在上一篇留下的 Dapper AOT 还有什么特别优化点的问题
在仔细阅读生成代码和源码之后,终于得到了答案
个人之前一直以为 Dapper AOT 只用了迭代器去实现,所以理应差不多实现代码却又极大差距,思维陷入了僵局,一度以为有什么黑魔法
结果 Dapper AOT 没有用迭代器去实现!!! 靠北啦,还以为迭代器有新姿势可以优化了
List<BenchmarkTest.Dog> results = new();
try
{
while (reader.Read())
{
results.Add(ReadOne(reader, readOnlyTokens));
}
return results;
}
当然就只能要求 用户必须使用 AsList
方法,因为 ToList
会导致复制list的问题, 导致负优化,
像这样
connection.Query<Dog>("select * from dog").AsList();
// AsList 实现
public static List<T> AsList<T>(this IEnumerable<T>? source) => source switch
{
null => null!,
List<T> list => list,
_ => Enumerable.ToList(source),
};
再没有了迭代器方法限制, span 就可以放飞自我,随意使用了
public static BenchmarkTest.Dog ReadOne(this IDataReader reader, ref ReadOnlySpan<int> ss)
{
var d = new BenchmarkTest.Dog();
for (int j = 0; j < ss.Length; j++)
{
public Span<int> GetTokens()
{
FieldCount = Reader!.FieldCount;
if (Tokens is null || Tokens.Length < FieldCount)
{
// no leased array, or existing lease is not big enough; rent a new array
if (Tokens is not null) ArrayPool<int>.Shared.Return(Tokens);
Tokens = ArrayPool<int>.Shared.Rent(FieldCount);
}
return MemoryMarshal.CreateSpan(ref MemoryMarshal.GetArrayDataReference(Tokens), FieldCount);
}
var s = reader.FieldCount <= 64 ? MemoryMarshal.CreateSpan(ref MemoryMarshal.GetReference(stackalloc int[reader.FieldCount]), reader.FieldCount) : state.GetTokens();
因为比较现在也并不耗时了, 所以 缓存也没有必要了, 也一并移除
public static void GenerateReadTokens(this IDataReader reader, Span<int> s)
{
for (int i = 0; i < reader.FieldCount; i++)
{
var name = reader.GetName(i);
var type = reader.GetFieldType(i);
switch (EntitiesGenerator.NormalizedHash(name))
{
case 742476188U:
s[i] = type == typeof(int) ? 1 : 2;
break;
case 2369371622U:
s[i] = type == typeof(string) ? 3 : 4;
break;
case 1352703673U:
s[i] = type == typeof(float) ? 5 : 6;
break;
default:
break;
}
}
}
这里特别说明一下
使用的 BenchmarkDotNet ,其本身已经考虑了 jit优化等等方面, 有预热,超多次执行,
结果值也是按照统计学有考虑结果集分布情况处理,移除变差大的值(比如少数的孤立的极大极小值), 差异不大情况,一般显示平均值,有大差异时还会显示 中位值
感兴趣的童鞋可以去 https://github.com/dotnet/BenchmarkDotNet 了解
chole 有点棘手,为了方便mock,所以 copy了部分源码,只比较实体映射部分
DapperAOT 和 纯 dapper 很难一起运行,所以不再比较了,反正 dapper 肯定慢
测试数据 正如之前说过, 采用 手动 mock 方式,避免 db 驱动 、db 执行、mock库 等等 带来的执行差异影响
非常简单的类,当然不能代表所有情况,不过简单测试够用了
public class Dog
{
public int? Age { get; set; }
public string Name { get; set; }
public float? Weight { get; set; }
}
public class TestDbConnection : DbConnection
{
public int RowCount { get; set; }
public IDbCommand CreateCommand()
{
return new TestDbCommand() { RowCount = RowCount };
}
}
public class TestDbCommand : DbCommand
{
public int RowCount { get; set; }
public IDataParameterCollection Parameters { get; } = new TestDataParameterCollection();
public IDbDataParameter CreateParameter()
{
return new TestDataParameter();
}
protected override DbDataReader ExecuteDbDataReader(CommandBehavior behavior)
{
return new TestDbDataReader() { RowCount = RowCount };
}
}
public class TestDbDataReader : DbDataReader
{
public int RowCount { get; set; }
private int calls = 0;
public override object this[int ordinal]
{
get
{
switch (ordinal)
{
case 0:
return "XX";
case 1:
return 2;
case 2:
return 3.3f;
default:
return null;
}
}
}
public override int FieldCount => 3;
public override Type GetFieldType(int ordinal)
{
switch (ordinal)
{
case 0:
return typeof(string);
case 1:
return typeof(int);
case 2:
return typeof(float);
default:
return null;
}
}
public override float GetFloat(int ordinal)
{
switch (ordinal)
{
case 2:
return 3.3f;
default:
return 0;
}
}
public override int GetInt32(int ordinal)
{
switch (ordinal)
{
case 1:
return 2;
default:
return 0;
}
}
public override string GetName(int ordinal)
{
switch (ordinal)
{
case 0:
return "Name";
case 1:
return "Age";
case 2:
return "Weight";
default:
return null;
}
}
public override string GetString(int ordinal)
{
switch (ordinal)
{
case 0:
return "XX";
default:
return null;
}
}
public override object GetValue(int ordinal)
{
switch (ordinal)
{
case 0:
return "XX";
case 1:
return 2;
case 2:
return 3.3f;
default:
return null;
}
}
public override bool Read()
{
calls++;
return calls <= RowCount;
}
}
[MemoryDiagnoser, Orderer(summaryOrderPolicy: SummaryOrderPolicy.FastestToSlowest), GroupBenchmarksBy(BenchmarkLogicalGroupRule.ByCategory), CategoriesColumn]
public class ObjectMappingTest
{
[Params(1, 1000, 10000, 100000, 1000000)]
public int RowCount { get; set; }
[Benchmark(Baseline = true)]
public void SetClass()
{
var connection = new TestDbConnection() { RowCount = RowCount };
var dogs = new List<Dog>();
try
{
connection.Open();
var cmd = connection.CreateCommand();
cmd.CommandText = "select ";
using (var reader = cmd.ExecuteReader(CommandBehavior.Default))
{
while (reader.Read())
{
var dog = new Dog();
dogs.Add(dog);
dog.Name = reader.GetString(0);
dog.Age = reader.GetInt32(1);
dog.Weight = reader.GetFloat(2);
}
}
}
finally
{
connection.Close();
}
}
[Benchmark]
public void DapperAOT()
{
var connection = new TestDbConnection() { RowCount = RowCount };
var dogs = connection.Query<Dog>("select * from dog").AsList();
}
[Benchmark]
public void SourceGenerator()
{
var connection = new TestDbConnection() { RowCount = RowCount };
List<Dog> dogs;
try
{
connection.Open();
var cmd = connection.CreateCommand();
cmd.CommandText = "select ";
using (var reader = cmd.ExecuteReader(CommandBehavior.Default))
{
dogs = reader.ReadTo<Dog>().AsList();
}
}
finally
{
connection.Close();
}
}
[Benchmark]
public void Chloe()
{
var connection = new TestDbConnection() { RowCount = RowCount };
try
{
connection.Open();
var cmd = connection.CreateCommand();
var dogs = new InternalSqlQuery<Dog>(cmd, "select").AsList();
}
finally
{
connection.Close();
}
}
}
完整代码可以参考 https://github.com/fs7744/SlowestEM
BenchmarkDotNet v0.13.12, Windows 10 (10.0.19045.4651/22H2/2022Update)
Intel Core i7-10700 CPU 2.90GHz, 1 CPU, 16 logical and 8 physical cores
