable environments. 2020-06-13 23:43:52 -07:00
6b5f7e25b7 Maybe finish the let/in code? 2020-06-01 00:23:41 -07:00
e7229e644f Start working on translation. 2020-05-31 18:52:52 -07:00
08c8aca144 Start working on a lifted version of a definition. 2020-05-31 14:37:33 -07:00
7f8dae74ac Adjust type output. 2020-05-31 00:50:58 -07:00
08503116ff Mark some definitions as global, so as not to capture them. 2020-05-31 00:34:12 -07:00
a1d679a59d No longer destroy the list of free variables.
It so happens that this list will tell us which variables
need to be captured.
2020-05-30 23:29:36 -07:00
4586bd0188 Check for free variables in the environment before generalizing. 2020-05-30 16:40:27 -07:00
a97b50f497 Add parsing of let/in. 2020-05-28 14:44:12 -07:00
c84ff11d0d Add typechecking to let/in expressions. 2020-05-26 00:52:54 -07:00
e966e74487 Extract ordering functionality into definition group. 2020-05-25 23:58:56 -07:00
3865abfb4d Add a struct to contain groups of mutually recursive definitions. 2020-05-25 22:11:45 -07:00
1905601aaa Fork off the 12th version of the compiler. 2020-05-25 21:20:41 -07:00
aacb9e2874 Prefer invisible text to unstyled text. 2020-05-25 20:55:20 -07:00
78f3b18969 Increase padding in post lists. 2020-05-17 22:00:10 -07:00
9f73ca2950 Remove bold font weight on post lists.
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-09 17:53:55 -07:00
035b98a602 Start using description meta.
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-09 17:29:57 -07:00
17f4ebc297 Add media screen to all stylesheets. 2020-05-09 17:02:13 -07:00
906e15674e Make minor edits to the content 2020-05-09 16:52:05 -07:00
85bd0b6c9c Switch to left-alignment on small screens to prevent ugly gaps.
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-09 01:48:22 -07:00
b19e8713e0 Center post titles and word counters. 2020-05-09 01:45:53 -07:00
68fb78e765 Only left-justify post titles 2020-05-09 01:41:10 -07:00
be8a0a4a3a Left jusify text in links to prevent random variation
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-04 19:55:48 -07:00
e883e3c60e Update link lists to visually indicate they're links.
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-04 19:25:03 -07:00
4ede62b39a Stop displaying links in bold.
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-04 03:56:21 -07:00
7d9f487a78 Stop using revert due to Chrome issues
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-04 03:47:53 -07:00
9da584ded4 Make page lists bold
Some checks failed
continuous-integration/drone/push Build is failing
2020-05-04 03:36:23 -07:00
9452c90cf3 Switch back to Raleway 2020-05-04 03:35:24 -07:00
a80064f40a Merge branch 'master' of https://dev.danilafe.com/Web-Projects/blog-static 2020-05-04 03:29:48 -07:00
49691803cc Stop displaying links as inline-block 2020-05-04 03:28:17 -07:00
ee4738b245 Add redesign CSS 2020-05-04 03:23:55 -07:00
b270fa78da Add draft of lazy evaluation post 2020-05-04 01:58:25 -07:00
18339d7e4d Build and test version 11 of the compiler instead of 10.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 21:28:34 -07:00
78563448fb Update to LLVM 10.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 21:24:24 -07:00
144d5f3324 Try to fix compiler build failure
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 21:05:28 -07:00
0fb315ec47 Prevent single-column tables in the case of highlighting
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 20:47:51 -07:00
1ff67341a1 Disable live uploading while the new server doesn't accept the files.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 19:57:50 -07:00
a441280812 Add article about Crystal and Nix with OpenSSL.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-26 19:56:15 -07:00
eda9bbb191 Add more to part 12 of compiler series 2020-04-25 18:07:32 -07:00
2d9da2899f Switch to no line breaks (for Ghostwriter support) 2020-04-25 15:45:15 -07:00
a95490d9d4 Add more content to part 12 of compiler series
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-22 00:30:42 -07:00
44135b1824 Add experimental figure-based styles 2020-04-22 00:30:22 -07:00
4a0367b401 Start work on part 12 of compiler series
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-21 20:44:02 -07:00
c1f0104edb Add missing arity checks to compiler series
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-14 19:21:33 -07:00
c9a7fbf6dd Finalize part 10 of compiler series
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-14 19:06:29 -07:00
1f00b6a3f8 Make compiler check for arity in part 11 of compiler series.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-14 16:20:12 -07:00
acb22c4119 Finish draft of part 11 of compiler series 2020-04-14 16:19:54 -07:00
be2b855ffe Rename inaccurately named rule 2020-04-14 15:58:25 -07:00
88c9418350 Update example to include polymorphism 2020-04-14 15:58:13 -07:00
2255543d94 Add more work on part 11 of compiler series
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-14 00:15:32 -07:00
b4c91d2dd4 Adjust grammar to allow for broader range of type inputs. 2020-04-13 23:17:34 -07:00
98c1b5a3b2 Make substitution replace types at every lookup step 2020-04-13 17:59:57 -07:00
122a1d73d3 Integrate new types into the rest of the project 2020-04-13 17:12:43 -07:00
74e6dba914 Add parsed type data structure, type application and arity. 2020-04-13 17:08:03 -07:00
d7846e0b32 Fork off code for part 11 of compiler series.
Some checks failed
continuous-integration/drone/push Build is failing
2020-04-09 23:48:53 -07:00
133 changed files with 8554 additions and 115 deletions

View File

@@ -10,7 +10,7 @@ steps:
- name: test-compiler - name: test-compiler
image: archlinux image: archlinux
commands: commands:
- pacman -Sy cmake gcc make llvm bison flex gettext --noconfirm - pacman -Sy cmake gcc make llvm bison flex gettext libffi --noconfirm
- cd code/compiler - cd code/compiler
- ./test.sh - ./test.sh
- name: build-live - name: build-live
@@ -22,17 +22,17 @@ steps:
path: /live-output path: /live-output
environment: environment:
HUGO_DESTINATION: /live-output HUGO_DESTINATION: /live-output
- name: upload-live # - name: upload-live
image: eeacms/rsync # image: eeacms/rsync
commands: # commands:
- eval `ssh-agent -s` # - eval `ssh-agent -s`
- echo "$CUSTOM_KEY" | ssh-add - # - echo "$CUSTOM_KEY" | ssh-add -
- mkdir -p ~/.ssh # - mkdir -p ~/.ssh
- echo -e "Host *\n\tStrictHostKeyChecking no\n\n" > ~/.ssh/config # - echo -e "Host *\n\tStrictHostKeyChecking no\n\n" > ~/.ssh/config
- rsync -rv -e "ssh -p 22" /live-output/ blog-live@danilafe.com:/var/www/blog-live/ --checksum # - rsync -rv -e "ssh -p 22" /live-output/ blog-live@danilafe.com:/var/www/blog-live/ --checksum
environment: # environment:
CUSTOM_KEY: # CUSTOM_KEY:
from_secret: live_ssh_key # from_secret: live_ssh_key
volumes: # volumes:
- name: live-output # - name: live-output
path: /live-output # path: /live-output

View File

@@ -0,0 +1,45 @@
cmake_minimum_required(VERSION 3.1)
project(compiler)
# Find all the required packages
find_package(BISON)
find_package(FLEX)
find_package(LLVM REQUIRED CONFIG)
# Set up the flex and bison targets
bison_target(parser
${CMAKE_CURRENT_SOURCE_DIR}/parser.y
${CMAKE_CURRENT_BINARY_DIR}/parser.cpp
COMPILE_FLAGS "-d")
flex_target(scanner
${CMAKE_CURRENT_SOURCE_DIR}/scanner.l
${CMAKE_CURRENT_BINARY_DIR}/scanner.cpp)
add_flex_bison_dependency(scanner parser)
# Find all the relevant LLVM components
llvm_map_components_to_libnames(LLVM_LIBS core x86asmparser x86codegen)
# Create compiler executable
add_executable(compiler
definition.cpp definition.hpp
parsed_type.cpp parsed_type.hpp
ast.cpp ast.hpp
llvm_context.cpp llvm_context.hpp
type_env.cpp type_env.hpp
env.cpp env.hpp
type.cpp type.hpp
error.cpp error.hpp
binop.cpp binop.hpp
instruction.cpp instruction.hpp
graph.cpp graph.hpp
${BISON_parser_OUTPUTS}
${FLEX_scanner_OUTPUTS}
main.cpp
)
# Configure compiler executable
target_include_directories(compiler PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
target_include_directories(compiler PUBLIC ${CMAKE_CURRENT_BINARY_DIR})
target_include_directories(compiler PUBLIC ${LLVM_INCLUDE_DIRS})
target_compile_definitions(compiler PUBLIC ${LLVM_DEFINITIONS})
target_link_libraries(compiler ${LLVM_LIBS})

269
code/compiler/11/ast.cpp Normal file
View File

@@ -0,0 +1,269 @@
#include "ast.hpp"
#include <ostream>
#include "binop.hpp"
#include "error.hpp"
#include "type_env.hpp"
static void print_indent(int n, std::ostream& to) {
while(n--) to << " ";
}
void ast_int::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "INT: " << value << std::endl;
}
void ast_int::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
}
type_ptr ast_int::typecheck(type_mgr& mgr) {
return type_ptr(new type_app(env->lookup_type("Int")));
}
void ast_int::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(new instruction_pushint(value)));
}
void ast_lid::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "LID: " << id << std::endl;
}
void ast_lid::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
if(env->lookup(id) == nullptr) into.insert(id);
}
type_ptr ast_lid::typecheck(type_mgr& mgr) {
return env->lookup(id)->instantiate(mgr);
}
void ast_lid::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(
env->has_variable(id) ?
(instruction*) new instruction_push(env->get_offset(id)) :
(instruction*) new instruction_pushglobal(id)));
}
void ast_uid::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "UID: " << id << std::endl;
}
void ast_uid::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
}
type_ptr ast_uid::typecheck(type_mgr& mgr) {
return env->lookup(id)->instantiate(mgr);
}
void ast_uid::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(new instruction_pushglobal(id)));
}
void ast_binop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "BINOP: " << op_name(op) << std::endl;
left->print(indent + 1, to);
right->print(indent + 1, to);
}
void ast_binop::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
left->find_free(mgr, env, into);
right->find_free(mgr, env, into);
}
type_ptr ast_binop::typecheck(type_mgr& mgr) {
type_ptr ltype = left->typecheck(mgr);
type_ptr rtype = right->typecheck(mgr);
type_ptr ftype = env->lookup(op_name(op))->instantiate(mgr);
if(!ftype) throw type_error(std::string("unknown binary operator ") + op_name(op));
type_ptr return_type = mgr.new_type();
type_ptr arrow_one = type_ptr(new type_arr(rtype, return_type));
type_ptr arrow_two = type_ptr(new type_arr(ltype, arrow_one));
mgr.unify(arrow_two, ftype);
return return_type;
}
void ast_binop::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
right->compile(env, into);
left->compile(env_ptr(new env_offset(1, env)), into);
into.push_back(instruction_ptr(new instruction_pushglobal(op_action(op))));
into.push_back(instruction_ptr(new instruction_mkapp()));
into.push_back(instruction_ptr(new instruction_mkapp()));
}
void ast_app::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "APP:" << std::endl;
left->print(indent + 1, to);
right->print(indent + 1, to);
}
void ast_app::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
left->find_free(mgr, env, into);
right->find_free(mgr, env, into);
}
type_ptr ast_app::typecheck(type_mgr& mgr) {
type_ptr ltype = left->typecheck(mgr);
type_ptr rtype = right->typecheck(mgr);
type_ptr return_type = mgr.new_type();
type_ptr arrow = type_ptr(new type_arr(rtype, return_type));
mgr.unify(arrow, ltype);
return return_type;
}
void ast_app::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
right->compile(env, into);
left->compile(env_ptr(new env_offset(1, env)), into);
into.push_back(instruction_ptr(new instruction_mkapp()));
}
void ast_case::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "CASE: " << std::endl;
for(auto& branch : branches) {
print_indent(indent + 1, to);
branch->pat->print(to);
to << std::endl;
branch->expr->print(indent + 2, to);
}
}
void ast_case::find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into) {
this->env = env;
of->find_free(mgr, env, into);
for(auto& branch : branches) {
type_env_ptr new_env = type_scope(env);
branch->pat->insert_bindings(mgr, new_env);
branch->expr->find_free(mgr, new_env, into);
}
}
type_ptr ast_case::typecheck(type_mgr& mgr) {
type_var* var;
type_ptr case_type = mgr.resolve(of->typecheck(mgr), var);
type_ptr branch_type = mgr.new_type();
for(auto& branch : branches) {
branch->pat->typecheck(case_type, mgr, branch->expr->env);
type_ptr curr_branch_type = branch->expr->typecheck(mgr);
mgr.unify(branch_type, curr_branch_type);
}
input_type = mgr.resolve(case_type, var);
type_app* app_type;
if(!(app_type = dynamic_cast<type_app*>(input_type.get())) ||
!dynamic_cast<type_data*>(app_type->constructor.get())) {
throw type_error("attempting case analysis of non-data type");
}
return branch_type;
}
void ast_case::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
type_app* app_type = dynamic_cast<type_app*>(input_type.get());
type_data* type = dynamic_cast<type_data*>(app_type->constructor.get());
of->compile(env, into);
into.push_back(instruction_ptr(new instruction_eval()));
instruction_jump* jump_instruction = new instruction_jump();
into.push_back(instruction_ptr(jump_instruction));
for(auto& branch : branches) {
std::vector<instruction_ptr> branch_instructions;
pattern_var* vpat;
pattern_constr* cpat;
if((vpat = dynamic_cast<pattern_var*>(branch->pat.get()))) {
branch->expr->compile(env_ptr(new env_offset(1, env)), branch_instructions);
for(auto& constr_pair : type->constructors) {
if(jump_instruction->tag_mappings.find(constr_pair.second.tag) !=
jump_instruction->tag_mappings.end())
break;
jump_instruction->tag_mappings[constr_pair.second.tag] =
jump_instruction->branches.size();
}
jump_instruction->branches.push_back(std::move(branch_instructions));
} else if((cpat = dynamic_cast<pattern_constr*>(branch->pat.get()))) {
env_ptr new_env = env;
for(auto it = cpat->params.rbegin(); it != cpat->params.rend(); it++) {
new_env = env_ptr(new env_var(*it, new_env));
}
branch_instructions.push_back(instruction_ptr(new instruction_split(
cpat->params.size())));
branch->expr->compile(new_env, branch_instructions);
branch_instructions.push_back(instruction_ptr(new instruction_slide(
cpat->params.size())));
int new_tag = type->constructors[cpat->constr].tag;
if(jump_instruction->tag_mappings.find(new_tag) !=
jump_instruction->tag_mappings.end())
throw type_error("technically not a type error: duplicate pattern");
jump_instruction->tag_mappings[new_tag] =
jump_instruction->branches.size();
jump_instruction->branches.push_back(std::move(branch_instructions));
}
}
for(auto& constr_pair : type->constructors) {
if(jump_instruction->tag_mappings.find(constr_pair.second.tag) ==
jump_instruction->tag_mappings.end())
throw type_error("non-total pattern");
}
}
void pattern_var::print(std::ostream& to) const {
to << var;
}
void pattern_var::insert_bindings(type_mgr& mgr, type_env_ptr& env) const {
env->bind(var, mgr.new_type());
}
void pattern_var::typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const {
mgr.unify(env->lookup(var)->instantiate(mgr), t);
}
void pattern_constr::print(std::ostream& to) const {
to << constr;
for(auto& param : params) {
to << " " << param;
}
}
void pattern_constr::insert_bindings(type_mgr& mgr, type_env_ptr& env) const {
for(auto& param : params) {
env->bind(param, mgr.new_type());
}
}
void pattern_constr::typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const {
type_ptr constructor_type = env->lookup(constr)->instantiate(mgr);
if(!constructor_type) {
throw type_error(std::string("pattern using unknown constructor ") + constr);
}
for(auto& param : params) {
type_arr* arr = dynamic_cast<type_arr*>(constructor_type.get());
if(!arr) throw type_error("too many parameters in constructor pattern");
mgr.unify(env->lookup(param)->instantiate(mgr), arr->left);
constructor_type = arr->right;
}
mgr.unify(t, constructor_type);
}

144
code/compiler/11/ast.hpp Normal file
View File

@@ -0,0 +1,144 @@
#pragma once
#include <memory>
#include <vector>
#include <set>
#include "type.hpp"
#include "type_env.hpp"
#include "binop.hpp"
#include "instruction.hpp"
#include "env.hpp"
struct ast {
type_env_ptr env;
virtual ~ast() = default;
virtual void print(int indent, std::ostream& to) const = 0;
virtual void find_free(type_mgr& mgr,
type_env_ptr& env, std::set<std::string>& into) = 0;
virtual type_ptr typecheck(type_mgr& mgr) = 0;
virtual void compile(const env_ptr& env,
std::vector<instruction_ptr>& into) const = 0;
};
using ast_ptr = std::unique_ptr<ast>;
struct pattern {
virtual ~pattern() = default;
virtual void print(std::ostream& to) const = 0;
virtual void insert_bindings(type_mgr& mgr, type_env_ptr& env) const = 0;
virtual void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const = 0;
};
using pattern_ptr = std::unique_ptr<pattern>;
struct branch {
pattern_ptr pat;
ast_ptr expr;
branch(pattern_ptr p, ast_ptr a)
: pat(std::move(p)), expr(std::move(a)) {}
};
using branch_ptr = std::unique_ptr<branch>;
struct ast_int : public ast {
int value;
explicit ast_int(int v)
: value(v) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_lid : public ast {
std::string id;
explicit ast_lid(std::string i)
: id(std::move(i)) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_uid : public ast {
std::string id;
explicit ast_uid(std::string i)
: id(std::move(i)) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_binop : public ast {
binop op;
ast_ptr left;
ast_ptr right;
ast_binop(binop o, ast_ptr l, ast_ptr r)
: op(o), left(std::move(l)), right(std::move(r)) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_app : public ast {
ast_ptr left;
ast_ptr right;
ast_app(ast_ptr l, ast_ptr r)
: left(std::move(l)), right(std::move(r)) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_case : public ast {
ast_ptr of;
type_ptr input_type;
std::vector<branch_ptr> branches;
ast_case(ast_ptr o, std::vector<branch_ptr> b)
: of(std::move(o)), branches(std::move(b)) {}
void print(int indent, std::ostream& to) const;
void find_free(type_mgr& mgr, type_env_ptr& env, std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct pattern_var : public pattern {
std::string var;
pattern_var(std::string v)
: var(std::move(v)) {}
void print(std::ostream &to) const;
void insert_bindings(type_mgr& mgr, type_env_ptr& env) const;
void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const;
};
struct pattern_constr : public pattern {
std::string constr;
std::vector<std::string> params;
pattern_constr(std::string c, std::vector<std::string> p)
: constr(std::move(c)), params(std::move(p)) {}
void print(std::ostream &to) const;
virtual void insert_bindings(type_mgr& mgr, type_env_ptr& env) const;
virtual void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const;
};

View File

@@ -0,0 +1,21 @@
#include "binop.hpp"
std::string op_name(binop op) {
switch(op) {
case PLUS: return "+";
case MINUS: return "-";
case TIMES: return "*";
case DIVIDE: return "/";
}
return "??";
}
std::string op_action(binop op) {
switch(op) {
case PLUS: return "plus";
case MINUS: return "minus";
case TIMES: return "times";
case DIVIDE: return "divide";
}
return "??";
}

View File

@@ -0,0 +1,12 @@
#pragma once
#include <string>
enum binop {
PLUS,
MINUS,
TIMES,
DIVIDE
};
std::string op_name(binop op);
std::string op_action(binop op);

View File

@@ -0,0 +1,109 @@
#include "definition.hpp"
#include "error.hpp"
#include "ast.hpp"
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "type.hpp"
#include "type_env.hpp"
#include <llvm/IR/DerivedTypes.h>
#include <llvm/IR/Function.h>
#include <llvm/IR/Type.h>
void definition_defn::find_free(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
var_env = type_scope(env);
return_type = mgr.new_type();
full_type = return_type;
for(auto it = params.rbegin(); it != params.rend(); it++) {
type_ptr param_type = mgr.new_type();
full_type = type_ptr(new type_arr(param_type, full_type));
var_env->bind(*it, param_type);
}
body->find_free(mgr, var_env, free_variables);
}
void definition_defn::insert_types(type_mgr& mgr) {
env->bind(name, full_type);
}
void definition_defn::typecheck(type_mgr& mgr) {
type_ptr body_type = body->typecheck(mgr);
mgr.unify(return_type, body_type);
}
void definition_defn::compile() {
env_ptr new_env = env_ptr(new env_offset(0, nullptr));
for(auto it = params.rbegin(); it != params.rend(); it++) {
new_env = env_ptr(new env_var(*it, new_env));
}
body->compile(new_env, instructions);
instructions.push_back(instruction_ptr(new instruction_update(params.size())));
instructions.push_back(instruction_ptr(new instruction_pop(params.size())));
}
void definition_defn::declare_llvm(llvm_context& ctx) {
generated_function = ctx.create_custom_function(name, params.size());
}
void definition_defn::generate_llvm(llvm_context& ctx) {
ctx.builder.SetInsertPoint(&generated_function->getEntryBlock());
for(auto& instruction : instructions) {
instruction->gen_llvm(ctx, generated_function);
}
ctx.builder.CreateRetVoid();
}
void definition_data::insert_types(type_env_ptr& env) {
this->env = env;
env->bind_type(name, type_ptr(new type_data(name, vars.size())));
}
void definition_data::insert_constructors() const {
type_ptr this_type_ptr = env->lookup_type(name);
type_data* this_type = static_cast<type_data*>(this_type_ptr.get());
int next_tag = 0;
std::set<std::string> var_set;
type_app* return_app = new type_app(std::move(this_type_ptr));
type_ptr return_type(return_app);
for(auto& var : vars) {
if(var_set.find(var) != var_set.end()) throw 0;
var_set.insert(var);
return_app->arguments.push_back(type_ptr(new type_var(var)));
}
for(auto& constructor : constructors) {
constructor->tag = next_tag;
this_type->constructors[constructor->name] = { next_tag++ };
type_ptr full_type = return_type;
for(auto it = constructor->types.rbegin(); it != constructor->types.rend(); it++) {
type_ptr type = (*it)->to_type(var_set, env);
full_type = type_ptr(new type_arr(type, full_type));
}
type_scheme_ptr full_scheme(new type_scheme(std::move(full_type)));
full_scheme->forall.insert(full_scheme->forall.begin(), vars.begin(), vars.end());
env->bind(constructor->name, full_scheme);
}
}
void definition_data::generate_llvm(llvm_context& ctx) {
for(auto& constructor : constructors) {
auto new_function =
ctx.create_custom_function(constructor->name, constructor->types.size());
std::vector<instruction_ptr> instructions;
instructions.push_back(instruction_ptr(
new instruction_pack(constructor->tag, constructor->types.size())
));
instructions.push_back(instruction_ptr(new instruction_update(0)));
ctx.builder.SetInsertPoint(&new_function->getEntryBlock());
for (auto& instruction : instructions) {
instruction->gen_llvm(ctx, new_function);
}
ctx.builder.CreateRetVoid();
}
}

