/root/doris/be/src/exec/partitioner/partitioner.h
Line | Count | Source |
1 | | // Licensed to the Apache Software Foundation (ASF) under one |
2 | | // or more contributor license agreements. See the NOTICE file |
3 | | // distributed with this work for additional information |
4 | | // regarding copyright ownership. The ASF licenses this file |
5 | | // to you under the Apache License, Version 2.0 (the |
6 | | // "License"); you may not use this file except in compliance |
7 | | // with the License. You may obtain a copy of the License at |
8 | | // |
9 | | // http://www.apache.org/licenses/LICENSE-2.0 |
10 | | // |
11 | | // Unless required by applicable law or agreed to in writing, |
12 | | // software distributed under the License is distributed on an |
13 | | // "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY |
14 | | // KIND, either express or implied. See the License for the |
15 | | // specific language governing permissions and limitations |
16 | | // under the License. |
17 | | |
18 | | #pragma once |
19 | | |
20 | | #include <algorithm> |
21 | | |
22 | | #include "core/block/block.h" |
23 | | #include "exprs/vexpr.h" |
24 | | #include "exprs/vexpr_context.h" |
25 | | |
26 | | namespace doris { |
27 | | #include "common/compile_check_begin.h" |
28 | | |
29 | | class PartitionerBase { |
30 | | public: |
31 | | using HashValType = uint32_t; |
32 | | |
33 | 56 | PartitionerBase(HashValType partition_count) : _partition_count(partition_count) {} |
34 | 56 | virtual ~PartitionerBase() = default; |
35 | | |
36 | | virtual Status init(const std::vector<TExpr>& texprs) = 0; |
37 | | |
38 | | virtual Status prepare(RuntimeState* state, const RowDescriptor& row_desc) = 0; |
39 | | |
40 | | virtual Status open(RuntimeState* state) = 0; |
41 | | |
42 | | virtual Status close(RuntimeState* state) = 0; |
43 | | |
44 | | virtual Status do_partitioning(RuntimeState* state, Block* block) const = 0; |
45 | | |
46 | | virtual const std::vector<HashValType>& get_channel_ids() const = 0; |
47 | | |
48 | | virtual Status clone(RuntimeState* state, std::unique_ptr<PartitionerBase>& partitioner) = 0; |
49 | | |
50 | | // use _partition_count as invalid sentinel value. since modulo operation result is [0, partition_count-1] |
51 | 6 | HashValType partition_count() const { return _partition_count; } |
52 | | // use a individual function to highlight its special meaning |
53 | 4 | HashValType invalid_sentinel() const { return partition_count(); } |
54 | | |
55 | | protected: |
56 | | const HashValType _partition_count; |
57 | | }; |
58 | | |
59 | | enum class ShuffleHashMethod { |
60 | | CRC32, |
61 | | CRC32C, |
62 | | }; |
63 | | |
64 | | class PartitionFunction { |
65 | | public: |
66 | | using HashValType = PartitionerBase::HashValType; |
67 | | |
68 | 10 | virtual ~PartitionFunction() = default; |
69 | | |
70 | | virtual Status init(const std::vector<TExpr>& texprs) = 0; |
71 | | |
72 | | virtual Status prepare(RuntimeState* state, const RowDescriptor& row_desc) = 0; |
73 | | |
74 | | virtual Status open(RuntimeState* state) = 0; |
75 | | |
76 | | virtual Status close(RuntimeState* state) = 0; |
77 | | |
78 | | virtual Status get_partitions(RuntimeState* state, Block* block, size_t partition_count, |
79 | | std::vector<HashValType>& partitions) const = 0; |
80 | | |
81 | | virtual HashValType partition_count() const = 0; |
82 | | |
83 | | virtual Status clone(RuntimeState* state, |
84 | | std::unique_ptr<PartitionFunction>& function) const = 0; |
85 | | }; |
86 | | |
87 | | // Adapts the standard Doris expression hash partitioner to the composable |
88 | | // PartitionFunction interface used by sink routing. |
89 | | class HashPartitionFunction final : public PartitionFunction { |
90 | | public: |
91 | | HashPartitionFunction(HashValType partition_count, ShuffleHashMethod hash_method); |
92 | | |
93 | | Status init(const std::vector<TExpr>& texprs) override; |
94 | | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override; |
95 | | Status open(RuntimeState* state) override; |
96 | | Status close(RuntimeState* state) override; |
97 | | Status get_partitions(RuntimeState* state, Block* block, size_t partition_count, |
98 | | std::vector<HashValType>& partitions) const override; |
99 | 0 | HashValType partition_count() const override { return _partition_count; } |
100 | | Status clone(RuntimeState* state, std::unique_ptr<PartitionFunction>& function) const override; |
101 | | |
102 | | private: |
103 | | HashValType _partition_count; |
104 | | ShuffleHashMethod _hash_method; |
105 | | std::unique_ptr<PartitionerBase> _partitioner; |
106 | | }; |
107 | | |
108 | | template <typename ChannelIds> |
109 | | class Crc32HashPartitioner : public PartitionerBase { |
110 | | public: |
