tursom/memgraph
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Pmr allocator unify (#1801)

Browse Source 
Query allocator and evaluation allocator were different.
After analysis, was determined they should be the same, this will help 
future development reduce TypeValue copies during queries.

Changes:
- Common allocator, PoolResource backed by MonotonicResource
- Optimized Pool, now O(1) alloc/dealloc as all chunks in Pool form a single 
  free list
- 2nd PoolResource, using bin sizing, not as perfect for memory usage but 
  O(1) bin selection
- Now have jemalloc's background thread to make sure decay and return 
  to OS happens
- Optimized ProperyValue to be faster at destruction/copy/move
- Less temporary memory allocations
  - CSV reader now maintains a common line buffer it reuses on line reads
  - Writing out bolt values, now reuses a values buffer
  - Evaluating an int no longer makes temporary strings for errors it most 
    likely never throws
  - ExpandVariable will reuse existing edge list in frame it one existed
This commit is contained in:
Gareth Andrew Lloyd 2024-03-14 18:21:59 +00:00 committed by GitHub
parent b0cdcd3483
commit 8bc8e867e4
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
23 changed files with 738 additions and 770 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
CMakeLists.txt
View File

@ -300,6 +300,19 @@ endif()
option(ENABLE_JEMALLOC "Use jemalloc" ON)
option(MG_MEMORY_PROFILE "If build should be setup for memory profiling" OFF)
if (MG_MEMORY_PROFILE AND ENABLE_JEMALLOC)
message(STATUS "Jemalloc has been disabled because MG_MEMORY_PROFILE is enabled")
set(ENABLE_JEMALLOC OFF)
endif ()
if (MG_MEMORY_PROFILE AND ASAN)
message(STATUS "ASAN has been disabled because MG_MEMORY_PROFILE is enabled")
set(ASAN OFF)
endif ()
if (MG_MEMORY_PROFILE)
add_compile_definitions(MG_MEMORY_PROFILE)
endif ()
if (ASAN)
message(WARNING "Disabling jemalloc as it doesn't work well with ASAN")
set(ENABLE_JEMALLOC OFF)

4
libs/setup.sh
View File

@ -268,13 +268,13 @@ repo_clone_try_double "${primary_urls[jemalloc]}" "${secondary_urls[jemalloc]}"
pushd jemalloc
./autogen.sh
MALLOC_CONF="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000" \
MALLOC_CONF="background_thread:true,retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000" \
./configure \
--disable-cxx \
--with-lg-page=12 \
--with-lg-hugepage=21 \
--enable-shared=no --prefix=$working_dir \
--with-malloc-conf="retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000"
--with-malloc-conf="background_thread:true,retain:false,percpu_arena:percpu,oversize_threshold:0,muzzy_decay_ms:5000,dirty_decay_ms:5000"
make -j$CPUS install
popd

4
src/csv/include/csv/parsing.hpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -119,6 +119,8 @@ class Reader {
auto GetHeader() const -> Header const &;
auto GetNextRow(utils::MemoryResource *mem) -> std::optional<Row>;
void Reset();
private:
// Some implementation issues that need clearing up, but this is mainly because
// I don't want `boost/iostreams/filtering_stream.hpp` included in this header file

37
src/csv/parsing.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -34,6 +34,10 @@ struct Reader::impl {
[[nodiscard]] bool HasHeader() const { return read_config_.with_header; }
[[nodiscard]] auto Header() const -> Header const & { return header_; }
void Reset() {
line_buffer_.clear();
line_buffer_.shrink_to_fit();
}
auto GetNextRow(utils::MemoryResource *mem) -> std::optional<Reader::Row>;
@ -42,7 +46,7 @@ struct Reader::impl {
void TryInitializeHeader();
std::optional<utils::pmr::string> GetNextLine(utils::MemoryResource *mem);
bool GetNextLine();
ParsingResult ParseHeader();
@ -55,6 +59,8 @@ struct Reader::impl {
Config read_config_;
uint64_t line_count_{1};
uint16_t number_of_columns_{0};
uint64_t estimated_number_of_columns_{0};
utils::pmr::string line_buffer_{memory_};
Reader::Header header_{memory_};
};
@ -129,17 +135,16 @@ void Reader::impl::InitializeStream() {
MG_ASSERT(csv_stream_.is_complete(), "Should be 'complete' for correct operation");
}
std::optional<utils::pmr::string> Reader::impl::GetNextLine(utils::MemoryResource *mem) {
utils::pmr::string line(mem);
if (!std::getline(csv_stream_, line)) {
bool Reader::impl::GetNextLine() {
if (!std::getline(csv_stream_, line_buffer_)) {
// reached end of file or an I/0 error occurred
if (!csv_stream_.good()) {
csv_stream_.reset(); // this will close the file_stream_ and clear the chain
}
return std::nullopt;
return false;
}
++line_count_;
return std::move(line);
return true;
}
Reader::ParsingResult Reader::impl::ParseHeader() {
@ -170,6 +175,8 @@ void Reader::impl::TryInitializeHeader() {
const Reader::Header &Reader::GetHeader() const { return pimpl->Header(); }
void Reader::Reset() { pimpl->Reset(); }
namespace {
enum class CsvParserState : uint8_t { INITIAL_FIELD, NEXT_FIELD, QUOTING, EXPECT_DELIMITER, DONE };
@ -179,6 +186,8 @@ Reader::ParsingResult Reader::impl::ParseRow(utils::MemoryResource *mem) {
utils::pmr::vector<utils::pmr::string> row(mem);
if (number_of_columns_ != 0) {
row.reserve(number_of_columns_);
} else if (estimated_number_of_columns_ != 0) {
row.reserve(estimated_number_of_columns_);
}
utils::pmr::string column(memory_);
@ -186,13 +195,12 @@ Reader::ParsingResult Reader::impl::ParseRow(utils::MemoryResource *mem) {
auto state = CsvParserState::INITIAL_FIELD;
do {
const auto maybe_line = GetNextLine(mem);
if (!maybe_line) {
if (!GetNextLine()) {
// The whole file was processed.
break;
}
std::string_view line_string_view = *maybe_line;
std::string_view line_string_view = line_buffer_;
// remove '\r' from the end in case we have dos file format
if (line_string_view.back() == '\r') {
@ -312,6 +320,11 @@ Reader::ParsingResult Reader::impl::ParseRow(utils::MemoryResource *mem) {
fmt::format("Expected {:d} columns in row {:d}, but got {:d}", number_of_columns_,
line_count_ - 1, row.size()));
}
// To avoid unessisary dynamic growth of the row, remember the number of
// columns for future calls
if (number_of_columns_ == 0 && estimated_number_of_columns_ == 0) {
estimated_number_of_columns_ = row.size();
}
return std::move(row);
}
@ -319,7 +332,7 @@ Reader::ParsingResult Reader::impl::ParseRow(utils::MemoryResource *mem) {
std::optional<Reader::Row> Reader::impl::GetNextRow(utils::MemoryResource *mem) {
auto row = ParseRow(mem);
if (row.HasError()) {
if (row.HasError()) [[unlikely]] {
if (!read_config_.ignore_bad) {
throw CsvReadException("CSV Reader: Bad row at line {:d}: {}", line_count_ - 1, row.GetError().message);
}
@ -333,7 +346,7 @@ std::optional<Reader::Row> Reader::impl::GetNextRow(utils::MemoryResource *mem)
} while (row.HasError());
}
if (row->empty()) {
if (row->empty()) [[unlikely]] {
// reached end of file
return std::nullopt;
}

21
src/glue/SessionHL.cpp
View File

@ -59,12 +59,14 @@ class TypedValueResultStreamBase {
public:
explicit TypedValueResultStreamBase(memgraph::storage::Storage *storage);
std::vector<memgraph::communication::bolt::Value> DecodeValues(
const std::vector<memgraph::query::TypedValue> &values) const;
void DecodeValues(const std::vector<memgraph::query::TypedValue> &values);
auto AccessValues() const -> std::vector<memgraph::communication::bolt::Value> const & { return decoded_values_; }
protected:
// NOTE: Needed only for ToBoltValue conversions
memgraph::storage::Storage *storage_;
std::vector<memgraph::communication::bolt::Value> decoded_values_;
};
/// Wrapper around TEncoder which converts TypedValue to Value
@ -75,16 +77,18 @@ class TypedValueResultStream : public TypedValueResultStreamBase {
TypedValueResultStream(TEncoder *encoder, memgraph::storage::Storage *storage)
: TypedValueResultStreamBase{storage}, encoder_(encoder) {}
void Result(const std::vector<memgraph::query::TypedValue> &values) { encoder_->MessageRecord(DecodeValues(values)); }
void Result(const std::vector<memgraph::query::TypedValue> &values) {
DecodeValues(values);
encoder_->MessageRecord(AccessValues());
}
private:
TEncoder *encoder_;
};
std::vector<memgraph::communication::bolt::Value> TypedValueResultStreamBase::DecodeValues(
const std::vector<memgraph::query::TypedValue> &values) const {
std::vector<memgraph::communication::bolt::Value> decoded_values;
decoded_values.reserve(values.size());
void TypedValueResultStreamBase::DecodeValues(const std::vector<memgraph::query::TypedValue> &values) {
decoded_values_.reserve(values.size());
decoded_values_.clear();
for (const auto &v : values) {
auto maybe_value = memgraph::glue::ToBoltValue(v, storage_, memgraph::storage::View::NEW);
if (maybe_value.HasError()) {
@ -99,9 +103,8 @@ std::vector<memgraph::communication::bolt::Value> TypedValueResultStreamBase::De
throw memgraph::communication::bolt::ClientError("Unexpected storage error when streaming results.");
}
}
decoded_values.emplace_back(std::move(*maybe_value));
decoded_values_.emplace_back(std::move(*maybe_value));
}
return decoded_values;
}
TypedValueResultStreamBase::TypedValueResultStreamBase(memgraph::storage::Storage *storage) : storage_(storage) {}

6
src/query/interpret/eval.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -13,12 +13,12 @@
namespace memgraph::query {
int64_t EvaluateInt(ExpressionEvaluator *evaluator, Expression *expr, const std::string &what) {
int64_t EvaluateInt(ExpressionEvaluator *evaluator, Expression *expr, std::string_view what) {
TypedValue value = expr->Accept(*evaluator);
try {
return value.ValueInt();
} catch (TypedValueException &e) {
throw QueryRuntimeException(what + " must be an int");
throw QueryRuntimeException(std::string(what) + " must be an int");
}
}

2
src/query/interpret/eval.hpp
View File

@ -1209,7 +1209,7 @@ class ExpressionEvaluator : public ExpressionVisitor<TypedValue> {
/// @param what - Name of what's getting evaluated. Used for user feedback (via
/// exception) when the evaluated value is not an int.
/// @throw QueryRuntimeException if expression doesn't evaluate to an int.
int64_t EvaluateInt(ExpressionEvaluator *evaluator, Expression *expr, const std::string &what);
int64_t EvaluateInt(ExpressionEvaluator *evaluator, Expression *expr, std::string_view what);
std::optional<size_t> EvaluateMemoryLimit(ExpressionVisitor<TypedValue> &eval, Expression *memory_limit,
size_t memory_scale);