View File

@@ -0,0 +1,72 @@
#pragma once
#include <memory>
#include <vector>
#include <set>
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "parsed_type.hpp"
#include "type_env.hpp"
struct ast;
using ast_ptr = std::unique_ptr<ast>;
struct constructor {
std::string name;
std::vector<parsed_type_ptr> types;
int8_t tag;
constructor(std::string n, std::vector<parsed_type_ptr> ts)
: name(std::move(n)), types(std::move(ts)) {}
};
using constructor_ptr = std::unique_ptr<constructor>;
struct definition_defn {
std::string name;
std::vector<std::string> params;
ast_ptr body;
type_env_ptr env;
type_env_ptr var_env;
std::set<std::string> free_variables;
type_ptr full_type;
type_ptr return_type;
std::vector<instruction_ptr> instructions;
llvm::Function* generated_function;
definition_defn(std::string n, std::vector<std::string> p, ast_ptr b)
: name(std::move(n)), params(std::move(p)), body(std::move(b)) {
}
void find_free(type_mgr& mgr, type_env_ptr& env);
void insert_types(type_mgr& mgr);
void typecheck(type_mgr& mgr);
void compile();
void declare_llvm(llvm_context& ctx);
void generate_llvm(llvm_context& ctx);
};
using definition_defn_ptr = std::unique_ptr<definition_defn>;
struct definition_data {
std::string name;
std::vector<std::string> vars;
std::vector<constructor_ptr> constructors;
type_env_ptr env;
definition_data(
std::string n,
std::vector<std::string> vs,
std::vector<constructor_ptr> cs)
: name(std::move(n)), vars(std::move(vs)), constructors(std::move(cs)) {}
void insert_types(type_env_ptr& env);
void insert_constructors() const;
void generate_llvm(llvm_context& ctx);
};
using definition_data_ptr = std::unique_ptr<definition_data>;

23
code/compiler/11/env.cpp Normal file
View File

@@ -0,0 +1,23 @@
#include "env.hpp"
int env_var::get_offset(const std::string& name) const {
if(name == this->name) return 0;
if(parent) return parent->get_offset(name) + 1;
throw 0;
}
bool env_var::has_variable(const std::string& name) const {
if(name == this->name) return true;
if(parent) return parent->has_variable(name);
return false;
}
int env_offset::get_offset(const std::string& name) const {
if(parent) return parent->get_offset(name) + offset;
throw 0;
}
bool env_offset::has_variable(const std::string& name) const {
if(parent) return parent->has_variable(name);
return false;
}

34
code/compiler/11/env.hpp Normal file
View File

@@ -0,0 +1,34 @@
#pragma once
#include <memory>
#include <string>
struct env {
virtual ~env() = default;
virtual int get_offset(const std::string& name) const = 0;
virtual bool has_variable(const std::string& name) const = 0;
};
using env_ptr = std::shared_ptr<env>;
struct env_var : public env {
std::string name;
env_ptr parent;
env_var(std::string& n, env_ptr p)
: name(std::move(n)), parent(std::move(p)) {}
int get_offset(const std::string& name) const;
bool has_variable(const std::string& name) const;
};
struct env_offset : public env {
int offset;
env_ptr parent;
env_offset(int o, env_ptr p)
: offset(o), parent(std::move(p)) {}
int get_offset(const std::string& name) const;
bool has_variable(const std::string& name) const;
};

View File

@@ -0,0 +1,5 @@
#include "error.hpp"
const char* type_error::what() const noexcept {
return "an error occured while checking the types of the program";
}

View File

@@ -0,0 +1,21 @@
#pragma once
#include <exception>
#include "type.hpp"
struct type_error : std::exception {
std::string description;
type_error(std::string d)
: description(std::move(d)) {}
const char* what() const noexcept override;
};
struct unification_error : public type_error {
type_ptr left;
type_ptr right;
unification_error(type_ptr l, type_ptr r)
: left(std::move(l)), right(std::move(r)),
type_error("failed to unify types") {}
};

View File

@@ -0,0 +1,2 @@
data Bool = { True, False }
defn main = { 3 + True }

View File

@@ -0,0 +1 @@
defn main = { 1 2 3 4 5 }

View File

@@ -0,0 +1,8 @@
data List = { Nil, Cons Int List }
defn head l = {
case l of {
Nil -> { 0 }
Cons x y z -> { x }
}
}

View File

@@ -0,0 +1,8 @@
data Bool = { True, False }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn main = { if (if True False True) 11 3 }

View File

@@ -0,0 +1,32 @@
data List a = { Nil, Cons a (List a) }
defn map f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { Cons (f x) (map f xs) }
}
}
defn foldl f b l = {
case l of {
Nil -> { b }
Cons x xs -> { foldl f (f b x) xs }
}
}
defn foldr f b l = {
case l of {
Nil -> { b }
Cons x xs -> { f x (foldr f b xs) }
}
}
defn list = { Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil))) }
defn add x y = { x + y }
defn sum l = { foldr add 0 l }
defn skipAdd x y = { y + 1 }
defn length l = { foldr skipAdd 0 l }
defn main = { sum list + length list }

View File

@@ -0,0 +1,25 @@
data Bool = { True, False }
data List = { Nil, Cons Int List }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn oddEven l e = {
case l of {
Nil -> { e }
Cons x xs -> { evenOdd xs e }
}
}
defn evenOdd l e = {
case l of {
Nil -> { e }
Cons x xs -> { oddEven xs e }
}
}
defn main = { if (oddEven (Cons 1 (Cons 2 (Cons 3 Nil))) True) (oddEven (Cons 1 (Cons 2 (Cons 3 Nil))) 1) 3 }

View File

@@ -0,0 +1,17 @@
data Pair a b = { MkPair a b }
defn fst p = {
case p of {
MkPair a b -> { a }
}
}
defn snd p = {
case p of {
MkPair a b -> { b }
}
}
defn pair = { MkPair 1 (MkPair 2 3) }
defn main = { fst pair + snd (snd pair) }

View File

@@ -0,0 +1,122 @@
data List = { Nil, Cons Nat List }
data Bool = { True, False }
data Nat = { O, S Nat }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn toInt n = {
case n of {
O -> { 0 }
S np -> { 1 + toInt np }
}
}
defn lte n m = {
case m of {
O -> {
case n of {
O -> { True }
S np -> { False }
}
}
S mp -> {
case n of {
O -> { True }
S np -> { lte np mp }
}
}
}
}
defn minus n m = {
case m of {
O -> { n }
S mp -> {
case n of {
O -> { O }
S np -> {
minus np mp
}
}
}
}
}
defn mod n m = {
if (lte m n) (mod (minus n m) m) n
}
defn notDivisibleBy n m = {
case (mod m n) of {
O -> { False }
S mp -> { True }
}
}
defn filter f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { if (f x) (Cons x (filter f xs)) (filter f xs) }
}
}
defn map f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { Cons (f x) (map f xs) }
}
}
defn nats = {
Cons (S (S O)) (map S nats)
}
defn primesRec l = {
case l of {
Nil -> { Nil }
Cons p xs -> { Cons p (primesRec (filter (notDivisibleBy p) xs)) }
}
}
defn primes = {
primesRec nats
}
defn take n l = {
case l of {
Nil -> { Nil }
Cons x xs -> {
case n of {
O -> { Nil }
S np -> { Cons x (take np xs) }
}
}
}
}
defn head l = {
case l of {
Nil -> { O }
Cons x xs -> { x }
}
}
defn reverseAcc a l = {
case l of {
Nil -> { a }
Cons x xs -> { reverseAcc (Cons x a) xs }
}
}
defn reverse l = {
reverseAcc Nil l
}
defn main = {
toInt (head (reverse (take ((S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S O))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))) primes)))
}

View File

@@ -0,0 +1,31 @@
#include "../runtime.h"
void f_add(struct stack* s) {
struct node_num* left = (struct node_num*) eval(stack_peek(s, 0));
struct node_num* right = (struct node_num*) eval(stack_peek(s, 1));
stack_push(s, (struct node_base*) alloc_num(left->value + right->value));
}
void f_main(struct stack* s) {
// PushInt 320
stack_push(s, (struct node_base*) alloc_num(320));
// PushInt 6
stack_push(s, (struct node_base*) alloc_num(6));
// PushGlobal f_add (the function for +)
stack_push(s, (struct node_base*) alloc_global(f_add, 2));
struct node_base* left;
struct node_base* right;
// MkApp
left = stack_pop(s);
right = stack_pop(s);
stack_push(s, (struct node_base*) alloc_app(left, right));
// MkApp
left = stack_pop(s);
right = stack_pop(s);
stack_push(s, (struct node_base*) alloc_app(left, right));
}

View File

@@ -0,0 +1,2 @@
defn main = { sum 320 6 }
defn sum x y = { x + y }

View File

@@ -0,0 +1,3 @@
defn add x y = { x + y }
defn double x = { add x x }
defn main = { double 163 }

View File

@@ -0,0 +1,9 @@
data List a = { Nil, Cons a (List a) }
data Bool = { True, False }
defn length l = {
case l of {
Nil -> { 0 }
Cons x xs -> { 1 + length xs }
}
}
defn main = { length (Cons 1 (Cons 2 (Cons 3 Nil))) + length (Cons True (Cons False (Cons True Nil))) }

View File

@@ -0,0 +1,16 @@
data List = { Nil, Cons Int List }
defn add x y = { x + y }
defn mul x y = { x * y }
defn foldr f b l = {
case l of {
Nil -> { b }
Cons x xs -> { f x (foldr f b xs) }
}
}
defn main = {
foldr add 0 (Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil)))) +
foldr mul 1 (Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil))))
}

View File

@@ -0,0 +1,17 @@
data List = { Nil, Cons Int List }
defn sumZip l m = {
case l of {
Nil -> { 0 }
Cons x xs -> {
case m of {
Nil -> { 0 }
Cons y ys -> { x + y + sumZip xs ys }
}
}
}
}
defn ones = { Cons 1 ones }
defn main = { sumZip ones (Cons 1 (Cons 2 (Cons 3 Nil))) }

114
code/compiler/11/graph.cpp Normal file
View File

@@ -0,0 +1,114 @@
#include "graph.hpp"
std::set<function_graph::edge> function_graph::compute_transitive_edges() {
std::set<edge> transitive_edges;
transitive_edges.insert(edges.begin(), edges.end());
for(auto& connector : adjacency_lists) {
for(auto& from : adjacency_lists) {
edge to_connector { from.first, connector.first };
for(auto& to : adjacency_lists) {
edge full_jump { from.first, to.first };
if(transitive_edges.find(full_jump) != transitive_edges.end()) continue;
edge from_connector { connector.first, to.first };
if(transitive_edges.find(to_connector) != transitive_edges.end() &&
transitive_edges.find(from_connector) != transitive_edges.end())
transitive_edges.insert(std::move(full_jump));
}
}
}
return transitive_edges;
}
void function_graph::create_groups(
const std::set<edge>& transitive_edges,
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
group_id id_counter = 0;
for(auto& vertex : adjacency_lists) {
if(group_ids.find(vertex.first) != group_ids.end())
continue;
data_ptr new_group(new group_data);
new_group->functions.insert(vertex.first);
group_data_map[id_counter] = new_group;
group_ids[vertex.first] = id_counter;
for(auto& other_vertex : adjacency_lists) {
if(transitive_edges.find({vertex.first, other_vertex.first}) != transitive_edges.end() &&
transitive_edges.find({other_vertex.first, vertex.first}) != transitive_edges.end()) {
group_ids[other_vertex.first] = id_counter;
new_group->functions.insert(other_vertex.first);
}
}
id_counter++;
}
}
void function_graph::create_edges(
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
std::set<std::pair<group_id, group_id>> group_edges;
for(auto& vertex : adjacency_lists) {
auto vertex_id = group_ids[vertex.first];
auto& vertex_data = group_data_map[vertex_id];
for(auto& other_vertex : vertex.second) {
auto other_id = group_ids[other_vertex];
if(vertex_id == other_id) continue;
if(group_edges.find({vertex_id, other_id}) != group_edges.end())
continue;
group_edges.insert({vertex_id, other_id});
vertex_data->adjacency_list.insert(other_id);
group_data_map[other_id]->indegree++;
}
}
}
std::vector<group_ptr> function_graph::generate_order(
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
std::queue<group_id> id_queue;
std::vector<group_ptr> output;
for(auto& group : group_data_map) {
if(group.second->indegree == 0) id_queue.push(group.first);
}
while(!id_queue.empty()) {
auto new_id = id_queue.front();
auto& group_data = group_data_map[new_id];
group_ptr output_group(new group);
output_group->members = std::move(group_data->functions);
id_queue.pop();
for(auto& adjacent_group : group_data->adjacency_list) {
if(--group_data_map[adjacent_group]->indegree == 0)
id_queue.push(adjacent_group);
}
output.push_back(std::move(output_group));
}
return output;
}
std::set<function>& function_graph::add_function(const function& f) {
auto adjacency_list_it = adjacency_lists.find(f);
if(adjacency_list_it != adjacency_lists.end()) {
return adjacency_list_it->second;
} else {
return adjacency_lists[f] = { };
}
}
void function_graph::add_edge(const function& from, const function& to) {
add_function(from).insert(to);
edges.insert({ from, to });
}
std::vector<group_ptr> function_graph::compute_order() {
std::set<edge> transitive_edges = compute_transitive_edges();
std::map<function, group_id> group_ids;
std::map<group_id, data_ptr> group_data_map;
create_groups(transitive_edges, group_ids, group_data_map);
create_edges(group_ids, group_data_map);
return generate_order(group_ids, group_data_map);
}

View File

@@ -0,0 +1,52 @@
#pragma once
#include <algorithm>
#include <cstddef>
#include <queue>
#include <set>
#include <string>
#include <map>
#include <memory>
#include <vector>
#include <iostream>
using function = std::string;
struct group {
std::set<function> members;
};
using group_ptr = std::unique_ptr<group>;
class function_graph {
using group_id = size_t;
struct group_data {
std::set<function> functions;
std::set<group_id> adjacency_list;
size_t indegree;
};
using data_ptr = std::shared_ptr<group_data>;
using edge = std::pair<function, function>;
using group_edge = std::pair<group_id, group_id>;
std::map<function, std::set<function>> adjacency_lists;
std::set<edge> edges;
std::set<edge> compute_transitive_edges();
void create_groups(
const std::set<edge>&,
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
void create_edges(
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
std::vector<group_ptr> generate_order(
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
public:
std::set<function>& add_function(const function& f);
void add_edge(const function& from, const function& to);
std::vector<group_ptr> compute_order();
};

View File

@@ -0,0 +1,177 @@
#include "instruction.hpp"
#include "llvm_context.hpp"
#include <llvm/IR/BasicBlock.h>
#include <llvm/IR/Function.h>
using namespace llvm;
static void print_indent(int n, std::ostream& to) {
while(n--) to << " ";
}
void instruction_pushint::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "PushInt(" << value << ")" << std::endl;
}
void instruction_pushint::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_push(f, ctx.create_num(f, ctx.create_i32(value)));
}
void instruction_pushglobal::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "PushGlobal(" << name << ")" << std::endl;
}
void instruction_pushglobal::gen_llvm(llvm_context& ctx, Function* f) const {
auto& global_f = ctx.custom_functions.at("f_" + name);
auto arity = ctx.create_i32(global_f->arity);
ctx.create_push(f, ctx.create_global(f, global_f->function, arity));
}
void instruction_push::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Push(" << offset << ")" << std::endl;
}
void instruction_push::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_push(f, ctx.create_peek(f, ctx.create_size(offset)));
}
void instruction_pop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Pop(" << count << ")" << std::endl;
}
void instruction_pop::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_popn(f, ctx.create_size(count));
}
void instruction_mkapp::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "MkApp()" << std::endl;
}
void instruction_mkapp::gen_llvm(llvm_context& ctx, Function* f) const {
auto left = ctx.create_pop(f);
auto right = ctx.create_pop(f);
ctx.create_push(f, ctx.create_app(f, left, right));
}
void instruction_update::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Update(" << offset << ")" << std::endl;
}
void instruction_update::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_update(f, ctx.create_size(offset));
}
void instruction_pack::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Pack(" << tag << ", " << size << ")" << std::endl;
}
void instruction_pack::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_pack(f, ctx.create_size(size), ctx.create_i8(tag));
}
void instruction_split::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Split()" << std::endl;
}
void instruction_split::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_split(f, ctx.create_size(size));
}
void instruction_jump::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Jump(" << std::endl;
for(auto& instruction_set : branches) {
for(auto& instruction : instruction_set) {
instruction->print(indent + 2, to);
}
to << std::endl;
}
print_indent(indent, to);
to << ")" << std::endl;
}
void instruction_jump::gen_llvm(llvm_context& ctx, Function* f) const {
auto top_node = ctx.create_peek(f, ctx.create_size(0));
auto tag = ctx.unwrap_data_tag(top_node);
auto safety_block = BasicBlock::Create(ctx.ctx, "safety", f);
auto switch_op = ctx.builder.CreateSwitch(tag, safety_block, tag_mappings.size());
std::vector<BasicBlock*> blocks;
for(auto& branch : branches) {
auto branch_block = BasicBlock::Create(ctx.ctx, "branch", f);
ctx.builder.SetInsertPoint(branch_block);
for(auto& instruction : branch) {
instruction->gen_llvm(ctx, f);
}
ctx.builder.CreateBr(safety_block);
blocks.push_back(branch_block);
}
for(auto& mapping : tag_mappings) {
switch_op->addCase(ctx.create_i8(mapping.first), blocks[mapping.second]);
}
ctx.builder.SetInsertPoint(safety_block);
}
void instruction_slide::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Slide(" << offset << ")" << std::endl;
}
void instruction_slide::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_slide(f, ctx.create_size(offset));
}
void instruction_binop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "BinOp(" << op_action(op) << ")" << std::endl;
}
void instruction_binop::gen_llvm(llvm_context& ctx, Function* f) const {
auto left_int = ctx.unwrap_num(ctx.create_pop(f));
auto right_int = ctx.unwrap_num(ctx.create_pop(f));
llvm::Value* result;
switch(op) {
case PLUS: result = ctx.builder.CreateAdd(left_int, right_int); break;
case MINUS: result = ctx.builder.CreateSub(left_int, right_int); break;
case TIMES: result = ctx.builder.CreateMul(left_int, right_int); break;
case DIVIDE: result = ctx.builder.CreateSDiv(left_int, right_int); break;
}
ctx.create_push(f, ctx.create_num(f, result));
}
void instruction_eval::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Eval()" << std::endl;
}
void instruction_eval::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_unwind(f);
}
void instruction_alloc::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Alloc(" << amount << ")" << std::endl;
}
void instruction_alloc::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_alloc(f, ctx.create_size(amount));
}
void instruction_unwind::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Unwind()" << std::endl;
}
void instruction_unwind::gen_llvm(llvm_context& ctx, Function* f) const {
// Nothing
}

View File

@@ -0,0 +1,142 @@
#pragma once
#include <llvm/IR/Function.h>
#include <string>
#include <memory>
#include <vector>
#include <map>
#include <ostream>
#include "binop.hpp"
#include "llvm_context.hpp"
struct instruction {
virtual ~instruction() = default;
virtual void print(int indent, std::ostream& to) const = 0;
virtual void gen_llvm(llvm_context& ctx, llvm::Function* f) const = 0;
};
using instruction_ptr = std::unique_ptr<instruction>;
struct instruction_pushint : public instruction {
int value;
instruction_pushint(int v)
: value(v) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pushglobal : public instruction {
std::string name;
instruction_pushglobal(std::string n)
: name(std::move(n)) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_push : public instruction {
int offset;
instruction_push(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pop : public instruction {
int count;
instruction_pop(int c)
: count(c) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_mkapp : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_update : public instruction {
int offset;
instruction_update(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pack : public instruction {
int tag;
int size;
instruction_pack(int t, int s)
: tag(t), size(s) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_split : public instruction {
int size;
instruction_split(int s)
: size(s) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_jump : public instruction {
std::vector<std::vector<instruction_ptr>> branches;
std::map<int, int> tag_mappings;
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_slide : public instruction {
int offset;
instruction_slide(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_binop : public instruction {
binop op;
instruction_binop(binop o)
: op(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_eval : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_alloc : public instruction {
int amount;
instruction_alloc(int a)
: amount(a) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_unwind : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};

View File

@@ -0,0 +1,278 @@
#include "llvm_context.hpp"
#include <llvm/IR/DerivedTypes.h>
using namespace llvm;
void llvm_context::create_types() {
stack_type = StructType::create(ctx, "stack");
gmachine_type = StructType::create(ctx, "gmachine");
stack_ptr_type = PointerType::getUnqual(stack_type);
gmachine_ptr_type = PointerType::getUnqual(gmachine_type);
tag_type = IntegerType::getInt8Ty(ctx);
struct_types["node_base"] = StructType::create(ctx, "node_base");
struct_types["node_app"] = StructType::create(ctx, "node_app");
struct_types["node_num"] = StructType::create(ctx, "node_num");
struct_types["node_global"] = StructType::create(ctx, "node_global");
struct_types["node_ind"] = StructType::create(ctx, "node_ind");
struct_types["node_data"] = StructType::create(ctx, "node_data");
node_ptr_type = PointerType::getUnqual(struct_types.at("node_base"));
function_type = FunctionType::get(Type::getVoidTy(ctx), { gmachine_ptr_type }, false);
gmachine_type->setBody(
stack_ptr_type,
node_ptr_type,
IntegerType::getInt64Ty(ctx),
IntegerType::getInt64Ty(ctx)
);
struct_types.at("node_base")->setBody(
IntegerType::getInt32Ty(ctx),
IntegerType::getInt8Ty(ctx),
node_ptr_type
);
struct_types.at("node_app")->setBody(
struct_types.at("node_base"),
node_ptr_type,
node_ptr_type
);
struct_types.at("node_num")->setBody(
struct_types.at("node_base"),
IntegerType::getInt32Ty(ctx)
);
struct_types.at("node_global")->setBody(
struct_types.at("node_base"),
FunctionType::get(Type::getVoidTy(ctx), { stack_ptr_type }, false)
);
struct_types.at("node_ind")->setBody(
struct_types.at("node_base"),
node_ptr_type
);
struct_types.at("node_data")->setBody(
struct_types.at("node_base"),
IntegerType::getInt8Ty(ctx),
PointerType::getUnqual(node_ptr_type)
);
}
void llvm_context::create_functions() {
auto void_type = Type::getVoidTy(ctx);
auto sizet_type = IntegerType::get(ctx, sizeof(size_t) * 8);
functions["stack_init"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_init",
&module
);
functions["stack_free"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_free",
&module
);
functions["stack_push"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_push",
&module
);
functions["stack_pop"] = Function::Create(
FunctionType::get(node_ptr_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_pop",
&module
);
functions["stack_peek"] = Function::Create(
FunctionType::get(node_ptr_type, { stack_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_peek",
&module
);
functions["stack_popn"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_popn",
&module
);
functions["gmachine_slide"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_slide",
&module
);
functions["gmachine_update"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_update",
&module
);
functions["gmachine_alloc"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_alloc",
&module
);
functions["gmachine_pack"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type, tag_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_pack",
&module
);
functions["gmachine_split"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_split",
&module
);
functions["gmachine_track"] = Function::Create(
FunctionType::get(node_ptr_type, { gmachine_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_track",
&module
);
auto int32_type = IntegerType::getInt32Ty(ctx);
functions["alloc_app"] = Function::Create(
FunctionType::get(node_ptr_type, { node_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_app",
&module
);
functions["alloc_num"] = Function::Create(
FunctionType::get(node_ptr_type, { int32_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_num",
&module
);
functions["alloc_global"] = Function::Create(
FunctionType::get(node_ptr_type, { function_type, int32_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_global",
&module
);
functions["alloc_ind"] = Function::Create(
FunctionType::get(node_ptr_type, { node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_ind",
&module
);
functions["unwind"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"unwind",
&module
);
}
ConstantInt* llvm_context::create_i8(int8_t i) {
return ConstantInt::get(ctx, APInt(8, i));
}
ConstantInt* llvm_context::create_i32(int32_t i) {
return ConstantInt::get(ctx, APInt(32, i));
}
ConstantInt* llvm_context::create_size(size_t i) {
return ConstantInt::get(ctx, APInt(sizeof(size_t) * 8, i));
}
Value* llvm_context::create_pop(Function* f) {
auto pop_f = functions.at("stack_pop");
return builder.CreateCall(pop_f, { unwrap_gmachine_stack_ptr(f->arg_begin()) });
}
Value* llvm_context::create_peek(Function* f, Value* off) {
auto peek_f = functions.at("stack_peek");
return builder.CreateCall(peek_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), off });
}
void llvm_context::create_push(Function* f, Value* v) {
auto push_f = functions.at("stack_push");
builder.CreateCall(push_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), v });
}
void llvm_context::create_popn(Function* f, Value* off) {
auto popn_f = functions.at("stack_popn");
builder.CreateCall(popn_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), off });
}
void llvm_context::create_update(Function* f, Value* off) {
auto update_f = functions.at("gmachine_update");
builder.CreateCall(update_f, { f->arg_begin(), off });
}
void llvm_context::create_pack(Function* f, Value* c, Value* t) {
auto pack_f = functions.at("gmachine_pack");
builder.CreateCall(pack_f, { f->arg_begin(), c, t });
}
void llvm_context::create_split(Function* f, Value* c) {
auto split_f = functions.at("gmachine_split");
builder.CreateCall(split_f, { f->arg_begin(), c });
}
void llvm_context::create_slide(Function* f, Value* off) {
auto slide_f = functions.at("gmachine_slide");
builder.CreateCall(slide_f, { f->arg_begin(), off });
}
void llvm_context::create_alloc(Function* f, Value* n) {
auto alloc_f = functions.at("gmachine_alloc");
builder.CreateCall(alloc_f, { f->arg_begin(), n });
}
Value* llvm_context::create_track(Function* f, Value* v) {
auto track_f = functions.at("gmachine_track");
return builder.CreateCall(track_f, { f->arg_begin(), v });
}
void llvm_context::create_unwind(Function* f) {
auto unwind_f = functions.at("unwind");
builder.CreateCall(unwind_f, { f->args().begin() });
}
Value* llvm_context::unwrap_gmachine_stack_ptr(Value* g) {
auto offset_0 = create_i32(0);
return builder.CreateGEP(g, { offset_0, offset_0 });
}
Value* llvm_context::unwrap_num(Value* v) {
auto num_ptr_type = PointerType::getUnqual(struct_types.at("node_num"));
auto cast = builder.CreatePointerCast(v, num_ptr_type);
auto offset_0 = create_i32(0);
auto offset_1 = create_i32(1);
auto int_ptr = builder.CreateGEP(cast, { offset_0, offset_1 });
return builder.CreateLoad(int_ptr);
}
Value* llvm_context::create_num(Function* f, Value* v) {
auto alloc_num_f = functions.at("alloc_num");
auto alloc_num_call = builder.CreateCall(alloc_num_f, { v });
return create_track(f, alloc_num_call);
}
Value* llvm_context::unwrap_data_tag(Value* v) {
auto data_ptr_type = PointerType::getUnqual(struct_types.at("node_data"));
auto cast = builder.CreatePointerCast(v, data_ptr_type);
auto offset_0 = create_i32(0);
auto offset_1 = create_i32(1);
auto tag_ptr = builder.CreateGEP(cast, { offset_0, offset_1 });
return builder.CreateLoad(tag_ptr);
}
Value* llvm_context::create_global(Function* f, Value* gf, Value* a) {
auto alloc_global_f = functions.at("alloc_global");
auto alloc_global_call = builder.CreateCall(alloc_global_f, { gf, a });
return create_track(f, alloc_global_call);
}
Value* llvm_context::create_app(Function* f, Value* l, Value* r) {
auto alloc_app_f = functions.at("alloc_app");
auto alloc_app_call = builder.CreateCall(alloc_app_f, { l, r });
return create_track(f, alloc_app_call);
}
llvm::Function* llvm_context::create_custom_function(std::string name, int32_t arity) {
auto void_type = llvm::Type::getVoidTy(ctx);
auto new_function = llvm::Function::Create(
function_type,
llvm::Function::LinkageTypes::ExternalLinkage,
"f_" + name,
&module
);
auto start_block = llvm::BasicBlock::Create(ctx, "entry", new_function);
auto new_custom_f = custom_function_ptr(new custom_function());
new_custom_f->arity = arity;
new_custom_f->function = new_function;
custom_functions["f_" + name] = std::move(new_custom_f);
return new_function;
}

