11629611787

Committed 01 Nov 2024 12:48PM UTC coverage: 77.319% (+0.3%) from 77.042%

Build # 11629611787

Build Type

push

github

Committed by

SuperFola

Commit Message

feat(tests): adding first test for IR generation and optimization

Run Details

5209 of 6737 relevant lines covered (77.32%)

9473.78 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

92.5

/src/arkreactor/Compiler/IntermediateRepresentation/IRCompiler.cpp

#include <Ark/Compiler/IntermediateRepresentation/IRCompiler.hpp>

#include <chrono>
#include <utility>
#include <unordered_map>
#include <picosha2.h>
#include <fmt/ostream.h>

#include <Ark/Constants.hpp>
#include <Ark/Literals.hpp>

namespace Ark::internal
{
    using namespace literals;

    IRCompiler::IRCompiler(const unsigned debug) :
        m_logger("IRCompiler", debug)
    {}

    void IRCompiler::process(const std::vector<IR::Block>& pages, const std::vector<std::string>& symbols, const std::vector<ValTableElem>& values)
    {
        m_logger.traceStart("process");
        pushFileHeader();
        pushSymAndValTables(symbols, values);

        m_ir = pages;
        compile();

        if (m_ir.empty())
        {
            // code segment with a single instruction
            m_bytecode.push_back(CODE_SEGMENT_START);
            m_bytecode.push_back(0_u8);
            m_bytecode.push_back(1_u8);

            m_bytecode.push_back(0_u8);
            m_bytecode.push_back(HALT);
            m_bytecode.push_back(0_u8);
            m_bytecode.push_back(0_u8);
        }

        constexpr std::size_t header_size = 18;

        // generate a hash of the tables + bytecode
        std::vector<unsigned char> hash_out(picosha2::k_digest_size);
        picosha2::hash256(m_bytecode.begin() + header_size, m_bytecode.end(), hash_out);
        m_bytecode.insert(m_bytecode.begin() + header_size, hash_out.begin(), hash_out.end());

        m_logger.traceEnd();
    }

    void IRCompiler::dumpToStream(std::ostream& stream) const
    {
        std::size_t index = 0;
        for (const auto& block : m_ir)
        {
            fmt::println(stream, "page_{}", index);
            for (const auto entity : block)
            {
                switch (entity.kind())
                {
                    case IR::Kind::Label:
                        fmt::println(stream, ".L{}:", entity.label());
                        break;

                    case IR::Kind::Goto:
                        fmt::println(stream, "\tGOTO L{}", entity.label());
                        break;

                    case IR::Kind::GotoIfTrue:
                        fmt::println(stream, "\tGOTO_IF_TRUE L{}", entity.label());
                        break;

                    case IR::Kind::GotoIfFalse:
                        fmt::println(stream, "\tGOTO_IF_FALSE L{}", entity.label());
                        break;

                    case IR::Kind::Opcode:
                        fmt::println(stream, "\t{} {}", InstructionNames[entity.inst()], entity.primaryArg());
                        break;

                    case IR::Kind::Opcode2Args:
                        fmt::println(stream, "\t{} {}, {}", InstructionNames[entity.inst()], entity.primaryArg(), entity.secondaryArg());
                        break;
                }
            }

            fmt::println(stream, "");
            ++index;
        }
    }

    const bytecode_t& IRCompiler::bytecode() const noexcept
    {
        return m_bytecode;
    }

    void IRCompiler::compile()
    {
        // push the different code segments
        for (std::size_t i = 0, end = m_ir.size(); i < end; ++i)
        {
            IR::Block& page = m_ir[i];
            // just in case we got too far, always add a HALT to be sure the
            // VM won't do anything crazy
            page.emplace_back(HALT);

            // push number of elements
            const auto page_size = std::ranges::count_if(page, [](const auto& a) {
                return a.kind() != IR::Kind::Label;
            });
            if (std::cmp_greater(page_size, std::numeric_limits<uint16_t>::max()))
                throw std::overflow_error(fmt::format("Size of page {} exceeds the maximum size of 2^16 - 1", i));

            m_bytecode.push_back(CODE_SEGMENT_START);
            m_bytecode.push_back(static_cast<uint8_t>((page_size & 0xff00) >> 8));
            m_bytecode.push_back(static_cast<uint8_t>(page_size & 0x00ff));