193
src/query/interpreter.cpp
View File

@ -246,27 +246,6 @@ std::optional<std::string> GetOptionalStringValue(query::Expression *expression,
return {};
};
bool IsAllShortestPathsQuery(const std::vector<memgraph::query::Clause *> &clauses) {
for (const auto &clause : clauses) {
if (clause->GetTypeInfo() != Match::kType) {
continue;
}
auto *match_clause = utils::Downcast<Match>(clause);
for (const auto &pattern : match_clause->patterns_) {
for (const auto &atom : pattern->atoms_) {
if (atom->GetTypeInfo() != EdgeAtom::kType) {
continue;
}
auto *edge_atom = utils::Downcast<EdgeAtom>(atom);
if (edge_atom->type_ == EdgeAtom::Type::ALL_SHORTEST_PATHS) {
return true;
}
}
}
}
return false;
}
inline auto convertFromCoordinatorToReplicationMode(const CoordinatorQuery::SyncMode &sync_mode)
-> replication_coordination_glue::ReplicationMode {
switch (sync_mode) {
@ -1733,8 +1712,7 @@ struct PullPlan {
std::shared_ptr<QueryUserOrRole> user_or_role, std::atomic<TransactionStatus> *transaction_status,
std::shared_ptr<utils::AsyncTimer> tx_timer,
TriggerContextCollector *trigger_context_collector = nullptr,
std::optional<size_t> memory_limit = {}, bool use_monotonic_memory = true,
FrameChangeCollector *frame_change_collector_ = nullptr);
std::optional<size_t> memory_limit = {}, FrameChangeCollector *frame_change_collector_ = nullptr);
std::optional<plan::ProfilingStatsWithTotalTime> Pull(AnyStream *stream, std::optional<int> n,
const std::vector<Symbol> &output_symbols,
@ -1759,26 +1737,17 @@ struct PullPlan {
// we have to keep track of any unsent results from previous `PullPlan::Pull`
// manually by using this flag.
bool has_unsent_results_ = false;
// In the case of LOAD CSV, we want to use only PoolResource without MonotonicMemoryResource
// to reuse allocated memory. As LOAD CSV is processing row by row
// it is possible to reduce memory usage significantly if MemoryResource deals with memory allocation
// can reuse memory that was allocated on processing the first row on all subsequent rows.
// This flag signals to `PullPlan::Pull` which MemoryResource to use
bool use_monotonic_memory_;
};
PullPlan::PullPlan(const std::shared_ptr<PlanWrapper> plan, const Parameters &parameters, const bool is_profile_query,
DbAccessor *dba, InterpreterContext *interpreter_context, utils::MemoryResource *execution_memory,
std::shared_ptr<QueryUserOrRole> user_or_role, std::atomic<TransactionStatus> *transaction_status,
std::shared_ptr<utils::AsyncTimer> tx_timer, TriggerContextCollector *trigger_context_collector,
const std::optional<size_t> memory_limit, bool use_monotonic_memory,
FrameChangeCollector *frame_change_collector)
const std::optional<size_t> memory_limit, FrameChangeCollector *frame_change_collector)
: plan_(plan),
cursor_(plan->plan().MakeCursor(execution_memory)),
frame_(plan->symbol_table().max_position(), execution_memory),
memory_limit_(memory_limit),
use_monotonic_memory_(use_monotonic_memory) {
memory_limit_(memory_limit) {
ctx_.db_accessor = dba;
ctx_.symbol_table = plan->symbol_table();
ctx_.evaluation_context.timestamp = QueryTimestamp();
@ -1804,6 +1773,7 @@ PullPlan::PullPlan(const std::shared_ptr<PlanWrapper> plan, const Parameters &pa
ctx_.is_profile_query = is_profile_query;
ctx_.trigger_context_collector = trigger_context_collector;
ctx_.frame_change_collector = frame_change_collector;
ctx_.evaluation_context.memory = execution_memory;
}
std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::Pull(AnyStream *stream, std::optional<int> n,
@ -1827,43 +1797,14 @@ std::optional<plan::ProfilingStatsWithTotalTime> PullPlan::Pull(AnyStream *strea
}
}};
// Set up temporary memory for a single Pull. Initial memory comes from the
// stack. 256 KiB should fit on the stack and should be more than enough for a
// single `Pull`.
static constexpr size_t stack_size = 256UL * 1024UL;
char stack_data[stack_size];
utils::ResourceWithOutOfMemoryException resource_with_exception;
utils::MonotonicBufferResource monotonic_memory{&stack_data[0], stack_size, &resource_with_exception};
std::optional<utils::PoolResource> pool_memory;
static constexpr auto kMaxBlockPerChunks = 128;
if (!use_monotonic_memory_) {
pool_memory.emplace(kMaxBlockPerChunks, kExecutionPoolMaxBlockSize, &resource_with_exception,
&resource_with_exception);
} else {
// We can throw on every query because a simple queries for deleting will use only
// the stack allocated buffer.
// Also, we want to throw only when the query engine requests more memory and not the storage
// so we add the exception to the allocator.
// TODO (mferencevic): Tune the parameters accordingly.
pool_memory.emplace(kMaxBlockPerChunks, 1024, &monotonic_memory, &resource_with_exception);
}
ctx_.evaluation_context.memory = &*pool_memory;
// Returns true if a result was pulled.
const auto pull_result = [&]() -> bool { return cursor_->Pull(frame_, ctx_); };
const auto stream_values = [&]() {
// TODO: The streamed values should also probably use the above memory.
std::vector<TypedValue> values;
values.reserve(output_symbols.size());
for (const auto &symbol : output_symbols) {
values.emplace_back(frame_[symbol]);
auto values = std::vector<TypedValue>(output_symbols.size());
const auto stream_values = [&] {
for (auto const i : ranges::views::iota(0UL, output_symbols.size())) {
values[i] = frame_[output_symbols[i]];
}
stream->Result(values);
};
@ -1973,7 +1914,6 @@ PreparedQuery Interpreter::PrepareTransactionQuery(std::string_view query_upper,
std::function<void()> handler;
if (query_upper == "BEGIN") {
ResetInterpreter();
// TODO: Evaluate doing move(extras). Currently the extras is very small, but this will be important if it ever
// becomes large.
handler = [this, extras = extras] {
@ -2051,30 +1991,6 @@ inline static void TryCaching(const AstStorage &ast_storage, FrameChangeCollecto
}
}
bool IsLoadCsvQuery(const std::vector<memgraph::query::Clause *> &clauses) {
return std::any_of(clauses.begin(), clauses.end(),
[](memgraph::query::Clause const *clause) { return clause->GetTypeInfo() == LoadCsv::kType; });
}
bool IsCallBatchedProcedureQuery(const std::vector<memgraph::query::Clause *> &clauses) {
EvaluationContext evaluation_context;
return std::ranges::any_of(clauses, [&evaluation_context](memgraph::query::Clause *clause) -> bool {
if (!(clause->GetTypeInfo() == CallProcedure::kType)) return false;
auto *call_procedure_clause = utils::Downcast<CallProcedure>(clause);
const auto &maybe_found = memgraph::query::procedure::FindProcedure(
procedure::gModuleRegistry, call_procedure_clause->procedure_name_, evaluation_context.memory);
if (!maybe_found) {
throw QueryRuntimeException("There is no procedure named '{}'.", call_procedure_clause->procedure_name_);
}
const auto &[module, proc] = *maybe_found;
if (!proc->info.is_batched) return false;
spdlog::trace("Using PoolResource for batched query procedure");
return true;
});
}
PreparedQuery PrepareCypherQuery(ParsedQuery parsed_query, std::map<std::string, TypedValue> *summary,
InterpreterContext *interpreter_context, CurrentDB &current_db,
utils::MemoryResource *execution_memory, std::vector<Notification> *notifications,
@ -2094,7 +2010,6 @@ PreparedQuery PrepareCypherQuery(ParsedQuery parsed_query, std::map<std::string,
spdlog::info("Running query with memory limit of {}", utils::GetReadableSize(*memory_limit));
}
auto clauses = cypher_query->single_query_->clauses_;
bool contains_csv = false;
if (std::any_of(clauses.begin(), clauses.end(),
[](const auto *clause) { return clause->GetTypeInfo() == LoadCsv::kType; })) {
notifications->emplace_back(
@ -2102,13 +2017,8 @@ PreparedQuery PrepareCypherQuery(ParsedQuery parsed_query, std::map<std::string,
"It's important to note that the parser parses the values as strings. It's up to the user to "
"convert the parsed row values to the appropriate type. This can be done using the built-in "
"conversion functions such as ToInteger, ToFloat, ToBoolean etc.");
contains_csv = true;
}
// If this is LOAD CSV query, use PoolResource without MonotonicMemoryResource as we want to reuse allocated memory
auto use_monotonic_memory =
!contains_csv && !IsCallBatchedProcedureQuery(clauses) && !IsAllShortestPathsQuery(clauses);
MG_ASSERT(current_db.execution_db_accessor_, "Cypher query expects a current DB transaction");
auto *dba =
&*current_db
@ -2147,7 +2057,7 @@ PreparedQuery PrepareCypherQuery(ParsedQuery parsed_query, std::map<std::string,
current_db.trigger_context_collector_ ? &*current_db.trigger_context_collector_ : nullptr;
auto pull_plan = std::make_shared<PullPlan>(
plan, parsed_query.parameters, false, dba, interpreter_context, execution_memory, std::move(user_or_role),
transaction_status, std::move(tx_timer), trigger_context_collector, memory_limit, use_monotonic_memory,
transaction_status, std::move(tx_timer), trigger_context_collector, memory_limit,
frame_change_collector->IsTrackingValues() ? frame_change_collector : nullptr);
return PreparedQuery{std::move(header), std::move(parsed_query.required_privileges),
[pull_plan = std::move(pull_plan), output_symbols = std::move(output_symbols), summary](
@ -2261,18 +2171,6 @@ PreparedQuery PrepareProfileQuery(ParsedQuery parsed_query, bool in_explicit_tra
auto *cypher_query = utils::Downcast<CypherQuery>(parsed_inner_query.query);
bool contains_csv = false;
auto clauses = cypher_query->single_query_->clauses_;
if (std::any_of(clauses.begin(), clauses.end(),
[](const auto *clause) { return clause->GetTypeInfo() == LoadCsv::kType; })) {
contains_csv = true;
}
// If this is LOAD CSV, BatchedProcedure or AllShortest query, use PoolResource without MonotonicMemoryResource as we
// want to reuse allocated memory
auto use_monotonic_memory =
!contains_csv && !IsCallBatchedProcedureQuery(clauses) && !IsAllShortestPathsQuery(clauses);
MG_ASSERT(cypher_query, "Cypher grammar should not allow other queries in PROFILE");
EvaluationContext evaluation_context;
evaluation_context.timestamp = QueryTimestamp();
@ -2306,14 +2204,14 @@ PreparedQuery PrepareProfileQuery(ParsedQuery parsed_query, bool in_explicit_tra
// We want to execute the query we are profiling lazily, so we delay
// the construction of the corresponding context.
stats_and_total_time = std::optional<plan::ProfilingStatsWithTotalTime>{},
pull_plan = std::shared_ptr<PullPlanVector>(nullptr), transaction_status, use_monotonic_memory,
frame_change_collector, tx_timer = std::move(tx_timer)](
AnyStream *stream, std::optional<int> n) mutable -> std::optional<QueryHandlerResult> {
pull_plan = std::shared_ptr<PullPlanVector>(nullptr), transaction_status, frame_change_collector,
tx_timer = std::move(tx_timer)](AnyStream *stream,
std::optional<int> n) mutable -> std::optional<QueryHandlerResult> {
// No output symbols are given so that nothing is streamed.
if (!stats_and_total_time) {
stats_and_total_time =
PullPlan(plan, parameters, true, dba, interpreter_context, execution_memory, std::move(user_or_role),
transaction_status, std::move(tx_timer), nullptr, memory_limit, use_monotonic_memory,
transaction_status, std::move(tx_timer), nullptr, memory_limit,
frame_change_collector->IsTrackingValues() ? frame_change_collector : nullptr)
.Pull(stream, {}, {}, summary);
pull_plan = std::make_shared<PullPlanVector>(ProfilingStatsToTable(*stats_and_total_time));
@ -4276,6 +4174,7 @@ PreparedQuery PrepareShowDatabasesQuery(ParsedQuery parsed_query, InterpreterCon
std::optional<uint64_t> Interpreter::GetTransactionId() const { return current_transaction_; }
void Interpreter::BeginTransaction(QueryExtras const &extras) {
ResetInterpreter();
const auto prepared_query = PrepareTransactionQuery("BEGIN", extras);
prepared_query.query_handler(nullptr, {});
}
@ -4310,12 +4209,12 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
const auto upper_case_query = utils::ToUpperCase(query_string);
const auto trimmed_query = utils::Trim(upper_case_query);
if (trimmed_query == "BEGIN" || trimmed_query == "COMMIT" || trimmed_query == "ROLLBACK") {
auto resource = utils::MonotonicBufferResource(kExecutionMemoryBlockSize);
auto prepared_query = PrepareTransactionQuery(trimmed_query, extras);
auto &query_execution =
query_executions_.emplace_back(QueryExecution::Create(std::move(resource), std::move(prepared_query)));
std::optional<int> qid =
in_explicit_transaction_ ? static_cast<int>(query_executions_.size() - 1) : std::optional<int>{};
if (trimmed_query == "BEGIN") {
ResetInterpreter();
}
auto &query_execution = query_executions_.emplace_back(QueryExecution::Create());
query_execution->prepared_query = PrepareTransactionQuery(trimmed_query, extras);
auto qid = in_explicit_transaction_ ? static_cast<int>(query_executions_.size() - 1) : std::optional<int>{};
return {query_execution->prepared_query->header, query_execution->prepared_query->privileges, qid, {}};
}
@ -4345,35 +4244,8 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
ParseQuery(query_string, params, &interpreter_context_->ast_cache, interpreter_context_->config.query);
auto parsing_time = parsing_timer.Elapsed().count();
CypherQuery const *const cypher_query = [&]() -> CypherQuery * {
if (auto *cypher_query = utils::Downcast<CypherQuery>(parsed_query.query)) {
return cypher_query;
}
if (auto *profile_query = utils::Downcast<ProfileQuery>(parsed_query.query)) {
return profile_query->cypher_query_;
}
return nullptr;
}(); // IILE
auto const [usePool, hasAllShortestPaths] = [&]() -> std::pair<bool, bool> {
if (!cypher_query) {
return {false, false};
}
auto const &clauses = cypher_query->single_query_->clauses_;
bool hasAllShortestPaths = IsAllShortestPathsQuery(clauses);
// Using PoolResource without MonotonicMemoryResouce for LOAD CSV reduces memory usage.
bool usePool = hasAllShortestPaths || IsCallBatchedProcedureQuery(clauses) || IsLoadCsvQuery(clauses);
return {usePool, hasAllShortestPaths};
}(); // IILE
// Setup QueryExecution
// its MemoryResource is mostly used for allocations done on Frame and storing `row`s
if (usePool) {
query_executions_.emplace_back(QueryExecution::Create(utils::PoolResource(128, kExecutionPoolMaxBlockSize)));
} else {
query_executions_.emplace_back(QueryExecution::Create(utils::MonotonicBufferResource(kExecutionMemoryBlockSize)));
}
query_executions_.emplace_back(QueryExecution::Create());
auto &query_execution = query_executions_.back();
query_execution_ptr = &query_execution;
@ -4442,9 +4314,7 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
utils::Timer planning_timer;
PreparedQuery prepared_query;
utils::MemoryResource *memory_resource =
std::visit([](auto &execution_memory) -> utils::MemoryResource * { return &execution_memory; },
query_execution->execution_memory);
utils::MemoryResource *memory_resource = query_execution->execution_memory.resource();
frame_change_collector_.reset();
frame_change_collector_.emplace();
if (utils::Downcast<CypherQuery>(parsed_query.query)) {
@ -4455,10 +4325,10 @@ Interpreter::PrepareResult Interpreter::Prepare(const std::string &query_string,
prepared_query = PrepareExplainQuery(std::move(parsed_query), &query_execution->summary,
&query_execution->notifications, interpreter_context_, current_db_);
} else if (utils::Downcast<ProfileQuery>(parsed_query.query)) {
prepared_query = PrepareProfileQuery(std::move(parsed_query), in_explicit_transaction_, &query_execution->summary,
&query_execution->notifications, interpreter_context_, current_db_,
&query_execution->execution_memory_with_exception, user_or_role_,
&transaction_status_, current_timeout_timer_, &*frame_change_collector_);
prepared_query =
PrepareProfileQuery(std::move(parsed_query), in_explicit_transaction_, &query_execution->summary,
&query_execution->notifications, interpreter_context_, current_db_, memory_resource,
user_or_role_, &transaction_status_, current_timeout_timer_, &*frame_change_collector_);
} else if (utils::Downcast<DumpQuery>(parsed_query.query)) {
prepared_query = PrepareDumpQuery(std::move(parsed_query), current_db_);
} else if (utils::Downcast<IndexQuery>(parsed_query.query)) {
@ -4660,7 +4530,7 @@ void RunTriggersAfterCommit(dbms::DatabaseAccess db_acc, InterpreterContext *int
std::atomic<TransactionStatus> *transaction_status) {
// Run the triggers
for (const auto &trigger : db_acc->trigger_store()->AfterCommitTriggers().access()) {
utils::MonotonicBufferResource execution_memory{kExecutionMemoryBlockSize};
QueryAllocator execution_memory{};
// create a new transaction for each trigger
auto tx_acc = db_acc->Access();
@ -4671,7 +4541,7 @@ void RunTriggersAfterCommit(dbms::DatabaseAccess db_acc, InterpreterContext *int
auto trigger_context = original_trigger_context;
trigger_context.AdaptForAccessor(&db_accessor);
try {
trigger.Execute(&db_accessor, &execution_memory, flags::run_time::GetExecutionTimeout(),
trigger.Execute(&db_accessor, execution_memory.resource(), flags::run_time::GetExecutionTimeout(),
&interpreter_context->is_shutting_down, transaction_status, trigger_context);
} catch (const utils::BasicException &exception) {
spdlog::warn("Trigger '{}' failed with exception:\n{}", trigger.Name(), exception.what());
@ -4825,11 +4695,12 @@ void Interpreter::Commit() {
if (trigger_context) {
// Run the triggers
for (const auto &trigger : db->trigger_store()->BeforeCommitTriggers().access()) {
utils::MonotonicBufferResource execution_memory{kExecutionMemoryBlockSize};
QueryAllocator execution_memory{};
AdvanceCommand();
try {
trigger.Execute(&*current_db_.execution_db_accessor_, &execution_memory, flags::run_time::GetExecutionTimeout(),
&interpreter_context_->is_shutting_down, &transaction_status_, *trigger_context);
trigger.Execute(&*current_db_.execution_db_accessor_, execution_memory.resource(),
flags::run_time::GetExecutionTimeout(), &interpreter_context_->is_shutting_down,
&transaction_status_, *trigger_context);
} catch (const utils::BasicException &e) {
throw utils::BasicException(
fmt::format("Trigger '{}' caused the transaction to fail.\nException: {}", trigger.Name(), e.what()));

88
src/query/interpreter.hpp
View File

@ -65,6 +65,54 @@ extern const Event SuccessfulQuery;
namespace memgraph::query {
struct QueryAllocator {
QueryAllocator() = default;
QueryAllocator(QueryAllocator const &) = delete;
QueryAllocator &operator=(QueryAllocator const &) = delete;
// No move addresses to pool & monotonic fields must be stable
QueryAllocator(QueryAllocator &&) = delete;
QueryAllocator &operator=(QueryAllocator &&) = delete;
auto resource() -> utils::MemoryResource * {
#ifndef MG_MEMORY_PROFILE
return &pool;
#else
return upstream_resource();
#endif
}
auto resource_without_pool() -> utils::MemoryResource * {
#ifndef MG_MEMORY_PROFILE
return &monotonic;
#else
return upstream_resource();
#endif
}
auto resource_without_pool_or_mono() -> utils::MemoryResource * { return upstream_resource(); }
private:
// At least one page to ensure not sharing page with other subsystems
static constexpr auto kMonotonicInitialSize = 4UL * 1024UL;
// TODO: need to profile to check for good defaults, also maybe PoolResource
// needs to be smarter. We expect more reuse of smaller objects than larger
// objects. 64*1024B is maybe wasteful, whereas 256*32B maybe sensible.
// Depends on number of small objects expected.
static constexpr auto kPoolBlockPerChunk = 64UL;
static constexpr auto kPoolMaxBlockSize = 1024UL;
static auto upstream_resource() -> utils::MemoryResource * {
// singleton ResourceWithOutOfMemoryException
// explicitly backed by NewDeleteResource
static auto upstream = utils::ResourceWithOutOfMemoryException{utils::NewDeleteResource()};
return &upstream;
}
#ifndef MG_MEMORY_PROFILE
memgraph::utils::MonotonicBufferResource monotonic{kMonotonicInitialSize, upstream_resource()};
memgraph::utils::PoolResource pool{kPoolBlockPerChunk, &monotonic, upstream_resource()};
#endif
};
struct InterpreterContext;
inline constexpr size_t kExecutionMemoryBlockSize = 1UL * 1024UL * 1024UL;
@ -304,45 +352,25 @@ class Interpreter final {
}
struct QueryExecution {
std::variant<utils::MonotonicBufferResource, utils::PoolResource> execution_memory;
utils::ResourceWithOutOfMemoryException execution_memory_with_exception;
std::optional<PreparedQuery> prepared_query;
QueryAllocator execution_memory; // NOTE: before all other fields which uses this memory
std::optional<PreparedQuery> prepared_query;
std::map<std::string, TypedValue> summary;
std::vector<Notification> notifications;
static auto Create(std::variant<utils::MonotonicBufferResource, utils::PoolResource> memory_resource,
std::optional<PreparedQuery> prepared_query = std::nullopt) -> std::unique_ptr<QueryExecution> {
return std::make_unique<QueryExecution>(std::move(memory_resource), std::move(prepared_query));
}
static auto Create() -> std::unique_ptr<QueryExecution> { return std::make_unique<QueryExecution>(); }
explicit QueryExecution(std::variant<utils::MonotonicBufferResource, utils::PoolResource> memory_resource,
std::optional<PreparedQuery> prepared_query)
: execution_memory(std::move(memory_resource)), prepared_query{std::move(prepared_query)} {
std::visit(
[&](auto &memory_resource) {
execution_memory_with_exception = utils::ResourceWithOutOfMemoryException(&memory_resource);
},
execution_memory);
};
explicit QueryExecution() = default;
QueryExecution(const QueryExecution &) = delete;
QueryExecution(QueryExecution &&) = default;
QueryExecution(QueryExecution &&) = delete;
QueryExecution &operator=(const QueryExecution &) = delete;
QueryExecution &operator=(QueryExecution &&) = default;
QueryExecution &operator=(QueryExecution &&) = delete;
~QueryExecution() {
// We should always release the execution memory AFTER we
// destroy the prepared query which is using that instance
// of execution memory.
prepared_query.reset();
std::visit([](auto &memory_resource) { memory_resource.Release(); }, execution_memory);
}
~QueryExecution() = default;
void CleanRuntimeData() {
if (prepared_query.has_value()) {
prepared_query.reset();
}
prepared_query.reset();
notifications.clear();
}
};
@ -413,9 +441,7 @@ std::map<std::string, TypedValue> Interpreter::Pull(TStream *result_stream, std:
try {
// Wrap the (statically polymorphic) stream type into a common type which
// the handler knows.
AnyStream stream{result_stream,
std::visit([](auto &execution_memory) -> utils::MemoryResource * { return &execution_memory; },
query_execution->execution_memory)};
AnyStream stream{result_stream, query_execution->execution_memory.resource()};
const auto maybe_res = query_execution->prepared_query->query_handler(&stream, n);
// Stream is using execution memory of the query_execution which
// can be deleted after its execution so the stream should be cleared

50
src/query/plan/operator.cpp
View File

@ -69,6 +69,7 @@
#include "utils/pmr/vector.hpp"
#include "utils/readable_size.hpp"
#include "utils/string.hpp"
#include "utils/tag.hpp"
#include "utils/temporal.hpp"
#include "utils/typeinfo.hpp"
@ -864,17 +865,15 @@ bool Expand::ExpandCursor::Pull(Frame &frame, ExecutionContext &context) {
SCOPED_PROFILE_OP_BY_REF(self_);
// A helper function for expanding a node from an edge.
auto pull_node = [this, &frame](const EdgeAccessor &new_edge, EdgeAtom::Direction direction) {
auto pull_node = [this, &frame]<EdgeAtom::Direction direction>(const EdgeAccessor &new_edge,
utils::tag_value<direction>) {
if (self_.common_.existing_node) return;
switch (direction) {
case EdgeAtom::Direction::IN:
frame[self_.common_.node_symbol] = new_edge.From();
break;
case EdgeAtom::Direction::OUT:
frame[self_.common_.node_symbol] = new_edge.To();
break;
case EdgeAtom::Direction::BOTH:
LOG_FATAL("Must indicate exact expansion direction here");
if constexpr (direction == EdgeAtom::Direction::IN) {
frame[self_.common_.node_symbol] = new_edge.From();
} else if constexpr (direction == EdgeAtom::Direction::OUT) {
frame[self_.common_.node_symbol] = new_edge.To();
} else {
LOG_FATAL("Must indicate exact expansion direction here");
}
};
@ -893,7 +892,7 @@ bool Expand::ExpandCursor::Pull(Frame &frame, ExecutionContext &context) {
#endif
frame[self_.common_.edge_symbol] = edge;
pull_node(edge, EdgeAtom::Direction::IN);
pull_node(edge, utils::tag_v<EdgeAtom::Direction::IN>);
return true;
}
@ -913,7 +912,7 @@ bool Expand::ExpandCursor::Pull(Frame &frame, ExecutionContext &context) {
}
#endif
frame[self_.common_.edge_symbol] = edge;
pull_node(edge, EdgeAtom::Direction::OUT);
pull_node(edge, utils::tag_v<EdgeAtom::Direction::OUT>);
return true;
}
@ -1007,12 +1006,12 @@ bool Expand::ExpandCursor::InitEdges(Frame &frame, ExecutionContext &context) {
auto existing_node = *expansion_info_.existing_node;
auto edges_result = UnwrapEdgesResult(vertex.InEdges(self_.view_, self_.common_.edge_types, existing_node));
in_edges_.emplace(edges_result.edges);
in_edges_.emplace(std::move(edges_result.edges));
num_expanded_first = edges_result.expanded_count;
}
} else {
auto edges_result = UnwrapEdgesResult(vertex.InEdges(self_.view_, self_.common_.edge_types));
in_edges_.emplace(edges_result.edges);
in_edges_.emplace(std::move(edges_result.edges));
num_expanded_first = edges_result.expanded_count;
}
if (in_edges_) {
@ -1026,12 +1025,12 @@ bool Expand::ExpandCursor::InitEdges(Frame &frame, ExecutionContext &context) {
if (expansion_info_.existing_node) {
auto existing_node = *expansion_info_.existing_node;
auto edges_result = UnwrapEdgesResult(vertex.OutEdges(self_.view_, self_.common_.edge_types, existing_node));
out_edges_.emplace(edges_result.edges);
out_edges_.emplace(std::move(edges_result.edges));
num_expanded_second = edges_result.expanded_count;
}
} else {
auto edges_result = UnwrapEdgesResult(vertex.OutEdges(self_.view_, self_.common_.edge_types));
out_edges_.emplace(edges_result.edges);
out_edges_.emplace(std::move(edges_result.edges));
num_expanded_second = edges_result.expanded_count;
}
if (out_edges_) {
@ -1117,14 +1116,14 @@ auto ExpandFromVertex(const VertexAccessor &vertex, EdgeAtom::Direction directio
if (direction != EdgeAtom::Direction::OUT) {
auto edges = UnwrapEdgesResult(vertex.InEdges(view, edge_types)).edges;
if (edges.begin() != edges.end()) {
if (!edges.empty()) {
chain_elements.emplace_back(wrapper(EdgeAtom::Direction::IN, std::move(edges)));
}
}
if (direction != EdgeAtom::Direction::IN) {
auto edges = UnwrapEdgesResult(vertex.OutEdges(view, edge_types)).edges;
if (edges.begin() != edges.end()) {
if (!edges.empty()) {
chain_elements.emplace_back(wrapper(EdgeAtom::Direction::OUT, std::move(edges)));
}
}
@ -1244,8 +1243,13 @@ class ExpandVariableCursor : public Cursor {
}
// reset the frame value to an empty edge list
auto *pull_memory = context.evaluation_context.memory;
frame[self_.common_.edge_symbol] = TypedValue::TVector(pull_memory);
if (frame[self_.common_.edge_symbol].IsList()) {
// Preserve the list capacity if possible
frame[self_.common_.edge_symbol].ValueList().clear();
} else {
auto *pull_memory = context.evaluation_context.memory;
frame[self_.common_.edge_symbol] = TypedValue::TVector(pull_memory);
}
return true;
}
@ -4474,9 +4478,8 @@ class UnwindCursor : public Cursor {
TypedValue input_value = self_.input_expression_->Accept(evaluator);
if (input_value.type() != TypedValue::Type::List)
throw QueryRuntimeException("Argument of UNWIND must be a list, but '{}' was provided.", input_value.type());
// Copy the evaluted input_value_list to our vector.
// eval memory != query memory
input_value_ = input_value.ValueList();
// Move the evaluted input_value_list to our vector.
input_value_ = std::move(input_value.ValueList());
input_value_it_ = input_value_.begin();
}
@ -5336,6 +5339,7 @@ class LoadCsvCursor : public Cursor {
"1");
}
did_pull_ = true;
reader_->Reset();
}
auto row = reader_->GetNextRow(context.evaluation_context.memory);

19
src/query/plan/operator.hpp
View File

@ -76,18 +76,13 @@ using UniqueCursorPtr = std::unique_ptr<Cursor, std::function<void(Cursor *)>>;
template <class TCursor, class... TArgs>
std::unique_ptr<Cursor, std::function<void(Cursor *)>> MakeUniqueCursorPtr(utils::Allocator<TCursor> allocator,
TArgs &&...args) {
auto *ptr = allocator.allocate(1);
try {
auto *cursor = new (ptr) TCursor(std::forward<TArgs>(args)...);
return std::unique_ptr<Cursor, std::function<void(Cursor *)>>(cursor, [allocator](Cursor *base_ptr) mutable {
auto *p = static_cast<TCursor *>(base_ptr);
p->~TCursor();
allocator.deallocate(p, 1);
});
} catch (...) {
allocator.deallocate(ptr, 1);
throw;
}
auto *cursor = allocator.template new_object<TCursor>(std::forward<TArgs>(args)...);
auto dtr = [allocator](Cursor *base_ptr) mutable {
auto *p = static_cast<TCursor *>(base_ptr);
allocator.delete_object(p);
};
// TODO: not std::function
return std::unique_ptr<Cursor, std::function<void(Cursor *)>>(cursor, std::move(dtr));
}
class Once;

23
src/query/trigger.cpp
View File

@ -191,9 +191,9 @@ std::shared_ptr<Trigger::TriggerPlan> Trigger::GetPlan(DbAccessor *db_accessor)
return trigger_plan_;
}
void Trigger::Execute(DbAccessor *dba, utils::MonotonicBufferResource *execution_memory,
const double max_execution_time_sec, std::atomic<bool> *is_shutting_down,
std::atomic<TransactionStatus> *transaction_status, const TriggerContext &context) const {
void Trigger::Execute(DbAccessor *dba, utils::MemoryResource *execution_memory, const double max_execution_time_sec,
std::atomic<bool> *is_shutting_down, std::atomic<TransactionStatus> *transaction_status,
const TriggerContext &context) const {
if (!context.ShouldEventTrigger(event_type_)) {
return;
}
@ -214,22 +214,7 @@ void Trigger::Execute(DbAccessor *dba, utils::MonotonicBufferResource *execution
ctx.is_shutting_down = is_shutting_down;
ctx.transaction_status = transaction_status;
ctx.is_profile_query = false;
// Set up temporary memory for a single Pull. Initial memory comes from the
// stack. 256 KiB should fit on the stack and should be more than enough for a
// single `Pull`.
static constexpr size_t stack_size = 256UL * 1024UL;
char stack_data[stack_size];
// We can throw on every query because a simple queries for deleting will use only
// the stack allocated buffer.
// Also, we want to throw only when the query engine requests more memory and not the storage
// so we add the exception to the allocator.
utils::ResourceWithOutOfMemoryException resource_with_exception;
utils::MonotonicBufferResource monotonic_memory(&stack_data[0], stack_size, &resource_with_exception);
// TODO (mferencevic): Tune the parameters accordingly.
utils::PoolResource pool_memory(128, 1024, &monotonic_memory);
ctx.evaluation_context.memory = &pool_memory;
ctx.evaluation_context.memory = execution_memory;
auto cursor = plan.plan().MakeCursor(execution_memory);
Frame frame{plan.symbol_table().max_position(), execution_memory};

2
src/query/trigger.hpp
View File

@ -39,7 +39,7 @@ struct Trigger {
utils::SkipList<QueryCacheEntry> *query_cache, DbAccessor *db_accessor,
const InterpreterConfig::Query &query_config, std::shared_ptr<QueryUserOrRole> owner);
void Execute(DbAccessor *dba, utils::MonotonicBufferResource *execution_memory, double max_execution_time_sec,
void Execute(DbAccessor *dba, utils::MemoryResource *execution_memory, double max_execution_time_sec,
std::atomic<bool> *is_shutting_down, std::atomic<TransactionStatus> *transaction_status,
const TriggerContext &context) const;

225
src/storage/v2/property_value.hpp
View File

@ -92,7 +92,28 @@ class PropertyValue {
// TODO: Implement copy assignment operators for primitive types.
// TODO: Implement copy and move assignment operators for non-primitive types.
~PropertyValue() { DestroyValue(); }
~PropertyValue() {
switch (type_) {
// destructor for primitive types does nothing
case Type::Null:
case Type::Bool:
case Type::Int:
case Type::Double:
case Type::TemporalData:
return;
// destructor for non primitive types since we used placement new
case Type::String:
std::destroy_at(&string_v.val_);
return;
case Type::List:
std::destroy_at(&list_v.val_);
return;
case Type::Map:
std::destroy_at(&map_v.val_);
return;
}
}
Type type() const { return type_; }
@ -189,8 +210,6 @@ class PropertyValue {
}
private:
void DestroyValue() noexcept;
// NOTE: this may look strange but it is for better data layout
// https://eel.is/c++draft/class.union#general-note-1
union {
@ -357,13 +376,13 @@ inline PropertyValue::PropertyValue(const PropertyValue &other) : type_(other.ty
this->double_v.val_ = other.double_v.val_;
return;
case Type::String:
new (&string_v.val_) std::string(other.string_v.val_);
std::construct_at(&string_v.val_, other.string_v.val_);
return;
case Type::List:
new (&list_v.val_) std::vector<PropertyValue>(other.list_v.val_);
std::construct_at(&list_v.val_, other.list_v.val_);
return;
case Type::Map:
new (&map_v.val_) std::map<std::string, PropertyValue>(other.map_v.val_);
std::construct_at(&map_v.val_, other.map_v.val_);
return;
case Type::TemporalData:
this->temporal_data_v.val_ = other.temporal_data_v.val_;
@ -371,7 +390,7 @@ inline PropertyValue::PropertyValue(const PropertyValue &other) : type_(other.ty
}
}
inline PropertyValue::PropertyValue(PropertyValue &&other) noexcept : type_(std::exchange(other.type_, Type::Null)) {
inline PropertyValue::PropertyValue(PropertyValue &&other) noexcept : type_(other.type_) {
switch (type_) {
case Type::Null:
break;
@ -386,15 +405,12 @@ inline PropertyValue::PropertyValue(PropertyValue &&other) noexcept : type_(std:
break;
case Type::String:
std::construct_at(&string_v.val_, std::move(other.string_v.val_));
std::destroy_at(&other.string_v.val_);
break;
case Type::List:
std::construct_at(&list_v.val_, std::move(other.list_v.val_));
std::destroy_at(&other.list_v.val_);
break;
case Type::Map:
std::construct_at(&map_v.val_, std::move(other.map_v.val_));
std::destroy_at(&other.map_v.val_);
break;
case Type::TemporalData:
temporal_data_v.val_ = other.temporal_data_v.val_;
@ -403,38 +419,88 @@ inline PropertyValue::PropertyValue(PropertyValue &&other) noexcept : type_(std:
}
inline PropertyValue &PropertyValue::operator=(const PropertyValue &other) {
if (this == &other) return *this;
if (type_ == other.type_) {
if (this == &other) return *this;
switch (other.type_) {
case Type::Null:
break;
case Type::Bool:
bool_v.val_ = other.bool_v.val_;
break;
case Type::Int:
int_v.val_ = other.int_v.val_;
break;
case Type::Double:
double_v.val_ = other.double_v.val_;
break;
case Type::String:
string_v.val_ = other.string_v.val_;
break;
case Type::List:
list_v.val_ = other.list_v.val_;
break;
case Type::Map:
map_v.val_ = other.map_v.val_;
break;
case Type::TemporalData:
temporal_data_v.val_ = other.temporal_data_v.val_;
break;
}
return *this;
} else {
// destroy
switch (type_) {
case Type::Null:
break;
case Type::Bool:
break;
case Type::Int:
break;
case Type::Double:
break;
case Type::String:
std::destroy_at(&string_v.val_);
break;
case Type::List:
std::destroy_at(&list_v.val_);
break;
case Type::Map:
std::destroy_at(&map_v.val_);
break;
case Type::TemporalData:
break;
}
// construct
auto *new_this = std::launder(this);
switch (other.type_) {
case Type::Null:
break;
case Type::Bool:
new_this->bool_v.val_ = other.bool_v.val_;
break;
case Type::Int:
new_this->int_v.val_ = other.int_v.val_;
break;
case Type::Double:
new_this->double_v.val_ = other.double_v.val_;
break;
case Type::String:
std::construct_at(&new_this->string_v.val_, other.string_v.val_);
break;
case Type::List:
std::construct_at(&new_this->list_v.val_, other.list_v.val_);
break;
case Type::Map:
std::construct_at(&new_this->map_v.val_, other.map_v.val_);
break;
case Type::TemporalData:
new_this->temporal_data_v.val_ = other.temporal_data_v.val_;
break;
}
DestroyValue();
type_ = other.type_;
switch (other.type_) {
case Type::Null:
break;
case Type::Bool:
this->bool_v.val_ = other.bool_v.val_;
break;
case Type::Int:
this->int_v.val_ = other.int_v.val_;
break;
case Type::Double:
this->double_v.val_ = other.double_v.val_;
break;
case Type::String:
new (&string_v.val_) std::string(other.string_v.val_);
break;
case Type::List:
new (&list_v.val_) std::vector<PropertyValue>(other.list_v.val_);
break;
case Type::Map:
new (&map_v.val_) std::map<std::string, PropertyValue>(other.map_v.val_);
break;
case Type::TemporalData:
this->temporal_data_v.val_ = other.temporal_data_v.val_;
break;
new_this->type_ = other.type_;
return *new_this;
}
return *this;
}
inline PropertyValue &PropertyValue::operator=(PropertyValue &&other) noexcept {
@ -456,48 +522,71 @@ inline PropertyValue &PropertyValue::operator=(PropertyValue &&other) noexcept {
break;
case Type::String:
string_v.val_ = std::move(other.string_v.val_);
std::destroy_at(&other.string_v.val_);
break;
case Type::List:
list_v.val_ = std::move(other.list_v.val_);
std::destroy_at(&other.list_v.val_);
break;
case Type::Map:
map_v.val_ = std::move(other.map_v.val_);
std::destroy_at(&other.map_v.val_);
break;
case Type::TemporalData:
temporal_data_v.val_ = other.temporal_data_v.val_;
break;
}
other.type_ = Type::Null;
return *this;
} else {
std::destroy_at(this);
return *std::construct_at(std::launder(this), std::move(other));
}
}
// destroy
switch (type_) {
case Type::Null:
break;
case Type::Bool:
break;
case Type::Int:
break;
case Type::Double:
break;
case Type::String:
std::destroy_at(&string_v.val_);
break;
case Type::List:
std::destroy_at(&list_v.val_);
break;
case Type::Map:
std::destroy_at(&map_v.val_);
break;
case Type::TemporalData:
break;
}
// construct (no need to destroy moved from type)
auto *new_this = std::launder(this);
switch (other.type_) {
case Type::Null:
break;
case Type::Bool:
new_this->bool_v.val_ = other.bool_v.val_;
break;
case Type::Int:
new_this->int_v.val_ = other.int_v.val_;
break;
case Type::Double:
new_this->double_v.val_ = other.double_v.val_;
break;
case Type::String:
std::construct_at(&new_this->string_v.val_, std::move(other.string_v.val_));
break;
case Type::List:
std::construct_at(&new_this->list_v.val_, std::move(other.list_v.val_));
break;
case Type::Map:
std::construct_at(&new_this->map_v.val_, std::move(other.map_v.val_));
break;
case Type::TemporalData:
new_this->temporal_data_v.val_ = other.temporal_data_v.val_;
break;
}
inline void PropertyValue::DestroyValue() noexcept {
switch (std::exchange(type_, Type::Null)) {
// destructor for primitive types does nothing
case Type::Null:
case Type::Bool:
case Type::Int:
case Type::Double:
case Type::TemporalData:
return;
// destructor for non primitive types since we used placement new
case Type::String:
std::destroy_at(&string_v.val_);
return;
case Type::List:
std::destroy_at(&list_v.val_);
return;
case Type::Map:
std::destroy_at(&map_v.val_);
return;
new_this->type_ = other.type_;
return *new_this;
}
}

317
src/utils/memory.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -150,128 +150,133 @@ void *MonotonicBufferResource::DoAllocate(size_t bytes, size_t alignment) {
namespace impl {
Pool::Pool(size_t block_size, unsigned char blocks_per_chunk, MemoryResource *memory)
: blocks_per_chunk_(blocks_per_chunk), block_size_(block_size), chunks_(memory) {}
Pool::~Pool() { MG_ASSERT(chunks_.empty(), "You need to call Release before destruction!"); }
void *Pool::Allocate() {
auto allocate_block_from_chunk = [this](Chunk *chunk) {
unsigned char *available_block = chunk->data + (chunk->first_available_block_ix * block_size_);
// Update free-list pointer (index in our case) by reading "next" from the
// available_block.
chunk->first_available_block_ix = *available_block;
--chunk->blocks_available;
return available_block;
};
if (last_alloc_chunk_ && last_alloc_chunk_->blocks_available > 0U)
return allocate_block_from_chunk(last_alloc_chunk_);
// Find a Chunk with available memory.
for (auto &chunk : chunks_) {
if (chunk.blocks_available > 0U) {
last_alloc_chunk_ = &chunk;
return allocate_block_from_chunk(last_alloc_chunk_);
}
}
// We haven't found a Chunk with available memory, so allocate a new one.
if (block_size_ > std::numeric_limits<size_t>::max() / blocks_per_chunk_) throw BadAlloc("Allocation size overflow");
size_t data_size = blocks_per_chunk_ * block_size_;
Pool::Pool(size_t block_size, unsigned char blocks_per_chunk, MemoryResource *chunk_memory)
: blocks_per_chunk_(blocks_per_chunk), block_size_(block_size), chunks_(chunk_memory) {
// Use the next pow2 of block_size_ as alignment, so that we cover alignment
// requests between 1 and block_size_. Users of this class should make sure
// that requested alignment of particular blocks is never greater than the
// block itself.
size_t alignment = Ceil2(block_size_);
if (alignment < block_size_) throw BadAlloc("Allocation alignment overflow");
auto *data = reinterpret_cast<unsigned char *>(GetUpstreamResource()->Allocate(data_size, alignment));
// Form a free-list of blocks in data.
for (unsigned char i = 0U; i < blocks_per_chunk_; ++i) {
*(data + (i * block_size_)) = i + 1U;
}
Chunk chunk{data, 0, blocks_per_chunk_};
// Insert the big block in the sorted position.
auto it = std::lower_bound(chunks_.begin(), chunks_.end(), chunk,
[](const auto &a, const auto &b) { return a.data < b.data; });
try {
it = chunks_.insert(it, chunk);
} catch (...) {
GetUpstreamResource()->Deallocate(data, data_size, alignment);
throw;
}
if (block_size_ > std::numeric_limits<size_t>::max() / blocks_per_chunk_) throw BadAlloc("Allocation size overflow");
}
last_alloc_chunk_ = &*it;
last_dealloc_chunk_ = &*it;
return allocate_block_from_chunk(last_alloc_chunk_);
Pool::~Pool() {
if (!chunks_.empty()) {
auto *resource = GetUpstreamResource();
auto const dataSize = blocks_per_chunk_ * block_size_;
auto const alignment = Ceil2(block_size_);
for (auto &chunk : chunks_) {
resource->Deallocate(chunk.raw_data, dataSize, alignment);
}
chunks_.clear();
}
free_list_ = nullptr;
}
void *Pool::Allocate() {
if (!free_list_) [[unlikely]] {
// need new chunk
auto const data_size = blocks_per_chunk_ * block_size_;
auto const alignment = Ceil2(block_size_);
auto *resource = GetUpstreamResource();
auto *data = reinterpret_cast<std::byte *>(resource->Allocate(data_size, alignment));
try {
auto &new_chunk = chunks_.emplace_front(data);
free_list_ = new_chunk.build_freelist(block_size_, blocks_per_chunk_);
} catch (...) {
resource->Deallocate(data, data_size, alignment);
throw;
}
}
return std::exchange(free_list_, *reinterpret_cast<std::byte **>(free_list_));
}
void Pool::Deallocate(void *p) {
MG_ASSERT(last_dealloc_chunk_, "No chunk to deallocate");
MG_ASSERT(!chunks_.empty(),
"Expected a call to Deallocate after at least a "
"single Allocate has been done.");
auto is_in_chunk = [this, p](const Chunk &chunk) {
auto ptr = reinterpret_cast<uintptr_t>(p);
size_t data_size = blocks_per_chunk_ * block_size_;
return reinterpret_cast<uintptr_t>(chunk.data) <= ptr && ptr < reinterpret_cast<uintptr_t>(chunk.data + data_size);
};
auto deallocate_block_from_chunk = [this, p](Chunk *chunk) {
// NOTE: This check is not enough to cover all double-free issues.
MG_ASSERT(chunk->blocks_available < blocks_per_chunk_,
"Deallocating more blocks than a chunk can contain, possibly a "
"double-free situation or we have a bug in the allocator.");
// Link the block into the free-list
auto *block = reinterpret_cast<unsigned char *>(p);
*block = chunk->first_available_block_ix;
chunk->first_available_block_ix = (block - chunk->data) / block_size_;
chunk->blocks_available++;
};
if (is_in_chunk(*last_dealloc_chunk_)) {
deallocate_block_from_chunk(last_dealloc_chunk_);
return;
}
// Find the chunk which served this allocation
Chunk chunk{reinterpret_cast<unsigned char *>(p) - blocks_per_chunk_ * block_size_, 0, 0};
auto it = std::lower_bound(chunks_.begin(), chunks_.end(), chunk,
[](const auto &a, const auto &b) { return a.data <= b.data; });
MG_ASSERT(it != chunks_.end(), "Failed deallocation in utils::Pool");
MG_ASSERT(is_in_chunk(*it), "Failed deallocation in utils::Pool");
// Update last_alloc_chunk_ as well because it now has a free block.
// Additionally this corresponds with C++ pattern of allocations and
// deallocations being done in reverse order.
last_alloc_chunk_ = &*it;
last_dealloc_chunk_ = &*it;
deallocate_block_from_chunk(last_dealloc_chunk_);
// TODO: We could release the Chunk to upstream memory
}
void Pool::Release() {
for (auto &chunk : chunks_) {
size_t data_size = blocks_per_chunk_ * block_size_;
size_t alignment = Ceil2(block_size_);
GetUpstreamResource()->Deallocate(chunk.data, data_size, alignment);
}
chunks_.clear();
last_alloc_chunk_ = nullptr;
last_dealloc_chunk_ = nullptr;
*reinterpret_cast<std::byte **>(p) = std::exchange(free_list_, reinterpret_cast<std::byte *>(p));
}
} // namespace impl
PoolResource::PoolResource(size_t max_blocks_per_chunk, size_t max_block_size, MemoryResource *memory_pools,
MemoryResource *memory_unpooled)
: pools_(memory_pools),
unpooled_(memory_unpooled),
max_blocks_per_chunk_(std::min(max_blocks_per_chunk, static_cast<size_t>(impl::Pool::MaxBlocksInChunk()))),
max_block_size_(max_block_size) {
MG_ASSERT(max_blocks_per_chunk_ > 0U, "Invalid number of blocks per chunk");
MG_ASSERT(max_block_size_ > 0U, "Invalid size of block");
struct NullMemoryResourceImpl final : public MemoryResource {
NullMemoryResourceImpl() = default;
NullMemoryResourceImpl(NullMemoryResourceImpl const &) = default;
NullMemoryResourceImpl &operator=(NullMemoryResourceImpl const &) = default;
NullMemoryResourceImpl(NullMemoryResourceImpl &&) = default;
NullMemoryResourceImpl &operator=(NullMemoryResourceImpl &&) = default;
~NullMemoryResourceImpl() override = default;
private:
void *DoAllocate(size_t /*bytes*/, size_t /*alignment*/) override {
throw BadAlloc{"NullMemoryResource doesn't allocate"};
}
void DoDeallocate(void * /*p*/, size_t /*bytes*/, size_t /*alignment*/) override {
throw BadAlloc{"NullMemoryResource doesn't deallocate"};
}
bool DoIsEqual(MemoryResource const &other) const noexcept override {
return dynamic_cast<NullMemoryResourceImpl const *>(&other) != nullptr;
}
};
MemoryResource *NullMemoryResource() noexcept {
static auto res = NullMemoryResourceImpl{};
return &res;
}
namespace impl {
/// 1 bit sensitivity test
static_assert(bin_index<1>(9U) == 0);
static_assert(bin_index<1>(10U) == 0);
static_assert(bin_index<1>(11U) == 0);
static_assert(bin_index<1>(12U) == 0);
static_assert(bin_index<1>(13U) == 0);
static_assert(bin_index<1>(14U) == 0);
static_assert(bin_index<1>(15U) == 0);
static_assert(bin_index<1>(16U) == 0);
static_assert(bin_index<1>(17U) == 1);
static_assert(bin_index<1>(18U) == 1);
static_assert(bin_index<1>(19U) == 1);
static_assert(bin_index<1>(20U) == 1);
static_assert(bin_index<1>(21U) == 1);
static_assert(bin_index<1>(22U) == 1);
static_assert(bin_index<1>(23U) == 1);
static_assert(bin_index<1>(24U) == 1);
static_assert(bin_index<1>(25U) == 1);
static_assert(bin_index<1>(26U) == 1);
static_assert(bin_index<1>(27U) == 1);
static_assert(bin_index<1>(28U) == 1);
static_assert(bin_index<1>(29U) == 1);
static_assert(bin_index<1>(30U) == 1);
static_assert(bin_index<1>(31U) == 1);
static_assert(bin_index<1>(32U) == 1);
/// 2 bit sensitivity test
static_assert(bin_index<2>(9U) == 0);
static_assert(bin_index<2>(10U) == 0);
static_assert(bin_index<2>(11U) == 0);
static_assert(bin_index<2>(12U) == 0);
static_assert(bin_index<2>(13U) == 1);
static_assert(bin_index<2>(14U) == 1);
static_assert(bin_index<2>(15U) == 1);
static_assert(bin_index<2>(16U) == 1);
static_assert(bin_index<2>(17U) == 2);
static_assert(bin_index<2>(18U) == 2);
static_assert(bin_index<2>(19U) == 2);
static_assert(bin_index<2>(20U) == 2);
static_assert(bin_index<2>(21U) == 2);
static_assert(bin_index<2>(22U) == 2);
static_assert(bin_index<2>(23U) == 2);
static_assert(bin_index<2>(24U) == 2);
} // namespace impl
void *PoolResource::DoAllocate(size_t bytes, size_t alignment) {
// Take the max of `bytes` and `alignment` so that we simplify handling
// alignment requests.
size_t block_size = std::max(bytes, alignment);
size_t block_size = std::max({bytes, alignment, 1UL});
// Check that we have received a regular allocation request with non-padded
// structs/classes in play. These will always have
// `sizeof(T) % alignof(T) == 0`. Special requests which don't have that
@ -279,80 +284,36 @@ void *PoolResource::DoAllocate(size_t bytes, size_t alignment) {
// have to write a general-purpose allocator which has to behave as complex
// as malloc/free.
if (block_size % alignment != 0) throw BadAlloc("Requested bytes must be a multiple of alignment");
if (block_size > max_block_size_) {
// Allocate a big block.
BigBlock big_block{bytes, alignment, GetUpstreamResourceBlocks()->Allocate(bytes, alignment)};
// Insert the big block in the sorted position.
auto it = std::lower_bound(unpooled_.begin(), unpooled_.end(), big_block,
[](const auto &a, const auto &b) { return a.data < b.data; });
try {
unpooled_.insert(it, big_block);
} catch (...) {
GetUpstreamResourceBlocks()->Deallocate(big_block.data, bytes, alignment);
throw;
}
return big_block.data;
}
// Allocate a regular block, first check if last_alloc_pool_ is suitable.
if (last_alloc_pool_ && last_alloc_pool_->GetBlockSize() == block_size) {
return last_alloc_pool_->Allocate();
}
// Find the pool with greater or equal block_size.
impl::Pool pool(block_size, max_blocks_per_chunk_, GetUpstreamResource());
auto it = std::lower_bound(pools_.begin(), pools_.end(), pool,
[](const auto &a, const auto &b) { return a.GetBlockSize() < b.GetBlockSize(); });
if (it != pools_.end() && it->GetBlockSize() == block_size) {
last_alloc_pool_ = &*it;
last_dealloc_pool_ = &*it;
return it->Allocate();
}
// We don't have a pool for this block_size, so insert it in the sorted
// position.
it = pools_.emplace(it, std::move(pool));
last_alloc_pool_ = &*it;
last_dealloc_pool_ = &*it;
return it->Allocate();
}
if (block_size <= 64) {
return mini_pools_[(block_size - 1UL) / 8UL].Allocate();
}
if (block_size <= 128) {
return pools_3bit_.allocate(block_size);
}
if (block_size <= 512) {
return pools_4bit_.allocate(block_size);
}
if (block_size <= 1024) {
return pools_5bit_.allocate(block_size);
}
return unpooled_memory_->Allocate(bytes, alignment);
}
void PoolResource::DoDeallocate(void *p, size_t bytes, size_t alignment) {
size_t block_size = std::max(bytes, alignment);
MG_ASSERT(block_size % alignment == 0,
"PoolResource shouldn't serve allocation requests where bytes aren't "
"a multiple of alignment");
if (block_size > max_block_size_) {
// Deallocate a big block.
BigBlock big_block{bytes, alignment, p};
auto it = std::lower_bound(unpooled_.begin(), unpooled_.end(), big_block,
[](const auto &a, const auto &b) { return a.data < b.data; });
MG_ASSERT(it != unpooled_.end(), "Failed deallocation");
MG_ASSERT(it->data == p && it->bytes == bytes && it->alignment == alignment, "Failed deallocation");
unpooled_.erase(it);
GetUpstreamResourceBlocks()->Deallocate(p, bytes, alignment);
return;
size_t block_size = std::max({bytes, alignment, 1UL});
DMG_ASSERT(block_size % alignment == 0);
if (block_size <= 64) {
mini_pools_[(block_size - 1UL) / 8UL].Deallocate(p);
} else if (block_size <= 128) {
pools_3bit_.deallocate(p, block_size);
} else if (block_size <= 512) {
pools_4bit_.deallocate(p, block_size);
} else if (block_size <= 1024) {
pools_5bit_.deallocate(p, block_size);
} else {
unpooled_memory_->Deallocate(p, bytes, alignment);
}
// Deallocate a regular block, first check if last_dealloc_pool_ is suitable.
if (last_dealloc_pool_ && last_dealloc_pool_->GetBlockSize() == block_size) return last_dealloc_pool_->Deallocate(p);
// Find the pool with equal block_size.
impl::Pool pool(block_size, max_blocks_per_chunk_, GetUpstreamResource());
auto it = std::lower_bound(pools_.begin(), pools_.end(), pool,
[](const auto &a, const auto &b) { return a.GetBlockSize() < b.GetBlockSize(); });
MG_ASSERT(it != pools_.end(), "Failed deallocation");
MG_ASSERT(it->GetBlockSize() == block_size, "Failed deallocation");
last_alloc_pool_ = &*it;
last_dealloc_pool_ = &*it;
return it->Deallocate(p);
}
void PoolResource::Release() {
for (auto &pool : pools_) pool.Release();
pools_.clear();
for (auto &big_block : unpooled_)
GetUpstreamResourceBlocks()->Deallocate(big_block.data, big_block.bytes, big_block.alignment);
unpooled_.clear();
last_alloc_pool_ = nullptr;
last_dealloc_pool_ = nullptr;
}
// PoolResource END
bool PoolResource::DoIsEqual(MemoryResource const &other) const noexcept { return this == &other; }
} // namespace memgraph::utils

298
src/utils/memory.hpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -15,7 +15,11 @@
#pragma once
#include <climits>
#include <cstddef>
#include <cstdint>
#include <forward_list>
#include <list>
#include <memory>
#include <mutex>
#include <new>
@ -248,6 +252,8 @@ bool operator!=(const Allocator<T> &a, const Allocator<U> &b) {
return !(a == b);
}
auto NullMemoryResource() noexcept -> MemoryResource *;
/// Wraps std::pmr::memory_resource for use with out MemoryResource
class StdMemoryResource final : public MemoryResource {
public:
@ -380,37 +386,45 @@ class MonotonicBufferResource final : public MemoryResource {
namespace impl {
template <class T>
using AList = std::forward_list<T, Allocator<T>>;
template <class T>
using AVector = std::vector<T, Allocator<T>>;
/// Holds a number of Chunks each serving blocks of particular size. When a
/// Chunk runs out of available blocks, a new Chunk is allocated. The naming is
/// taken from `libstdc++` implementation, but the implementation details are
/// more similar to `FixedAllocator` described in "Small Object Allocation" from
/// "Modern C++ Design".
/// Chunk runs out of available blocks, a new Chunk is allocated.
class Pool final {
/// Holds a pointer into a chunk of memory which consists of equal sized
/// blocks. Each Chunk can handle `std::numeric_limits<unsigned char>::max()`
/// number of blocks. Blocks form a "free-list", where each unused block has
/// an embedded index to the next unused block.
/// blocks. Blocks form a "free-list"
struct Chunk {
unsigned char *data;
unsigned char first_available_block_ix;
unsigned char blocks_available;
// TODO: make blocks_per_chunk a per chunk thing (ie. allow chunk growth)
std::byte *raw_data;
explicit Chunk(std::byte *rawData) : raw_data(rawData) {}
std::byte *build_freelist(std::size_t block_size, std::size_t blocks_in_chunk) {
auto current = raw_data;
std::byte *prev = nullptr;
auto end = current + (blocks_in_chunk * block_size);
while (current != end) {
std::byte **list_entry = reinterpret_cast<std::byte **>(current);
*list_entry = std::exchange(prev, current);
current += block_size;
}
DMG_ASSERT(prev != nullptr);
return prev;
}
};
unsigned char blocks_per_chunk_;
size_t block_size_;
AVector<Chunk> chunks_;
Chunk *last_alloc_chunk_{nullptr};
Chunk *last_dealloc_chunk_{nullptr};
std::byte *free_list_{nullptr};
uint8_t blocks_per_chunk_{};
std::size_t block_size_{};
AList<Chunk> chunks_; // TODO: do ourself so we can do fast Release (detect monotonic, do nothing)
public:
static constexpr auto MaxBlocksInChunk() {
return std::numeric_limits<decltype(Chunk::first_available_block_ix)>::max();
}
static constexpr auto MaxBlocksInChunk = std::numeric_limits<decltype(blocks_per_chunk_)>::max();
Pool(size_t block_size, unsigned char blocks_per_chunk, MemoryResource *memory);
Pool(size_t block_size, unsigned char blocks_per_chunk, MemoryResource *chunk_memory);
Pool(const Pool &) = delete;
Pool &operator=(const Pool &) = delete;
@ -430,8 +444,147 @@ class Pool final {
void *Allocate();
void Deallocate(void *p);
};
void Release();
// C++ overloads for clz
constexpr auto clz(unsigned int x) { return __builtin_clz(x); }
constexpr auto clz(unsigned long x) { return __builtin_clzl(x); }
constexpr auto clz(unsigned long long x) { return __builtin_clzll(x); }
template <typename T>
constexpr auto bits_sizeof = sizeof(T) * CHAR_BIT;
/// 0-based bit index of the most significant bit assumed that `n` != 0
template <typename T>
constexpr auto msb_index(T n) {
return bits_sizeof<T> - clz(n) - T(1);
}
/* This function will in O(1) time provide a bin index based on:
* B - the number of most significant bits to be sensitive to
* LB - the value that should be considered below the consideration for bin index of 0 (LB is exclusive)
*
* lets say we were:
* - sensitive to two bits (B == 2)
* - lowest bin is for 8 (LB == 8)
*
* our bin indexes would look like:
* 0 - 0000'1100 12
* 1 - 0001'0000 16
* 2 - 0001'1000 24
* 3 - 0010'0000 32
* 4 - 0011'0000 48
* 5 - 0100'0000 64
* 6 - 0110'0000 96
* 7 - 1000'0000 128
* 8 - 1100'0000 192
* ...
*
* Example:
* Given n == 70, we want to return the bin index to the first value which is
* larger than n.
* bin_index<2,8>(70) => 6, as 64 (index 5) < 70 and 70 <= 96 (index 6)
*/
template <std::size_t B = 2, std::size_t LB = 8>
constexpr std::size_t bin_index(std::size_t n) {
static_assert(B >= 1U, "Needs to be sensitive to at least one bit");
static_assert(LB != 0U, "Lower bound need to be non-zero");
DMG_ASSERT(n > LB);
// We will alway be sensitive to at least the MSB
// exponent tells us how many bits we need to use to select within a level
constexpr auto kExponent = B - 1U;
// 2^exponent gives the size of each level
constexpr auto kSize = 1U << kExponent;
// offset help adjust results down to be inline with bin_index(LB) == 0
constexpr auto kOffset = msb_index(LB);
auto const msb_idx = msb_index(n);
DMG_ASSERT(msb_idx != 0);
auto const mask = (1u << msb_idx) - 1u;
auto const under = n & mask;
auto const selector = under >> (msb_idx - kExponent);
auto const rest = under & (mask >> kExponent);
auto const no_overflow = rest == 0U;
auto const msb_level = kSize * (msb_idx - kOffset);
return msb_level + selector - no_overflow;
}
// This is the inverse opperation for bin_index
// bin_size(bin_index(X)-1) < X <= bin_size(bin_index(X))
template <std::size_t B = 2, std::size_t LB = 8>
std::size_t bin_size(std::size_t idx) {
constexpr auto kExponent = B - 1U;
constexpr auto kSize = 1U << kExponent;
constexpr auto kOffset = msb_index(LB);
// no need to optimise `/` or `%` compiler can see `kSize` is a power of 2
auto const level = (idx + 1) / kSize;
auto const sub_level = (idx + 1) % kSize;
return (1U << (level + kOffset)) | (sub_level << (level + kOffset - kExponent));
}
template <std::size_t Bits, std::size_t LB, std::size_t UB>
struct MultiPool {
static_assert(LB < UB, "lower bound must be less than upper bound");
static_assert(IsPow2(LB) && IsPow2(UB), "Design untested for non powers of 2");
static_assert((LB << Bits) % sizeof(void *) == 0, "Smallest pool must have space and alignment for freelist");
// upper bound is inclusive
static bool is_size_handled(std::size_t size) { return LB < size && size <= UB; }
static bool is_above_upper_bound(std::size_t size) { return UB < size; }
static constexpr auto n_bins = bin_index<Bits, LB>(UB) + 1U;
MultiPool(uint8_t blocks_per_chunk, MemoryResource *memory, MemoryResource *internal_memory)
: blocks_per_chunk_{blocks_per_chunk}, memory_{memory}, internal_memory_{internal_memory} {}
~MultiPool() {
if (pools_) {
auto pool_alloc = Allocator<Pool>(internal_memory_);
for (auto i = 0U; i != n_bins; ++i) {
pool_alloc.destroy(&pools_[i]);
}
pool_alloc.deallocate(pools_, n_bins);
}
}
void *allocate(std::size_t bytes) {
auto idx = bin_index<Bits, LB>(bytes);
if (!pools_) [[unlikely]] {
initialise_pools();
}
return pools_[idx].Allocate();
}
void deallocate(void *ptr, std::size_t bytes) {
auto idx = bin_index<Bits, LB>(bytes);
pools_[idx].Deallocate(ptr);
}
private:
void initialise_pools() {
auto pool_alloc = Allocator<Pool>(internal_memory_);
auto pools = pool_alloc.allocate(n_bins);
try {
for (auto i = 0U; i != n_bins; ++i) {
auto block_size = bin_size<Bits, LB>(i);
pool_alloc.construct(&pools[i], block_size, blocks_per_chunk_, memory_);
}
pools_ = pools;
} catch (...) {
pool_alloc.deallocate(pools, n_bins);
throw;
}
}
Pool *pools_{};
uint8_t blocks_per_chunk_{};
MemoryResource *memory_{};
MemoryResource *internal_memory_{};
};
} // namespace impl
@ -442,8 +595,6 @@ class Pool final {
///
/// This class has the following properties with regards to memory management.
///
/// * All allocated memory will be freed upon destruction, even if Deallocate
/// has not been called for some of the allocated blocks.
/// * It consists of a collection of impl::Pool instances, each serving
/// requests for different block sizes. Each impl::Pool manages a collection
/// of impl::Pool::Chunk instances which are divided into blocks of uniform
@ -452,91 +603,46 @@ class Pool final {
/// arbitrary alignment requests. Each requested block size must be a
/// multiple of alignment or smaller than the alignment value.
/// * An allocation request within the limits of the maximum block size will
/// find a Pool serving the requested size. If there's no Pool serving such
/// a request, a new one is instantiated.
/// find a Pool serving the requested size. Some requests will share a larger
/// pool size.
/// * When a Pool exhausts its Chunk, a new one is allocated with the size for
/// the maximum number of blocks.
/// * Allocation requests which exceed the maximum block size will be
/// forwarded to upstream MemoryResource.
/// * Maximum block size and maximum number of blocks per chunk can be tuned
/// by passing the arguments to the constructor.
/// * Maximum number of blocks per chunk can be tuned by passing the
/// arguments to the constructor.
class PoolResource final : public MemoryResource {
public:
/// Construct with given max_blocks_per_chunk, max_block_size and upstream
/// memory.
///
/// The implementation will use std::min(max_blocks_per_chunk,
/// impl::Pool::MaxBlocksInChunk()) as the real maximum number of blocks per
/// chunk. Allocation requests exceeding max_block_size are simply forwarded
/// to upstream memory.
PoolResource(size_t max_blocks_per_chunk, size_t max_block_size, MemoryResource *memory_pools = NewDeleteResource(),
MemoryResource *memory_unpooled = NewDeleteResource());
PoolResource(const PoolResource &) = delete;
PoolResource &operator=(const PoolResource &) = delete;
PoolResource(PoolResource &&) = default;
PoolResource &operator=(PoolResource &&) = default;
~PoolResource() override { Release(); }
MemoryResource *GetUpstreamResource() const { return pools_.get_allocator().GetMemoryResource(); }
MemoryResource *GetUpstreamResourceBlocks() const { return unpooled_.get_allocator().GetMemoryResource(); }
/// Release all allocated memory.
void Release();
PoolResource(uint8_t blocks_per_chunk, MemoryResource *memory = NewDeleteResource(),
MemoryResource *internal_memory = NewDeleteResource())
: mini_pools_{
impl::Pool{8, blocks_per_chunk, memory},
impl::Pool{16, blocks_per_chunk, memory},
impl::Pool{24, blocks_per_chunk, memory},
impl::Pool{32, blocks_per_chunk, memory},
impl::Pool{40, blocks_per_chunk, memory},
impl::Pool{48, blocks_per_chunk, memory},
impl::Pool{56, blocks_per_chunk, memory},
impl::Pool{64, blocks_per_chunk, memory},
},
pools_3bit_(blocks_per_chunk, memory, internal_memory),
pools_4bit_(blocks_per_chunk, memory, internal_memory),
pools_5bit_(blocks_per_chunk, memory, internal_memory),
unpooled_memory_{internal_memory} {}
~PoolResource() override = default;
private:
// Big block larger than max_block_size_, doesn't go into a pool.
struct BigBlock {
size_t bytes;
size_t alignment;
void *data;
};
// TODO: Potential memory optimization is replacing `std::vector` with our
// custom vector implementation which doesn't store a `MemoryResource *`.
// Currently we have vectors for `pools_` and `unpooled_`, as well as each
// `impl::Pool` stores a `chunks_` vector.
// Pools are sorted by bound_size_, ascending.
impl::AVector<impl::Pool> pools_;
impl::Pool *last_alloc_pool_{nullptr};
impl::Pool *last_dealloc_pool_{nullptr};
// Unpooled BigBlocks are sorted by data pointer.
impl::AVector<BigBlock> unpooled_;
size_t max_blocks_per_chunk_;
size_t max_block_size_;
void *DoAllocate(size_t bytes, size_t alignment) override;
void DoDeallocate(void *p, size_t bytes, size_t alignment) override;
bool DoIsEqual(const MemoryResource &other) const noexcept override { return this == &other; }
};
/// Like PoolResource but uses SpinLock for thread safe usage.
class SynchronizedPoolResource final : public MemoryResource {
public:
SynchronizedPoolResource(size_t max_blocks_per_chunk, size_t max_block_size,
MemoryResource *memory = NewDeleteResource())
: pool_memory_(max_blocks_per_chunk, max_block_size, memory) {}
bool DoIsEqual(MemoryResource const &other) const noexcept override;
private:
PoolResource pool_memory_;
SpinLock lock_;
void *DoAllocate(size_t bytes, size_t alignment) override {
std::lock_guard<SpinLock> guard(lock_);
return pool_memory_.Allocate(bytes, alignment);
}
void DoDeallocate(void *p, size_t bytes, size_t alignment) override {
std::lock_guard<SpinLock> guard(lock_);
pool_memory_.Deallocate(p, bytes, alignment);
}
bool DoIsEqual(const MemoryResource &other) const noexcept override { return this == &other; }
std::array<impl::Pool, 8> mini_pools_;
impl::MultiPool<3, 64, 128> pools_3bit_;
impl::MultiPool<4, 128, 512> pools_4bit_;
impl::MultiPool<5, 512, 1024> pools_5bit_;
MemoryResource *unpooled_memory_;
};
class MemoryTrackingResource final : public utils::MemoryResource {

32
src/utils/tag.hpp Normal file
View File

@ -0,0 +1,32 @@
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
// License, and you may not use this file except in compliance with the Business Source License.
//
// As of the Change Date specified in that file, in accordance with
// the Business Source License, use of this software will be governed
// by the Apache License, Version 2.0, included in the file
// licenses/APL.txt.
#pragma once
namespace memgraph::utils {
template <typename T>
struct tag_type {
using type = T;
};
template <auto V>
struct tag_value {
static constexpr auto value = V;
};
template <typename T>
auto tag_t = tag_type<T>{};
template <auto V>
auto tag_v = tag_value<V>{};
} // namespace memgraph::utils

4
tests/benchmark/query/execution.cpp
View File

@ -55,12 +55,12 @@ class NewDeleteResource final {
};
class PoolResource final {
memgraph::utils::PoolResource memory_{128, 4 * 1024};
memgraph::utils::PoolResource memory_{128};
public:
memgraph::utils::MemoryResource *get() { return &memory_; }
void Reset() { memory_.Release(); }
void Reset() {}
};
static void AddVertices(memgraph::storage::Storage *db, int vertex_count) {

14
tests/benchmark/skip_list_vs_stl.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2022 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -101,8 +101,7 @@ class StdSetWithPoolAllocatorInsertFixture : public benchmark::Fixture {
}
protected:
memgraph::utils::PoolResource memory_{256U /* max_blocks_per_chunk */, 1024U /* max_block_size */,
memgraph::utils::NewDeleteResource()};
memgraph::utils::PoolResource memory_{128U /* max_blocks_per_chunk */, memgraph::utils::NewDeleteResource()};
std::set<uint64_t, std::less<>, memgraph::utils::Allocator<uint64_t>> container{&memory_};
memgraph::utils::SpinLock lock;
};
@ -208,8 +207,7 @@ class StdSetWithPoolAllocatorFindFixture : public benchmark::Fixture {
}
protected:
memgraph::utils::PoolResource memory_{256U /* max_blocks_per_chunk */, 1024U /* max_block_size */,
memgraph::utils::NewDeleteResource()};
memgraph::utils::PoolResource memory_{128U /* max_blocks_per_chunk */, memgraph::utils::NewDeleteResource()};
std::set<uint64_t, std::less<>, memgraph::utils::Allocator<uint64_t>> container{&memory_};
memgraph::utils::SpinLock lock;
};
@ -325,8 +323,7 @@ class StdMapWithPoolAllocatorInsertFixture : public benchmark::Fixture {
}
protected:
memgraph::utils::PoolResource memory_{256U /* max_blocks_per_chunk */, 1024U /* max_block_size */,
memgraph::utils::NewDeleteResource()};
memgraph::utils::PoolResource memory_{128U /* max_blocks_per_chunk */, memgraph::utils::NewDeleteResource()};
std::map<uint64_t, uint64_t, std::less<>, memgraph::utils::Allocator<std::pair<const uint64_t, uint64_t>>> container{
&memory_};
memgraph::utils::SpinLock lock;
@ -433,8 +430,7 @@ class StdMapWithPoolAllocatorFindFixture : public benchmark::Fixture {
}
protected:
memgraph::utils::PoolResource memory_{256U /* max_blocks_per_chunk */, 1024U /* max_block_size */,
memgraph::utils::NewDeleteResource()};
memgraph::utils::PoolResource memory_{128U /* max_blocks_per_chunk */, memgraph::utils::NewDeleteResource()};
std::map<uint64_t, uint64_t, std::less<>, memgraph::utils::Allocator<std::pair<const uint64_t, uint64_t>>> container{
&memory_};
memgraph::utils::SpinLock lock;