111 | 37 | Crc32HashPartitioner(int partition_count) : PartitionerBase(partition_count) {}_ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEEC2Ei Line | Count | Source | 111 | 19 | Crc32HashPartitioner(int partition_count) : PartitionerBase(partition_count) {} |
_ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEEC2Ei Line | Count | Source | 111 | 7 | Crc32HashPartitioner(int partition_count) : PartitionerBase(partition_count) {} |
_ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEEC2Ei Line | Count | Source | 111 | 1 | Crc32HashPartitioner(int partition_count) : PartitionerBase(partition_count) {} |
_ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEEC2Ei Line | Count | Source | 111 | 10 | Crc32HashPartitioner(int partition_count) : PartitionerBase(partition_count) {} |
|
112 | 37 | ~Crc32HashPartitioner() override = default; _ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEED2Ev Line | Count | Source | 112 | 19 | ~Crc32HashPartitioner() override = default; |
_ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEED2Ev Line | Count | Source | 112 | 7 | ~Crc32HashPartitioner() override = default; |
_ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEED2Ev Line | Count | Source | 112 | 1 | ~Crc32HashPartitioner() override = default; |
_ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEED2Ev Line | Count | Source | 112 | 10 | ~Crc32HashPartitioner() override = default; |
|
113 | | |
114 | 22 | Status init(const std::vector<TExpr>& texprs) override { |
115 | 22 | return VExpr::create_expr_trees(texprs, _partition_expr_ctxs); |
116 | 22 | } _ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE4initERKSt6vectorINS_5TExprESaIS4_EE Line | Count | Source | 114 | 1 | Status init(const std::vector<TExpr>& texprs) override { | 115 | 1 | return VExpr::create_expr_trees(texprs, _partition_expr_ctxs); | 116 | 1 | } |
_ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE4initERKSt6vectorINS_5TExprESaIS4_EE Line | Count | Source | 114 | 14 | Status init(const std::vector<TExpr>& texprs) override { | 115 | 14 | return VExpr::create_expr_trees(texprs, _partition_expr_ctxs); | 116 | 14 | } |
_ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE4initERKSt6vectorINS_5TExprESaIS4_EE Line | Count | Source | 114 | 1 | Status init(const std::vector<TExpr>& texprs) override { | 115 | 1 | return VExpr::create_expr_trees(texprs, _partition_expr_ctxs); | 116 | 1 | } |
_ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE4initERKSt6vectorINS_5TExprESaIS4_EE Line | Count | Source | 114 | 6 | Status init(const std::vector<TExpr>& texprs) override { | 115 | 6 | return VExpr::create_expr_trees(texprs, _partition_expr_ctxs); | 116 | 6 | } |
|
117 | | |
118 | 19 | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override { |
119 | 19 | return VExpr::prepare(_partition_expr_ctxs, state, row_desc); |
120 | 19 | } _ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE7prepareEPNS_12RuntimeStateERKNS_13RowDescriptorE Line | Count | Source | 118 | 1 | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override { | 119 | 1 | return VExpr::prepare(_partition_expr_ctxs, state, row_desc); | 120 | 1 | } |
_ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE7prepareEPNS_12RuntimeStateERKNS_13RowDescriptorE Line | Count | Source | 118 | 11 | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override { | 119 | 11 | return VExpr::prepare(_partition_expr_ctxs, state, row_desc); | 120 | 11 | } |
_ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE7prepareEPNS_12RuntimeStateERKNS_13RowDescriptorE Line | Count | Source | 118 | 1 | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override { | 119 | 1 | return VExpr::prepare(_partition_expr_ctxs, state, row_desc); | 120 | 1 | } |
_ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE7prepareEPNS_12RuntimeStateERKNS_13RowDescriptorE Line | Count | Source | 118 | 6 | Status prepare(RuntimeState* state, const RowDescriptor& row_desc) override { | 119 | 6 | return VExpr::prepare(_partition_expr_ctxs, state, row_desc); | 120 | 6 | } |
|
121 | | |
122 | 14 | Status open(RuntimeState* state) override { return VExpr::open(_partition_expr_ctxs, state); }_ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE4openEPNS_12RuntimeStateE Line | Count | Source | 122 | 1 | Status open(RuntimeState* state) override { return VExpr::open(_partition_expr_ctxs, state); } |
_ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE4openEPNS_12RuntimeStateE Line | Count | Source | 122 | 6 | Status open(RuntimeState* state) override { return VExpr::open(_partition_expr_ctxs, state); } |
_ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE4openEPNS_12RuntimeStateE Line | Count | Source | 122 | 1 | Status open(RuntimeState* state) override { return VExpr::open(_partition_expr_ctxs, state); } |
_ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE4openEPNS_12RuntimeStateE Line | Count | Source | 122 | 6 | Status open(RuntimeState* state) override { return VExpr::open(_partition_expr_ctxs, state); } |
