You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@doris.apache.org by pa...@apache.org on 2023/10/26 09:23:16 UTC
[doris] 04/10: update dev_join to pre calculate bucket num (#25663)
This is an automated email from the ASF dual-hosted git repository.
panxiaolei pushed a commit to branch dev_join
in repository https://gitbox.apache.org/repos/asf/doris.git
commit d9ea1a1723ed0ec5a2ed9ab711645e1f9d7ad76a
Author: Pxl <px...@qq.com>
AuthorDate: Fri Oct 20 11:50:08 2023 +0800
update dev_join to pre calculate bucket num (#25663)
---
be/src/vec/common/hash_table/hash_map.h | 51 +++++++++++----------
be/src/vec/common/hash_table/hash_map_context.h | 8 ++++
be/src/vec/exec/join/process_hash_table_probe.h | 6 ---
.../vec/exec/join/process_hash_table_probe_impl.h | 53 ++--------------------
be/src/vec/exec/join/vhash_join_node.h | 2 +
5 files changed, 39 insertions(+), 81 deletions(-)
diff --git a/be/src/vec/common/hash_table/hash_map.h b/be/src/vec/common/hash_table/hash_map.h
index 53cb01dbfaa..89dfe7f8aac 100644
--- a/be/src/vec/common/hash_table/hash_map.h
+++ b/be/src/vec/common/hash_table/hash_map.h
@@ -216,18 +216,23 @@ public:
return phmap::priv::NormalizeCapacity(expect_bucket_size) + 1;
}
+ void reserve(int num_elem) {
+ bucket_size = calc_bucket_size(num_elem + 1);
+ first.resize(bucket_size, 0);
+ next.resize(num_elem);
+ }
+
void build(const Key* __restrict keys, const size_t* __restrict hash_values, size_t num_elem,
int batch_size) {
- max_batch_size = batch_size;
- bucket_size = calc_bucket_size(num_elem + 1);
+ _batch_size = batch_size;
+ bucket_size = calc_bucket_size(num_elem);
first.resize(bucket_size, 0);
next.resize(num_elem);
build_keys = keys;
for (size_t i = 1; i < num_elem; i++) {
- uint32_t bucket_num = hash_values[i] & (bucket_size - 1);
- next[i] = first[bucket_num];
- first[bucket_num] = i;
+ next[i] = first[hash_values[i]];
+ first[hash_values[i]] = i;
}
}
@@ -248,18 +253,16 @@ public:
return std::pair {0, 0};
}
+ size_t get_bucket_mask() { return bucket_size - 1; }
+
private:
template <int JoinOpType>
auto _find_batch_left_semi_anti(const Key* __restrict keys,
const size_t* __restrict hash_values, int probe_idx,
int probe_rows, std::vector<uint32_t>& probe_idxs) {
- auto matched_cnt = 0;
- const auto batch_size = max_batch_size;
-
- while (LIKELY(probe_idx < probe_rows && matched_cnt < batch_size)) {
- uint32_t bucket_num = hash_values[probe_idx] & (bucket_size - 1);
- auto build_idx = first[bucket_num];
-
+ int matched_cnt = 0;
+ while (LIKELY(probe_idx < probe_rows && matched_cnt < _batch_size)) {
+ uint32_t build_idx = first[hash_values[probe_idx]];
while (build_idx) {
if (keys[probe_idx] == build_keys[build_idx]) {
break;
@@ -279,12 +282,11 @@ private:
const size_t* __restrict hash_values, int probe_idx,
int probe_rows, std::vector<uint32_t>& probe_idxs,
std::vector<uint32_t>& build_idxs) {
- auto matched_cnt = 0;
- const auto batch_size = max_batch_size;
+ int matched_cnt = 0;
uint32_t build_idx = 0;
auto do_the_probe = [&]() {
- while (build_idx && LIKELY(matched_cnt < batch_size)) {
+ while (build_idx && LIKELY(matched_cnt < _batch_size)) {
if (keys[probe_idx] == build_keys[build_idx]) {
probe_idxs[matched_cnt] = probe_idx;
build_idxs[matched_cnt] = build_idx;
@@ -302,12 +304,7 @@ private:
}
}
- if (matched_cnt == max_batch_size && build_idx) {
- current_probe_idx = probe_idx;
- current_build_idx = build_idx;
- } else {
- probe_idx++;
- }
+ probe_idx++;
};
// some row over the batch_size, need dispose first
@@ -317,17 +314,21 @@ private:
current_build_idx = 0;
do_the_probe();
}
- while (LIKELY(probe_idx < probe_rows && matched_cnt < batch_size)) {
- uint32_t bucket_num = hash_values[probe_idx] & (bucket_size - 1);
- build_idx = first[bucket_num];
+ while (LIKELY(probe_idx < probe_rows && matched_cnt < _batch_size)) {
+ build_idx = first[hash_values[probe_idx]];
do_the_probe();
}
+
+ if (matched_cnt == _batch_size && build_idx) {
+ current_probe_idx = probe_idx - 1;
+ current_build_idx = build_idx;
+ }
return std::pair {probe_idx, matched_cnt};
}
const Key* __restrict build_keys;
uint32_t bucket_size = 0;
- int max_batch_size = 0;
+ int _batch_size = 0;
int current_probe_idx = -1;
uint32_t current_build_idx = 0;
diff --git a/be/src/vec/common/hash_table/hash_map_context.h b/be/src/vec/common/hash_table/hash_map_context.h
index 3664151c66a..6e8f9e0746c 100644
--- a/be/src/vec/common/hash_table/hash_map_context.h
+++ b/be/src/vec/common/hash_table/hash_map_context.h
@@ -85,6 +85,7 @@ struct MethodBase {
hash_values[k] = hash_table->hash(keys[k]);
}
}
+
void init_hash_values(size_t num_rows) {
hash_values.resize(num_rows);
for (size_t k = 0; k < num_rows; ++k) {
@@ -92,6 +93,13 @@ struct MethodBase {
}
}
+ void calculate_bucket(size_t num_rows) {
+ size_t mask = hash_table->get_bucket_mask();
+ for (int i = 0; i < num_rows; i++) {
+ hash_values[i] &= mask;
+ }
+ }
+
template <bool read>
void prefetch(int current) {
if (LIKELY(current + HASH_MAP_PREFETCH_DIST < hash_values.size())) {
diff --git a/be/src/vec/exec/join/process_hash_table_probe.h b/be/src/vec/exec/join/process_hash_table_probe.h
index eb1f6b78df5..ebd63f9f55d 100644
--- a/be/src/vec/exec/join/process_hash_table_probe.h
+++ b/be/src/vec/exec/join/process_hash_table_probe.h
@@ -75,17 +75,11 @@ struct ProcessHashTableProbe {
UInt8* __restrict null_map_data,
UInt8* __restrict filter_map, Block* output_block);
- void _emplace_element(int32_t block_row, int& current_offset);
-
template <typename HashTableType>
typename HashTableType::State _init_probe_side(HashTableType& hash_table_ctx, size_t probe_rows,
bool with_other_join_conjuncts,
const uint8_t* null_map);
- template <typename Mapped, bool with_other_join_conjuncts>
- ForwardIterator<Mapped>& _probe_row_match(int& current_offset, int& probe_index,
- size_t& probe_size, bool& all_match_one);
-
// Process full outer join/ right join / right semi/anti join to output the join result
// in hash table
template <typename HashTableType>
diff --git a/be/src/vec/exec/join/process_hash_table_probe_impl.h b/be/src/vec/exec/join/process_hash_table_probe_impl.h
index 3c6ab25c21d..f4bef996911 100644
--- a/be/src/vec/exec/join/process_hash_table_probe_impl.h
+++ b/be/src/vec/exec/join/process_hash_table_probe_impl.h
@@ -137,52 +137,18 @@ typename HashTableType::State ProcessHashTableProbe<JoinOpType, Parent>::_init_p
_visited_map.reserve(_batch_size * PROBE_SIDE_EXPLODE_RATE);
_same_to_prev.reserve(_batch_size * PROBE_SIDE_EXPLODE_RATE);
}
- _probe_indexs.reserve(_batch_size * PROBE_SIDE_EXPLODE_RATE);
- _build_block_rows.reserve(_batch_size * PROBE_SIDE_EXPLODE_RATE);
+ _probe_indexs.resize(_batch_size);
+ _build_block_rows.resize(_batch_size);
if (!_parent->_ready_probe) {
_parent->_ready_probe = true;
hash_table_ctx.reset();
hash_table_ctx.init_serialized_keys(_parent->_probe_columns, probe_rows, null_map);
+ hash_table_ctx.calculate_bucket(probe_rows);
}
return typename HashTableType::State(_parent->_probe_columns);
}
-template <int JoinOpType, typename Parent>
-template <typename Mapped, bool with_other_join_conjuncts>