            // register labels position
            uint16_t pos = 0;
            std::unordered_map<IR::label_t, uint16_t> label_to_position;
            for (auto inst : page)
            {
                switch (inst.kind())
                {
                    case IR::Kind::Label:
                        label_to_position[inst.label()] = pos;
                        break;

                    default:
                        ++pos;
                }
            }

            for (auto inst : page)
            {
                switch (inst.kind())
                {
                    case IR::Kind::Goto:
                        pushWord(Word(JUMP, label_to_position[inst.label()]));
                        break;

                    case IR::Kind::GotoIfTrue:
                        pushWord(Word(POP_JUMP_IF_TRUE, label_to_position[inst.label()]));
                        break;

                    case IR::Kind::GotoIfFalse:
                        pushWord(Word(POP_JUMP_IF_FALSE, label_to_position[inst.label()]));
                        break;

                    case IR::Kind::Opcode:
                        [[fallthrough]];
                    case IR::Kind::Opcode2Args:
                        pushWord(inst.bytecode());
                        break;

                    default:
                        break;
                }
            }
        }
    }

    void IRCompiler::pushWord(const Word& word)
    {
        m_bytecode.push_back(word.opcode);
        m_bytecode.push_back(word.byte_1);
        m_bytecode.push_back(word.byte_2);
        m_bytecode.push_back(word.byte_3);
    }

    void IRCompiler::pushFileHeader() noexcept
    {
        /*
            Generating headers:
                - lang name (to be sure we are executing an ArkScript file)
                    on 4 bytes (ark + padding)
                - version (major: 2 bytes, minor: 2 bytes, patch: 2 bytes)
                - timestamp (8 bytes, unix format)
        */

        m_bytecode.push_back('a');
        m_bytecode.push_back('r');
        m_bytecode.push_back('k');
        m_bytecode.push_back(0_u8);

        // push version
        for (const int n : std::array { ARK_VERSION_MAJOR, ARK_VERSION_MINOR, ARK_VERSION_PATCH })
        {
            m_bytecode.push_back(static_cast<uint8_t>((n & 0xff00) >> 8));
            m_bytecode.push_back(static_cast<uint8_t>(n & 0x00ff));
        }

        // push timestamp
        const long long timestamp = std::chrono::duration_cast<std::chrono::seconds>(
                                        std::chrono::system_clock::now().time_since_epoch())
                                        .count();
        for (long i = 0; i < 8; ++i)
        {
            const long shift = 8 * (7 - i);
            const auto ts_byte = static_cast<uint8_t>((timestamp & (0xffLL << shift)) >> shift);
            m_bytecode.push_back(ts_byte);
        }
    }

    void IRCompiler::pushSymAndValTables(const std::vector<std::string>& symbols, const std::vector<ValTableElem>& values)
    {
        const std::size_t symbol_size = symbols.size();
        if (symbol_size > std::numeric_limits<uint16_t>::max())
            throw std::overflow_error(fmt::format("Too many symbols: {}, exceeds the maximum size of 2^16 - 1", symbol_size));

        m_bytecode.push_back(SYM_TABLE_START);
        m_bytecode.push_back(static_cast<uint8_t>((symbol_size & 0xff00) >> 8));
        m_bytecode.push_back(static_cast<uint8_t>(symbol_size & 0x00ff));

        for (const auto& sym : symbols)
        {
            // push the string, null terminated
            std::ranges::transform(sym, std::back_inserter(m_bytecode), [](const char i) {
                return static_cast<uint8_t>(i);
            });
            m_bytecode.push_back(0_u8);
        }

        const std::size_t value_size = values.size();
        if (value_size > std::numeric_limits<uint16_t>::max())
            throw std::overflow_error(fmt::format("Too many values: {}, exceeds the maximum size of 2^16 - 1", value_size));

        m_bytecode.push_back(VAL_TABLE_START);
        m_bytecode.push_back(static_cast<uint8_t>((value_size & 0xff00) >> 8));
        m_bytecode.push_back(static_cast<uint8_t>(value_size & 0x00ff));