|
123 | | |
124 | 2 | Status close(RuntimeState* state) override { return Status::OK(); }_ZN5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE5closeEPNS_12RuntimeStateE Line | Count | Source | 124 | 1 | Status close(RuntimeState* state) override { return Status::OK(); } |
Unexecuted instantiation: _ZN5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE5closeEPNS_12RuntimeStateE _ZN5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE5closeEPNS_12RuntimeStateE Line | Count | Source | 124 | 1 | Status close(RuntimeState* state) override { return Status::OK(); } |
Unexecuted instantiation: _ZN5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE5closeEPNS_12RuntimeStateE |
125 | | |
126 | | Status do_partitioning(RuntimeState* state, Block* block) const override; |
127 | | |
128 | 38 | const std::vector<HashValType>& get_channel_ids() const override { return _hash_vals; }_ZNK5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE15get_channel_idsEv Line | Count | Source | 128 | 1 | const std::vector<HashValType>& get_channel_ids() const override { return _hash_vals; } |
_ZNK5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE15get_channel_idsEv Line | Count | Source | 128 | 6 | const std::vector<HashValType>& get_channel_ids() const override { return _hash_vals; } |
_ZNK5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE15get_channel_idsEv Line | Count | Source | 128 | 23 | const std::vector<HashValType>& get_channel_ids() const override { return _hash_vals; } |
_ZNK5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE15get_channel_idsEv Line | Count | Source | 128 | 8 | const std::vector<HashValType>& get_channel_ids() const override { return _hash_vals; } |
|
129 | | |
130 | | Status clone(RuntimeState* state, std::unique_ptr<PartitionerBase>& partitioner) override; |
131 | | |
132 | | protected: |
133 | 38 | Status _get_partition_column_result(Block* block, std::vector<int>& result) const { |
134 | 38 | int counter = 0; |
135 | 38 | for (auto ctx : _partition_expr_ctxs) { |
136 | 37 | RETURN_IF_ERROR(ctx->execute(block, &result[counter++])); |
137 | 37 | } |
138 | 38 | return Status::OK(); |
139 | 38 | } _ZNK5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE28_get_partition_column_resultEPNS_5BlockERSt6vectorIiSaIiEE Line | Count | Source | 133 | 23 | Status _get_partition_column_result(Block* block, std::vector<int>& result) const { | 134 | 23 | int counter = 0; | 135 | 23 | for (auto ctx : _partition_expr_ctxs) { | 136 | 22 | RETURN_IF_ERROR(ctx->execute(block, &result[counter++])); | 137 | 22 | } | 138 | 23 | return Status::OK(); | 139 | 23 | } |
_ZNK5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE28_get_partition_column_resultEPNS_5BlockERSt6vectorIiSaIiEE Line | Count | Source | 133 | 6 | Status _get_partition_column_result(Block* block, std::vector<int>& result) const { | 134 | 6 | int counter = 0; | 135 | 6 | for (auto ctx : _partition_expr_ctxs) { | 136 | 6 | RETURN_IF_ERROR(ctx->execute(block, &result[counter++])); | 137 | 6 | } | 138 | 6 | return Status::OK(); | 139 | 6 | } |
_ZNK5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE28_get_partition_column_resultEPNS_5BlockERSt6vectorIiSaIiEE Line | Count | Source | 133 | 8 | Status _get_partition_column_result(Block* block, std::vector<int>& result) const { | 134 | 8 | int counter = 0; | 135 | 8 | for (auto ctx : _partition_expr_ctxs) { | 136 | 8 | RETURN_IF_ERROR(ctx->execute(block, &result[counter++])); | 137 | 8 | } | 138 | 8 | return Status::OK(); | 139 | 8 | } |
_ZNK5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE28_get_partition_column_resultEPNS_5BlockERSt6vectorIiSaIiEE Line | Count | Source | 133 | 1 | Status _get_partition_column_result(Block* block, std::vector<int>& result) const { | 134 | 1 | int counter = 0; | 135 | 1 | for (auto ctx : _partition_expr_ctxs) { | 136 | 1 | RETURN_IF_ERROR(ctx->execute(block, &result[counter++])); | 137 | 1 | } | 138 | 1 | return Status::OK(); | 139 | 1 | } |
|
140 | | |
141 | 9 | Status _clone_expr_ctxs(RuntimeState* state, VExprContextSPtrs& new_partition_expr_ctxs) const { |
142 | 9 | new_partition_expr_ctxs.resize(_partition_expr_ctxs.size()); |
143 | 18 | for (size_t i = 0; i < _partition_expr_ctxs.size(); i++) { |
144 | 9 | RETURN_IF_ERROR(_partition_expr_ctxs[i]->clone(state, new_partition_expr_ctxs[i])); |
145 | 9 | } |
146 | 9 | return Status::OK(); |
147 | 9 | } Unexecuted instantiation: _ZNK5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE16_clone_expr_ctxsEPNS_12RuntimeStateERSt6vectorISt10shared_ptrINS_12VExprContextEESaIS8_EE _ZNK5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE16_clone_expr_ctxsEPNS_12RuntimeStateERSt6vectorISt10shared_ptrINS_12VExprContextEESaIS8_EE Line | Count | Source | 141 | 5 | Status _clone_expr_ctxs(RuntimeState* state, VExprContextSPtrs& new_partition_expr_ctxs) const { | 142 | 5 | new_partition_expr_ctxs.resize(_partition_expr_ctxs.size()); | 143 | 10 | for (size_t i = 0; i < _partition_expr_ctxs.size(); i++) { | 144 | 5 | RETURN_IF_ERROR(_partition_expr_ctxs[i]->clone(state, new_partition_expr_ctxs[i])); | 145 | 5 | } | 146 | 5 | return Status::OK(); | 147 | 5 | } |
