Skip to content

reading indexUpdateScheduled under the lock guard #813

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 27, 2025
Merged

reading indexUpdateScheduled under the lock guard #813

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0374363
reading indexUpdateScheduled under the lock
meiravgri Oct 26, 2025
6d446ec
dont lock twice
meiravgri Oct 27, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
17 changes: 10 additions & 7 deletions src/VecSim/algorithms/svs/svs_tiered.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -874,14 +874,17 @@ class TieredSVSIndex : public VecSimTieredIndex<DataType, float> {
VecSimIndexDebugInfo debugInfo() const override {
auto info = Base::debugInfo();

SvsTieredInfo svsTieredInfo = {.trainingTriggerThreshold = this->trainingTriggerThreshold,
.updateTriggerThreshold = this->updateTriggerThreshold,
.updateJobWaitTime = this->updateJobWaitTime,
.indexUpdateScheduled =
static_cast<bool>(this->indexUpdateScheduled.test())};
SvsTieredInfo svsTieredInfo = {
.trainingTriggerThreshold = this->trainingTriggerThreshold,
.updateTriggerThreshold = this->updateTriggerThreshold,
.updateJobWaitTime = this->updateJobWaitTime,
};
{
std::lock_guard<std::mutex> lock(this->updateJobMutex);
svsTieredInfo.indexUpdateScheduled =
this->indexUpdateScheduled.test() == VecSimBool_TRUE;
}
info.tieredInfo.specificTieredBackendInfo.svsTieredInfo = svsTieredInfo;
// prevent parallel updates
std::lock_guard<std::mutex> lock(this->updateJobMutex);
info.tieredInfo.backgroundIndexing =
svsTieredInfo.indexUpdateScheduled && info.tieredInfo.frontendCommonInfo.indexSize > 0
? VecSimBool_TRUE
Expand Down
79 changes: 32 additions & 47 deletions tests/unit/test_svs_tiered.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -500,9 +500,11 @@ TYPED_TEST(SVSTieredIndexTest, CreateIndexInstance) {
ASSERT_EQ(tiered_index->GetBackendIndex()->indexSize(), 1);
}

TYPED_TEST(SVSTieredIndexTest, addVector) {
TYPED_TEST(SVSTieredIndexTest, background_indexing_check) {
// Create TieredSVS index instance with a mock queue.
size_t dim = 4;
size_t dim = 2;
constexpr size_t training_th = DEFAULT_BLOCK_SIZE;
constexpr size_t update_th = DEFAULT_BLOCK_SIZE;
SVSParams params = {.type = TypeParam::get_index_type(),
.dim = dim,
.metric = VecSimMetric_L2,
Expand All @@ -511,59 +513,42 @@ TYPED_TEST(SVSTieredIndexTest, addVector) {

auto mock_thread_pool = tieredIndexMock();

auto tiered_params = this->CreateTieredSVSParams(svs_params, mock_thread_pool, 1, 1);
auto tiered_params =
this->CreateTieredSVSParams(svs_params, mock_thread_pool, training_th, update_th);
auto *tiered_index = this->CreateTieredSVSIndex(tiered_params, mock_thread_pool);
ASSERT_INDEX(tiered_index);

// Get the allocator from the tiered index.
auto allocator = tiered_index->getAllocator();
mock_thread_pool.init_threads();

BFParams bf_params = {.type = TypeParam::get_index_type(),
.dim = dim,
.metric = VecSimMetric_L2,
.multi = TypeParam::isMulti()};
for (size_t i = 0; i < training_th; i++) {
TEST_DATA_T vector[dim];
GenerateVector<TEST_DATA_T>(vector, dim, i);
VecSimIndex_AddVector(tiered_index, vector, i);
}

size_t expected_mem = TieredFactory::EstimateInitialSize(&tiered_params);
ASSERT_LE(expected_mem, tiered_index->getAllocationSize());
ASSERT_GE(expected_mem * 1.02, tiered_index->getAllocationSize());
ASSERT_EQ(mock_thread_pool.jobQ.size(), 0);
while (tiered_index->debugInfo().tieredInfo.backgroundIndexing != VecSimBool_FALSE) {
std::this_thread::sleep_for(std::chrono::milliseconds(100));
}

// Create a vector and add it to the tiered index.
labelType vec_label = 1;
TEST_DATA_T vector[dim];
GenerateVector<TEST_DATA_T>(vector, dim, vec_label);
VecSimIndex_AddVector(tiered_index, vector, vec_label);
// Validate that the vector was inserted to the flat buffer properly.
ASSERT_EQ(tiered_index->indexSize(), 1);
ASSERT_EQ(tiered_index->GetFlatIndex()->indexSize(), 1);
ASSERT_EQ(tiered_index->GetBackendIndex()->indexSize(), 0);
ASSERT_EQ(tiered_index->GetFlatIndex()->indexCapacity(), DEFAULT_BLOCK_SIZE);
ASSERT_EQ(tiered_index->indexCapacity(), DEFAULT_BLOCK_SIZE);
ASSERT_EQ(tiered_index->GetFlatIndex()->getDistanceFrom_Unsafe(vec_label, vector), 0);
ASSERT_EQ(mock_thread_pool.jobQ.size(), mock_thread_pool.thread_pool_size);
ASSERT_EQ(tiered_index->GetBackendIndex()->indexSize(), training_th);
ASSERT_EQ(tiered_index->GetFlatIndex()->indexSize(), 0);
ASSERT_EQ(tiered_index->indexSize(), training_th);

// Account for the allocation of a new block due to the vector insertion.
expected_mem += (BruteForceFactory::EstimateElementSize(&bf_params)) * DEFAULT_BLOCK_SIZE;
// Account for the memory that was allocated in the labelToId map (approx.)
expected_mem += sizeof(vecsim_stl::unordered_map<labelType, idType>::value_type) +
sizeof(void *) + sizeof(size_t);
// Account for the insert job that was created.
expected_mem +=
SVSMultiThreadJob::estimateSize(mock_thread_pool.thread_pool_size) + sizeof(size_t);
auto actual_mem = tiered_index->getAllocationSize();
ASSERT_GE(expected_mem * 1.02, tiered_index->getAllocationSize());
ASSERT_LE(expected_mem, tiered_index->getAllocationSize());

if constexpr (TypeParam::isMulti()) {
// Add another vector under the same label
VecSimIndex_AddVector(tiered_index, vector, vec_label);
ASSERT_EQ(tiered_index->indexSize(), 2);
ASSERT_EQ(tiered_index->indexLabelCount(), 1);
ASSERT_EQ(tiered_index->GetBackendIndex()->indexSize(), 0);
ASSERT_EQ(tiered_index->GetFlatIndex()->indexSize(), 2);
// Validate that there still 1 update jobs set
ASSERT_EQ(mock_thread_pool.jobQ.size(), mock_thread_pool.thread_pool_size);
constexpr size_t second_batch = 2500;

for (size_t i = 0; i < second_batch; i++) {
TEST_DATA_T vector[dim];
GenerateVector<TEST_DATA_T>(vector, dim, i);
VecSimIndex_AddVector(tiered_index, vector, training_th + i);
}

while (tiered_index->debugInfo().tieredInfo.backgroundIndexing != VecSimBool_FALSE) {
std::this_thread::sleep_for(std::chrono::milliseconds(100));
}

ASSERT_GT(tiered_index->GetBackendIndex()->indexSize(), training_th + second_batch / update_th);
ASSERT_LT(tiered_index->GetFlatIndex()->indexSize(), update_th);
ASSERT_EQ(tiered_index->indexSize(), second_batch + training_th);
}

TYPED_TEST(SVSTieredIndexTest, insertJob) {
Expand Down
Loading