View File

@@ -0,0 +1,72 @@
#pragma once
#include <llvm/IR/DerivedTypes.h>
#include <llvm/IR/Function.h>
#include <llvm/IR/LLVMContext.h>
#include <llvm/IR/IRBuilder.h>
#include <llvm/IR/Module.h>
#include <llvm/IR/Value.h>
#include <map>
struct llvm_context {
struct custom_function {
llvm::Function* function;
int32_t arity;
};
using custom_function_ptr = std::unique_ptr<custom_function>;
llvm::LLVMContext ctx;
llvm::IRBuilder<> builder;
llvm::Module module;
std::map<std::string, custom_function_ptr> custom_functions;
std::map<std::string, llvm::Function*> functions;
std::map<std::string, llvm::StructType*> struct_types;
llvm::StructType* stack_type;
llvm::StructType* gmachine_type;
llvm::PointerType* stack_ptr_type;
llvm::PointerType* gmachine_ptr_type;
llvm::PointerType* node_ptr_type;
llvm::IntegerType* tag_type;
llvm::FunctionType* function_type;
llvm_context()
: builder(ctx), module("bloglang", ctx) {
create_types();
create_functions();
}
void create_types();
void create_functions();
llvm::ConstantInt* create_i8(int8_t);
llvm::ConstantInt* create_i32(int32_t);
llvm::ConstantInt* create_size(size_t);
llvm::Value* create_pop(llvm::Function*);
llvm::Value* create_peek(llvm::Function*, llvm::Value*);
void create_push(llvm::Function*, llvm::Value*);
void create_popn(llvm::Function*, llvm::Value*);
void create_update(llvm::Function*, llvm::Value*);
void create_pack(llvm::Function*, llvm::Value*, llvm::Value*);
void create_split(llvm::Function*, llvm::Value*);
void create_slide(llvm::Function*, llvm::Value*);
void create_alloc(llvm::Function*, llvm::Value*);
llvm::Value* create_track(llvm::Function*, llvm::Value*);
void create_unwind(llvm::Function*);
llvm::Value* unwrap_gmachine_stack_ptr(llvm::Value*);
llvm::Value* unwrap_num(llvm::Value*);
llvm::Value* create_num(llvm::Function*, llvm::Value*);
llvm::Value* unwrap_data_tag(llvm::Value*);
llvm::Value* create_global(llvm::Function*, llvm::Value*, llvm::Value*);
llvm::Value* create_app(llvm::Function*, llvm::Value*, llvm::Value*);
llvm::Function* create_custom_function(std::string name, int32_t arity);
};

205
code/compiler/11/main.cpp Normal file
View File

@@ -0,0 +1,205 @@
#include "ast.hpp"
#include <iostream>
#include "binop.hpp"
#include "definition.hpp"
#include "graph.hpp"
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "parser.hpp"
#include "error.hpp"
#include "type.hpp"
#include "llvm/IR/LegacyPassManager.h"
#include "llvm/IR/Verifier.h"
#include "llvm/Support/TargetSelect.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Target/TargetMachine.h"
void yy::parser::error(const std::string& msg) {
std::cout << "An error occured: " << msg << std::endl;
}
extern std::map<std::string, definition_data_ptr> defs_data;
extern std::map<std::string, definition_defn_ptr> defs_defn;
void typecheck_program(
const std::map<std::string, definition_data_ptr>& defs_data,
const std::map<std::string, definition_defn_ptr>& defs_defn,
type_mgr& mgr, type_env_ptr& env) {
type_ptr int_type = type_ptr(new type_base("Int"));
env->bind_type("Int", int_type);
type_ptr int_type_app = type_ptr(new type_app(int_type));
type_ptr binop_type = type_ptr(new type_arr(
int_type_app,
type_ptr(new type_arr(int_type_app, int_type_app))));
env->bind("+", binop_type);
env->bind("-", binop_type);
env->bind("*", binop_type);
env->bind("/", binop_type);
for(auto& def_data : defs_data) {
def_data.second->insert_types(env);
}
for(auto& def_data : defs_data) {
def_data.second->insert_constructors();
}
function_graph dependency_graph;
for(auto& def_defn : defs_defn) {
def_defn.second->find_free(mgr, env);
dependency_graph.add_function(def_defn.second->name);
for(auto& dependency : def_defn.second->free_variables) {
if(defs_defn.find(dependency) == defs_defn.end())
throw 0;
dependency_graph.add_edge(def_defn.second->name, dependency);
}
}
std::vector<group_ptr> groups = dependency_graph.compute_order();
for(auto it = groups.rbegin(); it != groups.rend(); it++) {
auto& group = *it;
for(auto& def_defnn_name : group->members) {
auto& def_defn = defs_defn.find(def_defnn_name)->second;
def_defn->insert_types(mgr);
}
for(auto& def_defnn_name : group->members) {
auto& def_defn = defs_defn.find(def_defnn_name)->second;
def_defn->typecheck(mgr);
}
for(auto& def_defnn_name : group->members) {
env->generalize(def_defnn_name, mgr);
}
}
for(auto& pair : env->names) {
std::cout << pair.first << ": ";
pair.second->print(mgr, std::cout);
std::cout << std::endl;
}
}
void compile_program(const std::map<std::string, definition_defn_ptr>& defs_defn) {
for(auto& def_defn : defs_defn) {
def_defn.second->compile();
for(auto& instruction : def_defn.second->instructions) {
instruction->print(0, std::cout);
}
std::cout << std::endl;
}
}
void gen_llvm_internal_op(llvm_context& ctx, binop op) {
auto new_function = ctx.create_custom_function(op_action(op), 2);
std::vector<instruction_ptr> instructions;
instructions.push_back(instruction_ptr(new instruction_push(1)));
instructions.push_back(instruction_ptr(new instruction_eval()));
instructions.push_back(instruction_ptr(new instruction_push(1)));
instructions.push_back(instruction_ptr(new instruction_eval()));
instructions.push_back(instruction_ptr(new instruction_binop(op)));
instructions.push_back(instruction_ptr(new instruction_update(2)));
instructions.push_back(instruction_ptr(new instruction_pop(2)));
ctx.builder.SetInsertPoint(&new_function->getEntryBlock());
for(auto& instruction : instructions) {
instruction->gen_llvm(ctx, new_function);
}
ctx.builder.CreateRetVoid();
}
void output_llvm(llvm_context& ctx, const std::string& filename) {
std::string targetTriple = llvm::sys::getDefaultTargetTriple();
llvm::InitializeNativeTarget();
llvm::InitializeNativeTargetAsmParser();
llvm::InitializeNativeTargetAsmPrinter();
std::string error;
const llvm::Target* target =
llvm::TargetRegistry::lookupTarget(targetTriple, error);
if (!target) {
std::cerr << error << std::endl;
} else {
std::string cpu = "generic";
std::string features = "";
llvm::TargetOptions options;
llvm::TargetMachine* targetMachine =
target->createTargetMachine(targetTriple, cpu, features,
options, llvm::Optional<llvm::Reloc::Model>());
ctx.module.setDataLayout(targetMachine->createDataLayout());
ctx.module.setTargetTriple(targetTriple);
std::error_code ec;
llvm::raw_fd_ostream file(filename, ec, llvm::sys::fs::F_None);
if (ec) {
throw 0;
} else {
llvm::CodeGenFileType type = llvm::CGFT_ObjectFile;
llvm::legacy::PassManager pm;
if (targetMachine->addPassesToEmitFile(pm, file, NULL, type)) {
throw 0;
} else {
pm.run(ctx.module);
file.close();
}
}
}
}
void gen_llvm(
const std::map<std::string, definition_data_ptr>& defs_data,
const std::map<std::string, definition_defn_ptr>& defs_defn) {
llvm_context ctx;
gen_llvm_internal_op(ctx, PLUS);
gen_llvm_internal_op(ctx, MINUS);
gen_llvm_internal_op(ctx, TIMES);
gen_llvm_internal_op(ctx, DIVIDE);
for(auto& def_data : defs_data) {
def_data.second->generate_llvm(ctx);
}
for(auto& def_defn : defs_defn) {
def_defn.second->declare_llvm(ctx);
}
for(auto& def_defn : defs_defn) {
def_defn.second->generate_llvm(ctx);
}
ctx.module.print(llvm::outs(), nullptr);
output_llvm(ctx, "program.o");
}
int main() {
yy::parser parser;
type_mgr mgr;
type_env_ptr env(new type_env);
parser.parse();
for(auto& def_defn : defs_defn) {
std::cout << def_defn.second->name;
for(auto& param : def_defn.second->params) std::cout << " " << param;
std::cout << ":" << std::endl;
def_defn.second->body->print(1, std::cout);
}
try {
typecheck_program(defs_data, defs_defn, mgr, env);
compile_program(defs_defn);
gen_llvm(defs_data, defs_defn);
} catch(unification_error& err) {
std::cout << "failed to unify types: " << std::endl;
std::cout << " (1) \033[34m";
err.left->print(mgr, std::cout);
std::cout << "\033[0m" << std::endl;
std::cout << " (2) \033[32m";
err.right->print(mgr, std::cout);
std::cout << "\033[0m" << std::endl;
} catch(type_error& err) {
std::cout << "failed to type check program: " << err.description << std::endl;
}
}

View File

@@ -0,0 +1,36 @@
#include "parsed_type.hpp"
#include "type.hpp"
#include "type_env.hpp"
type_ptr parsed_type_app::to_type(
const std::set<std::string>& vars,
const type_env& e) const {
auto parent_type = e.lookup_type(name);
if(parent_type == nullptr) throw 0;
type_base* base_type;
if(!(base_type = dynamic_cast<type_base*>(parent_type.get()))) throw 0;
if(base_type->arity != arguments.size()) throw 0;
type_app* new_app = new type_app(std::move(parent_type));
type_ptr to_return(new_app);
for(auto& arg : arguments) {
new_app->arguments.push_back(arg->to_type(vars, e));
}
return to_return;
}
type_ptr parsed_type_var::to_type(
const std::set<std::string>& vars,
const type_env& e) const {
if(vars.find(var) == vars.end()) throw 0;
return type_ptr(new type_var(var));
}
type_ptr parsed_type_arr::to_type(
const std::set<std::string>& vars,
const type_env& env) const {
auto new_left = left->to_type(vars, env);
auto new_right = right->to_type(vars, env);
return type_ptr(new type_arr(std::move(new_left), std::move(new_right)));
}

View File

@@ -0,0 +1,43 @@
#pragma once
#include <memory>
#include <set>
#include <string>
#include "type_env.hpp"
struct parsed_type {
virtual type_ptr to_type(
const std::set<std::string>& vars,
const type_env& env) const = 0;
};
using parsed_type_ptr = std::unique_ptr<parsed_type>;
struct parsed_type_app : parsed_type {
std::string name;
std::vector<parsed_type_ptr> arguments;
parsed_type_app(
std::string n,
std::vector<parsed_type_ptr> as)
: name(std::move(n)), arguments(std::move(as)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};
struct parsed_type_var : parsed_type {
std::string var;
parsed_type_var(std::string v) : var(std::move(v)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};
struct parsed_type_arr : parsed_type {
parsed_type_ptr left;
parsed_type_ptr right;
parsed_type_arr(parsed_type_ptr l, parsed_type_ptr r)
: left(std::move(l)), right(std::move(r)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};

163
code/compiler/11/parser.y Normal file
View File

@@ -0,0 +1,163 @@
%{
#include <string>
#include <iostream>
#include <map>
#include "ast.hpp"
#include "definition.hpp"
#include "parser.hpp"
#include "parsed_type.hpp"
std::map<std::string, definition_data_ptr> defs_data;
std::map<std::string, definition_defn_ptr> defs_defn;
extern yy::parser::symbol_type yylex();
%}
%token PLUS
%token TIMES
%token MINUS
%token DIVIDE
%token <int> INT
%token DEFN
%token DATA
%token CASE
%token OF
%token OCURLY
%token CCURLY
%token OPAREN
%token CPAREN
%token COMMA
%token ARROW
%token EQUAL
%token <std::string> LID
%token <std::string> UID
%language "c++"
%define api.value.type variant
%define api.token.constructor
%type <std::vector<std::string>> lowercaseParams
%type <std::vector<branch_ptr>> branches
%type <std::vector<constructor_ptr>> constructors
%type <std::vector<parsed_type_ptr>> typeList
%type <parsed_type_ptr> type nonArrowType typeListElement
%type <ast_ptr> aAdd aMul case app appBase
%type <definition_data_ptr> data
%type <definition_defn_ptr> defn
%type <branch_ptr> branch
%type <pattern_ptr> pattern
%type <constructor_ptr> constructor
%start program
%%
program
: definitions { }
;
definitions
: definitions definition { }
| definition { }
;
definition
: defn { auto name = $1->name; defs_defn[name] = std::move($1); }
| data { auto name = $1->name; defs_data[name] = std::move($1); }
;
defn
: DEFN LID lowercaseParams EQUAL OCURLY aAdd CCURLY
{ $$ = definition_defn_ptr(
new definition_defn(std::move($2), std::move($3), std::move($6))); }
;
lowercaseParams
: %empty { $$ = std::vector<std::string>(); }
| lowercaseParams LID { $$ = std::move($1); $$.push_back(std::move($2)); }
;
aAdd
: aAdd PLUS aMul { $$ = ast_ptr(new ast_binop(PLUS, std::move($1), std::move($3))); }
| aAdd MINUS aMul { $$ = ast_ptr(new ast_binop(MINUS, std::move($1), std::move($3))); }
| aMul { $$ = std::move($1); }
;
aMul
: aMul TIMES app { $$ = ast_ptr(new ast_binop(TIMES, std::move($1), std::move($3))); }
| aMul DIVIDE app { $$ = ast_ptr(new ast_binop(DIVIDE, std::move($1), std::move($3))); }
| app { $$ = std::move($1); }
;
app
: app appBase { $$ = ast_ptr(new ast_app(std::move($1), std::move($2))); }
| appBase { $$ = std::move($1); }
;
appBase
: INT { $$ = ast_ptr(new ast_int($1)); }
| LID { $$ = ast_ptr(new ast_lid(std::move($1))); }
| UID { $$ = ast_ptr(new ast_uid(std::move($1))); }
| OPAREN aAdd CPAREN { $$ = std::move($2); }
| case { $$ = std::move($1); }
;
case
: CASE aAdd OF OCURLY branches CCURLY
{ $$ = ast_ptr(new ast_case(std::move($2), std::move($5))); }
;
branches
: branches branch { $$ = std::move($1); $$.push_back(std::move($2)); }
| branch { $$ = std::vector<branch_ptr>(); $$.push_back(std::move($1));}
;
branch
: pattern ARROW OCURLY aAdd CCURLY
{ $$ = branch_ptr(new branch(std::move($1), std::move($4))); }
;
pattern
: LID { $$ = pattern_ptr(new pattern_var(std::move($1))); }
| UID lowercaseParams
{ $$ = pattern_ptr(new pattern_constr(std::move($1), std::move($2))); }
;
data
: DATA UID lowercaseParams EQUAL OCURLY constructors CCURLY
{ $$ = definition_data_ptr(new definition_data(std::move($2), std::move($3), std::move($6))); }
;
constructors
: constructors COMMA constructor { $$ = std::move($1); $$.push_back(std::move($3)); }
| constructor
{ $$ = std::vector<constructor_ptr>(); $$.push_back(std::move($1)); }
;
constructor
: UID typeList
{ $$ = constructor_ptr(new constructor(std::move($1), std::move($2))); }
;
type
: nonArrowType ARROW type { $$ = parsed_type_ptr(new parsed_type_arr(std::move($1), std::move($3))); }
| nonArrowType { $$ = std::move($1); }
;
nonArrowType
: UID typeList { $$ = parsed_type_ptr(new parsed_type_app(std::move($1), std::move($2))); }
| LID { $$ = parsed_type_ptr(new parsed_type_var(std::move($1))); }
| OPAREN type CPAREN { $$ = std::move($2); }
;
typeListElement
: OPAREN type CPAREN { $$ = std::move($2); }
| UID { $$ = parsed_type_ptr(new parsed_type_app(std::move($1), {})); }
| LID { $$ = parsed_type_ptr(new parsed_type_var(std::move($1))); }
;
typeList
: %empty { $$ = std::vector<parsed_type_ptr>(); }
| typeList typeListElement { $$ = std::move($1); $$.push_back(std::move($2)); }
;

269
code/compiler/11/runtime.c Normal file
View File

@@ -0,0 +1,269 @@
#include <stdint.h>
#include <assert.h>
#include <memory.h>
#include <stdio.h>
#include "runtime.h"
struct node_base* alloc_node() {
struct node_base* new_node = malloc(sizeof(struct node_app));
new_node->gc_next = NULL;
new_node->gc_reachable = 0;
assert(new_node != NULL);
return new_node;
}
struct node_app* alloc_app(struct node_base* l, struct node_base* r) {
struct node_app* node = (struct node_app*) alloc_node();
node->base.tag = NODE_APP;
node->left = l;
node->right = r;
return node;
}
struct node_num* alloc_num(int32_t n) {
struct node_num* node = (struct node_num*) alloc_node();
node->base.tag = NODE_NUM;
node->value = n;
return node;
}
struct node_global* alloc_global(void (*f)(struct gmachine*), int32_t a) {
struct node_global* node = (struct node_global*) alloc_node();
node->base.tag = NODE_GLOBAL;
node->arity = a;
node->function = f;
return node;
}
struct node_ind* alloc_ind(struct node_base* n) {
struct node_ind* node = (struct node_ind*) alloc_node();
node->base.tag = NODE_IND;
node->next = n;
return node;
}
void free_node_direct(struct node_base* n) {
if(n->tag == NODE_DATA) {
free(((struct node_data*) n)->array);
}
}
void gc_visit_node(struct node_base* n) {
if(n->gc_reachable) return;
n->gc_reachable = 1;
if(n->tag == NODE_APP) {
struct node_app* app = (struct node_app*) n;
gc_visit_node(app->left);
gc_visit_node(app->right);
} if(n->tag == NODE_IND) {
struct node_ind* ind = (struct node_ind*) n;
gc_visit_node(ind->next);
} if(n->tag == NODE_DATA) {
struct node_data* data = (struct node_data*) n;
struct node_base** to_visit = data->array;
while(*to_visit) {
gc_visit_node(*to_visit);
to_visit++;
}
}
}
void stack_init(struct stack* s) {
s->size = 4;
s->count = 0;
s->data = malloc(sizeof(*s->data) * s->size);
assert(s->data != NULL);
}
void stack_free(struct stack* s) {
free(s->data);
}
void stack_push(struct stack* s, struct node_base* n) {
while(s->count >= s->size) {
s->data = realloc(s->data, sizeof(*s->data) * (s->size *= 2));
assert(s->data != NULL);
}
s->data[s->count++] = n;
}
struct node_base* stack_pop(struct stack* s) {
assert(s->count > 0);
return s->data[--s->count];
}
struct node_base* stack_peek(struct stack* s, size_t o) {
assert(s->count > o);
return s->data[s->count - o - 1];
}
void stack_popn(struct stack* s, size_t n) {
assert(s->count >= n);
s->count -= n;
}
void gmachine_init(struct gmachine* g) {
stack_init(&g->stack);
g->gc_nodes = NULL;
g->gc_node_count = 0;
g->gc_node_threshold = 128;
}
void gmachine_free(struct gmachine* g) {
stack_free(&g->stack);
struct node_base* to_free = g->gc_nodes;
struct node_base* next;
while(to_free) {
next = to_free->gc_next;
free_node_direct(to_free);
free(to_free);
to_free = next;
}
}
void gmachine_slide(struct gmachine* g, size_t n) {
assert(g->stack.count > n);
g->stack.data[g->stack.count - n - 1] = g->stack.data[g->stack.count - 1];
g->stack.count -= n;
}
void gmachine_update(struct gmachine* g, size_t o) {
assert(g->stack.count > o + 1);
struct node_ind* ind =
(struct node_ind*) g->stack.data[g->stack.count - o - 2];
ind->base.tag = NODE_IND;
ind->next = g->stack.data[g->stack.count -= 1];
}
void gmachine_alloc(struct gmachine* g, size_t o) {
while(o--) {
stack_push(&g->stack,
gmachine_track(g, (struct node_base*) alloc_ind(NULL)));
}
}
void gmachine_pack(struct gmachine* g, size_t n, int8_t t) {
assert(g->stack.count >= n);
struct node_base** data = malloc(sizeof(*data) * (n + 1));
assert(data != NULL);
memcpy(data, &g->stack.data[g->stack.count - n], n * sizeof(*data));
data[n] = NULL;
struct node_data* new_node = (struct node_data*) alloc_node();
new_node->array = data;
new_node->base.tag = NODE_DATA;
new_node->tag = t;
stack_popn(&g->stack, n);
stack_push(&g->stack, gmachine_track(g, (struct node_base*) new_node));
}
void gmachine_split(struct gmachine* g, size_t n) {
struct node_data* node = (struct node_data*) stack_pop(&g->stack);
for(size_t i = 0; i < n; i++) {
stack_push(&g->stack, node->array[i]);
}
}
struct node_base* gmachine_track(struct gmachine* g, struct node_base* b) {
g->gc_node_count++;
b->gc_next = g->gc_nodes;
g->gc_nodes = b;
if(g->gc_node_count >= g->gc_node_threshold) {
uint64_t nodes_before = g->gc_node_count;
gc_visit_node(b);
gmachine_gc(g);
g->gc_node_threshold = g->gc_node_count * 2;
}
return b;
}
void gmachine_gc(struct gmachine* g) {
for(size_t i = 0; i < g->stack.count; i++) {
gc_visit_node(g->stack.data[i]);
}
struct node_base** head_ptr = &g->gc_nodes;
while(*head_ptr) {
if((*head_ptr)->gc_reachable) {
(*head_ptr)->gc_reachable = 0;
head_ptr = &(*head_ptr)->gc_next;
} else {
struct node_base* to_free = *head_ptr;
*head_ptr = to_free->gc_next;
free_node_direct(to_free);
free(to_free);
g->gc_node_count--;
}
}
}
void unwind(struct gmachine* g) {
struct stack* s = &g->stack;
while(1) {
struct node_base* peek = stack_peek(s, 0);
if(peek->tag == NODE_APP) {
struct node_app* n = (struct node_app*) peek;
stack_push(s, n->left);
} else if(peek->tag == NODE_GLOBAL) {
struct node_global* n = (struct node_global*) peek;
assert(s->count > n->arity);
for(size_t i = 1; i <= n->arity; i++) {
s->data[s->count - i]
= ((struct node_app*) s->data[s->count - i - 1])->right;
}
n->function(g);
} else if(peek->tag == NODE_IND) {
struct node_ind* n = (struct node_ind*) peek;
stack_pop(s);
stack_push(s, n->next);
} else {
break;
}
}
}
extern void f_main(struct gmachine* s);
void print_node(struct node_base* n) {
if(n->tag == NODE_APP) {
struct node_app* app = (struct node_app*) n;
print_node(app->left);
putchar(' ');
print_node(app->right);
} else if(n->tag == NODE_DATA) {
printf("(Packed)");
} else if(n->tag == NODE_GLOBAL) {
struct node_global* global = (struct node_global*) n;
printf("(Global: %p)", global->function);
} else if(n->tag == NODE_IND) {
print_node(((struct node_ind*) n)->next);
} else if(n->tag == NODE_NUM) {
struct node_num* num = (struct node_num*) n;
printf("%d", num->value);
}
}
int main(int argc, char** argv) {
struct gmachine gmachine;
struct node_global* first_node = alloc_global(f_main, 0);
struct node_base* result;
gmachine_init(&gmachine);
gmachine_track(&gmachine, (struct node_base*) first_node);
stack_push(&gmachine.stack, (struct node_base*) first_node);
unwind(&gmachine);
result = stack_pop(&gmachine.stack);
printf("Result: ");
print_node(result);
putchar('\n');
gmachine_free(&gmachine);
}