_ZNK5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE16_clone_expr_ctxsEPNS_12RuntimeStateERSt6vectorISt10shared_ptrINS_12VExprContextEESaIS8_EE Line | Count | Source | 141 | 4 | Status _clone_expr_ctxs(RuntimeState* state, VExprContextSPtrs& new_partition_expr_ctxs) const { | 142 | 4 | new_partition_expr_ctxs.resize(_partition_expr_ctxs.size()); | 143 | 8 | for (size_t i = 0; i < _partition_expr_ctxs.size(); i++) { | 144 | 4 | RETURN_IF_ERROR(_partition_expr_ctxs[i]->clone(state, new_partition_expr_ctxs[i])); | 145 | 4 | } | 146 | 4 | return Status::OK(); | 147 | 4 | } |
Unexecuted instantiation: _ZNK5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE16_clone_expr_ctxsEPNS_12RuntimeStateERSt6vectorISt10shared_ptrINS_12VExprContextEESaIS8_EE |
148 | | |
149 | | virtual void _do_hash(const ColumnPtr& column, HashValType* __restrict result, int idx) const; |
150 | 37 | virtual void _initialize_hash_vals(size_t rows) const { |
151 | 37 | _hash_vals.resize(rows); |
152 | 37 | std::ranges::fill(_hash_vals, 0); |
153 | 37 | } Unexecuted instantiation: _ZNK5doris20Crc32HashPartitionerINS_15ShiftChannelIdsEE21_initialize_hash_valsEm _ZNK5doris20Crc32HashPartitionerINS_24SpillPartitionChannelIdsEE21_initialize_hash_valsEm Line | Count | Source | 150 | 6 | virtual void _initialize_hash_vals(size_t rows) const { | 151 | 6 | _hash_vals.resize(rows); | 152 | 6 | std::ranges::fill(_hash_vals, 0); | 153 | 6 | } |
_ZNK5doris20Crc32HashPartitionerINS_17ShuffleChannelIdsEE21_initialize_hash_valsEm Line | Count | Source | 150 | 23 | virtual void _initialize_hash_vals(size_t rows) const { | 151 | 23 | _hash_vals.resize(rows); | 152 | 23 | std::ranges::fill(_hash_vals, 0); | 153 | 23 | } |
_ZNK5doris20Crc32HashPartitionerINS_26SpillRePartitionChannelIdsEE21_initialize_hash_valsEm Line | Count | Source | 150 | 8 | virtual void _initialize_hash_vals(size_t rows) const { | 151 | 8 | _hash_vals.resize(rows); | 152 | 8 | std::ranges::fill(_hash_vals, 0); | 153 | 8 | } |
|
154 | | |
155 | | VExprContextSPtrs _partition_expr_ctxs; |
156 | | mutable std::vector<HashValType> _hash_vals; |
157 | | }; |
158 | | |
159 | | struct ShuffleChannelIds { |
160 | | using HashValType = PartitionerBase::HashValType; |
161 | 217 | HashValType operator()(HashValType l, size_t r) { return l % r; } |
162 | | }; |
163 | | |
164 | | struct SpillPartitionChannelIds { |
165 | | using HashValType = PartitionerBase::HashValType; |
166 | | // Default spill partition mapping used by level-0 partitioning: |
167 | | // rotate hash bits and apply modulo to get a channel id directly. |
168 | 3.14M | HashValType operator()(HashValType l, size_t r) { return ((l >> 16) | (l << 16)) % r; } |
169 | | }; |
170 | | |
171 | | struct SpillRePartitionChannelIds { |
172 | | using HashValType = PartitionerBase::HashValType; |
173 | | |
174 | | // Repartition mode: return the raw hash value without modulo. |
175 | | // The caller (SpillRepartitioner) will apply level-aware hash mixing and |
176 | | // final channel mapping, so repartition behavior can vary by level. |
177 | 20 | HashValType operator()(HashValType l, size_t /*r*/) { return l; } |
178 | | }; |
179 | | |
180 | 655 | static inline PartitionerBase::HashValType crc32c_shuffle_mix(PartitionerBase::HashValType h) { |
181 | | // Step 1: fold high entropy into low bits |
182 | 655 | h ^= h >> 16; |
183 | | // Step 2: odd multiplicative scramble (cheap avalanche) |
184 | 655 | h *= 0xA5B35705U; |
185 | | // Step 3: final fold to break remaining linearity |
186 | 655 | h ^= h >> 13; |
187 | 655 | return h; |
188 | 655 | } Unexecuted instantiation: merge_partitioner_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_sink_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_writer_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: agg_operator_group_by_limit_opt_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: agg_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: analytic_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: assert_nums_rows_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: datagen_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: distinct_streaming_aggregation_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: empty_set_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_source_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hash_join_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hashjoin_build_sink_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hashjoin_probe_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: join_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_merge_sort_source_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: materialization_shared_state_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: nested_loop_join_build_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: operator_projection_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partition_sort_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_aggregation_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_aggregation_source_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_aggregation_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_hash_join_probe_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_hash_join_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_hash_join_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: query_cache_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: repeat_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scan_normalize_predicate_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: set_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: sort_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_sort_sink_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_sort_source_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_sort_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spillable_operator_test_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: streaming_agg_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: table_function_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: union_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: external_table_sink_hash_partitioner_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_exchanger_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: multi_cast_data_streamer_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline_task_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vdata_stream_recvr_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_consumer_helper_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_consumer_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_merger_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_mgr_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_producer_helper_cross_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_producer_helper_set_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_producer_helper_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_producer_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vruntimefilter_wrapper_sampling_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: file_scanner_v2_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scan_operator_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scanner_context_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scanner_late_arrival_rf_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vfile_scanner_exception_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: tablet_sink_hash_partitioner_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vrow_distribution_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vtablet_writer_v2_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: orc_convert_to_orc_literal_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: orc_read_lines.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: parquet_read_lines.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: wal_manager_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: cached_remote_file_reader_peer_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: memory_reclamation_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_predicate_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: workload_sched_policy_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: profile_spec_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_repartitioner_test.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_writer.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_exchange_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_exchange_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_exchanger.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vdata_stream_recvr.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vdata_stream_sender.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: aggregation_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: aggregation_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: analytic_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: analytic_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: assert_num_rows_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: cache_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: cache_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: datagen_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: distinct_streaming_aggregation_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: empty_set_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_sink_buffer.