20
tests/e2e/memory/workloads.yaml
View File

@ -52,26 +52,26 @@ in_memory_query_limit_cluster: &in_memory_query_limit_cluster
setup_queries: []
validation_queries: []
args_450_MiB_limit: &args_450_MiB_limit
args_350_MiB_limit: &args_350_MiB_limit
- "--bolt-port"
- *bolt_port
- "--memory-limit=450"
- "--memory-limit=350"
- "--storage-gc-cycle-sec=180"
- "--log-level=INFO"
in_memory_450_MiB_limit_cluster: &in_memory_450_MiB_limit_cluster
in_memory_350_MiB_limit_cluster: &in_memory_350_MiB_limit_cluster
cluster:
main:
args: *args_450_MiB_limit
args: *args_350_MiB_limit
log_file: "memory-e2e.log"
setup_queries: []
validation_queries: []
disk_450_MiB_limit_cluster: &disk_450_MiB_limit_cluster
disk_350_MiB_limit_cluster: &disk_350_MiB_limit_cluster
cluster:
main:
args: *args_450_MiB_limit
args: *args_350_MiB_limit
log_file: "memory-e2e.log"
setup_queries: []
validation_queries: []
@ -192,22 +192,22 @@ workloads:
- name: "Memory control for accumulation"
binary: "tests/e2e/memory/memgraph__e2e__memory__limit_accumulation"
args: ["--bolt-port", *bolt_port]
<<: *in_memory_450_MiB_limit_cluster
<<: *in_memory_350_MiB_limit_cluster
- name: "Memory control for accumulation on disk storage"
binary: "tests/e2e/memory/memgraph__e2e__memory__limit_accumulation"
args: ["--bolt-port", *bolt_port]
<<: *disk_450_MiB_limit_cluster
<<: *disk_350_MiB_limit_cluster
- name: "Memory control for edge create"
binary: "tests/e2e/memory/memgraph__e2e__memory__limit_edge_create"
args: ["--bolt-port", *bolt_port]
<<: *in_memory_450_MiB_limit_cluster
<<: *in_memory_350_MiB_limit_cluster
- name: "Memory control for edge create on disk storage"
binary: "tests/e2e/memory/memgraph__e2e__memory__limit_edge_create"
args: ["--bolt-port", *bolt_port]
<<: *disk_450_MiB_limit_cluster
<<: *disk_350_MiB_limit_cluster
- name: "Memory control for memory limit global thread alloc"
binary: "tests/e2e/memory/memgraph__e2e__memory_limit_global_thread_alloc_proc"

2
tests/mgbench/runners.py
View File

@ -416,6 +416,7 @@ class Memgraph(BaseRunner):
def __init__(self, benchmark_context: BenchmarkContext):
super().__init__(benchmark_context=benchmark_context)
self._memgraph_binary = benchmark_context.vendor_binary
self._bolt_num_workers = benchmark_context.num_workers_for_benchmark
self._performance_tracking = benchmark_context.performance_tracking
self._directory = tempfile.TemporaryDirectory(dir=benchmark_context.temporary_directory)
self._vendor_args = benchmark_context.vendor_args
@ -440,6 +441,7 @@ class Memgraph(BaseRunner):
kwargs["bolt_port"] = self._bolt_port
kwargs["data_directory"] = data_directory
kwargs["storage_properties_on_edges"] = True
kwargs["bolt_num_workers"] = self._bolt_num_workers
for key, value in self._vendor_args.items():
kwargs[key] = value
return _convert_args_to_flags(self._memgraph_binary, **kwargs)

4
tests/unit/property_value_v2.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2022 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -570,7 +570,6 @@ TEST(PropertyValue, MoveConstructor) {
for (auto &item : data) {
memgraph::storage::PropertyValue copy(item);
memgraph::storage::PropertyValue pv(std::move(item));
ASSERT_EQ(item.type(), memgraph::storage::PropertyValue::Type::Null);
ASSERT_EQ(pv.type(), copy.type());
switch (copy.type()) {
case memgraph::storage::PropertyValue::Type::Null:
@ -668,7 +667,6 @@ TEST(PropertyValue, MoveAssignment) {
memgraph::storage::PropertyValue copy(item);
memgraph::storage::PropertyValue pv(123);
pv = std::move(item);
ASSERT_EQ(item.type(), memgraph::storage::PropertyValue::Type::Null);
ASSERT_EQ(pv.type(), copy.type());
switch (copy.type()) {
case memgraph::storage::PropertyValue::Type::Null:

130
tests/unit/utils_memory.cpp
View File

@ -1,4 +1,4 @@
// Copyright 2023 Memgraph Ltd.
// Copyright 2024 Memgraph Ltd.
//
// Use of this software is governed by the Business Source License
// included in the file licenses/BSL.txt; by using this file, you agree to be bound by the terms of the Business Source
@ -195,134 +195,6 @@ TEST(MonotonicBufferResource, AllocationWithInitialBufferOnStack) {
}
}
// NOLINTNEXTLINE(hicpp-special-member-functions)
TEST(PoolResource, SingleSmallBlockAllocations) {
TestMemory test_mem;
const size_t max_blocks_per_chunk = 3U;
const size_t max_block_size = 64U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, max_block_size, &test_mem);
// Fill the first chunk.
CheckAllocation(&mem, 64U, 1U);
// May allocate more than once due to bookkeeping.
EXPECT_GE(test_mem.new_count_, 1U);
// Reset tracking and continue filling the first chunk.
test_mem.new_count_ = 0U;
CheckAllocation(&mem, 64U, 64U);
CheckAllocation(&mem, 64U);
EXPECT_EQ(test_mem.new_count_, 0U);
// Reset tracking and fill the second chunk
test_mem.new_count_ = 0U;
CheckAllocation(&mem, 64U, 32U);
auto *ptr1 = CheckAllocation(&mem, 32U, 64U); // this will become 64b block
auto *ptr2 = CheckAllocation(&mem, 64U, 32U);
// We expect one allocation for chunk and at most one for bookkeeping.
EXPECT_TRUE(test_mem.new_count_ >= 1U && test_mem.new_count_ <= 2U);
test_mem.delete_count_ = 0U;
mem.Deallocate(ptr1, 32U, 64U);
mem.Deallocate(ptr2, 64U, 32U);
EXPECT_EQ(test_mem.delete_count_, 0U);
mem.Release();
EXPECT_GE(test_mem.delete_count_, 2U);
CheckAllocation(&mem, 64U, 1U);
}
// NOLINTNEXTLINE(hicpp-special-member-functions)
TEST(PoolResource, MultipleSmallBlockAllocations) {
TestMemory test_mem;
const size_t max_blocks_per_chunk = 1U;
const size_t max_block_size = 64U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, max_block_size, &test_mem);
CheckAllocation(&mem, 64U);
CheckAllocation(&mem, 18U, 2U);
CheckAllocation(&mem, 24U, 8U);
// May allocate more than once per chunk due to bookkeeping.
EXPECT_GE(test_mem.new_count_, 3U);
// Reset tracking and fill the second chunk
test_mem.new_count_ = 0U;
CheckAllocation(&mem, 64U);
CheckAllocation(&mem, 18U, 2U);
CheckAllocation(&mem, 24U, 8U);
// We expect one allocation for chunk and at most one for bookkeeping.
EXPECT_TRUE(test_mem.new_count_ >= 3U && test_mem.new_count_ <= 6U);
mem.Release();
EXPECT_GE(test_mem.delete_count_, 6U);
CheckAllocation(&mem, 64U);
}
// NOLINTNEXTLINE(hicpp-special-member-functions)
TEST(PoolResource, BigBlockAllocations) {
TestMemory test_mem;
TestMemory test_mem_unpooled;
const size_t max_blocks_per_chunk = 3U;
const size_t max_block_size = 64U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, max_block_size, &test_mem, &test_mem_unpooled);
CheckAllocation(&mem, max_block_size + 1, 1U);
// May allocate more than once per block due to bookkeeping.
EXPECT_GE(test_mem_unpooled.new_count_, 1U);
CheckAllocation(&mem, max_block_size + 1, 1U);
EXPECT_GE(test_mem_unpooled.new_count_, 2U);
auto *ptr = CheckAllocation(&mem, max_block_size * 2, 1U);
EXPECT_GE(test_mem_unpooled.new_count_, 3U);
mem.Deallocate(ptr, max_block_size * 2, 1U);
EXPECT_GE(test_mem_unpooled.delete_count_, 1U);
mem.Release();
EXPECT_GE(test_mem_unpooled.delete_count_, 3U);
CheckAllocation(&mem, max_block_size + 1, 1U);
}
// NOLINTNEXTLINE(hicpp-special-member-functions)
TEST(PoolResource, BlockSizeIsNotMultipleOfAlignment) {
const size_t max_blocks_per_chunk = 3U;
const size_t max_block_size = 64U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, max_block_size);
EXPECT_THROW(mem.Allocate(64U, 24U), std::bad_alloc);
EXPECT_THROW(mem.Allocate(63U), std::bad_alloc);
EXPECT_THROW(mem.Allocate(max_block_size + 1, max_block_size), std::bad_alloc);
}
// NOLINTNEXTLINE(hicpp-special-member-functions)
TEST(PoolResource, AllocationWithOverflow) {
{
const size_t max_blocks_per_chunk = 2U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, std::numeric_limits<size_t>::max());
EXPECT_THROW(mem.Allocate(std::numeric_limits<size_t>::max(), 1U), std::bad_alloc);
// Throws because initial chunk block is aligned to
// memgraph::utils::Ceil2(block_size), which wraps in this case.
EXPECT_THROW(mem.Allocate((std::numeric_limits<size_t>::max() - 1U) / max_blocks_per_chunk, 1U), std::bad_alloc);
}
{
const size_t max_blocks_per_chunk = memgraph::utils::impl::Pool::MaxBlocksInChunk();
memgraph::utils::PoolResource mem(max_blocks_per_chunk, std::numeric_limits<size_t>::max());
EXPECT_THROW(mem.Allocate(std::numeric_limits<size_t>::max(), 1U), std::bad_alloc);
// Throws because initial chunk block is aligned to
// memgraph::utils::Ceil2(block_size), which wraps in this case.
EXPECT_THROW(mem.Allocate((std::numeric_limits<size_t>::max() - 1U) / max_blocks_per_chunk, 1U), std::bad_alloc);
}
}
TEST(PoolResource, BlockDeallocation) {
TestMemory test_mem;
const size_t max_blocks_per_chunk = 2U;
const size_t max_block_size = 64U;
memgraph::utils::PoolResource mem(max_blocks_per_chunk, max_block_size, &test_mem);
auto *ptr = CheckAllocation(&mem, max_block_size);
test_mem.new_count_ = 0U;
// Do another allocation before deallocating `ptr`, so that we are sure that
// the chunk of 2 blocks is still alive and therefore `ptr` may be reused when
// it's deallocated. If we deallocate now, the implementation may choose to
// free the whole chunk, and we do not want that for the purposes of this
// test.
CheckAllocation(&mem, max_block_size);
EXPECT_EQ(test_mem.new_count_, 0U);
EXPECT_EQ(test_mem.delete_count_, 0U);
mem.Deallocate(ptr, max_block_size);
EXPECT_EQ(test_mem.delete_count_, 0U);
// CheckAllocation(&mem, max_block_size) will fail as PoolResource should
// reuse free blocks.
EXPECT_EQ(ptr, mem.Allocate(max_block_size));
EXPECT_EQ(test_mem.new_count_, 0U);
}
class AllocationTrackingMemory final : public memgraph::utils::MemoryResource {
public:
std::vector<size_t> allocated_sizes_;