View File

@@ -0,0 +1,84 @@
#pragma once
#include <stdlib.h>
struct gmachine;
enum node_tag {
NODE_APP,
NODE_NUM,
NODE_GLOBAL,
NODE_IND,
NODE_DATA
};
struct node_base {
enum node_tag tag;
int8_t gc_reachable;
struct node_base* gc_next;
};
struct node_app {
struct node_base base;
struct node_base* left;
struct node_base* right;
};
struct node_num {
struct node_base base;
int32_t value;
};
struct node_global {
struct node_base base;
int32_t arity;
void (*function)(struct gmachine*);
};
struct node_ind {
struct node_base base;
struct node_base* next;
};
struct node_data {
struct node_base base;
int8_t tag;
struct node_base** array;
};
struct node_base* alloc_node();
struct node_app* alloc_app(struct node_base* l, struct node_base* r);
struct node_num* alloc_num(int32_t n);
struct node_global* alloc_global(void (*f)(struct gmachine*), int32_t a);
struct node_ind* alloc_ind(struct node_base* n);
void free_node_direct(struct node_base*);
void gc_visit_node(struct node_base*);
struct stack {
size_t size;
size_t count;
struct node_base** data;
};
void stack_init(struct stack* s);
void stack_free(struct stack* s);
void stack_push(struct stack* s, struct node_base* n);
struct node_base* stack_pop(struct stack* s);
struct node_base* stack_peek(struct stack* s, size_t o);
void stack_popn(struct stack* s, size_t n);
struct gmachine {
struct stack stack;
struct node_base* gc_nodes;
int64_t gc_node_count;
int64_t gc_node_threshold;
};
void gmachine_init(struct gmachine* g);
void gmachine_free(struct gmachine* g);
void gmachine_slide(struct gmachine* g, size_t n);
void gmachine_update(struct gmachine* g, size_t o);
void gmachine_alloc(struct gmachine* g, size_t o);
void gmachine_pack(struct gmachine* g, size_t n, int8_t t);
void gmachine_split(struct gmachine* g, size_t n);
struct node_base* gmachine_track(struct gmachine* g, struct node_base* b);
void gmachine_gc(struct gmachine* g);

View File

@@ -0,0 +1,35 @@
%option noyywrap
%{
#include <iostream>
#include "ast.hpp"
#include "definition.hpp"
#include "parser.hpp"
#define YY_DECL yy::parser::symbol_type yylex()
%}
%%
[ \n]+ {}
\+ { return yy::parser::make_PLUS(); }
\* { return yy::parser::make_TIMES(); }
- { return yy::parser::make_MINUS(); }
\/ { return yy::parser::make_DIVIDE(); }
[0-9]+ { return yy::parser::make_INT(atoi(yytext)); }
defn { return yy::parser::make_DEFN(); }
data { return yy::parser::make_DATA(); }
case { return yy::parser::make_CASE(); }
of { return yy::parser::make_OF(); }
\{ { return yy::parser::make_OCURLY(); }
\} { return yy::parser::make_CCURLY(); }
\( { return yy::parser::make_OPAREN(); }
\) { return yy::parser::make_CPAREN(); }
, { return yy::parser::make_COMMA(); }
-> { return yy::parser::make_ARROW(); }
= { return yy::parser::make_EQUAL(); }
[a-z][a-zA-Z]* { return yy::parser::make_LID(std::string(yytext)); }
[A-Z][a-zA-Z]* { return yy::parser::make_UID(std::string(yytext)); }
%%

23
code/compiler/11/test.cpp Normal file
View File

@@ -0,0 +1,23 @@
#include "graph.hpp"
int main() {
function_graph graph;
graph.add_edge("f", "g");
graph.add_edge("g", "h");
graph.add_edge("h", "f");
graph.add_edge("i", "j");
graph.add_edge("j", "i");
graph.add_edge("j", "f");
graph.add_edge("x", "f");
graph.add_edge("x", "i");
for(auto& group : graph.compute_order()) {
std::cout << "Group: " << std::endl;
for(auto& member : group->members) {
std::cout << member << std::endl;
}
}
}

187
code/compiler/11/type.cpp Normal file
View File

@@ -0,0 +1,187 @@
#include "type.hpp"
#include <ostream>
#include <sstream>
#include <algorithm>
#include <vector>
#include "error.hpp"
void type_scheme::print(const type_mgr& mgr, std::ostream& to) const {
if(forall.size() != 0) {
to << "forall ";
for(auto& var : forall) {
to << var << " ";
}
to << ". ";
}
monotype->print(mgr, to);
}
type_ptr type_scheme::instantiate(type_mgr& mgr) const {
if(forall.size() == 0) return monotype;
std::map<std::string, type_ptr> subst;
for(auto& var : forall) {
subst[var] = mgr.new_type();
}
return mgr.substitute(subst, monotype);
}
void type_var::print(const type_mgr& mgr, std::ostream& to) const {
auto it = mgr.types.find(name);
if(it != mgr.types.end()) {
it->second->print(mgr, to);
} else {
to << name;
}
}
void type_base::print(const type_mgr& mgr, std::ostream& to) const {
to << name;
}
void type_arr::print(const type_mgr& mgr, std::ostream& to) const {
left->print(mgr, to);
to << " -> (";
right->print(mgr, to);
to << ")";
}
void type_app::print(const type_mgr& mgr, std::ostream& to) const {
constructor->print(mgr, to);
to << "* ";
for(auto& arg : arguments) {
to << " ";
arg->print(mgr, to);
}
}
std::string type_mgr::new_type_name() {
int temp = last_id++;
std::string str = "";
while(temp != -1) {
str += (char) ('a' + (temp % 26));
temp = temp / 26 - 1;
}
std::reverse(str.begin(), str.end());
return str;
}
type_ptr type_mgr::new_type() {
return type_ptr(new type_var(new_type_name()));
}
type_ptr type_mgr::new_arrow_type() {
return type_ptr(new type_arr(new_type(), new_type()));
}
type_ptr type_mgr::resolve(type_ptr t, type_var*& var) const {
type_var* cast;
var = nullptr;
while((cast = dynamic_cast<type_var*>(t.get()))) {
auto it = types.find(cast->name);
if(it == types.end()) {
var = cast;
break;
}
t = it->second;
}
return t;
}
void type_mgr::unify(type_ptr l, type_ptr r) {
type_var *lvar, *rvar;
type_arr *larr, *rarr;
type_base *lid, *rid;
type_app *lapp, *rapp;
l = resolve(l, lvar);
r = resolve(r, rvar);
if(lvar) {
bind(lvar->name, r);
return;
} else if(rvar) {
bind(rvar->name, l);
return;
} else if((larr = dynamic_cast<type_arr*>(l.get())) &&
(rarr = dynamic_cast<type_arr*>(r.get()))) {
unify(larr->left, rarr->left);
unify(larr->right, rarr->right);
return;
} else if((lid = dynamic_cast<type_base*>(l.get())) &&
(rid = dynamic_cast<type_base*>(r.get()))) {
if(lid->name == rid->name && lid->arity == rid->arity) return;
} else if((lapp = dynamic_cast<type_app*>(l.get())) &&
(rapp = dynamic_cast<type_app*>(r.get()))) {
unify(lapp->constructor, rapp->constructor);
auto left_it = lapp->arguments.begin();
auto right_it = rapp->arguments.begin();
while(left_it != lapp->arguments.end() &&
right_it != rapp->arguments.end()) {
unify(*left_it, *right_it);
left_it++, right_it++;
}
return;
}
throw unification_error(l, r);
}
type_ptr type_mgr::substitute(const std::map<std::string, type_ptr>& subst, const type_ptr& t) const {
type_ptr temp = t;
while(type_var* var = dynamic_cast<type_var*>(temp.get())) {
auto subst_it = subst.find(var->name);
if(subst_it != subst.end()) return subst_it->second;
auto var_it = types.find(var->name);
if(var_it == types.end()) return t;
temp = var_it->second;
}
if(type_arr* arr = dynamic_cast<type_arr*>(temp.get())) {
auto left_result = substitute(subst, arr->left);
auto right_result = substitute(subst, arr->right);
if(left_result == arr->left && right_result == arr->right) return t;
return type_ptr(new type_arr(left_result, right_result));
} else if(type_app* app = dynamic_cast<type_app*>(temp.get())) {
auto constructor_result = substitute(subst, app->constructor);
bool arg_changed = false;
std::vector<type_ptr> new_args;
for(auto& arg : app->arguments) {
auto arg_result = substitute(subst, arg);
arg_changed |= arg_result != arg;
new_args.push_back(std::move(arg_result));
}
if(constructor_result == app->constructor && !arg_changed) return t;
type_app* new_app = new type_app(std::move(constructor_result));
std::swap(new_app->arguments, new_args);
return type_ptr(new_app);
}
return t;
}
void type_mgr::bind(const std::string& s, type_ptr t) {
type_var* other = dynamic_cast<type_var*>(t.get());
if(other && other->name == s) return;
types[s] = t;
}
void type_mgr::find_free(const type_ptr& t, std::set<std::string>& into) const {
type_var* var;
type_ptr resolved = resolve(t, var);
if(var) {
into.insert(var->name);
} else if(type_arr* arr = dynamic_cast<type_arr*>(resolved.get())) {
find_free(arr->left, into);
find_free(arr->right, into);
} else if(type_app* app = dynamic_cast<type_app*>(resolved.get())) {
find_free(app->constructor, into);
for(auto& arg : app->arguments) find_free(arg, into);
}
}

95
code/compiler/11/type.hpp Normal file
View File

@@ -0,0 +1,95 @@
#pragma once
#include <memory>
#include <map>
#include <string>
#include <vector>
#include <set>
struct type_mgr;
struct type {
virtual ~type() = default;
virtual void print(const type_mgr& mgr, std::ostream& to) const = 0;
};
using type_ptr = std::shared_ptr<type>;
struct type_scheme {
std::vector<std::string> forall;
type_ptr monotype;
type_scheme(type_ptr type) : forall(), monotype(std::move(type)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
type_ptr instantiate(type_mgr& mgr) const;
};
using type_scheme_ptr = std::shared_ptr<type_scheme>;
struct type_var : public type {
std::string name;
type_var(std::string n)
: name(std::move(n)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_base : public type {
std::string name;
int32_t arity;
type_base(std::string n, int32_t a = 0)
: name(std::move(n)), arity(a) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_data : public type_base {
struct constructor {
int tag;
};
std::map<std::string, constructor> constructors;
type_data(std::string n, int32_t a = 0)
: type_base(std::move(n), a) {}
};
struct type_arr : public type {
type_ptr left;
type_ptr right;
type_arr(type_ptr l, type_ptr r)
: left(std::move(l)), right(std::move(r)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_app : public type {
type_ptr constructor;
std::vector<type_ptr> arguments;
type_app(type_ptr c)
: constructor(std::move(c)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_mgr {
int last_id = 0;
std::map<std::string, type_ptr> types;
std::string new_type_name();
type_ptr new_type();
type_ptr new_arrow_type();
void unify(type_ptr l, type_ptr r);
type_ptr substitute(
const std::map<std::string, type_ptr>& subst,
const type_ptr& t) const;
type_ptr resolve(type_ptr t, type_var*& var) const;
void bind(const std::string& s, type_ptr t);
void find_free(const type_ptr& t, std::set<std::string>& into) const;
};

View File

@@ -0,0 +1,45 @@
#include "type_env.hpp"
#include "type.hpp"
type_scheme_ptr type_env::lookup(const std::string& name) const {
auto it = names.find(name);
if(it != names.end()) return it->second;
if(parent) return parent->lookup(name);
return nullptr;
}
type_ptr type_env::lookup_type(const std::string& name) const {
auto it = type_names.find(name);
if(it != type_names.end()) return it->second;
if(parent) return parent->lookup_type(name);
return nullptr;
}
void type_env::bind(const std::string& name, type_ptr t) {
names[name] = type_scheme_ptr(new type_scheme(t));
}
void type_env::bind(const std::string& name, type_scheme_ptr t) {
names[name] = t;
}
void type_env::bind_type(const std::string& type_name, type_ptr t) {
if(lookup_type(type_name) != nullptr) throw 0;
type_names[type_name] = t;
}
void type_env::generalize(const std::string& name, type_mgr& mgr) {
auto names_it = names.find(name);
if(names_it == names.end()) throw 0;
if(names_it->second->forall.size() > 0) throw 0;
std::set<std::string> free_variables;
mgr.find_free(names_it->second->monotype, free_variables);
for(auto& free : free_variables) {
names_it->second->forall.push_back(free);
}
}
type_env_ptr type_scope(type_env_ptr parent) {
return type_env_ptr(new type_env(std::move(parent)));
}

View File

@@ -0,0 +1,26 @@
#pragma once
#include <map>
#include <string>
#include "type.hpp"
struct type_env;
using type_env_ptr = std::shared_ptr<type_env>;
struct type_env {
type_env_ptr parent;
std::map<std::string, type_scheme_ptr> names;
std::map<std::string, type_ptr> type_names;
type_env(type_env_ptr p) : parent(std::move(p)) {}
type_env() : type_env(nullptr) {}
type_scheme_ptr lookup(const std::string& name) const;
type_ptr lookup_type(const std::string& name) const;
void bind(const std::string& name, type_ptr t);
void bind(const std::string& name, type_scheme_ptr t);
void bind_type(const std::string& type_name, type_ptr t);
void generalize(const std::string& name, type_mgr& mgr);
};
type_env_ptr type_scope(type_env_ptr parent);

View File

@@ -0,0 +1,46 @@
cmake_minimum_required(VERSION 3.1)
project(compiler)
# Find all the required packages
find_package(BISON)
find_package(FLEX)
find_package(LLVM REQUIRED CONFIG)
# Set up the flex and bison targets
bison_target(parser
${CMAKE_CURRENT_SOURCE_DIR}/parser.y
${CMAKE_CURRENT_BINARY_DIR}/parser.cpp
COMPILE_FLAGS "-d")
flex_target(scanner
${CMAKE_CURRENT_SOURCE_DIR}/scanner.l
${CMAKE_CURRENT_BINARY_DIR}/scanner.cpp)
add_flex_bison_dependency(scanner parser)
# Find all the relevant LLVM components
llvm_map_components_to_libnames(LLVM_LIBS core x86asmparser x86codegen)
# Create compiler executable
add_executable(compiler
definition.cpp definition.hpp
parsed_type.cpp parsed_type.hpp
ast.cpp ast.hpp
llvm_context.cpp llvm_context.hpp
type_env.cpp type_env.hpp
env.cpp env.hpp
type.cpp type.hpp
error.cpp error.hpp
binop.cpp binop.hpp
instruction.cpp instruction.hpp
graph.cpp graph.hpp
global_scope.cpp global_scope.hpp
${BISON_parser_OUTPUTS}
${FLEX_scanner_OUTPUTS}
main.cpp
)
# Configure compiler executable
target_include_directories(compiler PUBLIC ${CMAKE_CURRENT_SOURCE_DIR})
target_include_directories(compiler PUBLIC ${CMAKE_CURRENT_BINARY_DIR})
target_include_directories(compiler PUBLIC ${LLVM_INCLUDE_DIRS})
target_compile_definitions(compiler PUBLIC ${LLVM_DEFINITIONS})
target_link_libraries(compiler ${LLVM_LIBS})

