Alexander Guzhva c4b37fb285
enhance: Custom bitset and bitsetview prototypes (#30454)
Issue: #31285 

Basically, I've replaced `FixedVector<bool>` and `boost::dynamic_bitset`
with custom bitset and bitsetview in order to reduce the memory
bandwidth & increase performance for the filtering.

This PR is for internal use only. 

Current progress (numbers are for GCC 9.5.0 on Ubuntu 22.04 LTS;
clang-17 produces better performance numbers):
Baseline:
```
[ RUN      ] CApiTest.AssembeChunkPerfTest
start test
cost: 17903us
[       OK ] CApiTest.AssembeChunkPerfTest (183 ms)

[ RUN      ] Expr.TestMultiLogicalExprsOptimization
cost: 1391us
cost: 5us
cost: 4us
cost: 4us
cost: 6us
cost: 4us
cost: 4us
cost: 4us
cost: 4us
cost: 4us
143
cost: 10us
cost: 8us
cost: 10us
cost: 8us
cost: 8us
cost: 8us
cost: 8us
cost: 8us
cost: 8us
cost: 9us
8
/home/ubuntu/zilliz/milvus4/milvus/internal/core/unittest/test_expr.cpp:1561: Failure
Expected: (cost_op) < (cost_no_op), actual: 143 vs 8
[  FAILED  ] Expr.TestMultiLogicalExprsOptimization (7 ms)
[ RUN      ] Expr.TestExprs
start test
3cost: 889us
start test
10cost: 2us
start test
20cost: 2us
start test
30cost: 2us
start test
50cost: 3us
start test
100cost: 7us
start test
200cost: 16us
[       OK ] Expr.TestExprs (9 ms)

[ RUN      ] Expr.TestUnaryBenchTest
start test type:2
 cost: 124.8us
start test type:3
 cost: 163.1us
start test type:4
 cost: 275.9us
start test type:5
 cost: 590.9us
start test type:10
 cost: 62.7us
start test type:11
 cost: 65.9us
[       OK ] Expr.TestUnaryBenchTest (1153 ms)
[ RUN      ] Expr.TestBinaryRangeBenchTest
start test type:2
 cost: 151.4us
start test type:3
 cost: 198.4us
start test type:4
 cost: 361.9us
start test type:5
 cost: 753.9us
start test type:10
 cost: 64.6us
start test type:11
 cost: 62.2us
[       OK ] Expr.TestBinaryRangeBenchTest (1151 ms)
[ RUN      ] Expr.TestLogicalUnaryBenchTest
start test type:2
 cost: 121.14us
start test type:3
 cost: 156.84us
start test type:4
 cost: 249.76us
start test type:5
 cost: 534.44us
start test type:10
 cost: 82.2us
start test type:11
 cost: 83.52us
[       OK ] Expr.TestLogicalUnaryBenchTest (1202 ms)
[ RUN      ] Expr.TestBinaryLogicalBenchTest
start test type:2
 cost: 80.64us
start test type:3
 cost: 78.22us
start test type:4
 cost: 255.76us
start test type:5
 cost: 532.04us
start test type:10
 cost: 89.26us
start test type:11
 cost: 90us
[       OK ] Expr.TestBinaryLogicalBenchTest (1198 ms)
[ RUN      ] Expr.TestBinaryArithOpEvalRangeBenchExpr
start test type:2
 cost: 401.7us
start test type:3
 cost: 420.96us
start test type:4
 cost: 418.04us
start test type:5
 cost: 470.54us
start test type:10
 cost: 250.32us
start test type:11
 cost: 850.08us
[       OK ] Expr.TestBinaryArithOpEvalRangeBenchExpr (1273 ms)
[ RUN      ] Expr.TestCompareExprBenchTest
start test type:2
 cost: 162us
start test type:3
 cost: 142us
start test type:4
 cost: 374us
start test type:5
 cost: 674us
start test type:10
 cost: 366us
start test type:11
 cost: 645us
[       OK ] Expr.TestCompareExprBenchTest (1214 ms)
[ RUN      ] Expr.TestRefactorExprs
start test
3cost: 1253us
start test
10cost: 1060us
start test
20cost: 681us
start test
30cost: 522us
start test
50cost: 511us
start test
100cost: 506us
start test
200cost: 497us
[       OK ] Expr.TestRefactorExprs (1142 ms)

