IntrinsicsPlayground Save

My toys to play with SSE/AVX in pure C# (.NET Core 2.1)

Project README

IntrinsicsPlayground

My toys to play with intrinsics in pure C# (see System.Runtime.Intrinsics.Experimental)

Environment:


BenchmarkDotNet=v0.10.14, OS=Windows 10.0.17134
Intel Core i7-8700K CPU 3.70GHz (Coffee Lake), 1 CPU, 12 logical and 6 physical cores
Frequency=3609374 Hz, Resolution=277.0564 ns, Timer=TSC
.NET Core SDK=2.1.301
  [Host]     : .NET Core 2.1.1 (CoreCLR 4.6.26606.02, CoreFX 4.6.26606.05), 64bit RyuJIT
  DefaultJob : .NET Core 2.1.1 (CoreCLR 4.6.26606.02, CoreFX 4.6.26606.05), 64bit RyuJIT

1. ArraySum

Sum of elements in an array of floats

var sum = arrayOfFloats.Sum();
Method TestArrayLength Mean Scaled
Sum_LINQ 10 45.629 ns 4.81
Sum_Simple 10 2.879 ns 0.30
Sum_LinqFasterLib 10 15.100 ns 1.59
Sum_Avx 10 9.496 ns 1.00
Sum_LINQ 121 439.021 ns 25.20
Sum_Simple 121 78.111 ns 4.48
Sum_LinqFasterLib 121 22.432 ns 1.29
Sum_Avx 121 17.423 ns 1.00
Sum_LINQ 32768 109,128.390 ns 31.87
Sum_Simple 32768 27,345.722 ns 7.99
Sum_LinqFasterLib 32768 3,432.010 ns 1.00
Sum_Avx 32768 3,423.826 ns 1.00

2. ArrayEqual

Are arrays of floats equal?

var equal = arrayOfFloats1.SequenceEqual(arrayOfFloats2);
Method TestArrayLength Mean Scaled
ArrayEqual_LINQ 10 116.175 ns 30.22
ArrayEqual_Simple 10 6.130 ns 1.59
ArrayEqual_AVX2 10 3.845 ns 1.00
ArrayEqual_LINQ 121 783.744 ns 50.82
ArrayEqual_Simple 121 60.981 ns 3.95
ArrayEqual_AVX2 121 15.425 ns 1.00
ArrayEqual_LINQ 32768 198,072.829 ns 52.85
ArrayEqual_Simple 32768 13,773.698 ns 3.67
ArrayEqual_AVX2 32768 3,748.025 ns 1.00

3. ArrayMax

Max element in an array of ints

var max = arrayOfInts.Max();
Method TestArrayLength Mean Scaled
Max_LINQ 10 44.932 ns 3.38
Max_Simple 10 3.125 ns 0.23
Max_LinqFasterLib 10 14.071 ns 1.06
Max_Avx 10 13.307 ns 1.00
Max_LINQ 121 447.998 ns 23.34
Max_Simple 121 42.731 ns 2.23
Max_LinqFasterLib 121 19.230 ns 1.00
Max_Avx 121 19.219 ns 1.00
Max_LINQ 32768 116,163.414 ns 69.46
Max_Simple 32768 10,300.011 ns 6.16
Max_LinqFasterLib 32768 1,759.705 ns 1.05
Max_Avx 32768 1,672.383 ns 1.00

4. ArrayIsSorted

Check if an array is sorted or not

bool IsSorted(int[] array)
{
    if (array.Length < 2)
        return true;

    for (int i = 0; i < array.Length - 1; i++)
    {
        if (array[i] > array[i + 1])
            return false;
    }
    return true;
}
Method TestArrayLength Mean Scaled
IsSorted_Simple 10 6.309 ns 2.51
IsSorted_Simple_Optimized 10 4.784 ns 1.90
IsSorted_Sse41 10 2.691 ns 1.07
IsSorted_Avx2 10 2.513 ns 1.00
IsSorted_Simple 121 80.036 ns 6.38
IsSorted_Simple_Optimized 121 54.508 ns 4.35
IsSorted_Sse41 121 22.927 ns 1.83
IsSorted_Avx2 121 12.539 ns 1.00
IsSorted_Simple 32768 19,928.318 ns 5.31
IsSorted_Simple_Optimized 32768 13,705.714 ns 3.65
IsSorted_Sse41 32768 6,190.854 ns 1.65
IsSorted_Avx2 32768 3,755.430 ns 1.00

5. ArrayReverse

Reverse all elements in an array

Array.Reverse(arrayOfInts);
Method TestArrayLength Mean Scaled
Reverse_BCL 10 46.72 ns 0.96
Reverse_SSE 10 48.70 ns 1.00
Reverse_BCL 121 102.90 ns 1.27
Reverse_SSE 121 80.82 ns 1.00
Reverse_BCL 32768 18,083.44 ns 1.56
Reverse_SSE 32768 11,577.33 ns 1.00

6. ArrayIndexOf

Searche for the specified element and returns the index of its first occurrence

int index = Array.IndexOf(arrayOfInts, element);
Method TestArrayLength Mean Scaled
IndexOf_BCL 10 7.701 ns 1.37
IndexOf_Avx2 10 5.620 ns 1.00
IndexOf_BCL 121 31.721 ns 3.59
IndexOf_Avx2 121 8.825 ns 1.00
IndexOf_BCL 32768 6,961.574 ns 3.72
IndexOf_Avx2 32768 1,870.329 ns 1.00

7. MatrixSum

Add two matrices (Matrix4x4)

var matrix3 = matrix1 + matrix2;
Method Mean Scaled
MatrixSum_BCL 25.81 us 1.30
MatrixSum_Avx 19.89 us 1.00

8. MatrixNegate

Negate a matrix (Matrix4x4)

matrix = -matrix;
Method Mean Scaled
MatrixNegate_BCL 34.31 us 1.38
MatrixNegate_Avx 24.87 us 1.00

9. MatrixMultiply

Multiply a matrix by a vector

matrix = matrix * vec;
Method Mean Scaled
MatrixMultiply_ByVector_BCL 28.01 us 1.04
MatrixMultiply_ByVector_Avx 26.83 us 1.00

Misc

1. Array.Sort vs DualPivotQuickSort

Dual pivot quick sort is used in Java for primitive types (Arrays.sort(int[])). alt text

Open Source Agenda is not affiliated with "IntrinsicsPlayground" Project. README Source: EgorBo/IntrinsicsPlayground
Stars
64
Open Issues
3
Last Commit
5 years ago
License
MIT

Open Source Agenda Badge

Open Source Agenda Rating