437
code/compiler/12/ast.cpp Normal file
View File

@@ -0,0 +1,437 @@
#include "ast.hpp"
#include <ostream>
#include "binop.hpp"
#include "error.hpp"
#include "type_env.hpp"
#include "env.hpp"
static void print_indent(int n, std::ostream& to) {
while(n--) to << " ";
}
void ast_int::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "INT: " << value << std::endl;
}
void ast_int::find_free(std::set<std::string>& into) {
}
type_ptr ast_int::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
return type_ptr(new type_app(env->lookup_type("Int")));
}
void ast_int::translate(global_scope& scope) {
}
void ast_int::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(new instruction_pushint(value)));
}
void ast_lid::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "LID: " << id << std::endl;
}
void ast_lid::find_free(std::set<std::string>& into) {
into.insert(id);
}
type_ptr ast_lid::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
return env->lookup(id)->instantiate(mgr);
}
void ast_lid::translate(global_scope& scope) {
}
void ast_lid::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
auto mangled_name = this->env->get_mangled_name(id);
into.push_back(instruction_ptr(
(env->has_variable(mangled_name) && !this->env->is_global(id)) ?
(instruction*) new instruction_push(env->get_offset(mangled_name)) :
(instruction*) new instruction_pushglobal(mangled_name)));
}
void ast_uid::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "UID: " << id << std::endl;
}
void ast_uid::find_free(std::set<std::string>& into) {
}
type_ptr ast_uid::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
return env->lookup(id)->instantiate(mgr);
}
void ast_uid::translate(global_scope& scope) {
}
void ast_uid::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(
new instruction_pushglobal(this->env->get_mangled_name(id))));
}
void ast_binop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "BINOP: " << op_name(op) << std::endl;
left->print(indent + 1, to);
right->print(indent + 1, to);
}
void ast_binop::find_free(std::set<std::string>& into) {
left->find_free(into);
right->find_free(into);
}
type_ptr ast_binop::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
type_ptr ltype = left->typecheck(mgr, env);
type_ptr rtype = right->typecheck(mgr, env);
type_ptr ftype = env->lookup(op_name(op))->instantiate(mgr);
if(!ftype) throw type_error(std::string("unknown binary operator ") + op_name(op));
type_ptr return_type = mgr.new_type();
type_ptr arrow_one = type_ptr(new type_arr(rtype, return_type));
type_ptr arrow_two = type_ptr(new type_arr(ltype, arrow_one));
mgr.unify(arrow_two, ftype);
return return_type;
}
void ast_binop::translate(global_scope& scope) {
left->translate(scope);
right->translate(scope);
}
void ast_binop::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
right->compile(env, into);
left->compile(env_ptr(new env_offset(1, env)), into);
into.push_back(instruction_ptr(new instruction_pushglobal(op_action(op))));
into.push_back(instruction_ptr(new instruction_mkapp()));
into.push_back(instruction_ptr(new instruction_mkapp()));
}
void ast_app::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "APP:" << std::endl;
left->print(indent + 1, to);
right->print(indent + 1, to);
}
void ast_app::find_free(std::set<std::string>& into) {
left->find_free(into);
right->find_free(into);
}
type_ptr ast_app::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
type_ptr ltype = left->typecheck(mgr, env);
type_ptr rtype = right->typecheck(mgr, env);
type_ptr return_type = mgr.new_type();
type_ptr arrow = type_ptr(new type_arr(rtype, return_type));
mgr.unify(arrow, ltype);
return return_type;
}
void ast_app::translate(global_scope& scope) {
left->translate(scope);
right->translate(scope);
}
void ast_app::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
right->compile(env, into);
left->compile(env_ptr(new env_offset(1, env)), into);
into.push_back(instruction_ptr(new instruction_mkapp()));
}
void ast_case::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "CASE: " << std::endl;
for(auto& branch : branches) {
print_indent(indent + 1, to);
branch->pat->print(to);
to << std::endl;
branch->expr->print(indent + 2, to);
}
}
void ast_case::find_free(std::set<std::string>& into) {
of->find_free(into);
for(auto& branch : branches) {
std::set<std::string> free_in_branch;
std::set<std::string> pattern_variables;
branch->pat->find_variables(pattern_variables);
branch->expr->find_free(free_in_branch);
for(auto& free : free_in_branch) {
if(pattern_variables.find(free) == pattern_variables.end())
into.insert(free);
}
}
}
type_ptr ast_case::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
type_var* var;
type_ptr case_type = mgr.resolve(of->typecheck(mgr, env), var);
type_ptr branch_type = mgr.new_type();
for(auto& branch : branches) {
type_env_ptr new_env = type_scope(env);
branch->pat->typecheck(case_type, mgr, new_env);
type_ptr curr_branch_type = branch->expr->typecheck(mgr, new_env);
mgr.unify(branch_type, curr_branch_type);
}
input_type = mgr.resolve(case_type, var);
type_app* app_type;
if(!(app_type = dynamic_cast<type_app*>(input_type.get())) ||
!dynamic_cast<type_data*>(app_type->constructor.get())) {
throw type_error("attempting case analysis of non-data type");
}
return branch_type;
}
void ast_case::translate(global_scope& scope) {
of->translate(scope);
for(auto& branch : branches) {
branch->expr->translate(scope);
}
}
void ast_case::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
type_app* app_type = dynamic_cast<type_app*>(input_type.get());
type_data* type = dynamic_cast<type_data*>(app_type->constructor.get());
of->compile(env, into);
into.push_back(instruction_ptr(new instruction_eval()));
instruction_jump* jump_instruction = new instruction_jump();
into.push_back(instruction_ptr(jump_instruction));
for(auto& branch : branches) {
std::vector<instruction_ptr> branch_instructions;
pattern_var* vpat;
pattern_constr* cpat;
if((vpat = dynamic_cast<pattern_var*>(branch->pat.get()))) {
branch->expr->compile(env_ptr(new env_offset(1, env)), branch_instructions);
for(auto& constr_pair : type->constructors) {
if(jump_instruction->tag_mappings.find(constr_pair.second.tag) !=
jump_instruction->tag_mappings.end())
break;
jump_instruction->tag_mappings[constr_pair.second.tag] =
jump_instruction->branches.size();
}
jump_instruction->branches.push_back(std::move(branch_instructions));
} else if((cpat = dynamic_cast<pattern_constr*>(branch->pat.get()))) {
env_ptr new_env = env;
for(auto it = cpat->params.rbegin(); it != cpat->params.rend(); it++) {
new_env = env_ptr(new env_var(branch->expr->env->get_mangled_name(*it), new_env));
}
branch_instructions.push_back(instruction_ptr(new instruction_split(
cpat->params.size())));
branch->expr->compile(new_env, branch_instructions);
branch_instructions.push_back(instruction_ptr(new instruction_slide(
cpat->params.size())));
int new_tag = type->constructors[cpat->constr].tag;
if(jump_instruction->tag_mappings.find(new_tag) !=
jump_instruction->tag_mappings.end())
throw type_error("technically not a type error: duplicate pattern");
jump_instruction->tag_mappings[new_tag] =
jump_instruction->branches.size();
jump_instruction->branches.push_back(std::move(branch_instructions));
}
}
for(auto& constr_pair : type->constructors) {
if(jump_instruction->tag_mappings.find(constr_pair.second.tag) ==
jump_instruction->tag_mappings.end())
throw type_error("non-total pattern");
}
}
void ast_let::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "LET: " << std::endl;
in->print(indent + 1, to);
}
void ast_let::find_free(std::set<std::string>& into) {
definitions.find_free(into);
std::set<std::string> all_free;
in->find_free(all_free);
for(auto& free_var : all_free) {
if(definitions.defs_defn.find(free_var) == definitions.defs_defn.end())
into.insert(free_var);
}
}
type_ptr ast_let::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
definitions.typecheck(mgr, env);
return in->typecheck(mgr, definitions.env);
}
void ast_let::translate(global_scope& scope) {
for(auto& def : definitions.defs_data) {
def.second->into_globals(scope);
}
for(auto& def : definitions.defs_defn) {
size_t original_params = def.second->params.size();
std::string original_name = def.second->name;
auto& global_definition = def.second->into_global(scope);
size_t captured = global_definition.params.size() - original_params;
type_env_ptr mangled_env = type_scope(env);
mangled_env->bind(def.first, env->lookup(def.first), visibility::global);
mangled_env->set_mangled_name(def.first, global_definition.name);
ast_ptr global_app(new ast_lid(original_name));
global_app->env = mangled_env;
for(auto& param : global_definition.params) {
if(!(captured--)) break;
ast_ptr new_arg(new ast_lid(param));
new_arg->env = env;
global_app = ast_ptr(new ast_app(std::move(global_app), std::move(new_arg)));
global_app->env = env;
}
translated_definitions.push_back({ def.first, std::move(global_app) });
}
in->translate(scope);
}
void ast_let::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
into.push_back(instruction_ptr(new instruction_alloc(translated_definitions.size())));
env_ptr new_env = env;
for(auto& def : translated_definitions) {
new_env = env_ptr(new env_var(definitions.env->get_mangled_name(def.first), std::move(new_env)));
}
int offset = translated_definitions.size() - 1;
for(auto& def : translated_definitions) {
def.second->compile(new_env, into);
into.push_back(instruction_ptr(new instruction_update(offset--)));
}
in->compile(new_env, into);
into.push_back(instruction_ptr(new instruction_slide(translated_definitions.size())));
}
void ast_lambda::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "LAMBDA";
for(auto& param : params) {
to << " " << param;
}
to << std::endl;
body->print(indent+1, to);
}
void ast_lambda::find_free(std::set<std::string>& into) {
body->find_free(free_variables);
for(auto& param : params) {
free_variables.erase(param);
}
into.insert(free_variables.begin(), free_variables.end());
}
type_ptr ast_lambda::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = env;
var_env = type_scope(env);
type_ptr return_type = mgr.new_type();
type_ptr full_type = return_type;
for(auto it = params.rbegin(); it != params.rend(); it++) {
type_ptr param_type = mgr.new_type();
var_env->bind(*it, param_type);
full_type = type_ptr(new type_arr(std::move(param_type), full_type));
}
mgr.unify(return_type, body->typecheck(mgr, var_env));
return full_type;
}
void ast_lambda::translate(global_scope& scope) {
std::vector<std::string> function_params;
for(auto& free_variable : free_variables) {
if(env->is_global(free_variable)) continue;
function_params.push_back(free_variable);
}
size_t captured_count = function_params.size();
function_params.insert(function_params.end(), params.begin(), params.end());
auto& new_function = scope.add_function("lambda", std::move(function_params), std::move(body));
type_env_ptr mangled_env = type_scope(env);
mangled_env->bind("lambda", type_scheme_ptr(nullptr), visibility::global);
mangled_env->set_mangled_name("lambda", new_function.name);
ast_ptr new_application = ast_ptr(new ast_lid("lambda"));
new_application->env = mangled_env;
for(auto& param : new_function.params) {
if(!(captured_count--)) break;
ast_ptr new_arg = ast_ptr(new ast_lid(param));
new_arg->env = env;
new_application = ast_ptr(new ast_app(std::move(new_application), std::move(new_arg)));
new_application->env = env;
}
translated = std::move(new_application);
}
void ast_lambda::compile(const env_ptr& env, std::vector<instruction_ptr>& into) const {
translated->compile(env, into);
}
void pattern_var::print(std::ostream& to) const {
to << var;
}
void pattern_var::find_variables(std::set<std::string>& into) const {
into.insert(var);
}
void pattern_var::typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const {
env->bind(var, t);
}
void pattern_constr::print(std::ostream& to) const {
to << constr;
for(auto& param : params) {
to << " " << param;
}
}
void pattern_constr::find_variables(std::set<std::string>& into) const {
into.insert(params.begin(), params.end());
}
void pattern_constr::typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const {
type_scheme_ptr constructor_type_scheme = env->lookup(constr);
if(!constructor_type_scheme) {
throw type_error(std::string("pattern using unknown constructor ") + constr);
}
type_ptr constructor_type = constructor_type_scheme->instantiate(mgr);
for(auto& param : params) {
type_arr* arr = dynamic_cast<type_arr*>(constructor_type.get());
if(!arr) throw type_error("too many parameters in constructor pattern");
env->bind(param, arr->left);
constructor_type = arr->right;
}
mgr.unify(t, constructor_type);
}

189
code/compiler/12/ast.hpp Normal file
View File

@@ -0,0 +1,189 @@
#pragma once
#include <memory>
#include <vector>
#include <set>
#include "type.hpp"
#include "type_env.hpp"
#include "binop.hpp"
#include "instruction.hpp"
#include "env.hpp"
#include "definition.hpp"
#include "global_scope.hpp"
struct ast {
type_env_ptr env;
virtual ~ast() = default;
virtual void print(int indent, std::ostream& to) const = 0;
virtual void find_free(std::set<std::string>& into) = 0;
virtual type_ptr typecheck(type_mgr& mgr, type_env_ptr& env) = 0;
virtual void translate(global_scope& scope) = 0;
virtual void compile(const env_ptr& env,
std::vector<instruction_ptr>& into) const = 0;
};
using ast_ptr = std::unique_ptr<ast>;
struct pattern {
virtual ~pattern() = default;
virtual void print(std::ostream& to) const = 0;
virtual void find_variables(std::set<std::string>& into) const = 0;
virtual void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const = 0;
};
using pattern_ptr = std::unique_ptr<pattern>;
struct branch {
pattern_ptr pat;
ast_ptr expr;
branch(pattern_ptr p, ast_ptr a)
: pat(std::move(p)), expr(std::move(a)) {}
};
using branch_ptr = std::unique_ptr<branch>;
struct ast_int : public ast {
int value;
explicit ast_int(int v)
: value(v) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_lid : public ast {
std::string id;
explicit ast_lid(std::string i)
: id(std::move(i)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_uid : public ast {
std::string id;
explicit ast_uid(std::string i)
: id(std::move(i)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_binop : public ast {
binop op;
ast_ptr left;
ast_ptr right;
ast_binop(binop o, ast_ptr l, ast_ptr r)
: op(o), left(std::move(l)), right(std::move(r)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_app : public ast {
ast_ptr left;
ast_ptr right;
ast_app(ast_ptr l, ast_ptr r)
: left(std::move(l)), right(std::move(r)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_case : public ast {
ast_ptr of;
type_ptr input_type;
std::vector<branch_ptr> branches;
ast_case(ast_ptr o, std::vector<branch_ptr> b)
: of(std::move(o)), branches(std::move(b)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_let : public ast {
using basic_definition = std::pair<std::string, ast_ptr>;
definition_group definitions;
ast_ptr in;
std::vector<basic_definition> translated_definitions;
ast_let(definition_group g, ast_ptr i)
: definitions(std::move(g)), in(std::move(i)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct ast_lambda : public ast {
std::vector<std::string> params;
ast_ptr body;
type_env_ptr var_env;
std::set<std::string> free_variables;
ast_ptr translated;
ast_lambda(std::vector<std::string> ps, ast_ptr b)
: params(std::move(ps)), body(std::move(b)) {}
void print(int indent, std::ostream& to) const;
void find_free(std::set<std::string>& into);
type_ptr typecheck(type_mgr& mgr, type_env_ptr& env);
void translate(global_scope& scope);
void compile(const env_ptr& env, std::vector<instruction_ptr>& into) const;
};
struct pattern_var : public pattern {
std::string var;
pattern_var(std::string v)
: var(std::move(v)) {}
void print(std::ostream &to) const;
void find_variables(std::set<std::string>& into) const;
void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const;
};
struct pattern_constr : public pattern {
std::string constr;
std::vector<std::string> params;
pattern_constr(std::string c, std::vector<std::string> p)
: constr(std::move(c)), params(std::move(p)) {}
void print(std::ostream &to) const;
void find_variables(std::set<std::string>& into) const;
virtual void typecheck(type_ptr t, type_mgr& mgr, type_env_ptr& env) const;
};

View File

@@ -0,0 +1,21 @@
#include "binop.hpp"
std::string op_name(binop op) {
switch(op) {
case PLUS: return "+";
case MINUS: return "-";
case TIMES: return "*";
case DIVIDE: return "/";
}
return "??";
}
std::string op_action(binop op) {
switch(op) {
case PLUS: return "plus";
case MINUS: return "minus";
case TIMES: return "times";
case DIVIDE: return "divide";
}
return "??";
}

View File

@@ -0,0 +1,12 @@
#pragma once
#include <string>
enum binop {
PLUS,
MINUS,
TIMES,
DIVIDE
};
std::string op_name(binop op);
std::string op_action(binop op);

View File

@@ -0,0 +1,145 @@
#include "definition.hpp"
#include "error.hpp"
#include "ast.hpp"
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "type.hpp"
#include "type_env.hpp"
#include "graph.hpp"
#include <llvm/IR/DerivedTypes.h>
#include <llvm/IR/Function.h>
#include <llvm/IR/Type.h>
void definition_defn::find_free() {
body->find_free(free_variables);
for(auto& param : params) {
free_variables.erase(param);
}
}
void definition_defn::insert_types(type_mgr& mgr, type_env_ptr& env, visibility v) {
this->env = env;
var_env = type_scope(env);
return_type = mgr.new_type();
full_type = return_type;
for(auto it = params.rbegin(); it != params.rend(); it++) {
type_ptr param_type = mgr.new_type();
full_type = type_ptr(new type_arr(param_type, full_type));
var_env->bind(*it, param_type);
}
env->bind(name, full_type, v);
}
void definition_defn::typecheck(type_mgr& mgr) {
type_ptr body_type = body->typecheck(mgr, var_env);
mgr.unify(return_type, body_type);
}
global_function& definition_defn::into_global(global_scope& scope) {
std::vector<std::string> all_params;
for(auto& free : free_variables) {
if(env->is_global(free)) continue;
all_params.push_back(free);
}
all_params.insert(all_params.end(), params.begin(), params.end());
body->translate(scope);
return scope.add_function(name, std::move(all_params), std::move(body));
}
void definition_data::insert_types(type_env_ptr& env) {
this->env = env;
env->bind_type(name, type_ptr(new type_data(name, vars.size())));
}
void definition_data::insert_constructors() const {
type_ptr this_type_ptr = env->lookup_type(name);
type_data* this_type = static_cast<type_data*>(this_type_ptr.get());
int next_tag = 0;
std::set<std::string> var_set;
type_app* return_app = new type_app(std::move(this_type_ptr));
type_ptr return_type(return_app);
for(auto& var : vars) {
if(var_set.find(var) != var_set.end()) throw 0;
var_set.insert(var);
return_app->arguments.push_back(type_ptr(new type_var(var)));
}
for(auto& constructor : constructors) {
constructor->tag = next_tag;
this_type->constructors[constructor->name] = { next_tag++ };
type_ptr full_type = return_type;
for(auto it = constructor->types.rbegin(); it != constructor->types.rend(); it++) {
type_ptr type = (*it)->to_type(var_set, env);
full_type = type_ptr(new type_arr(type, full_type));
}
type_scheme_ptr full_scheme(new type_scheme(std::move(full_type)));
full_scheme->forall.insert(full_scheme->forall.begin(), vars.begin(), vars.end());
env->bind(constructor->name, full_scheme);
}
}
void definition_data::into_globals(global_scope& scope) {
for(auto& constructor : constructors) {
global_constructor& c = scope.add_constructor(
constructor->name, constructor->tag, constructor->types.size());
env->set_mangled_name(constructor->name, c.name);
}
}
void definition_group::find_free(std::set<std::string>& into) {
for(auto& def_pair : defs_defn) {
def_pair.second->find_free();
for(auto& free_var : def_pair.second->free_variables) {
if(defs_defn.find(free_var) == defs_defn.end()) {
into.insert(free_var);
} else {
def_pair.second->nearby_variables.insert(free_var);
}
}
}
}
void definition_group::typecheck(type_mgr& mgr, type_env_ptr& env) {
this->env = type_scope(env);
for(auto& def_data : defs_data) {
def_data.second->insert_types(this->env);
}
for(auto& def_data : defs_data) {
def_data.second->insert_constructors();
}
function_graph dependency_graph;
for(auto& def_defn : defs_defn) {
def_defn.second->find_free();
dependency_graph.add_function(def_defn.second->name);
for(auto& dependency : def_defn.second->nearby_variables) {
if(defs_defn.find(dependency) == defs_defn.end())
throw 0;
dependency_graph.add_edge(def_defn.second->name, dependency);
}
}
std::vector<group_ptr> groups = dependency_graph.compute_order();
for(auto it = groups.rbegin(); it != groups.rend(); it++) {
auto& group = *it;
for(auto& def_defnn_name : group->members) {
auto& def_defn = defs_defn.find(def_defnn_name)->second;
def_defn->insert_types(mgr, this->env, vis);
}
for(auto& def_defnn_name : group->members) {
auto& def_defn = defs_defn.find(def_defnn_name)->second;
def_defn->typecheck(mgr);
}
for(auto& def_defnn_name : group->members) {
this->env->generalize(def_defnn_name, *group, mgr);
}
}
}

View File

@@ -0,0 +1,83 @@
#pragma once
#include <memory>
#include <vector>
#include <map>
#include <set>
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "parsed_type.hpp"
#include "type_env.hpp"
#include "global_scope.hpp"
struct ast;
using ast_ptr = std::unique_ptr<ast>;
struct constructor {
std::string name;
std::vector<parsed_type_ptr> types;
int8_t tag;
constructor(std::string n, std::vector<parsed_type_ptr> ts)
: name(std::move(n)), types(std::move(ts)) {}
};
using constructor_ptr = std::unique_ptr<constructor>;
struct definition_defn {
std::string name;
std::vector<std::string> params;
ast_ptr body;
type_env_ptr env;
type_env_ptr var_env;
std::set<std::string> free_variables;
std::set<std::string> nearby_variables;
type_ptr full_type;
type_ptr return_type;
definition_defn(std::string n, std::vector<std::string> p, ast_ptr b)
: name(std::move(n)), params(std::move(p)), body(std::move(b)) {
}
void find_free();
void insert_types(type_mgr& mgr, type_env_ptr& env, visibility v);
void typecheck(type_mgr& mgr);
global_function& into_global(global_scope& scope);
};
using definition_defn_ptr = std::unique_ptr<definition_defn>;
struct definition_data {
std::string name;
std::vector<std::string> vars;
std::vector<constructor_ptr> constructors;
type_env_ptr env;
definition_data(
std::string n,
std::vector<std::string> vs,
std::vector<constructor_ptr> cs)
: name(std::move(n)), vars(std::move(vs)), constructors(std::move(cs)) {}
void insert_types(type_env_ptr& env);
void insert_constructors() const;
void into_globals(global_scope& scope);
};
using definition_data_ptr = std::unique_ptr<definition_data>;
struct definition_group {
std::map<std::string, definition_data_ptr> defs_data;
std::map<std::string, definition_defn_ptr> defs_defn;
visibility vis;
type_env_ptr env;
definition_group(visibility v = visibility::local) : vis(v) {}
void find_free(std::set<std::string>& into);
void typecheck(type_mgr& mgr, type_env_ptr& env);
};

23
code/compiler/12/env.cpp Normal file
View File

@@ -0,0 +1,23 @@
#include "env.hpp"
int env_var::get_offset(const std::string& name) const {
if(name == this->name) return 0;
if(parent) return parent->get_offset(name) + 1;
throw 0;
}
bool env_var::has_variable(const std::string& name) const {
if(name == this->name) return true;
if(parent) return parent->has_variable(name);
return false;
}
int env_offset::get_offset(const std::string& name) const {
if(parent) return parent->get_offset(name) + offset;
throw 0;
}
bool env_offset::has_variable(const std::string& name) const {
if(parent) return parent->has_variable(name);
return false;
}

34
code/compiler/12/env.hpp Normal file
View File

@@ -0,0 +1,34 @@
#pragma once
#include <memory>
#include <string>
struct env {
virtual ~env() = default;
virtual int get_offset(const std::string& name) const = 0;
virtual bool has_variable(const std::string& name) const = 0;
};
using env_ptr = std::shared_ptr<env>;
struct env_var : public env {
std::string name;
env_ptr parent;
env_var(std::string n, env_ptr p)
: name(std::move(n)), parent(std::move(p)) {}
int get_offset(const std::string& name) const;
bool has_variable(const std::string& name) const;
};
struct env_offset : public env {
int offset;
env_ptr parent;
env_offset(int o, env_ptr p)
: offset(o), parent(std::move(p)) {}
int get_offset(const std::string& name) const;
bool has_variable(const std::string& name) const;
};

View File

@@ -0,0 +1,5 @@
#include "error.hpp"
const char* type_error::what() const noexcept {
return "an error occured while checking the types of the program";
}

View File

@@ -0,0 +1,21 @@
#pragma once
#include <exception>
#include "type.hpp"
struct type_error : std::exception {
std::string description;
type_error(std::string d)
: description(std::move(d)) {}
const char* what() const noexcept override;
};
struct unification_error : public type_error {
type_ptr left;
type_ptr right;
unification_error(type_ptr l, type_ptr r)
: left(std::move(l)), right(std::move(r)),
type_error("failed to unify types") {}
};

View File

@@ -0,0 +1,2 @@
data Bool = { True, False }
defn main = { 3 + True }

View File

@@ -0,0 +1 @@
defn main = { 1 2 3 4 5 }

View File

@@ -0,0 +1,8 @@
data List = { Nil, Cons Int List }
defn head l = {
case l of {
Nil -> { 0 }
Cons x y z -> { x }
}
}

View File

@@ -0,0 +1,17 @@
data List a = { Nil, Cons a (List a) }
defn fix f = { let { defn x = { f x } } in { x } }
defn fixpointOnes fo = { Cons 1 fo }
defn sumTwo l = {
case l of {
Nil -> { 0 }
Cons x xs -> {
x + case xs of {
Nil -> { 0 }
Cons y ys -> { y }
}
}
}
}
defn main = { sumTwo (fix fixpointOnes) }

View File

@@ -0,0 +1,8 @@
data Bool = { True, False }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn main = { if (if True False True) 11 3 }

View File

@@ -0,0 +1,19 @@
data List a = { Nil, Cons a (List a) }
defn sum l = {
case l of {
Nil -> { 0 }
Cons x xs -> { x + sum xs}
}
}
defn map f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { Cons (f x) (map f xs) }
}
}
defn main = {
sum (map \x -> { x * x } (map (\x -> { x + x }) (Cons 1 (Cons 2 (Cons 3 Nil)))))
}

View File

@@ -0,0 +1,47 @@
data Bool = { True, False }
data List a = { Nil, Cons a (List a) }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn mergeUntil l r p = {
let {
defn mergeLeft nl nr = {
case nl of {
Nil -> { Nil }
Cons x xs -> { if (p x) (Cons x (mergeRight xs nr)) Nil }
}
}
defn mergeRight nl nr = {
case nr of {
Nil -> { Nil }
Cons x xs -> { if (p x) (Cons x (mergeLeft nl xs)) Nil }
}
}
} in {
mergeLeft l r
}
}
defn const x y = { x }
defn sum l = {
case l of {
Nil -> { 0 }
Cons x xs -> { x + sum xs }
}
}
defn main = {
let {
defn firstList = { Cons 1 (Cons 3 (Cons 5 Nil)) }
defn secondList = { Cons 2 (Cons 4 (Cons 6 Nil)) }
} in {
sum (mergeUntil firstList secondList (const True))
}
}

View File

@@ -0,0 +1,32 @@
data List a = { Nil, Cons a (List a) }
defn map f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { Cons (f x) (map f xs) }
}
}
defn foldl f b l = {
case l of {
Nil -> { b }
Cons x xs -> { foldl f (f b x) xs }
}
}
defn foldr f b l = {
case l of {
Nil -> { b }
Cons x xs -> { f x (foldr f b xs) }
}
}
defn list = { Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil))) }
defn add x y = { x + y }
defn sum l = { foldr add 0 l }
defn skipAdd x y = { y + 1 }
defn length l = { foldr skipAdd 0 l }
defn main = { sum list + length list }

View File

@@ -0,0 +1,25 @@
data Bool = { True, False }
data List = { Nil, Cons Int List }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn oddEven l e = {
case l of {
Nil -> { e }
Cons x xs -> { evenOdd xs e }
}
}
defn evenOdd l e = {
case l of {
Nil -> { e }
Cons x xs -> { oddEven xs e }
}
}
defn main = { if (oddEven (Cons 1 (Cons 2 (Cons 3 Nil))) True) (oddEven (Cons 1 (Cons 2 (Cons 3 Nil))) 1) 3 }

View File

@@ -0,0 +1,23 @@
data Pair a b = { Pair a b }
defn packer = {
let {
data Packed a = { Packed a }
defn pack a = { Packed a }
defn unpack p = {
case p of {
Packed a -> { a }
}
}
} in {
Pair pack unpack
}
}
defn main = {
case packer of {
Pair pack unpack -> {
unpack (pack 3)
}
}
}

View File

@@ -0,0 +1,17 @@
data Pair a b = { MkPair a b }
defn fst p = {
case p of {
MkPair a b -> { a }
}
}
defn snd p = {
case p of {
MkPair a b -> { b }
}
}
defn pair = { MkPair 1 (MkPair 2 3) }
defn main = { fst pair + snd (snd pair) }

View File

@@ -0,0 +1,122 @@
data List = { Nil, Cons Nat List }
data Bool = { True, False }
data Nat = { O, S Nat }
defn if c t e = {
case c of {
True -> { t }
False -> { e }
}
}
defn toInt n = {
case n of {
O -> { 0 }
S np -> { 1 + toInt np }
}
}
defn lte n m = {
case m of {
O -> {
case n of {
O -> { True }
S np -> { False }
}
}
S mp -> {
case n of {
O -> { True }
S np -> { lte np mp }
}
}
}
}
defn minus n m = {
case m of {
O -> { n }
S mp -> {
case n of {
O -> { O }
S np -> {
minus np mp
}
}
}
}
}
defn mod n m = {
if (lte m n) (mod (minus n m) m) n
}
defn notDivisibleBy n m = {
case (mod m n) of {
O -> { False }
S mp -> { True }
}
}
defn filter f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { if (f x) (Cons x (filter f xs)) (filter f xs) }
}
}
defn map f l = {
case l of {
Nil -> { Nil }
Cons x xs -> { Cons (f x) (map f xs) }
}
}
defn nats = {
Cons (S (S O)) (map S nats)
}
defn primesRec l = {
case l of {
Nil -> { Nil }
Cons p xs -> { Cons p (primesRec (filter (notDivisibleBy p) xs)) }
}
}
defn primes = {
primesRec nats
}
defn take n l = {
case l of {
Nil -> { Nil }
Cons x xs -> {
case n of {
O -> { Nil }
S np -> { Cons x (take np xs) }
}
}
}
}
defn head l = {
case l of {
Nil -> { O }
Cons x xs -> { x }
}
}
defn reverseAcc a l = {
case l of {
Nil -> { a }
Cons x xs -> { reverseAcc (Cons x a) xs }
}
}
defn reverse l = {
reverseAcc Nil l
}
defn main = {
toInt (head (reverse (take ((S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S (S O))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))))) primes)))
}

View File

@@ -0,0 +1,31 @@
#include "../runtime.h"
void f_add(struct stack* s) {
struct node_num* left = (struct node_num*) eval(stack_peek(s, 0));
struct node_num* right = (struct node_num*) eval(stack_peek(s, 1));
stack_push(s, (struct node_base*) alloc_num(left->value + right->value));
}
void f_main(struct stack* s) {
// PushInt 320
stack_push(s, (struct node_base*) alloc_num(320));
// PushInt 6
stack_push(s, (struct node_base*) alloc_num(6));
// PushGlobal f_add (the function for +)
stack_push(s, (struct node_base*) alloc_global(f_add, 2));
struct node_base* left;
struct node_base* right;
// MkApp
left = stack_pop(s);
right = stack_pop(s);
stack_push(s, (struct node_base*) alloc_app(left, right));
// MkApp
left = stack_pop(s);
right = stack_pop(s);
stack_push(s, (struct node_base*) alloc_app(left, right));
}

View File

@@ -0,0 +1,2 @@
defn main = { sum 320 6 }
defn sum x y = { x + y }

View File

@@ -0,0 +1,3 @@
defn add x y = { x + y }
defn double x = { add x x }
defn main = { double 163 }

View File

@@ -0,0 +1,9 @@
data List a = { Nil, Cons a (List a) }
data Bool = { True, False }
defn length l = {
case l of {
Nil -> { 0 }
Cons x xs -> { 1 + length xs }
}
}
defn main = { length (Cons 1 (Cons 2 (Cons 3 Nil))) + length (Cons True (Cons False (Cons True Nil))) }

View File

@@ -0,0 +1,16 @@
data List = { Nil, Cons Int List }
defn add x y = { x + y }
defn mul x y = { x * y }
defn foldr f b l = {
case l of {
Nil -> { b }
Cons x xs -> { f x (foldr f b xs) }
}
}
defn main = {
foldr add 0 (Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil)))) +
foldr mul 1 (Cons 1 (Cons 2 (Cons 3 (Cons 4 Nil))))
}

View File

@@ -0,0 +1,17 @@
data List = { Nil, Cons Int List }
defn sumZip l m = {
case l of {
Nil -> { 0 }
Cons x xs -> {
case m of {
Nil -> { 0 }
Cons y ys -> { x + y + sumZip xs ys }
}
}
}
}
defn ones = { Cons 1 ones }
defn main = { sumZip ones (Cons 1 (Cons 2 (Cons 3 Nil))) }

View File

@@ -0,0 +1,83 @@
#include "global_scope.hpp"
#include "ast.hpp"
void global_function::compile() {
env_ptr new_env = env_ptr(new env_offset(0, nullptr));
for(auto it = params.rbegin(); it != params.rend(); it++) {
new_env = env_ptr(new env_var(*it, new_env));
}
body->compile(new_env, instructions);
instructions.push_back(instruction_ptr(new instruction_update(params.size())));
instructions.push_back(instruction_ptr(new instruction_pop(params.size())));
}
void global_function::declare_llvm(llvm_context& ctx) {
generated_function = ctx.create_custom_function(name, params.size());
}
void global_function::generate_llvm(llvm_context& ctx) {
ctx.builder.SetInsertPoint(&generated_function->getEntryBlock());
for(auto& instruction : instructions) {
instruction->gen_llvm(ctx, generated_function);
}
ctx.builder.CreateRetVoid();
}
void global_constructor::generate_llvm(llvm_context& ctx) {
auto new_function =
ctx.create_custom_function(name, arity);
std::vector<instruction_ptr> instructions;
instructions.push_back(instruction_ptr(new instruction_pack(tag, arity)));
instructions.push_back(instruction_ptr(new instruction_update(0)));
ctx.builder.SetInsertPoint(&new_function->getEntryBlock());
for (auto& instruction : instructions) {
instruction->gen_llvm(ctx, new_function);
}
ctx.builder.CreateRetVoid();
}
global_function& global_scope::add_function(std::string n, std::vector<std::string> ps, ast_ptr b) {
global_function* new_function = new global_function(mangle_name(n), std::move(ps), std::move(b));
functions.push_back(global_function_ptr(new_function));
return *new_function;
}
global_constructor& global_scope::add_constructor(std::string n, int8_t t, size_t a) {
global_constructor* new_constructor = new global_constructor(mangle_name(n), t, a);
constructors.push_back(global_constructor_ptr(new_constructor));
return *new_constructor;
}
void global_scope::compile() {
for(auto& function : functions) {
function->compile();
}
}
void global_scope::generate_llvm(llvm_context& ctx) {
for(auto& constructor : constructors) {
constructor->generate_llvm(ctx);
}
for(auto& function : functions) {
function->declare_llvm(ctx);
}
for(auto& function : functions) {
function->generate_llvm(ctx);
}
}
std::string global_scope::mangle_name(const std::string& n) {
auto occurence_it = occurence_count.find(n);
int occurence = 0;
if(occurence_it != occurence_count.end()) {
occurence = occurence_it->second + 1;
}
occurence_count[n] = occurence;
std::string final_name = n;
if (occurence != 0) {
final_name += "_";
final_name += std::to_string(occurence);
}
return final_name;
}

View File

@@ -0,0 +1,55 @@
#pragma once
#include <memory>
#include <string>
#include <vector>
#include <llvm/IR/Function.h>
#include "instruction.hpp"
struct ast;
using ast_ptr = std::unique_ptr<ast>;
struct global_function {
std::string name;
std::vector<std::string> params;
ast_ptr body;
std::vector<instruction_ptr> instructions;
llvm::Function* generated_function;
global_function(std::string n, std::vector<std::string> ps, ast_ptr b)
: name(std::move(n)), params(std::move(ps)), body(std::move(b)) {}
void compile();
void declare_llvm(llvm_context& ctx);
void generate_llvm(llvm_context& ctx);
};
using global_function_ptr = std::unique_ptr<global_function>;
struct global_constructor {
std::string name;
int8_t tag;
size_t arity;
global_constructor(std::string n, int8_t t, size_t a)
: name(std::move(n)), tag(t), arity(a) {}
void generate_llvm(llvm_context& ctx);
};
using global_constructor_ptr = std::unique_ptr<global_constructor>;
struct global_scope {
std::map<std::string, int> occurence_count;
std::vector<global_function_ptr> functions;
std::vector<global_constructor_ptr> constructors;
global_function& add_function(std::string n, std::vector<std::string> ps, ast_ptr b);
global_constructor& add_constructor(std::string n, int8_t t, size_t a);
void compile();
void generate_llvm(llvm_context& ctx);
private:
std::string mangle_name(const std::string& n);
};

114
code/compiler/12/graph.cpp Normal file
View File

@@ -0,0 +1,114 @@
#include "graph.hpp"
std::set<function_graph::edge> function_graph::compute_transitive_edges() {
std::set<edge> transitive_edges;
transitive_edges.insert(edges.begin(), edges.end());
for(auto& connector : adjacency_lists) {
for(auto& from : adjacency_lists) {
edge to_connector { from.first, connector.first };
for(auto& to : adjacency_lists) {
edge full_jump { from.first, to.first };
if(transitive_edges.find(full_jump) != transitive_edges.end()) continue;
edge from_connector { connector.first, to.first };
if(transitive_edges.find(to_connector) != transitive_edges.end() &&
transitive_edges.find(from_connector) != transitive_edges.end())
transitive_edges.insert(std::move(full_jump));
}
}
}
return transitive_edges;
}
void function_graph::create_groups(
const std::set<edge>& transitive_edges,
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
group_id id_counter = 0;
for(auto& vertex : adjacency_lists) {
if(group_ids.find(vertex.first) != group_ids.end())
continue;
data_ptr new_group(new group_data);
new_group->functions.insert(vertex.first);
group_data_map[id_counter] = new_group;
group_ids[vertex.first] = id_counter;
for(auto& other_vertex : adjacency_lists) {
if(transitive_edges.find({vertex.first, other_vertex.first}) != transitive_edges.end() &&
transitive_edges.find({other_vertex.first, vertex.first}) != transitive_edges.end()) {
group_ids[other_vertex.first] = id_counter;
new_group->functions.insert(other_vertex.first);
}
}
id_counter++;
}
}
void function_graph::create_edges(
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
std::set<std::pair<group_id, group_id>> group_edges;
for(auto& vertex : adjacency_lists) {
auto vertex_id = group_ids[vertex.first];
auto& vertex_data = group_data_map[vertex_id];
for(auto& other_vertex : vertex.second) {
auto other_id = group_ids[other_vertex];
if(vertex_id == other_id) continue;
if(group_edges.find({vertex_id, other_id}) != group_edges.end())
continue;
group_edges.insert({vertex_id, other_id});
vertex_data->adjacency_list.insert(other_id);
group_data_map[other_id]->indegree++;
}
}
}
std::vector<group_ptr> function_graph::generate_order(
std::map<function, group_id>& group_ids,
std::map<group_id, data_ptr>& group_data_map) {
std::queue<group_id> id_queue;
std::vector<group_ptr> output;
for(auto& group : group_data_map) {
if(group.second->indegree == 0) id_queue.push(group.first);
}
while(!id_queue.empty()) {
auto new_id = id_queue.front();
auto& group_data = group_data_map[new_id];
group_ptr output_group(new group);
output_group->members = std::move(group_data->functions);
id_queue.pop();
for(auto& adjacent_group : group_data->adjacency_list) {
if(--group_data_map[adjacent_group]->indegree == 0)
id_queue.push(adjacent_group);
}
output.push_back(std::move(output_group));
}
return output;
}
std::set<function>& function_graph::add_function(const function& f) {
auto adjacency_list_it = adjacency_lists.find(f);
if(adjacency_list_it != adjacency_lists.end()) {
return adjacency_list_it->second;
} else {
return adjacency_lists[f] = { };
}
}
void function_graph::add_edge(const function& from, const function& to) {
add_function(from).insert(to);
edges.insert({ from, to });
}
std::vector<group_ptr> function_graph::compute_order() {
std::set<edge> transitive_edges = compute_transitive_edges();
std::map<function, group_id> group_ids;
std::map<group_id, data_ptr> group_data_map;
create_groups(transitive_edges, group_ids, group_data_map);
create_edges(group_ids, group_data_map);
return generate_order(group_ids, group_data_map);
}

View File

@@ -0,0 +1,51 @@
#pragma once
#include <algorithm>
#include <cstddef>
#include <queue>
#include <set>
#include <string>
#include <map>
#include <memory>
#include <vector>
using function = std::string;
struct group {
std::set<function> members;
};
using group_ptr = std::unique_ptr<group>;
class function_graph {
using group_id = size_t;
struct group_data {
std::set<function> functions;
std::set<group_id> adjacency_list;
size_t indegree;
};
using data_ptr = std::shared_ptr<group_data>;
using edge = std::pair<function, function>;
using group_edge = std::pair<group_id, group_id>;
std::map<function, std::set<function>> adjacency_lists;
std::set<edge> edges;
std::set<edge> compute_transitive_edges();
void create_groups(
const std::set<edge>&,
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
void create_edges(
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
std::vector<group_ptr> generate_order(
std::map<function, group_id>&,
std::map<group_id, data_ptr>&);
public:
std::set<function>& add_function(const function& f);
void add_edge(const function& from, const function& to);
std::vector<group_ptr> compute_order();
};

View File

@@ -0,0 +1,177 @@
#include "instruction.hpp"
#include "llvm_context.hpp"
#include <llvm/IR/BasicBlock.h>
#include <llvm/IR/Function.h>
using namespace llvm;
static void print_indent(int n, std::ostream& to) {
while(n--) to << " ";
}
void instruction_pushint::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "PushInt(" << value << ")" << std::endl;
}
void instruction_pushint::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_push(f, ctx.create_num(f, ctx.create_i32(value)));
}
void instruction_pushglobal::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "PushGlobal(" << name << ")" << std::endl;
}
void instruction_pushglobal::gen_llvm(llvm_context& ctx, Function* f) const {
auto& global_f = ctx.custom_functions.at("f_" + name);
auto arity = ctx.create_i32(global_f->arity);
ctx.create_push(f, ctx.create_global(f, global_f->function, arity));
}
void instruction_push::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Push(" << offset << ")" << std::endl;
}
void instruction_push::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_push(f, ctx.create_peek(f, ctx.create_size(offset)));
}
void instruction_pop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Pop(" << count << ")" << std::endl;
}
void instruction_pop::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_popn(f, ctx.create_size(count));
}
void instruction_mkapp::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "MkApp()" << std::endl;
}
void instruction_mkapp::gen_llvm(llvm_context& ctx, Function* f) const {
auto left = ctx.create_pop(f);
auto right = ctx.create_pop(f);
ctx.create_push(f, ctx.create_app(f, left, right));
}
void instruction_update::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Update(" << offset << ")" << std::endl;
}
void instruction_update::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_update(f, ctx.create_size(offset));
}
void instruction_pack::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Pack(" << tag << ", " << size << ")" << std::endl;
}
void instruction_pack::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_pack(f, ctx.create_size(size), ctx.create_i8(tag));
}
void instruction_split::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Split()" << std::endl;
}
void instruction_split::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_split(f, ctx.create_size(size));
}
void instruction_jump::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Jump(" << std::endl;
for(auto& instruction_set : branches) {
for(auto& instruction : instruction_set) {
instruction->print(indent + 2, to);
}
to << std::endl;
}
print_indent(indent, to);
to << ")" << std::endl;
}
void instruction_jump::gen_llvm(llvm_context& ctx, Function* f) const {
auto top_node = ctx.create_peek(f, ctx.create_size(0));
auto tag = ctx.unwrap_data_tag(top_node);
auto safety_block = BasicBlock::Create(ctx.ctx, "safety", f);
auto switch_op = ctx.builder.CreateSwitch(tag, safety_block, tag_mappings.size());
std::vector<BasicBlock*> blocks;
for(auto& branch : branches) {
auto branch_block = BasicBlock::Create(ctx.ctx, "branch", f);
ctx.builder.SetInsertPoint(branch_block);
for(auto& instruction : branch) {
instruction->gen_llvm(ctx, f);
}
ctx.builder.CreateBr(safety_block);
blocks.push_back(branch_block);
}
for(auto& mapping : tag_mappings) {
switch_op->addCase(ctx.create_i8(mapping.first), blocks[mapping.second]);
}
ctx.builder.SetInsertPoint(safety_block);
}
void instruction_slide::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Slide(" << offset << ")" << std::endl;
}
void instruction_slide::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_slide(f, ctx.create_size(offset));
}
void instruction_binop::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "BinOp(" << op_action(op) << ")" << std::endl;
}
void instruction_binop::gen_llvm(llvm_context& ctx, Function* f) const {
auto left_int = ctx.unwrap_num(ctx.create_pop(f));
auto right_int = ctx.unwrap_num(ctx.create_pop(f));
llvm::Value* result;
switch(op) {
case PLUS: result = ctx.builder.CreateAdd(left_int, right_int); break;
case MINUS: result = ctx.builder.CreateSub(left_int, right_int); break;
case TIMES: result = ctx.builder.CreateMul(left_int, right_int); break;
case DIVIDE: result = ctx.builder.CreateSDiv(left_int, right_int); break;
}
ctx.create_push(f, ctx.create_num(f, result));
}
void instruction_eval::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Eval()" << std::endl;
}
void instruction_eval::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_unwind(f);
}
void instruction_alloc::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Alloc(" << amount << ")" << std::endl;
}
void instruction_alloc::gen_llvm(llvm_context& ctx, Function* f) const {
ctx.create_alloc(f, ctx.create_size(amount));
}
void instruction_unwind::print(int indent, std::ostream& to) const {
print_indent(indent, to);
to << "Unwind()" << std::endl;
}
void instruction_unwind::gen_llvm(llvm_context& ctx, Function* f) const {
// Nothing
}

View File

@@ -0,0 +1,142 @@
#pragma once
#include <llvm/IR/Function.h>
#include <string>
#include <memory>
#include <vector>
#include <map>
#include <ostream>
#include "binop.hpp"
#include "llvm_context.hpp"
struct instruction {
virtual ~instruction() = default;
virtual void print(int indent, std::ostream& to) const = 0;
virtual void gen_llvm(llvm_context& ctx, llvm::Function* f) const = 0;
};
using instruction_ptr = std::unique_ptr<instruction>;
struct instruction_pushint : public instruction {
int value;
instruction_pushint(int v)
: value(v) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pushglobal : public instruction {
std::string name;
instruction_pushglobal(std::string n)
: name(std::move(n)) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_push : public instruction {
int offset;
instruction_push(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pop : public instruction {
int count;
instruction_pop(int c)
: count(c) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_mkapp : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_update : public instruction {
int offset;
instruction_update(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_pack : public instruction {
int tag;
int size;
instruction_pack(int t, int s)
: tag(t), size(s) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_split : public instruction {
int size;
instruction_split(int s)
: size(s) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_jump : public instruction {
std::vector<std::vector<instruction_ptr>> branches;
std::map<int, int> tag_mappings;
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_slide : public instruction {
int offset;
instruction_slide(int o)
: offset(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_binop : public instruction {
binop op;
instruction_binop(binop o)
: op(o) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_eval : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_alloc : public instruction {
int amount;
instruction_alloc(int a)
: amount(a) {}
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};
struct instruction_unwind : public instruction {
void print(int indent, std::ostream& to) const;
void gen_llvm(llvm_context& ctx, llvm::Function* f) const;
};

View File

@@ -0,0 +1,278 @@
#include "llvm_context.hpp"
#include <llvm/IR/DerivedTypes.h>
using namespace llvm;
void llvm_context::create_types() {
stack_type = StructType::create(ctx, "stack");
gmachine_type = StructType::create(ctx, "gmachine");
stack_ptr_type = PointerType::getUnqual(stack_type);
gmachine_ptr_type = PointerType::getUnqual(gmachine_type);
tag_type = IntegerType::getInt8Ty(ctx);
struct_types["node_base"] = StructType::create(ctx, "node_base");
struct_types["node_app"] = StructType::create(ctx, "node_app");
struct_types["node_num"] = StructType::create(ctx, "node_num");
struct_types["node_global"] = StructType::create(ctx, "node_global");
struct_types["node_ind"] = StructType::create(ctx, "node_ind");
struct_types["node_data"] = StructType::create(ctx, "node_data");
node_ptr_type = PointerType::getUnqual(struct_types.at("node_base"));
function_type = FunctionType::get(Type::getVoidTy(ctx), { gmachine_ptr_type }, false);
gmachine_type->setBody(
stack_ptr_type,
node_ptr_type,
IntegerType::getInt64Ty(ctx),
IntegerType::getInt64Ty(ctx)
);
struct_types.at("node_base")->setBody(
IntegerType::getInt32Ty(ctx),
IntegerType::getInt8Ty(ctx),
node_ptr_type
);
struct_types.at("node_app")->setBody(
struct_types.at("node_base"),
node_ptr_type,
node_ptr_type
);
struct_types.at("node_num")->setBody(
struct_types.at("node_base"),
IntegerType::getInt32Ty(ctx)
);
struct_types.at("node_global")->setBody(
struct_types.at("node_base"),
FunctionType::get(Type::getVoidTy(ctx), { stack_ptr_type }, false)
);
struct_types.at("node_ind")->setBody(
struct_types.at("node_base"),
node_ptr_type
);
struct_types.at("node_data")->setBody(
struct_types.at("node_base"),
IntegerType::getInt8Ty(ctx),
PointerType::getUnqual(node_ptr_type)
);
}
void llvm_context::create_functions() {
auto void_type = Type::getVoidTy(ctx);
auto sizet_type = IntegerType::get(ctx, sizeof(size_t) * 8);
functions["stack_init"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_init",
&module
);
functions["stack_free"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_free",
&module
);
functions["stack_push"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_push",
&module
);
functions["stack_pop"] = Function::Create(
FunctionType::get(node_ptr_type, { stack_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_pop",
&module
);
functions["stack_peek"] = Function::Create(
FunctionType::get(node_ptr_type, { stack_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_peek",
&module
);
functions["stack_popn"] = Function::Create(
FunctionType::get(void_type, { stack_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"stack_popn",
&module
);
functions["gmachine_slide"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_slide",
&module
);
functions["gmachine_update"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_update",
&module
);
functions["gmachine_alloc"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_alloc",
&module
);
functions["gmachine_pack"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type, tag_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_pack",
&module
);
functions["gmachine_split"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type, sizet_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_split",
&module
);
functions["gmachine_track"] = Function::Create(
FunctionType::get(node_ptr_type, { gmachine_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"gmachine_track",
&module
);
auto int32_type = IntegerType::getInt32Ty(ctx);
functions["alloc_app"] = Function::Create(
FunctionType::get(node_ptr_type, { node_ptr_type, node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_app",
&module
);
functions["alloc_num"] = Function::Create(
FunctionType::get(node_ptr_type, { int32_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_num",
&module
);
functions["alloc_global"] = Function::Create(
FunctionType::get(node_ptr_type, { function_type, int32_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_global",
&module
);
functions["alloc_ind"] = Function::Create(
FunctionType::get(node_ptr_type, { node_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"alloc_ind",
&module
);
functions["unwind"] = Function::Create(
FunctionType::get(void_type, { gmachine_ptr_type }, false),
Function::LinkageTypes::ExternalLinkage,
"unwind",
&module
);
}
ConstantInt* llvm_context::create_i8(int8_t i) {
return ConstantInt::get(ctx, APInt(8, i));
}
ConstantInt* llvm_context::create_i32(int32_t i) {
return ConstantInt::get(ctx, APInt(32, i));
}
ConstantInt* llvm_context::create_size(size_t i) {
return ConstantInt::get(ctx, APInt(sizeof(size_t) * 8, i));
}
Value* llvm_context::create_pop(Function* f) {
auto pop_f = functions.at("stack_pop");
return builder.CreateCall(pop_f, { unwrap_gmachine_stack_ptr(f->arg_begin()) });
}
Value* llvm_context::create_peek(Function* f, Value* off) {
auto peek_f = functions.at("stack_peek");
return builder.CreateCall(peek_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), off });
}
void llvm_context::create_push(Function* f, Value* v) {
auto push_f = functions.at("stack_push");
builder.CreateCall(push_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), v });
}
void llvm_context::create_popn(Function* f, Value* off) {
auto popn_f = functions.at("stack_popn");
builder.CreateCall(popn_f, { unwrap_gmachine_stack_ptr(f->arg_begin()), off });
}
void llvm_context::create_update(Function* f, Value* off) {
auto update_f = functions.at("gmachine_update");
builder.CreateCall(update_f, { f->arg_begin(), off });
}
void llvm_context::create_pack(Function* f, Value* c, Value* t) {
auto pack_f = functions.at("gmachine_pack");
builder.CreateCall(pack_f, { f->arg_begin(), c, t });
}
void llvm_context::create_split(Function* f, Value* c) {
auto split_f = functions.at("gmachine_split");
builder.CreateCall(split_f, { f->arg_begin(), c });
}
void llvm_context::create_slide(Function* f, Value* off) {
auto slide_f = functions.at("gmachine_slide");
builder.CreateCall(slide_f, { f->arg_begin(), off });
}
void llvm_context::create_alloc(Function* f, Value* n) {
auto alloc_f = functions.at("gmachine_alloc");
builder.CreateCall(alloc_f, { f->arg_begin(), n });
}
Value* llvm_context::create_track(Function* f, Value* v) {
auto track_f = functions.at("gmachine_track");
return builder.CreateCall(track_f, { f->arg_begin(), v });
}
void llvm_context::create_unwind(Function* f) {
auto unwind_f = functions.at("unwind");
builder.CreateCall(unwind_f, { f->args().begin() });
}
Value* llvm_context::unwrap_gmachine_stack_ptr(Value* g) {
auto offset_0 = create_i32(0);
return builder.CreateGEP(g, { offset_0, offset_0 });
}
Value* llvm_context::unwrap_num(Value* v) {
auto num_ptr_type = PointerType::getUnqual(struct_types.at("node_num"));
auto cast = builder.CreatePointerCast(v, num_ptr_type);
auto offset_0 = create_i32(0);
auto offset_1 = create_i32(1);
auto int_ptr = builder.CreateGEP(cast, { offset_0, offset_1 });
return builder.CreateLoad(int_ptr);
}
Value* llvm_context::create_num(Function* f, Value* v) {
auto alloc_num_f = functions.at("alloc_num");
auto alloc_num_call = builder.CreateCall(alloc_num_f, { v });
return create_track(f, alloc_num_call);
}
Value* llvm_context::unwrap_data_tag(Value* v) {
auto data_ptr_type = PointerType::getUnqual(struct_types.at("node_data"));
auto cast = builder.CreatePointerCast(v, data_ptr_type);
auto offset_0 = create_i32(0);
auto offset_1 = create_i32(1);
auto tag_ptr = builder.CreateGEP(cast, { offset_0, offset_1 });
return builder.CreateLoad(tag_ptr);
}
Value* llvm_context::create_global(Function* f, Value* gf, Value* a) {
auto alloc_global_f = functions.at("alloc_global");
auto alloc_global_call = builder.CreateCall(alloc_global_f, { gf, a });
return create_track(f, alloc_global_call);
}
Value* llvm_context::create_app(Function* f, Value* l, Value* r) {
auto alloc_app_f = functions.at("alloc_app");
auto alloc_app_call = builder.CreateCall(alloc_app_f, { l, r });
return create_track(f, alloc_app_call);
}
llvm::Function* llvm_context::create_custom_function(std::string name, int32_t arity) {
auto void_type = llvm::Type::getVoidTy(ctx);
auto new_function = llvm::Function::Create(
function_type,
llvm::Function::LinkageTypes::ExternalLinkage,
"f_" + name,
&module
);
auto start_block = llvm::BasicBlock::Create(ctx, "entry", new_function);
auto new_custom_f = custom_function_ptr(new custom_function());
new_custom_f->arity = arity;
new_custom_f->function = new_function;
custom_functions["f_" + name] = std::move(new_custom_f);
return new_function;
}

View File

@@ -0,0 +1,72 @@
#pragma once
#include <llvm/IR/DerivedTypes.h>
#include <llvm/IR/Function.h>
#include <llvm/IR/LLVMContext.h>
#include <llvm/IR/IRBuilder.h>
#include <llvm/IR/Module.h>
#include <llvm/IR/Value.h>
#include <map>
struct llvm_context {
struct custom_function {
llvm::Function* function;
int32_t arity;
};
using custom_function_ptr = std::unique_ptr<custom_function>;
llvm::LLVMContext ctx;
llvm::IRBuilder<> builder;
llvm::Module module;
std::map<std::string, custom_function_ptr> custom_functions;
std::map<std::string, llvm::Function*> functions;
std::map<std::string, llvm::StructType*> struct_types;
llvm::StructType* stack_type;
llvm::StructType* gmachine_type;
llvm::PointerType* stack_ptr_type;
llvm::PointerType* gmachine_ptr_type;
llvm::PointerType* node_ptr_type;
llvm::IntegerType* tag_type;
llvm::FunctionType* function_type;
llvm_context()
: builder(ctx), module("bloglang", ctx) {
create_types();
create_functions();
}
void create_types();
void create_functions();
llvm::ConstantInt* create_i8(int8_t);
llvm::ConstantInt* create_i32(int32_t);
llvm::ConstantInt* create_size(size_t);
llvm::Value* create_pop(llvm::Function*);
llvm::Value* create_peek(llvm::Function*, llvm::Value*);
void create_push(llvm::Function*, llvm::Value*);
void create_popn(llvm::Function*, llvm::Value*);
void create_update(llvm::Function*, llvm::Value*);
void create_pack(llvm::Function*, llvm::Value*, llvm::Value*);
void create_split(llvm::Function*, llvm::Value*);
void create_slide(llvm::Function*, llvm::Value*);
void create_alloc(llvm::Function*, llvm::Value*);
llvm::Value* create_track(llvm::Function*, llvm::Value*);
void create_unwind(llvm::Function*);
llvm::Value* unwrap_gmachine_stack_ptr(llvm::Value*);
llvm::Value* unwrap_num(llvm::Value*);
llvm::Value* create_num(llvm::Function*, llvm::Value*);
llvm::Value* unwrap_data_tag(llvm::Value*);
llvm::Value* create_global(llvm::Function*, llvm::Value*, llvm::Value*);
llvm::Value* create_app(llvm::Function*, llvm::Value*, llvm::Value*);
llvm::Function* create_custom_function(std::string name, int32_t arity);
};

163
code/compiler/12/main.cpp Normal file
View File

@@ -0,0 +1,163 @@
#include "ast.hpp"
#include <iostream>
#include "binop.hpp"
#include "definition.hpp"
#include "graph.hpp"
#include "instruction.hpp"
#include "llvm_context.hpp"
#include "parser.hpp"
#include "error.hpp"
#include "type.hpp"
#include "llvm/IR/LegacyPassManager.h"
#include "llvm/IR/Verifier.h"
#include "llvm/Support/TargetSelect.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Support/FileSystem.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Target/TargetMachine.h"
void yy::parser::error(const std::string& msg) {
std::cout << "An error occured: " << msg << std::endl;
}
extern definition_group global_defs;
void typecheck_program(
definition_group& defs,
type_mgr& mgr, type_env_ptr& env) {
type_ptr int_type = type_ptr(new type_base("Int"));
env->bind_type("Int", int_type);
type_ptr int_type_app = type_ptr(new type_app(int_type));
type_ptr binop_type = type_ptr(new type_arr(
int_type_app,
type_ptr(new type_arr(int_type_app, int_type_app))));
env->bind("+", binop_type, visibility::global);
env->bind("-", binop_type, visibility::global);
env->bind("*", binop_type, visibility::global);
env->bind("/", binop_type, visibility::global);
std::set<std::string> free;
defs.find_free(free);
defs.typecheck(mgr, env);
for(auto& pair : defs.env->names) {
std::cout << pair.first << ": ";
pair.second.type->print(mgr, std::cout);
std::cout << std::endl;
}
}
global_scope translate_program(definition_group& group) {
global_scope scope;
for(auto& data : group.defs_data) {
data.second->into_globals(scope);
}
for(auto& defn : group.defs_defn) {
auto& function = defn.second->into_global(scope);
function.body->env->parent->set_mangled_name(defn.first, function.name);
}
return scope;
}
void gen_llvm_internal_op(llvm_context& ctx, binop op) {
auto new_function = ctx.create_custom_function(op_action(op), 2);
std::vector<instruction_ptr> instructions;
instructions.push_back(instruction_ptr(new instruction_push(1)));
instructions.push_back(instruction_ptr(new instruction_eval()));
instructions.push_back(instruction_ptr(new instruction_push(1)));
instructions.push_back(instruction_ptr(new instruction_eval()));
instructions.push_back(instruction_ptr(new instruction_binop(op)));
instructions.push_back(instruction_ptr(new instruction_update(2)));
instructions.push_back(instruction_ptr(new instruction_pop(2)));
ctx.builder.SetInsertPoint(&new_function->getEntryBlock());
for(auto& instruction : instructions) {
instruction->gen_llvm(ctx, new_function);
}
ctx.builder.CreateRetVoid();
}
void output_llvm(llvm_context& ctx, const std::string& filename) {
std::string targetTriple = llvm::sys::getDefaultTargetTriple();
llvm::InitializeNativeTarget();
llvm::InitializeNativeTargetAsmParser();
llvm::InitializeNativeTargetAsmPrinter();
std::string error;
const llvm::Target* target =
llvm::TargetRegistry::lookupTarget(targetTriple, error);
if (!target) {
std::cerr << error << std::endl;
} else {
std::string cpu = "generic";
std::string features = "";
llvm::TargetOptions options;
llvm::TargetMachine* targetMachine =
target->createTargetMachine(targetTriple, cpu, features,
options, llvm::Optional<llvm::Reloc::Model>());
ctx.module.setDataLayout(targetMachine->createDataLayout());
ctx.module.setTargetTriple(targetTriple);
std::error_code ec;
llvm::raw_fd_ostream file(filename, ec, llvm::sys::fs::F_None);
if (ec) {
throw 0;
} else {
llvm::CodeGenFileType type = llvm::CGFT_ObjectFile;
llvm::legacy::PassManager pm;
if (targetMachine->addPassesToEmitFile(pm, file, NULL, type)) {
throw 0;
} else {
pm.run(ctx.module);
file.close();
}
}
}
}
void gen_llvm(global_scope& scope) {
llvm_context ctx;
gen_llvm_internal_op(ctx, PLUS);
gen_llvm_internal_op(ctx, MINUS);
gen_llvm_internal_op(ctx, TIMES);
gen_llvm_internal_op(ctx, DIVIDE);
scope.generate_llvm(ctx);
ctx.module.print(llvm::outs(), nullptr);
output_llvm(ctx, "program.o");
}
int main() {
yy::parser parser;
type_mgr mgr;
type_env_ptr env(new type_env);
parser.parse();
for(auto& def_defn : global_defs.defs_defn) {
std::cout << def_defn.second->name;
for(auto& param : def_defn.second->params) std::cout << " " << param;
std::cout << ":" << std::endl;
def_defn.second->body->print(1, std::cout);
}
try {
typecheck_program(global_defs, mgr, env);
global_scope scope = translate_program(global_defs);
scope.compile();
gen_llvm(scope);
} catch(unification_error& err) {
std::cout << "failed to unify types: " << std::endl;
std::cout << " (1) \033[34m";
err.left->print(mgr, std::cout);
std::cout << "\033[0m" << std::endl;
std::cout << " (2) \033[32m";
err.right->print(mgr, std::cout);
std::cout << "\033[0m" << std::endl;
} catch(type_error& err) {
std::cout << "failed to type check program: " << err.description << std::endl;
}
}

View File

@@ -0,0 +1,36 @@
#include "parsed_type.hpp"
#include "type.hpp"
#include "type_env.hpp"
type_ptr parsed_type_app::to_type(
const std::set<std::string>& vars,
const type_env& e) const {
auto parent_type = e.lookup_type(name);
if(parent_type == nullptr) throw 0;
type_base* base_type;
if(!(base_type = dynamic_cast<type_base*>(parent_type.get()))) throw 0;
if(base_type->arity != arguments.size()) throw 0;
type_app* new_app = new type_app(std::move(parent_type));
type_ptr to_return(new_app);
for(auto& arg : arguments) {
new_app->arguments.push_back(arg->to_type(vars, e));
}
return to_return;
}
type_ptr parsed_type_var::to_type(
const std::set<std::string>& vars,
const type_env& e) const {
if(vars.find(var) == vars.end()) throw 0;
return type_ptr(new type_var(var));
}
type_ptr parsed_type_arr::to_type(
const std::set<std::string>& vars,
const type_env& env) const {
auto new_left = left->to_type(vars, env);
auto new_right = right->to_type(vars, env);
return type_ptr(new type_arr(std::move(new_left), std::move(new_right)));
}

View File

@@ -0,0 +1,43 @@
#pragma once
#include <memory>
#include <set>
#include <string>
#include "type_env.hpp"
struct parsed_type {
virtual type_ptr to_type(
const std::set<std::string>& vars,
const type_env& env) const = 0;
};
using parsed_type_ptr = std::unique_ptr<parsed_type>;
struct parsed_type_app : parsed_type {
std::string name;
std::vector<parsed_type_ptr> arguments;
parsed_type_app(
std::string n,
std::vector<parsed_type_ptr> as)
: name(std::move(n)), arguments(std::move(as)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};
struct parsed_type_var : parsed_type {
std::string var;
parsed_type_var(std::string v) : var(std::move(v)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};
struct parsed_type_arr : parsed_type {
parsed_type_ptr left;
parsed_type_ptr right;
parsed_type_arr(parsed_type_ptr l, parsed_type_ptr r)
: left(std::move(l)), right(std::move(r)) {}
type_ptr to_type(const std::set<std::string>& vars, const type_env& env) const;
};

174
code/compiler/12/parser.y Normal file
View File

@@ -0,0 +1,174 @@
%{
#include <string>
#include <iostream>
#include <map>
#include "ast.hpp"
#include "definition.hpp"
#include "parser.hpp"
#include "parsed_type.hpp"
definition_group global_defs;
extern yy::parser::symbol_type yylex();
%}
%token BACKSLASH
%token PLUS
%token TIMES
%token MINUS
%token DIVIDE
%token <int> INT
%token DEFN
%token DATA
%token CASE
%token OF
%token LET
%token IN
%token OCURLY
%token CCURLY
%token OPAREN
%token CPAREN
%token COMMA
%token ARROW
%token EQUAL
%token <std::string> LID
%token <std::string> UID
%language "c++"
%define api.value.type variant
%define api.token.constructor
%type <std::vector<std::string>> lowercaseParams
%type <std::vector<branch_ptr>> branches
%type <std::vector<constructor_ptr>> constructors
%type <std::vector<parsed_type_ptr>> typeList
%type <definition_group> definitions
%type <parsed_type_ptr> type nonArrowType typeListElement
%type <ast_ptr> aAdd aMul case let lambda app appBase
%type <definition_data_ptr> data
%type <definition_defn_ptr> defn
%type <branch_ptr> branch
%type <pattern_ptr> pattern
%type <constructor_ptr> constructor
%start program
%%
program
: definitions { global_defs = std::move($1); global_defs.vis = visibility::global; }
;
definitions
: definitions defn { $$ = std::move($1); auto name = $2->name; $$.defs_defn[name] = std::move($2); }
| definitions data { $$ = std::move($1); auto name = $2->name; $$.defs_data[name] = std::move($2); }
| %empty { $$ = definition_group(); }
;
defn
: DEFN LID lowercaseParams EQUAL OCURLY aAdd CCURLY
{ $$ = definition_defn_ptr(
new definition_defn(std::move($2), std::move($3), std::move($6))); }
;
lowercaseParams
: %empty { $$ = std::vector<std::string>(); }
| lowercaseParams LID { $$ = std::move($1); $$.push_back(std::move($2)); }
;
aAdd
: aAdd PLUS aMul { $$ = ast_ptr(new ast_binop(PLUS, std::move($1), std::move($3))); }
| aAdd MINUS aMul { $$ = ast_ptr(new ast_binop(MINUS, std::move($1), std::move($3))); }
| aMul { $$ = std::move($1); }
;
aMul
: aMul TIMES app { $$ = ast_ptr(new ast_binop(TIMES, std::move($1), std::move($3))); }
| aMul DIVIDE app { $$ = ast_ptr(new ast_binop(DIVIDE, std::move($1), std::move($3))); }
| app { $$ = std::move($1); }
;
app
: app appBase { $$ = ast_ptr(new ast_app(std::move($1), std::move($2))); }
| appBase { $$ = std::move($1); }
;
appBase
: INT { $$ = ast_ptr(new ast_int($1)); }
| LID { $$ = ast_ptr(new ast_lid(std::move($1))); }
| UID { $$ = ast_ptr(new ast_uid(std::move($1))); }
| OPAREN aAdd CPAREN { $$ = std::move($2); }
| case { $$ = std::move($1); }
| let { $$ = std::move($1); }
| lambda { $$ = std::move($1); }
;
let
: LET OCURLY definitions CCURLY IN OCURLY aAdd CCURLY
{ $$ = ast_ptr(new ast_let(std::move($3), std::move($7))); }
;
lambda
: BACKSLASH lowercaseParams ARROW OCURLY aAdd CCURLY
{ $$ = ast_ptr(new ast_lambda(std::move($2), std::move($5))); }
;
case
: CASE aAdd OF OCURLY branches CCURLY
{ $$ = ast_ptr(new ast_case(std::move($2), std::move($5))); }
;
branches
: branches branch { $$ = std::move($1); $$.push_back(std::move($2)); }
| branch { $$ = std::vector<branch_ptr>(); $$.push_back(std::move($1));}
;
branch
: pattern ARROW OCURLY aAdd CCURLY
{ $$ = branch_ptr(new branch(std::move($1), std::move($4))); }
;
pattern
: LID { $$ = pattern_ptr(new pattern_var(std::move($1))); }
| UID lowercaseParams
{ $$ = pattern_ptr(new pattern_constr(std::move($1), std::move($2))); }
;
data
: DATA UID lowercaseParams EQUAL OCURLY constructors CCURLY
{ $$ = definition_data_ptr(new definition_data(std::move($2), std::move($3), std::move($6))); }
;
constructors
: constructors COMMA constructor { $$ = std::move($1); $$.push_back(std::move($3)); }
| constructor
{ $$ = std::vector<constructor_ptr>(); $$.push_back(std::move($1)); }
;
constructor
: UID typeList
{ $$ = constructor_ptr(new constructor(std::move($1), std::move($2))); }
;
type
: nonArrowType ARROW type { $$ = parsed_type_ptr(new parsed_type_arr(std::move($1), std::move($3))); }
| nonArrowType { $$ = std::move($1); }
;
nonArrowType
: UID typeList { $$ = parsed_type_ptr(new parsed_type_app(std::move($1), std::move($2))); }
| LID { $$ = parsed_type_ptr(new parsed_type_var(std::move($1))); }
| OPAREN type CPAREN { $$ = std::move($2); }
;
typeListElement
: OPAREN type CPAREN { $$ = std::move($2); }
| UID { $$ = parsed_type_ptr(new parsed_type_app(std::move($1), {})); }
| LID { $$ = parsed_type_ptr(new parsed_type_var(std::move($1))); }
;
typeList
: %empty { $$ = std::vector<parsed_type_ptr>(); }
| typeList typeListElement { $$ = std::move($1); $$.push_back(std::move($2)); }
;

269
code/compiler/12/runtime.c Normal file
View File

@@ -0,0 +1,269 @@
#include <stdint.h>
#include <assert.h>
#include <memory.h>
#include <stdio.h>
#include "runtime.h"
struct node_base* alloc_node() {
struct node_base* new_node = malloc(sizeof(struct node_app));
new_node->gc_next = NULL;
new_node->gc_reachable = 0;
assert(new_node != NULL);
return new_node;
}
struct node_app* alloc_app(struct node_base* l, struct node_base* r) {
struct node_app* node = (struct node_app*) alloc_node();
node->base.tag = NODE_APP;
node->left = l;
node->right = r;
return node;
}
struct node_num* alloc_num(int32_t n) {
struct node_num* node = (struct node_num*) alloc_node();
node->base.tag = NODE_NUM;
node->value = n;
return node;
}
struct node_global* alloc_global(void (*f)(struct gmachine*), int32_t a) {
struct node_global* node = (struct node_global*) alloc_node();
node->base.tag = NODE_GLOBAL;
node->arity = a;
node->function = f;
return node;
}
struct node_ind* alloc_ind(struct node_base* n) {
struct node_ind* node = (struct node_ind*) alloc_node();
node->base.tag = NODE_IND;
node->next = n;
return node;
}
void free_node_direct(struct node_base* n) {
if(n->tag == NODE_DATA) {
free(((struct node_data*) n)->array);
}
}
void gc_visit_node(struct node_base* n) {
if(n->gc_reachable) return;
n->gc_reachable = 1;
if(n->tag == NODE_APP) {
struct node_app* app = (struct node_app*) n;
gc_visit_node(app->left);
gc_visit_node(app->right);
} if(n->tag == NODE_IND) {
struct node_ind* ind = (struct node_ind*) n;
gc_visit_node(ind->next);
} if(n->tag == NODE_DATA) {
struct node_data* data = (struct node_data*) n;
struct node_base** to_visit = data->array;
while(*to_visit) {
gc_visit_node(*to_visit);
to_visit++;
}
}
}
void stack_init(struct stack* s) {
s->size = 4;
s->count = 0;
s->data = malloc(sizeof(*s->data) * s->size);
assert(s->data != NULL);
}
void stack_free(struct stack* s) {
free(s->data);
}
void stack_push(struct stack* s, struct node_base* n) {
while(s->count >= s->size) {
s->data = realloc(s->data, sizeof(*s->data) * (s->size *= 2));
assert(s->data != NULL);
}
s->data[s->count++] = n;
}
struct node_base* stack_pop(struct stack* s) {
assert(s->count > 0);
return s->data[--s->count];
}
struct node_base* stack_peek(struct stack* s, size_t o) {
assert(s->count > o);
return s->data[s->count - o - 1];
}
void stack_popn(struct stack* s, size_t n) {
assert(s->count >= n);
s->count -= n;
}
void gmachine_init(struct gmachine* g) {
stack_init(&g->stack);
g->gc_nodes = NULL;
g->gc_node_count = 0;
g->gc_node_threshold = 128;
}
void gmachine_free(struct gmachine* g) {
stack_free(&g->stack);
struct node_base* to_free = g->gc_nodes;
struct node_base* next;
while(to_free) {
next = to_free->gc_next;
free_node_direct(to_free);
free(to_free);
to_free = next;
}
}
void gmachine_slide(struct gmachine* g, size_t n) {
assert(g->stack.count > n);
g->stack.data[g->stack.count - n - 1] = g->stack.data[g->stack.count - 1];
g->stack.count -= n;
}
void gmachine_update(struct gmachine* g, size_t o) {
assert(g->stack.count > o + 1);
struct node_ind* ind =
(struct node_ind*) g->stack.data[g->stack.count - o - 2];
ind->base.tag = NODE_IND;
ind->next = g->stack.data[g->stack.count -= 1];
}
void gmachine_alloc(struct gmachine* g, size_t o) {
while(o--) {
stack_push(&g->stack,
gmachine_track(g, (struct node_base*) alloc_ind(NULL)));
}
}
void gmachine_pack(struct gmachine* g, size_t n, int8_t t) {
assert(g->stack.count >= n);
struct node_base** data = malloc(sizeof(*data) * (n + 1));
assert(data != NULL);
memcpy(data, &g->stack.data[g->stack.count - n], n * sizeof(*data));
data[n] = NULL;
struct node_data* new_node = (struct node_data*) alloc_node();
new_node->array = data;
new_node->base.tag = NODE_DATA;
new_node->tag = t;
stack_popn(&g->stack, n);
stack_push(&g->stack, gmachine_track(g, (struct node_base*) new_node));
}
void gmachine_split(struct gmachine* g, size_t n) {
struct node_data* node = (struct node_data*) stack_pop(&g->stack);
for(size_t i = 0; i < n; i++) {
stack_push(&g->stack, node->array[i]);
}
}
struct node_base* gmachine_track(struct gmachine* g, struct node_base* b) {
g->gc_node_count++;
b->gc_next = g->gc_nodes;
g->gc_nodes = b;
if(g->gc_node_count >= g->gc_node_threshold) {
uint64_t nodes_before = g->gc_node_count;
gc_visit_node(b);
gmachine_gc(g);
g->gc_node_threshold = g->gc_node_count * 2;
}
return b;
}
void gmachine_gc(struct gmachine* g) {
for(size_t i = 0; i < g->stack.count; i++) {
gc_visit_node(g->stack.data[i]);
}
struct node_base** head_ptr = &g->gc_nodes;
while(*head_ptr) {
if((*head_ptr)->gc_reachable) {
(*head_ptr)->gc_reachable = 0;
head_ptr = &(*head_ptr)->gc_next;
} else {
struct node_base* to_free = *head_ptr;
*head_ptr = to_free->gc_next;
free_node_direct(to_free);
free(to_free);
g->gc_node_count--;
}
}
}
void unwind(struct gmachine* g) {
struct stack* s = &g->stack;
while(1) {
struct node_base* peek = stack_peek(s, 0);
if(peek->tag == NODE_APP) {
struct node_app* n = (struct node_app*) peek;
stack_push(s, n->left);
} else if(peek->tag == NODE_GLOBAL) {
struct node_global* n = (struct node_global*) peek;
assert(s->count > n->arity);
for(size_t i = 1; i <= n->arity; i++) {
s->data[s->count - i]
= ((struct node_app*) s->data[s->count - i - 1])->right;
}
n->function(g);
} else if(peek->tag == NODE_IND) {
struct node_ind* n = (struct node_ind*) peek;
stack_pop(s);
stack_push(s, n->next);
} else {
break;
}
}
}
extern void f_main(struct gmachine* s);
void print_node(struct node_base* n) {
if(n->tag == NODE_APP) {
struct node_app* app = (struct node_app*) n;
print_node(app->left);
putchar(' ');
print_node(app->right);
} else if(n->tag == NODE_DATA) {
printf("(Packed)");
} else if(n->tag == NODE_GLOBAL) {
struct node_global* global = (struct node_global*) n;
printf("(Global: %p)", global->function);
} else if(n->tag == NODE_IND) {
print_node(((struct node_ind*) n)->next);
} else if(n->tag == NODE_NUM) {
struct node_num* num = (struct node_num*) n;
printf("%d", num->value);
}
}
int main(int argc, char** argv) {
struct gmachine gmachine;
struct node_global* first_node = alloc_global(f_main, 0);
struct node_base* result;
gmachine_init(&gmachine);
gmachine_track(&gmachine, (struct node_base*) first_node);
stack_push(&gmachine.stack, (struct node_base*) first_node);
unwind(&gmachine);
result = stack_pop(&gmachine.stack);
printf("Result: ");
print_node(result);
putchar('\n');
gmachine_free(&gmachine);
}