```

Candidate:
```
[ RUN      ] CApiTest.AssembeChunkPerfTest
start test
cost: 6099us
[       OK ] CApiTest.AssembeChunkPerfTest (153 ms)

[ RUN      ] Expr.TestMultiLogicalExprsOptimization
cost: 42us
cost: 15us
cost: 15us
cost: 14us
cost: 15us
cost: 15us
cost: 15us
cost: 15us
cost: 15us
cost: 15us
17
cost: 41us
cost: 39us
cost: 33us
cost: 33us
cost: 33us
cost: 33us
cost: 34us
cost: 41us
cost: 34us
cost: 34us
35
[       OK ] Expr.TestMultiLogicalExprsOptimization (6 ms)
[ RUN      ] Expr.TestExprs
start test
3cost: 20us
start test
10cost: 2us
start test
20cost: 2us
start test
30cost: 2us
start test
50cost: 4us
start test
100cost: 8us
start test
200cost: 15us
[       OK ] Expr.TestExprs (8 ms)

[ RUN      ] Expr.TestUnaryBenchTest
start test type:2
 cost: 55.7us
start test type:3
 cost: 79.8us
start test type:4
 cost: 177.6us
start test type:5
 cost: 337.2us
start test type:10
 cost: 16.9us
start test type:11
 cost: 15.7us
[       OK ] Expr.TestUnaryBenchTest (1140 ms)
[ RUN      ] Expr.TestBinaryRangeBenchTest
start test type:2
 cost: 57.1us
start test type:3
 cost: 87us
start test type:4
 cost: 177.5us
start test type:5
 cost: 342.7us
start test type:10
 cost: 17.9us
start test type:11
 cost: 16.7us
[       OK ] Expr.TestBinaryRangeBenchTest (1152 ms)
[ RUN      ] Expr.TestLogicalUnaryBenchTest
start test type:2
 cost: 34.58us
start test type:3
 cost: 68.86us
start test type:4
 cost: 151.38us
start test type:5
 cost: 286.8us
start test type:10
 cost: 16.54us
start test type:11
 cost: 16.7us
[       OK ] Expr.TestLogicalUnaryBenchTest (1165 ms)
[ RUN      ] Expr.TestBinaryLogicalBenchTest
start test type:2
 cost: 20us
start test type:3
 cost: 17.1us
start test type:4
 cost: 154.12us
start test type:5
 cost: 286.1us
start test type:10
 cost: 19.6us
start test type:11
 cost: 19.24us
[       OK ] Expr.TestBinaryLogicalBenchTest (1188 ms)
[ RUN      ] Expr.TestBinaryArithOpEvalRangeBenchExpr
start test type:2
 cost: 125.7us
start test type:3
 cost: 111.34us
start test type:4
 cost: 148.02us
start test type:5
 cost: 306.7us
start test type:10
 cost: 149.3us
start test type:11
 cost: 282.94us
[       OK ] Expr.TestBinaryArithOpEvalRangeBenchExpr (1221 ms)
[ RUN      ] Expr.TestCompareExprBenchTest
start test type:2
 cost: 89us
start test type:3
 cost: 79us
start test type:4
 cost: 323us
start test type:5
 cost: 629us
start test type:10
 cost: 313us
start test type:11
 cost: 591us
[       OK ] Expr.TestCompareExprBenchTest (1228 ms)
[ RUN      ] Expr.TestRefactorExprs
start test
3cost: 874us
start test
10cost: 611us
start test
20cost: 290us
start test
30cost: 294us
start test
50cost: 272us
start test
100cost: 278us
start test
200cost: 279us
[       OK ] Expr.TestRefactorExprs (1149 ms)

```

Signed-off-by: Alexandr Guzhva <alexanderguzhva@gmail.com>
2024-03-24 21:49:07 +08:00

73 lines
2.5 KiB
C++

// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "ExistsExpr.h"
#include "common/Json.h"
namespace milvus {
namespace exec {
void
PhyExistsFilterExpr::Eval(EvalCtx& context, VectorPtr& result) {
switch (expr_->column_.data_type_) {
case DataType::JSON: {
if (is_index_mode_) {
PanicInfo(ExprInvalid,
"exists expr for json index mode not supportted");
}
result = EvalJsonExistsForDataSegment();
break;
}
default:
PanicInfo(DataTypeInvalid,
"unsupported data type: {}",
expr_->column_.data_type_);
}
}
VectorPtr
PhyExistsFilterExpr::EvalJsonExistsForDataSegment() {
auto real_batch_size = GetNextBatchSize();
if (real_batch_size == 0) {
return nullptr;
}
auto res_vec =
std::make_shared<ColumnVector>(TargetBitmap(real_batch_size));
TargetBitmapView res(res_vec->GetRawData(), real_batch_size);
auto pointer = milvus::Json::pointer(expr_->column_.nested_path_);
auto execute_sub_batch = [](const milvus::Json* data,
const int size,
TargetBitmapView res,
const std::string& pointer) {
for (int i = 0; i < size; ++i) {
res[i] = data[i].exist(pointer);
}
};
int64_t processed_size = ProcessDataChunks<Json>(
execute_sub_batch, std::nullptr_t{}, res, pointer);
AssertInfo(processed_size == real_batch_size,
"internal error: expr processed rows {} not equal "
"expect batch size {}",
processed_size,
real_batch_size);
return res_vec;
}
} //namespace exec
} // namespace milvus