SourceGenerator 生成db to class代码优化结果记录 二

victor.x.qu2024-08-03 14:21

优化

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

不再使用迭代器

csharp 复制代码 
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

csharp 复制代码 
 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),
};

使用 span

再没有了迭代器方法限制， span 就可以放飞自我，随意使用了

csharp 复制代码 
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++)
    {

使用 ArrayPool 减少内存占用

csharp 复制代码 
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);
}

数据小时使用栈分配

csharp 复制代码 
 var s = reader.FieldCount <= 64 ? MemoryMarshal.CreateSpan(ref MemoryMarshal.GetReference(stackalloc int[reader.FieldCount]), reader.FieldCount) :  state.GetTokens();

提前生成部分 hashcode 进行比较

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

csharp 复制代码 
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;
        }
    }
}

性能测试说明

BenchmarkDotNet

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

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

class

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

csharp 复制代码 
public class Dog
{
    public int? Age { get; set; }
    public string Name { get; set; }
    public float? Weight { get; set; }
}

mock 数据

csharp 复制代码 
 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;
        }
}

Benchmark 代码

csharp 复制代码 
    [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

测试结果

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

热门推荐
01【一文读懂】NTN（非地面网络）技术介绍02优化手机性能，解决卡顿问题：关闭这3个微信开关，释放内存空间03玄机平台应急响应—webshell查杀04Cube-Studio：开源大模型全链路一站式中台05基于大语言模型（LLM）的合成数据生成、策展和评估的综述06(欧拉)openEuler系统添加网卡文件配置流程、(欧拉)openEuler系统手动配置ipv6地址流程、(欧拉)openEuler系统网络管理说明0719个Web前端交互式3D JavaScript框架和库08HCIA-datacom数通题库和录播视频资料09Docker 夺命连环 15 问10组基轨迹建模 GBTM的介绍与实现（Stata 或 R）