C#•16mo ago

✅ How to create char[] faster?

I have following class:

[MemoryDiagnoser(true)]
public class Generator
{
    [Benchmark]
    public char[] StartGeneration1()
    {
        var random = new Random();

        Span<char> str = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            str[i] = ((char)(random.Next(26) + 'a'));

        return str.ToArray();
    }

    [Benchmark]
    public char[] StartGeneration2()
    {
        var random = new Random();

        var str = new char[20];

        for (var i = 0; i < 20; i++)
            str[i] = (char)(random.Next(26) + 'a');

        return str;
    }
}

[MemoryDiagnoser(true)]
public class Generator
{
    [Benchmark]
    public char[] StartGeneration1()
    {
        var random = new Random();

        Span<char> str = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            str[i] = ((char)(random.Next(26) + 'a'));

        return str.ToArray();
    }

    [Benchmark]
    public char[] StartGeneration2()
    {
        var random = new Random();

        var str = new char[20];

        for (var i = 0; i < 20; i++)
            str[i] = (char)(random.Next(26) + 'a');

        return str;
    }
}

cs I don't understand why but second version is a little bit slower: | Method | Mean | Error | StdDev | Gen0 | Allocated | |----------------- |---------:|--------:|--------:|-------:|----------:| | StartGeneration1 | 174.6 ns | 3.41 ns | 3.65 ns | 0.0143 | 136 B | | StartGeneration2 | 170.3 ns | 3.36 ns | 6.47 ns | 0.0143 | 136 B | why? is it possible to create char[] without allocating memory?

33 Replies

ero•16mo ago

the results say the second version is faster? and you can't allocate an array without allocating memory the array itself is the allocation

333fred•16mo ago

The results say absolutely nothing You're within margin of error in either scenario

ero•16mo ago

i guess they're within error yeah

SWEETPONYOP•16mo ago

but 174 > 170 so StartGeneration1 is slower

333fred•16mo ago

No It's 174 + or - 3.4, and 170 + or - 3.36 When you look at your error, you need to consider whether there's any overlap In this case, you overlap between 171 (min for result 1) and 173 (max for result 2) Because there's this overlap, you can draw no conclusions from this test Run it again and you may get the exact opposite result

SWEETPONYOP•16mo ago

I wanted to create random string as fast as possible so I decided to use stackalloc but maybe it is wrong

333fred•16mo ago

And, frankly, I wouldn't expect there to be any statistically significant speed difference between these tests So that result is exactly what I would have expected you to get You are nearly certainly prematurely optimizing here

ero•16mo ago

it's probably faster if you get all the bytes and then cast to int or whatever that's assuming this was just for fun you wouldn't do that in actual code unless it's insanely performance critical

SWEETPONYOP•16mo ago

it is only for fun

Aaron•16mo ago

if you want a string, then you should use string.Create

ero•16mo ago

sumpin like that

Aaron•16mo ago

there's also this method in net8 you could try https://learn.microsoft.com/en-us/dotnet/api/system.security.cryptography.randomnumbergenerator.getstring?view=net-8.0

RandomNumberGenerator.GetString(ReadOnlySpan, Int32) Method (System...

Creates a string populated with characters chosen at random from choices.

SWEETPONYOP•16mo ago

didn't know about that, thanks by the way! I found the fastest solution, one moment

public class Generator
{
    [Benchmark]
    public char[] StartGeneration1()
    {
        var random = new Random();

        Span<char> str = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            str[i] = ((char)(random.Next(26) + 'a'));

        return str.ToArray();
    }

    [Benchmark]
    public char[] StartGeneration2()
    {
        var random = new Random();

        var str = new char[20];

        for (var i = 0; i < 20; i++)
            str[i] = (char)(random.Next(26) + 'a');

        return str;
    }

    [Benchmark]
    public unsafe char* StartGeneration3()
    {
        var random = new Random();

        var array = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            array[i] = (char)(random.Next(26) + 'a');

        return array;
    }
}

public class Generator
{
    [Benchmark]
    public char[] StartGeneration1()
    {
        var random = new Random();

        Span<char> str = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            str[i] = ((char)(random.Next(26) + 'a'));

        return str.ToArray();
    }

    [Benchmark]
    public char[] StartGeneration2()
    {
        var random = new Random();

        var str = new char[20];

        for (var i = 0; i < 20; i++)
            str[i] = (char)(random.Next(26) + 'a');

        return str;
    }

    [Benchmark]
    public unsafe char* StartGeneration3()
    {
        var random = new Random();

        var array = stackalloc char[20];

        for (var i = 0; i < 20; i++)
            array[i] = (char)(random.Next(26) + 'a');

        return array;
    }
}

third one is the fastest

Aaron•16mo ago

that doesn't work the third one returns a pointer to free'd memory

SWEETPONYOP•16mo ago

ah..

Aaron•16mo ago

it was allocated in that method's stack frame, and that method's stack frame is gone after you return

333fred•16mo ago

I would also love to see what output you got from the benchmark

SWEETPONYOP•16mo ago

| Method | Mean | Error | StdDev | Gen0 | Allocated | |----------------- |---------:|--------:|--------:|-------:|----------:| | StartGeneration1 | 172.0 ns | 1.60 ns | 1.42 ns | 0.0143 | 136 B | | StartGeneration2 | 170.9 ns | 3.39 ns | 5.48 ns | 0.0143 | 136 B | | StartGeneration3 | 164.5 ns | 3.11 ns | 3.05 ns | 0.0076 | 72 B | // * Legends * Mean : Arithmetic mean of all measurements Error : Half of 99.9% confidence interval StdDev : Standard deviation of all measurements Gen0 : GC Generation 0 collects per 1000 operations Allocated : Allocated memory per single operation (managed only, inclusive, 1KB = 1024B) 1 ns : 1 Nanosecond (0.000000001 sec)

333fred•16mo ago

Gotcha. Still not statistically signficiant

Aaron•16mo ago

almost all the time here is spent doing Random.Next not in how you make and return the array

333fred•16mo ago

Max of 3 is 167.61, min of 2 is 167.51 (There's also that point) (And also the fact that 3 returns a dangling pointer to invalid memory)

SWEETPONYOP•16mo ago

hm ok we found a bottleneck! I will rewrite rnd.next

333fred•16mo ago

You won't get faster I applaud the spirit of learning, but I really don't think this is it

ero•16mo ago

You can get a good bit faster

var bytes = new byte[length * sizeof(int)];
Random.Shared.NextBytes(bytes);
var ints = MemoryMarshal.Cast<byte, int>(bytes);

var bytes = new byte[length * sizeof(int)];
Random.Shared.NextBytes(bytes);
var ints = MemoryMarshal.Cast<byte, int>(bytes);

Something like that Probably better ways out there

Kouhai•16mo ago

I assume they want a valid ascii char

SWEETPONYOP•16mo ago

hm I thinked about this:

var buffer = new byte[20];
random.NextBytes(buffer);

var buffer = new byte[20];
random.NextBytes(buffer);

and then

for (var i = 0; i < 20; i++)
    str[i] = (char)(buffer[i] % 26 + 'a');

for (var i = 0; i < 20; i++)
    str[i] = (char)(buffer[i] % 26 + 'a');

and random is RandomNumberGenerator maybe I will test this harold

Aaron•16mo ago

I believe RandomNumberGenerator is slower than Random is since it's cryptographic randomness instead of a prng

SWEETPONYOP•16mo ago

SWEETPONYOP•16mo ago

but maybe second one is wrong, I will test the output works fine thanks guys

jcotton42•16mo ago

@ivefifthsence use string.Create

Tarcisio•16mo ago

how can i learn to be smart like you

Denis•16mo ago

Work hard

GrabYourPitchforks•16mo ago

The second one introduces a slight bias toward earlier letters of the alphabet. For example, A will appear in the output more than Z, and this will be statistically significant.

Gaming

Programming

✅ How to create char[] faster?