View File

@@ -0,0 +1,84 @@
#pragma once
#include <stdlib.h>
struct gmachine;
enum node_tag {
NODE_APP,
NODE_NUM,
NODE_GLOBAL,
NODE_IND,
NODE_DATA
};
struct node_base {
enum node_tag tag;
int8_t gc_reachable;
struct node_base* gc_next;
};
struct node_app {
struct node_base base;
struct node_base* left;
struct node_base* right;
};
struct node_num {
struct node_base base;
int32_t value;
};
struct node_global {
struct node_base base;
int32_t arity;
void (*function)(struct gmachine*);
};
struct node_ind {
struct node_base base;
struct node_base* next;
};
struct node_data {
struct node_base base;
int8_t tag;
struct node_base** array;
};
struct node_base* alloc_node();
struct node_app* alloc_app(struct node_base* l, struct node_base* r);
struct node_num* alloc_num(int32_t n);
struct node_global* alloc_global(void (*f)(struct gmachine*), int32_t a);
struct node_ind* alloc_ind(struct node_base* n);
void free_node_direct(struct node_base*);
void gc_visit_node(struct node_base*);
struct stack {
size_t size;
size_t count;
struct node_base** data;
};
void stack_init(struct stack* s);
void stack_free(struct stack* s);
void stack_push(struct stack* s, struct node_base* n);
struct node_base* stack_pop(struct stack* s);
struct node_base* stack_peek(struct stack* s, size_t o);
void stack_popn(struct stack* s, size_t n);
struct gmachine {
struct stack stack;
struct node_base* gc_nodes;
int64_t gc_node_count;
int64_t gc_node_threshold;
};
void gmachine_init(struct gmachine* g);
void gmachine_free(struct gmachine* g);
void gmachine_slide(struct gmachine* g, size_t n);
void gmachine_update(struct gmachine* g, size_t o);
void gmachine_alloc(struct gmachine* g, size_t o);
void gmachine_pack(struct gmachine* g, size_t n, int8_t t);
void gmachine_split(struct gmachine* g, size_t n);
struct node_base* gmachine_track(struct gmachine* g, struct node_base* b);
void gmachine_gc(struct gmachine* g);

View File

@@ -0,0 +1,38 @@
%option noyywrap
%{
#include <iostream>
#include "ast.hpp"
#include "definition.hpp"
#include "parser.hpp"
#define YY_DECL yy::parser::symbol_type yylex()
%}
%%
[ \n]+ {}
\\ { return yy::parser::make_BACKSLASH(); }
\+ { return yy::parser::make_PLUS(); }
\* { return yy::parser::make_TIMES(); }
- { return yy::parser::make_MINUS(); }
\/ { return yy::parser::make_DIVIDE(); }
[0-9]+ { return yy::parser::make_INT(atoi(yytext)); }
defn { return yy::parser::make_DEFN(); }
data { return yy::parser::make_DATA(); }
case { return yy::parser::make_CASE(); }
of { return yy::parser::make_OF(); }
let { return yy::parser::make_LET(); }
in { return yy::parser::make_IN(); }
\{ { return yy::parser::make_OCURLY(); }
\} { return yy::parser::make_CCURLY(); }
\( { return yy::parser::make_OPAREN(); }
\) { return yy::parser::make_CPAREN(); }
, { return yy::parser::make_COMMA(); }
-> { return yy::parser::make_ARROW(); }
= { return yy::parser::make_EQUAL(); }
[a-z][a-zA-Z]* { return yy::parser::make_LID(std::string(yytext)); }
[A-Z][a-zA-Z]* { return yy::parser::make_UID(std::string(yytext)); }
%%

23
code/compiler/12/test.cpp Normal file
View File

@@ -0,0 +1,23 @@
#include "graph.hpp"
int main() {
function_graph graph;
graph.add_edge("f", "g");
graph.add_edge("g", "h");
graph.add_edge("h", "f");
graph.add_edge("i", "j");
graph.add_edge("j", "i");
graph.add_edge("j", "f");
graph.add_edge("x", "f");
graph.add_edge("x", "i");
for(auto& group : graph.compute_order()) {
std::cout << "Group: " << std::endl;
for(auto& member : group->members) {
std::cout << member << std::endl;
}
}
}

219
code/compiler/12/type.cpp Normal file
View File

@@ -0,0 +1,219 @@
#include "type.hpp"
#include <ostream>
#include <sstream>
#include <algorithm>
#include <vector>
#include "error.hpp"
bool type::is_arrow(const type_mgr& mgr) const { return false; }
void type_scheme::print(const type_mgr& mgr, std::ostream& to) const {
if(forall.size() != 0) {
to << "forall ";
for(auto& var : forall) {
to << var << " ";
}
to << ". ";
}
monotype->print(mgr, to);
}
type_ptr type_scheme::instantiate(type_mgr& mgr) const {
if(forall.size() == 0) return monotype;
std::map<std::string, type_ptr> subst;
for(auto& var : forall) {
subst[var] = mgr.new_type();
}
return mgr.substitute(subst, monotype);
}
void type_var::print(const type_mgr& mgr, std::ostream& to) const {
auto it = mgr.types.find(name);
if(it != mgr.types.end()) {
it->second->print(mgr, to);
} else {
to << name;
}
}
bool type_var::is_arrow(const type_mgr& mgr) const {
auto it = mgr.types.find(name);
if(it != mgr.types.end()) {
return it->second->is_arrow(mgr);
} else {
return false;
}
}
void type_base::print(const type_mgr& mgr, std::ostream& to) const {
to << name;
}
void type_arr::print(const type_mgr& mgr, std::ostream& to) const {
bool print_parenths = left->is_arrow(mgr);
if(print_parenths) to << "(";
left->print(mgr, to);
if(print_parenths) to << ")";
to << " -> ";
right->print(mgr, to);
}
bool type_arr::is_arrow(const type_mgr& mgr) const {
return true;
}
void type_app::print(const type_mgr& mgr, std::ostream& to) const {
constructor->print(mgr, to);
to << "*";
for(auto& arg : arguments) {
to << " ";
arg->print(mgr, to);
}
}
std::string type_mgr::new_type_name() {
int temp = last_id++;
std::string str = "";
while(temp != -1) {
str += (char) ('a' + (temp % 26));
temp = temp / 26 - 1;
}
std::reverse(str.begin(), str.end());
return str;
}
type_ptr type_mgr::new_type() {
return type_ptr(new type_var(new_type_name()));
}
type_ptr type_mgr::new_arrow_type() {
return type_ptr(new type_arr(new_type(), new_type()));
}
type_ptr type_mgr::resolve(type_ptr t, type_var*& var) const {
type_var* cast;
var = nullptr;
while((cast = dynamic_cast<type_var*>(t.get()))) {
auto it = types.find(cast->name);
if(it == types.end()) {
var = cast;
break;
}
t = it->second;
}
return t;
}
void type_mgr::unify(type_ptr l, type_ptr r) {
type_var *lvar, *rvar;
type_arr *larr, *rarr;
type_base *lid, *rid;
type_app *lapp, *rapp;
l = resolve(l, lvar);
r = resolve(r, rvar);
if(lvar) {
bind(lvar->name, r);
return;
} else if(rvar) {
bind(rvar->name, l);
return;
} else if((larr = dynamic_cast<type_arr*>(l.get())) &&
(rarr = dynamic_cast<type_arr*>(r.get()))) {
unify(larr->left, rarr->left);
unify(larr->right, rarr->right);
return;
} else if((lid = dynamic_cast<type_base*>(l.get())) &&
(rid = dynamic_cast<type_base*>(r.get()))) {
if(lid->name == rid->name && lid->arity == rid->arity) return;
} else if((lapp = dynamic_cast<type_app*>(l.get())) &&
(rapp = dynamic_cast<type_app*>(r.get()))) {
unify(lapp->constructor, rapp->constructor);
auto left_it = lapp->arguments.begin();
auto right_it = rapp->arguments.begin();
while(left_it != lapp->arguments.end() &&
right_it != rapp->arguments.end()) {
unify(*left_it, *right_it);
left_it++, right_it++;
}
return;
}
throw unification_error(l, r);
}
type_ptr type_mgr::substitute(const std::map<std::string, type_ptr>& subst, const type_ptr& t) const {
type_ptr temp = t;
while(type_var* var = dynamic_cast<type_var*>(temp.get())) {
auto subst_it = subst.find(var->name);
if(subst_it != subst.end()) return subst_it->second;
auto var_it = types.find(var->name);
if(var_it == types.end()) return t;
temp = var_it->second;
}
if(type_arr* arr = dynamic_cast<type_arr*>(temp.get())) {
auto left_result = substitute(subst, arr->left);
auto right_result = substitute(subst, arr->right);
if(left_result == arr->left && right_result == arr->right) return t;
return type_ptr(new type_arr(left_result, right_result));
} else if(type_app* app = dynamic_cast<type_app*>(temp.get())) {
auto constructor_result = substitute(subst, app->constructor);
bool arg_changed = false;
std::vector<type_ptr> new_args;
for(auto& arg : app->arguments) {
auto arg_result = substitute(subst, arg);
arg_changed |= arg_result != arg;
new_args.push_back(std::move(arg_result));
}
if(constructor_result == app->constructor && !arg_changed) return t;
type_app* new_app = new type_app(std::move(constructor_result));
std::swap(new_app->arguments, new_args);
return type_ptr(new_app);
}
return t;
}
void type_mgr::bind(const std::string& s, type_ptr t) {
type_var* other = dynamic_cast<type_var*>(t.get());
if(other && other->name == s) return;
types[s] = t;
}
void type_mgr::find_free(const type_ptr& t, std::set<std::string>& into) const {
type_var* var;
type_ptr resolved = resolve(t, var);
if(var) {
into.insert(var->name);
} else if(type_arr* arr = dynamic_cast<type_arr*>(resolved.get())) {
find_free(arr->left, into);
find_free(arr->right, into);
} else if(type_app* app = dynamic_cast<type_app*>(resolved.get())) {
find_free(app->constructor, into);
for(auto& arg : app->arguments) find_free(arg, into);
}
}
void type_mgr::find_free(const type_scheme_ptr& t, std::set<std::string>& into) const {
std::set<std::string> monotype_free;
type_mgr limited_mgr;
for(auto& binding : types) {
auto existing_position = std::find(t->forall.begin(), t->forall.end(), binding.first);
if(existing_position != t->forall.end()) continue;
limited_mgr.types[binding.first] = binding.second;
}
limited_mgr.find_free(t->monotype, monotype_free);
for(auto& not_free : t->forall) {
monotype_free.erase(not_free);
}
into.insert(monotype_free.begin(), monotype_free.end());
}

99
code/compiler/12/type.hpp Normal file
View File

@@ -0,0 +1,99 @@
#pragma once
#include <memory>
#include <map>
#include <string>
#include <vector>
#include <set>
struct type_mgr;
struct type {
virtual ~type() = default;
virtual void print(const type_mgr& mgr, std::ostream& to) const = 0;
virtual bool is_arrow(const type_mgr& mgr) const;
};
using type_ptr = std::shared_ptr<type>;
struct type_scheme {
std::vector<std::string> forall;
type_ptr monotype;
type_scheme(type_ptr type) : forall(), monotype(std::move(type)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
type_ptr instantiate(type_mgr& mgr) const;
};
using type_scheme_ptr = std::shared_ptr<type_scheme>;
struct type_var : public type {
std::string name;
type_var(std::string n)
: name(std::move(n)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
bool is_arrow(const type_mgr& mgr) const;
};
struct type_base : public type {
std::string name;
int32_t arity;
type_base(std::string n, int32_t a = 0)
: name(std::move(n)), arity(a) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_data : public type_base {
struct constructor {
int tag;
};
std::map<std::string, constructor> constructors;
type_data(std::string n, int32_t a = 0)
: type_base(std::move(n), a) {}
};
struct type_arr : public type {
type_ptr left;
type_ptr right;
type_arr(type_ptr l, type_ptr r)
: left(std::move(l)), right(std::move(r)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
bool is_arrow(const type_mgr& mgr) const;
};
struct type_app : public type {
type_ptr constructor;
std::vector<type_ptr> arguments;
type_app(type_ptr c)
: constructor(std::move(c)) {}
void print(const type_mgr& mgr, std::ostream& to) const;
};
struct type_mgr {
int last_id = 0;
std::map<std::string, type_ptr> types;
std::string new_type_name();
type_ptr new_type();
type_ptr new_arrow_type();
void unify(type_ptr l, type_ptr r);
type_ptr substitute(
const std::map<std::string, type_ptr>& subst,
const type_ptr& t) const;
type_ptr resolve(type_ptr t, type_var*& var) const;
void bind(const std::string& s, type_ptr t);
void find_free(const type_ptr& t, std::set<std::string>& into) const;
void find_free(const type_scheme_ptr& t, std::set<std::string>& into) const;
};

View File

@@ -0,0 +1,85 @@
#include "type_env.hpp"
#include "type.hpp"
void type_env::find_free(const type_mgr& mgr, std::set<std::string>& into) const {
if(parent != nullptr) parent->find_free(mgr, into);
for(auto& binding : names) {
mgr.find_free(binding.second.type, into);
}
}
void type_env::find_free_except(const type_mgr& mgr, const group& avoid,
std::set<std::string>& into) const {
if(parent != nullptr) parent->find_free(mgr, into);
for(auto& binding : names) {
if(avoid.members.find(binding.first) != avoid.members.end()) continue;
mgr.find_free(binding.second.type, into);
}
}
type_scheme_ptr type_env::lookup(const std::string& name) const {
auto it = names.find(name);
if(it != names.end()) return it->second.type;
if(parent) return parent->lookup(name);
return nullptr;
}
bool type_env::is_global(const std::string& name) const {
auto it = names.find(name);
if(it != names.end()) return it->second.vis == visibility::global;
if(parent) return parent->is_global(name);
return false;
}
void type_env::set_mangled_name(const std::string& name, const std::string& mangled) {
auto it = names.find(name);
if(it != names.end()) it->second.mangled_name = mangled;
}
const std::string& type_env::get_mangled_name(const std::string& name) const {
auto it = names.find(name);
if(it != names.end())
return (it->second.mangled_name != "") ? it->second.mangled_name : name;
if(parent) return parent->get_mangled_name(name);
return name;
}
type_ptr type_env::lookup_type(const std::string& name) const {
auto it = type_names.find(name);
if(it != type_names.end()) return it->second;
if(parent) return parent->lookup_type(name);
return nullptr;
}
void type_env::bind(const std::string& name, type_ptr t, visibility v) {
type_scheme_ptr new_scheme(new type_scheme(std::move(t)));
names[name] = variable_data(std::move(new_scheme), v, "");
}
void type_env::bind(const std::string& name, type_scheme_ptr t, visibility v) {
names[name] = variable_data(std::move(t), v, "");
}
void type_env::bind_type(const std::string& type_name, type_ptr t) {
if(lookup_type(type_name) != nullptr) throw 0;
type_names[type_name] = t;
}
void type_env::generalize(const std::string& name, const group& grp, type_mgr& mgr) {
auto names_it = names.find(name);
if(names_it == names.end()) throw 0;
if(names_it->second.type->forall.size() > 0) throw 0;
std::set<std::string> free_in_type;
std::set<std::string> free_in_env;
mgr.find_free(names_it->second.type->monotype, free_in_type);
find_free_except(mgr, grp, free_in_env);
for(auto& free : free_in_type) {
if(free_in_env.find(free) != free_in_env.end()) continue;
names_it->second.type->forall.push_back(free);
}
}
type_env_ptr type_scope(type_env_ptr parent) {
return type_env_ptr(new type_env(std::move(parent)));
}

View File

@@ -0,0 +1,49 @@
#pragma once
#include <map>
#include <string>
#include <set>
#include "graph.hpp"
#include "type.hpp"
struct type_env;
using type_env_ptr = std::shared_ptr<type_env>;
enum class visibility { global,local };
struct type_env {
struct variable_data {
type_scheme_ptr type;
visibility vis;
std::string mangled_name;
variable_data()
: variable_data(nullptr, visibility::local, "") {}
variable_data(type_scheme_ptr t, visibility v, std::string n)
: type(std::move(t)), vis(v), mangled_name(std::move(n)) {}
};
type_env_ptr parent;
std::map<std::string, variable_data> names;
std::map<std::string, type_ptr> type_names;
type_env(type_env_ptr p) : parent(std::move(p)) {}
type_env() : type_env(nullptr) {}
void find_free(const type_mgr& mgr, std::set<std::string>& into) const;
void find_free_except(const type_mgr& mgr, const group& avoid,
std::set<std::string>& into) const;
type_scheme_ptr lookup(const std::string& name) const;
bool is_global(const std::string& name) const;
void set_mangled_name(const std::string& name, const std::string& mangled);
const std::string& get_mangled_name(const std::string& name) const;
type_ptr lookup_type(const std::string& name) const;
void bind(const std::string& name, type_ptr t,
visibility v = visibility::local);
void bind(const std::string& name, type_scheme_ptr t,
visibility v = visibility::local);
void bind_type(const std::string& type_name, type_ptr t);
void generalize(const std::string& name, const group& grp, type_mgr& mgr);
};
type_env_ptr type_scope(type_env_ptr parent);

View File

@@ -1,4 +1,4 @@
cd 10 cd 11
mkdir -p build && cd build mkdir -p build && cd build
cmake .. cmake ..
make -j8 make -j8

View File

@@ -4,3 +4,10 @@ title = "Daniel's Blog"
theme = "vanilla" theme = "vanilla"
pygmentsCodeFences = true pygmentsCodeFences = true
pygmentsStyle = "github" pygmentsStyle = "github"
summaryLength = 20
[markup]
[markup.tableOfContents]
endLevel = 4
ordered = false
startLevel = 3

View File

@@ -1,5 +1,6 @@
--- ---
title: Daniel's Blog title: Daniel's Blog
description: Daniel Fedorin's personal blog, covering topics such as functional programming, compiler development, and more!
--- ---
## Hello! ## Hello!
Welcome to my blog. Here, I write about various subjects, including (but not limited to) Welcome to my blog. Here, I write about various subjects, including (but not limited to)

View File

@@ -2,6 +2,7 @@
title: Compiling a Functional Language Using C++, Part 0 - Intro title: Compiling a Functional Language Using C++, Part 0 - Intro
date: 2019-08-03T01:02:30-07:00 date: 2019-08-03T01:02:30-07:00
tags: ["C and C++", "Functional Languages", "Compilers"] tags: ["C and C++", "Functional Languages", "Compilers"]
description: "In this first post of a larger series, we embark on a journey of developing a compiler for a lazily evaluated functional language."
--- ---
During my last academic term, I was enrolled in a compilers course. During my last academic term, I was enrolled in a compilers course.
We had a final project - develop a compiler for a basic Python subset, We had a final project - develop a compiler for a basic Python subset,
@@ -142,3 +143,6 @@ Here are the posts that I've written so far for this series:
* [LLVM]({{< relref "08_compiler_llvm.md" >}}) * [LLVM]({{< relref "08_compiler_llvm.md" >}})
* [Garbage Collection]({{< relref "09_compiler_garbage_collection.md" >}}) * [Garbage Collection]({{< relref "09_compiler_garbage_collection.md" >}})
* [Polymorphism]({{< relref "10_compiler_polymorphism.md" >}}) * [Polymorphism]({{< relref "10_compiler_polymorphism.md" >}})
* [Polymorphic Data Types]({{< relref "11_compiler_polymorphic_data_types.md" >}})
* [Let/In and Lambdas]({{< relref "12_compiler_let_in_lambda/index.md" >}})

View File

@@ -2,6 +2,7 @@
title: Compiling a Functional Language Using C++, Part 1 - Tokenizing title: Compiling a Functional Language Using C++, Part 1 - Tokenizing
date: 2019-08-03T01:02:30-07:00 date: 2019-08-03T01:02:30-07:00
tags: ["C and C++", "Functional Languages", "Compilers"] tags: ["C and C++", "Functional Languages", "Compilers"]
description: "In this post, we tackle the first component of our compiler: tokenizing."
--- ---
It makes sense to build a compiler bit by bit, following the stages we outlined in It makes sense to build a compiler bit by bit, following the stages we outlined in
the first post of the series. This is because these stages are essentially a pipeline, the first post of the series. This is because these stages are essentially a pipeline,

View File

@@ -2,6 +2,7 @@
title: Compiling a Functional Language Using C++, Part 2 - Parsing title: Compiling a Functional Language Using C++, Part 2 - Parsing
date: 2019-08-03T01:02:30-07:00 date: 2019-08-03T01:02:30-07:00
tags: ["C and C++", "Functional Languages", "Compilers"] tags: ["C and C++", "Functional Languages", "Compilers"]
description: "In this post, we combine our compiler's tokenizer with a parser, allowing us to extract structure from input source code."
--- ---
In the previous post, we covered tokenizing. We learned how to convert an input string into logical segments, and even wrote up a tokenizer to do it according to the rules of our language. Now, it's time to make sense of the tokens, and parse our language. In the previous post, we covered tokenizing. We learned how to convert an input string into logical segments, and even wrote up a tokenizer to do it according to the rules of our language. Now, it's time to make sense of the tokens, and parse our language.

View File

@@ -2,6 +2,7 @@
title: Compiling a Functional Language Using C++, Part 3 - Type Checking title: Compiling a Functional Language Using C++, Part 3 - Type Checking
date: 2019-08-06T14:26:38-07:00 date: 2019-08-06T14:26:38-07:00
tags: ["C and C++", "Functional Languages", "Compilers"] tags: ["C and C++", "Functional Languages", "Compilers"]
description: "In this post, we allow our compiler to throw away invalid programs, detected using a monomorphic typechecking algorithm."
--- ---
I think tokenizing and parsing are boring. The thing is, looking at syntax I think tokenizing and parsing are boring. The thing is, looking at syntax
is a pretty shallow measure of how interesting a language is. It's like is a pretty shallow measure of how interesting a language is. It's like

Some files were not shown because too many files have changed in this diff Show More