Use indices instead of pointers for saves/instruction in ThreadedRegexVM

Performance seems unaffacted, but memory usage should be lowered
as the Thread struct is 4 bytes instead of 16.
This commit is contained in:
Maxime Coste 2018-04-27 08:18:04 +10:00
parent 869736359a
commit f10eb9faa3
2 changed files with 39 additions and 38 deletions

View File

@ -551,7 +551,7 @@ private:
NodeIndex new_node(ParsedRegex::Op op, Codepoint value = -1,
ParsedRegex::Quantifier quantifier = {ParsedRegex::Quantifier::One})
{
constexpr auto max_nodes = std::numeric_limits<uint16_t>::max();
constexpr auto max_nodes = std::numeric_limits<int16_t>::max();
const NodeIndex res = m_parsed_regex.nodes.size();
if (res == max_nodes)
parse_error(format("regex parsed to more than {} ast nodes", max_nodes));
@ -837,7 +837,7 @@ private:
uint32_t push_inst(CompiledRegex::Op op, uint32_t param = 0)
{
constexpr auto max_instructions = std::numeric_limits<uint16_t>::max();
constexpr auto max_instructions = std::numeric_limits<int16_t>::max();
const uint32_t res = m_program.instructions.size();
if (res > max_instructions)
throw regex_error(format("regex compiled to more than {} instructions", max_instructions));

View File

@ -220,8 +220,8 @@ public:
ArrayView<const Iterator> captures() const
{
if (m_captures)
return { m_captures->pos, m_program.save_count };
if (m_captures >= 0)
return { m_saves[m_captures]->pos, m_program.save_count };
return {};
}
@ -231,25 +231,26 @@ private:
union // ref count when in use, next_free when in free list
{
int refcount;
Saves* next_free;
int16_t next_free;
};
Iterator pos[1];
};
template<bool copy>
Saves* new_saves(Iterator* pos)
int16_t new_saves(Iterator* pos)
{
kak_assert(not copy or pos != nullptr);
const auto count = m_program.save_count;
if (m_first_free != nullptr)
if (m_first_free >= 0)
{
Saves* res = m_first_free;
m_first_free = res->next_free;
res->refcount = 1;
const int16_t res = m_first_free;
Saves* save = m_saves[res];
m_first_free = save->next_free;
save->refcount = 1;
if (copy)
std::copy(pos, pos + count, res->pos);
std::copy(pos, pos + count, save->pos);
else
std::fill(res->pos, res->pos + count, Iterator{});
std::fill(save->pos, save->pos + count, Iterator{});
return res;
}
@ -259,22 +260,22 @@ private:
for (size_t i = 1; i < count; ++i)
new (&saves->pos[i]) Iterator{copy ? pos[i] : Iterator{}};
m_saves.push_back(saves);
return saves;
return static_cast<int16_t>(m_saves.size() - 1);
}
void release_saves(Saves* saves)
void release_saves(int16_t saves)
{
if (saves and --saves->refcount == 0)
if (saves >= 0 and --m_saves[saves]->refcount == 0)
{
saves->next_free = m_first_free;
m_saves[saves]->next_free = m_first_free;
m_first_free = saves;
}
};
struct Thread
{
const CompiledRegex::Instruction* inst;
Saves* saves;
int16_t inst;
int16_t saves;
};
using Utf8It = utf8::iterator<Iterator>;
@ -300,7 +301,7 @@ private:
auto* instructions = m_program.instructions.data();
while (true)
{
auto& inst = *thread.inst++;
auto& inst = instructions[thread.inst++];
// if this instruction was already executed for this step in another thread,
// then this thread is redundant and can be dropped
if (inst.last_step == current_step)
@ -320,35 +321,35 @@ private:
case CompiledRegex::AnyChar:
return StepResult::Consumed;
case CompiledRegex::Jump:
thread.inst = instructions + inst.param;
thread.inst = static_cast<int16_t>(inst.param);
break;
case CompiledRegex::Split_PrioritizeParent:
{
if (thread.saves)
++thread.saves->refcount;
m_current_threads.push_back({instructions + inst.param, thread.saves});
if (thread.saves >= 0)
++m_saves[thread.saves]->refcount;
m_current_threads.push_back({static_cast<int16_t>(inst.param), thread.saves});
break;
}
case CompiledRegex::Split_PrioritizeChild:
{
if (thread.saves)
++thread.saves->refcount;
if (thread.saves >= 0)
++m_saves[thread.saves]->refcount;
m_current_threads.push_back({thread.inst, thread.saves});
thread.inst = instructions + inst.param;
thread.inst = static_cast<uint16_t>(inst.param);
break;
}
case CompiledRegex::Save:
{
if (no_saves)
break;
if (not thread.saves)
if (thread.saves < 0)
thread.saves = new_saves<false>(nullptr);
else if (thread.saves->refcount > 1)
else if (m_saves[thread.saves]->refcount > 1)
{
--thread.saves->refcount;
thread.saves = new_saves<true>(thread.saves->pos);
--m_saves[thread.saves]->refcount;
thread.saves = new_saves<true>(m_saves[thread.saves]->pos);
}
thread.saves->pos[inst.param] = get_base(pos);
m_saves[thread.saves]->pos[inst.param] = get_base(pos);
break;
}
case CompiledRegex::Class:
@ -425,8 +426,8 @@ private:
{
kak_assert(m_current_threads.empty() and m_next_threads.empty());
release_saves(m_captures);
m_captures = nullptr;
m_current_threads.push_back({config.instructions.begin(), nullptr});
m_captures = -1;
m_current_threads.push_back({static_cast<int16_t>(&config.instructions[0] - &m_program.instructions[0]), -1});
const auto& start_desc = direction == MatchDirection::Forward ? m_program.forward_start_desc
: m_program.backward_start_desc;
@ -471,12 +472,12 @@ private:
release_saves(thread.saves);
break;
case StepResult::Consumed:
if (thread.inst->scheduled)
if (m_program.instructions[thread.inst].scheduled)
{
release_saves(thread.saves);
continue;
}
thread.inst->scheduled = true;
m_program.instructions[thread.inst].scheduled = true;
m_next_threads.push_back(thread);
break;
case StepResult::FindNextStart:
@ -487,7 +488,7 @@ private:
}
kak_assert(m_current_threads.empty());
for (auto& thread : m_next_threads)
thread.inst->scheduled = false;
m_program.instructions[thread.inst].scheduled = false;
if (pos == config.end or m_next_threads.empty() or
(found_match and (config.flags & RegexExecFlags::AnyMatch)))
@ -580,8 +581,8 @@ private:
Vector<Thread, MemoryDomain::Regex> m_next_threads;
Vector<Saves*, MemoryDomain::Regex> m_saves;
Saves* m_first_free = nullptr;
Saves* m_captures = nullptr;
int16_t m_first_free = -1;
int16_t m_captures = -1;
};
}