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exchange_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: file_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hashjoin_build_sink.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hashjoin_probe_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: asof_left_inner_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: asof_left_outer_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: full_outer_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: inner_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: left_anti_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: left_outer_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: left_semi_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: null_aware_left_anti_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: null_aware_left_semi_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: right_anti_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: right_outer_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: right_semi_join_impl.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: join_build_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: join_probe_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: local_merge_sort_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: materialization_opertor.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: multi_cast_data_streamer.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: nested_loop_join_build_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: nested_loop_join_probe_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: olap_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: memory_scratch_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: hive_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: iceberg_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: iceberg_delete_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: iceberg_merge_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: maxcompute_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: blackhole_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: multi_cast_data_stream_sink.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: group_commit_block_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: dict_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: group_commit_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: jdbc_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: es_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: multi_cast_data_stream_source.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: meta_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: jdbc_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: paimon_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partition_sort_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partition_sort_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_aggregation_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_aggregation_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_hash_join_probe_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: partitioned_hash_join_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: rec_cte_anchor_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: rec_cte_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: rec_cte_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: repeat_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: result_file_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: result_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: schema_scan_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: set_probe_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: set_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: set_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: sort_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: sort_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_iceberg_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_sort_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: spill_sort_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: streaming_aggregation_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: table_function_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: tvf_table_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: union_sink_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: union_source_operator.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: external_table_sink_hash_partitioner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: external_partition_function_factory.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: paimon_fixed_bucket_partition_function.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: paimon_row_hash_partition_function.cpp:_ZN5dorisL18crc32c_shuffle_mixEj partitioner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Line | Count | Source | 180 | 5 | static inline PartitionerBase::HashValType crc32c_shuffle_mix(PartitionerBase::HashValType h) { | 181 | | // Step 1: fold high entropy into low bits | 182 | 5 | h ^= h >> 16; | 183 | | // Step 2: odd multiplicative scramble (cheap avalanche) | 184 | 5 | h *= 0xA5B35705U; | 185 | | // Step 3: final fold to break remaining linearity | 186 | 5 | h ^= h >> 13; | 187 | 5 | return h; | 188 | 5 | } |
Unexecuted instantiation: dependency.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline_fragment_context.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline_task.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: task_queue.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: task_scheduler.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: rowid_fetcher.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_filter_producer_helper.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: es_scanner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: file_scanner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: file_scanner_v2.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: jdbc_scanner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: olap_scanner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: parallel_scanner_builder.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scanner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scanner_context.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: scanner_scheduler.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: tablet_sink_hash_partitioner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vfile_result_writer.cpp:_ZN5dorisL18crc32c_shuffle_mixEj spill_repartitioner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Line | Count | Source | 180 | 650 | static inline PartitionerBase::HashValType crc32c_shuffle_mix(PartitionerBase::HashValType h) { | 181 | | // Step 1: fold high entropy into low bits | 182 | 650 | h ^= h >> 16; | 183 | | // Step 2: odd multiplicative scramble (cheap avalanche) | 184 | 650 | h *= 0xA5B35705U; | 185 | | // Step 3: final fold to break remaining linearity | 186 | 650 | h ^= h >> 13; | 187 | 650 | return h; | 188 | 650 | } |
Unexecuted instantiation: vectorized_fn_call.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vexpr.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: vparquet_reader.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: iceberg_partition_function.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: merge_partitioner.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: exec_env_init.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: fragment_mgr.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: query_context.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: runtime_state.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: workload_group.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: workload_group_manager.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: query_task_controller.cpp:_ZN5dorisL18crc32c_shuffle_mixEj Unexecuted instantiation: pipeline_task_action.cpp:_ZN5dorisL18crc32c_shuffle_mixEj |
189 | | |
190 | | // use high 16 bits as channel id to avoid conflict with crc32c hash table |
191 | | // shuffle hash function same with crc32c hash table(eg join hash table) will lead bad performance |
192 | | // hash table offten use low 16 bits as bucket index, so we shift 16 bits to high bits to avoid conflict |
193 | | struct ShiftChannelIds { |
194 | | using HashValType = PartitionerBase::HashValType; |
195 | 5 | HashValType operator()(HashValType l, size_t r) { return crc32c_shuffle_mix(l) % r; } |
196 | | }; |
197 | | |
198 | | class Crc32CHashPartitioner : public Crc32HashPartitioner<ShiftChannelIds> { |
199 | | public: |
200 | | Crc32CHashPartitioner(int partition_count) |
201 | 1 | : Crc32HashPartitioner<ShiftChannelIds>(partition_count) {} |
202 | | |
203 | | Status clone(RuntimeState* state, std::unique_ptr<PartitionerBase>& partitioner) override; |
204 | | |
205 | | private: |
206 | | void _do_hash(const ColumnPtr& column, HashValType* __restrict result, int idx) const override; |
207 | | |
208 | 1 | void _initialize_hash_vals(size_t rows) const override { |
209 | 1 | _hash_vals.resize(rows); |
210 | | // use golden ratio to initialize hash values to avoid collision with hash table's hash function |
211 | 1 | constexpr HashValType CRC32C_SHUFFLE_SEED = 0x9E3779B9U; |
212 | 1 | std::ranges::fill(_hash_vals, CRC32C_SHUFFLE_SEED); |
213 | 1 | } |
214 | | }; |
215 | | |
216 | | #include "common/compile_check_end.h" |
217 | | } // namespace doris |