        for (const ValTableElem& val : values)
        {
            switch (val.type)
            {
                case ValTableElemType::Number:
                {
                    m_bytecode.push_back(NUMBER_TYPE);
                    const auto n = std::get<double>(val.value);
                    std::string t = std::to_string(n);
                    std::ranges::transform(t, std::back_inserter(m_bytecode), [](const char i) {
                        return static_cast<uint8_t>(i);
                    });
                    break;
                }

                case ValTableElemType::String:
                {
                    m_bytecode.push_back(STRING_TYPE);
                    auto t = std::get<std::string>(val.value);
                    std::ranges::transform(t, std::back_inserter(m_bytecode), [](const char i) {
                        return static_cast<uint8_t>(i);
                    });
                    break;
                }

                case ValTableElemType::PageAddr:
                {
                    m_bytecode.push_back(FUNC_TYPE);
                    const std::size_t addr = std::get<std::size_t>(val.value);
                    m_bytecode.push_back(static_cast<uint8_t>((addr & 0xff00) >> 8));
                    m_bytecode.push_back(static_cast<uint8_t>(addr & 0x00ff));
                    break;
                }
            }

            m_bytecode.push_back(0_u8);
        }
    }
}

1	#include <Ark/Compiler/IntermediateRepresentation/IRCompiler.hpp>
2
3	#include <chrono>
4	#include <utility>
5	#include <unordered_map>
6	#include <picosha2.h>
7	#include <fmt/ostream.h>
8
9	#include <Ark/Constants.hpp>
10	#include <Ark/Literals.hpp>
11
12	namespace Ark::internal
13	{
14	using namespace literals;
15
16	IRCompiler::IRCompiler(const unsigned debug) :	200✔
17	m_logger("IRCompiler", debug)	100✔
18	{}	200✔
19
20	void IRCompiler::process(const std::vector<IR::Block>& pages, const std::vector<std::string>& symbols, const std::vector<ValTableElem>& values)	27✔
21	{	27✔
22	m_logger.traceStart("process");	27✔
23	pushFileHeader();	27✔
24	pushSymAndValTables(symbols, values);	27✔
25
26	m_ir = pages;	27✔
27	compile();	27✔
28
29	if (m_ir.empty())	27✔
30	{
31	// code segment with a single instruction
32	m_bytecode.push_back(CODE_SEGMENT_START);	×
33	m_bytecode.push_back(0_u8);	×
34	m_bytecode.push_back(1_u8);	×
35
36	m_bytecode.push_back(0_u8);	×
37	m_bytecode.push_back(HALT);	×
38	m_bytecode.push_back(0_u8);	×
39	m_bytecode.push_back(0_u8);	×
40	}	×
41
42	constexpr std::size_t header_size = 18;	27✔
43
44	// generate a hash of the tables + bytecode
45	std::vector<unsigned char> hash_out(picosha2::k_digest_size);	27✔
46	picosha2::hash256(m_bytecode.begin() + header_size, m_bytecode.end(), hash_out);	27✔
47	m_bytecode.insert(m_bytecode.begin() + header_size, hash_out.begin(), hash_out.end());	27✔
48
49	m_logger.traceEnd();	27✔
50	}	27✔
51
52	void IRCompiler::dumpToStream(std::ostream& stream) const	1✔
53	{	1✔
54	std::size_t index = 0;	1✔
55	for (const auto& block : m_ir)	3✔
56	{
57	fmt::println(stream, "page_{}", index);	2✔
58	for (const auto entity : block)	37✔
59	{
60	switch (entity.kind())	35✔
61	{	4✔
62	case IR::Kind::Label:
63	fmt::println(stream, ".L{}:", entity.label());	4✔
64	break;	6✔
65
66	case IR::Kind::Goto:
67	fmt::println(stream, "\tGOTO L{}", entity.label());	2✔
68	break;	4✔
69
70	case IR::Kind::GotoIfTrue:
71	fmt::println(stream, "\tGOTO_IF_TRUE L{}", entity.label());	2✔
72	break;	2✔
73
74	case IR::Kind::GotoIfFalse:
75	fmt::println(stream, "\tGOTO_IF_FALSE L{}", entity.label());	×
76	break;	24✔
77
78	case IR::Kind::Opcode:
79	fmt::println(stream, "\t{} {}", InstructionNames[entity.inst()], entity.primaryArg());	24✔
80	break;	27✔
81
82	case IR::Kind::Opcode2Args:
83	fmt::println(stream, "\t{} {}, {}", InstructionNames[entity.inst()], entity.primaryArg(), entity.secondaryArg());	3✔
84	break;	3✔
85	}	35✔
86	}	35✔
87
88	fmt::println(stream, "");	2✔
89	++index;	2✔
90	}	2✔
91	}	1✔
92
93	const bytecode_t& IRCompiler::bytecode() const noexcept	27✔
94	{	27✔
95	return m_bytecode;	27✔
96	}
97
98	void IRCompiler::compile()	27✔
99	{	27✔
100	// push the different code segments
101	for (std::size_t i = 0, end = m_ir.size(); i < end; ++i)	160✔
102	{
103	IR::Block& page = m_ir[i];	133✔
104	// just in case we got too far, always add a HALT to be sure the
105	// VM won't do anything crazy
106	page.emplace_back(HALT);	133✔
107
108	// push number of elements
109	const auto page_size = std::ranges::count_if(page, [](const auto& a) {	8,340✔
110	return a.kind() != IR::Kind::Label;	8,207✔
111	});
112	if (std::cmp_greater(page_size, std::numeric_limits<uint16_t>::max()))	133✔
113	throw std::overflow_error(fmt::format("Size of page {} exceeds the maximum size of 2^16 - 1", i));	×
114
115	m_bytecode.push_back(CODE_SEGMENT_START);	133✔
116	m_bytecode.push_back(static_cast<uint8_t>((page_size & 0xff00) >> 8));	133✔
117	m_bytecode.push_back(static_cast<uint8_t>(page_size & 0x00ff));	133✔
118
119	// register labels position
120	uint16_t pos = 0;	133✔
121	std::unordered_map<IR::label_t, uint16_t> label_to_position;	133✔
122	for (auto inst : page)	8,340✔
123	{
124	switch (inst.kind())	8,207✔
125	{	680✔
126	case IR::Kind::Label:
127	label_to_position[inst.label()] = pos;	680✔
128	break;	8,207✔
129
130	default:
131	++pos;	7,527✔
132	}	8,207✔
133	}	8,207✔
134
135	for (auto inst : page)	8,340✔
136	{
137	switch (inst.kind())	8,207✔
138	{	329✔
139	case IR::Kind::Goto:
140	pushWord(Word(JUMP, label_to_position[inst.label()]));	329✔
141	break;	637✔
142
143	case IR::Kind::GotoIfTrue:
144	pushWord(Word(POP_JUMP_IF_TRUE, label_to_position[inst.label()]));	308✔
145	break;	359✔
146
147	case IR::Kind::GotoIfFalse:
148	pushWord(Word(POP_JUMP_IF_FALSE, label_to_position[inst.label()]));	51✔
149	break;	6,890✔
150
151	case IR::Kind::Opcode:
152	[[fallthrough]];
153	case IR::Kind::Opcode2Args:
154	pushWord(inst.bytecode());	6,839✔
155	break;	7,519✔
156
157	default:
158	break;	680✔
159	}	8,207✔
160	}	8,207✔
161	}	133✔
162	}	27✔
163
164	void IRCompiler::pushWord(const Word& word)	7,527✔
165	{	7,527✔
166	m_bytecode.push_back(word.opcode);	7,527✔
167	m_bytecode.push_back(word.byte_1);	7,527✔
168	m_bytecode.push_back(word.byte_2);	7,527✔
169	m_bytecode.push_back(word.byte_3);	7,527✔
170	}	7,527✔
171
172	void IRCompiler::pushFileHeader() noexcept	27✔
173	{	27✔
174	/*
175	Generating headers:
176	- lang name (to be sure we are executing an ArkScript file)
177	on 4 bytes (ark + padding)
178	- version (major: 2 bytes, minor: 2 bytes, patch: 2 bytes)
179	- timestamp (8 bytes, unix format)
180	*/
181
182	m_bytecode.push_back('a');	27✔
183	m_bytecode.push_back('r');	27✔