.NET SDK 9.0.100-preview.5.24307.3
[Host] : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2
DefaultJob : .NET 8.0.6 (8.0.624.26715), X64 RyuJIT AVX2
Method | RowCount | Mean | Error | StdDev | Ratio | RatioSD | Gen0 | Gen1 | Gen2 | Allocated | Alloc Ratio |
---|---|---|---|---|---|---|---|---|---|---|---|
DapperAOT | 1 | 446.3 ns | 8.81 ns | 8.65 ns | 0.60 | 0.03 | 0.0525 | 0.0515 | - | 440 B | 1.00 |
SourceGenerator | 1 | 690.0 ns | 13.72 ns | 32.34 ns | 0.95 | 0.07 | 0.0525 | 0.0515 | - | 440 B | 1.00 |
SetClass | 1 | 728.3 ns | 14.59 ns | 37.41 ns | 1.00 | 0.00 | 0.0525 | 0.0515 | - | 440 B | 1.00 |
Chloe | 1 | 909.7 ns | 17.49 ns | 22.75 ns | 1.25 | 0.06 | 0.1020 | 0.1011 | - | 856 B | 1.95 |
SetClass | 1000 | 8,593.3 ns | 169.90 ns | 390.38 ns | 1.00 | 0.00 | 6.7902 | 1.6937 | - | 56912 B | 1.00 |
SourceGenerator | 1000 | 16,967.8 ns | 310.02 ns | 258.88 ns | 1.91 | 0.08 | 6.7749 | 1.6785 | - | 56912 B | 1.00 |
DapperAOT | 1000 | 18,299.7 ns | 267.72 ns | 250.43 ns | 2.06 | 0.09 | 6.7749 | 1.3428 | - | 56912 B | 1.00 |
Chloe | 1000 | 116,049.4 ns | 297.71 ns | 263.91 ns | 13.06 | 0.54 | 6.8359 | 1.7090 | - | 57328 B | 1.01 |
SetClass | 10000 | 309,255.1 ns | 3,945.26 ns | 3,294.47 ns | 1.00 | 0.00 | 83.0078 | 82.5195 | 41.5039 | 662782 B | 1.00 |
DapperAOT | 10000 | 402,700.7 ns | 7,676.45 ns | 7,180.56 ns | 1.31 | 0.03 | 83.0078 | 82.5195 | 41.5039 | 662782 B | 1.00 |
SourceGenerator | 10000 | 414,226.2 ns | 8,149.22 ns | 10,007.97 ns | 1.34 | 0.04 | 83.0078 | 82.5195 | 41.5039 | 662782 B | 1.00 |
Chloe | 10000 | 1,453,166.1 ns | 19,660.10 ns | 17,428.16 ns | 4.70 | 0.07 | 82.0313 | 80.0781 | 41.0156 | 663199 B | 1.00 |
SetClass | 100000 | 2,176,860.4 ns | 42,449.84 ns | 63,536.93 ns | 1.00 | 0.00 | 496.0938 | 496.0938 | 496.0938 | 6098015 B | 1.00 |
SourceGenerator | 100000 | 3,045,760.4 ns | 59,378.23 ns | 63,534.04 ns | 1.39 | 0.05 | 496.0938 | 496.0938 | 496.0938 | 6098015 B | 1.00 |
DapperAOT | 100000 | 3,053,510.0 ns | 35,015.61 ns | 29,239.62 ns | 1.40 | 0.04 | 496.0938 | 496.0938 | 496.0938 | 6098015 B | 1.00 |
Chloe | 100000 | 13,152,653.6 ns | 65,400.49 ns | 51,060.40 ns | 6.02 | 0.14 | 484.3750 | 484.3750 | 484.3750 | 6098433 B | 1.00 |
SetClass | 1000000 | 105,420,410.0 ns | 2,093,734.23 ns | 3,380,990.50 ns | 1.00 | 0.00 | 6800.0000 | 6800.0000 | 2200.0000 | 56780029 B | 1.00 |
SourceGenerator | 1000000 | 115,534,043.8 ns | 1,828,036.86 ns | 1,795,376.62 ns | 1.09 | 0.03 | 6800.0000 | 6800.0000 | 2200.0000 | 56780118 B | 1.00 |
DapperAOT | 1000000 | 115,751,485.5 ns | 2,120,239.39 ns | 2,603,844.38 ns | 1.10 | 0.04 | 6800.0000 | 6800.0000 | 2200.0000 | 56780029 B | 1.00 |
Chloe | 1000000 | 208,295,919.3 ns | 4,031,590.18 ns | 4,481,101.81 ns | 1.97 | 0.06 | 6666.6667 | 6666.6667 | 2333.3333 | 56781907 B | 1.00 |
SourceGenerator 基本等同 DapperAOT 了, 除了没有使用 Interceptor, 以及各种情况细节没有考虑之外, 两者性能一样
SourceGenerator 肯定现在性能优化最佳方式,毕竟可以生成代码文件,上手难度其实比 emit 之类小多了