-ForwardIterator<Mapped>& ProcessHashTableProbe<JoinOpType, Parent>::_probe_row_match(
- int& current_offset, int& probe_index, size_t& probe_size, bool& all_match_one) {
- auto& probe_row_match_iter = std::get<ForwardIterator<Mapped>>(_parent->_probe_row_match_iter);
- if (!probe_row_match_iter.ok()) {
- return probe_row_match_iter;
- }
-
- SCOPED_TIMER(_search_hashtable_timer);
- for (; probe_row_match_iter.ok() && current_offset < _batch_size; ++probe_row_match_iter) {
- _emplace_element(probe_row_match_iter->row_num, current_offset);
- _probe_indexs.emplace_back(probe_index);
- if constexpr (with_other_join_conjuncts) {
- _visited_map.emplace_back(&probe_row_match_iter->visited);
- }
- }
-
- _row_count_from_last_probe = current_offset;
- all_match_one &= (current_offset == 1);
- if (!probe_row_match_iter.ok()) {
- ++probe_index;
- }
- probe_size = 1;
-
- return probe_row_match_iter;
-}
-
-template <int JoinOpType, typename Parent>
-void ProcessHashTableProbe<JoinOpType, Parent>::_emplace_element(int32_t block_row,
- int& current_offset) {
- _build_block_rows.emplace_back(block_row);
- current_offset++;
-}
-
template <int JoinOpType, typename Parent>
template <bool need_null_map_for_probe, bool ignore_null, typename HashTableType,
bool with_other_conjuncts, bool is_mark_join>
@@ -193,8 +159,6 @@ Status ProcessHashTableProbe<JoinOpType, Parent>::do_process(HashTableType& hash
size_t probe_rows) {
auto& probe_index = _parent->_probe_index;
- using Mapped = typename HashTableType::Mapped;
-
_init_probe_side<HashTableType>(hash_table_ctx, probe_rows, with_other_conjuncts,
need_null_map_for_probe ? null_map->data() : nullptr);
@@ -206,9 +170,6 @@ Status ProcessHashTableProbe<JoinOpType, Parent>::do_process(HashTableType& hash
bool all_match_one = false;
size_t probe_size = 0;
- auto& probe_row_match_iter = _probe_row_match<Mapped, with_other_conjuncts>(
- current_offset, probe_index, probe_size, all_match_one);
-
// If not(which means it excceed batch size), probe_index is not increased and
// remaining matched rows for the current probe row will be
// handled in the next call of this function
@@ -217,14 +178,6 @@ Status ProcessHashTableProbe<JoinOpType, Parent>::do_process(HashTableType& hash
// Is the last sub block of splitted block
bool is_the_last_sub_block = false;
- if (with_other_conjuncts && probe_size != 0) {
- is_the_last_sub_block = !probe_row_match_iter.ok();
- _same_to_prev.emplace_back(false);
- for (int i = 0; i < current_offset - 1; ++i) {
- _same_to_prev.emplace_back(true);
- }
- }
-
std::unique_ptr<ColumnFilterHelper> mark_column;
if (is_mark_join) {
mark_column = std::make_unique<ColumnFilterHelper>(*mcol[mcol.size() - 1]);
diff --git a/be/src/vec/exec/join/vhash_join_node.h b/be/src/vec/exec/join/vhash_join_node.h
index ee941fa5c1b..c15e3674642 100644
--- a/be/src/vec/exec/join/vhash_join_node.h
+++ b/be/src/vec/exec/join/vhash_join_node.h
@@ -131,8 +131,10 @@ struct ProcessHashTableBuild {
if (!_parent->runtime_filter_descs().empty()) {
_parent->_inserted_blocks.insert(&_acquired_block);
}
+ hash_table_ctx.hash_table->reserve(_rows);
hash_table_ctx.init_serialized_keys(_build_raw_ptrs, _rows,
null_map ? null_map->data() : nullptr);
+ hash_table_ctx.calculate_bucket(_rows);
SCOPED_TIMER(_parent->_build_table_insert_timer);
hash_table_ctx.hash_table->build(hash_table_ctx.keys, hash_table_ctx.hash_values.data(),
_rows, _state->batch_size());
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@doris.apache.org
For additional commands, e-mail: commits-help@doris.apache.org