184	m_bytecode.push_back('k');	27✔
185	m_bytecode.push_back(0_u8);	27✔
186
187	// push version
188	for (const int n : std::array { ARK_VERSION_MAJOR, ARK_VERSION_MINOR, ARK_VERSION_PATCH })	108✔
189	{
190	m_bytecode.push_back(static_cast<uint8_t>((n & 0xff00) >> 8));	81✔
191	m_bytecode.push_back(static_cast<uint8_t>(n & 0x00ff));	81✔
192	}	81✔
193
194	// push timestamp
195	const long long timestamp = std::chrono::duration_cast<std::chrono::seconds>(	27✔
196	std::chrono::system_clock::now().time_since_epoch())	27✔
197	.count();	27✔
198	for (long i = 0; i < 8; ++i)	243✔
199	{
200	const long shift = 8 * (7 - i);	216✔
201	const auto ts_byte = static_cast<uint8_t>((timestamp & (0xffLL << shift)) >> shift);	216✔
202	m_bytecode.push_back(ts_byte);	216✔
203	}	216✔
204	}	27✔
205
206	void IRCompiler::pushSymAndValTables(const std::vector<std::string>& symbols, const std::vector<ValTableElem>& values)	27✔
207	{	27✔
208	const std::size_t symbol_size = symbols.size();	27✔
209	if (symbol_size > std::numeric_limits<uint16_t>::max())	27✔
210	throw std::overflow_error(fmt::format("Too many symbols: {}, exceeds the maximum size of 2^16 - 1", symbol_size));	×
211
212	m_bytecode.push_back(SYM_TABLE_START);	27✔
213	m_bytecode.push_back(static_cast<uint8_t>((symbol_size & 0xff00) >> 8));	27✔
214	m_bytecode.push_back(static_cast<uint8_t>(symbol_size & 0x00ff));	27✔
215
216	for (const auto& sym : symbols)	293✔
217	{
218	// push the string, null terminated
219	std::ranges::transform(sym, std::back_inserter(m_bytecode), [](const char i) {	2,150✔
220	return static_cast<uint8_t>(i);	1,884✔
221	});
222	m_bytecode.push_back(0_u8);	266✔
223	}	266✔
224
225	const std::size_t value_size = values.size();	27✔
226	if (value_size > std::numeric_limits<uint16_t>::max())	27✔
227	throw std::overflow_error(fmt::format("Too many values: {}, exceeds the maximum size of 2^16 - 1", value_size));	×
228
229	m_bytecode.push_back(VAL_TABLE_START);	27✔
230	m_bytecode.push_back(static_cast<uint8_t>((value_size & 0xff00) >> 8));	27✔
231	m_bytecode.push_back(static_cast<uint8_t>(value_size & 0x00ff));	27✔
232
233	for (const ValTableElem& val : values)	682✔
234	{
235	switch (val.type)	655✔
236	{	111✔
237	case ValTableElemType::Number:
238	{
239	m_bytecode.push_back(NUMBER_TYPE);	111✔
240	const auto n = std::get<double>(val.value);	111✔
241	std::string t = std::to_string(n);	111✔
242	std::ranges::transform(t, std::back_inserter(m_bytecode), [](const char i) {	1,032✔
243	return static_cast<uint8_t>(i);	921✔
244	});
245	break;
246	}	549✔
247
248	case ValTableElemType::String:
249	{
250	m_bytecode.push_back(STRING_TYPE);	438✔
251	auto t = std::get<std::string>(val.value);	438✔
252	std::ranges::transform(t, std::back_inserter(m_bytecode), [](const char i) {	6,539✔
253	return static_cast<uint8_t>(i);	6,101✔
254	});
255	break;
256	}	544✔
257
258	case ValTableElemType::PageAddr:
259	{
260	m_bytecode.push_back(FUNC_TYPE);	106✔
261	const std::size_t addr = std::get<std::size_t>(val.value);	106✔
262	m_bytecode.push_back(static_cast<uint8_t>((addr & 0xff00) >> 8));	106✔
263	m_bytecode.push_back(static_cast<uint8_t>(addr & 0x00ff));	106✔
264	break;
265	}	106✔
266	}	655✔
267
268	m_bytecode.push_back(0_u8);	655✔
269	}	655✔
270	}	27✔
271	}

ArkScript-lang / Ark / 11629611787

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous