Open OoLunar opened 1 year ago
You added bin and obj folders. I don't think they should be included in git. 😉
Whoops! You're right
Good job @OoLunar , I agree your correcations.
in fact would be better to use top-level statements to show beauty of modern c# for the doubters hehe
also i'd use Task WhenAll(IEnumerable<Task> tasks)
overload so i dont allocate array by myself
my code would look like this:
const int delayMs = 1000 * 10;
var taskCount = args.Length > 0 ? int.Parse(args[0]) : 100000;
var tasks = Enumerable
.Range(0, taskCount)
.Select(_ => Task.Delay(delayMs));
await Task.WhenAll(tasks);
in fact would be better to use top-level statements to show beauty of modern c# for the doubters hehe also i'd use
Task WhenAll(IEnumerable<Task> tasks)
overload so i dont allocate array by myselfmy code would look like this:
const int delayMs = 1000 * 10; var taskCount = args.Length > 0 ? int.Parse(args[0]) : 100000; var tasks = Enumerable .Range(0, taskCount) .Select(_ => Task.Delay(delayMs)); await Task.WhenAll(tasks);
And also do top level namespace
Have either one of y'all benchmarked your new code? While I agree, the code could look appealing to non-C# users, the goal here is to provide efficiency, not readability.
i'd use Task WhenAll(IEnumerable
tasks) overload so i dont allocate array by myself
What... do you think an enumerable is, and how do you think it is internally represented? Arrays are quite efficient upon iteration, indexing and long lifetimes. Enumerables are a read-only interface which hides implementation from user. The goal here is to create X amount of tasks and wait for all of them to finish executing, thus there is no reason to hide the implementation via enumerables since we are the only people who are able to access the implementation.
@OoLunar of course i did benchmark different solutions, as far as i remember there was no noticeable difference between your and my code, but original code was somewhat inefficient.
I pretty much know what Enumerable is, dont tell me that.
by saying use overload
i do not "hide the implementation"
, i use existing overload of Task.WhenAll
method. also what implementation did i hide? filling an array? isnt it just your implementation?
There's no noticeable difference because the bulk of the benchmark is waitinh on the tasks to complete, however, in a less contrived and more realistic scenario, LINQ vs manual construction can make a huge difference - and if your goal is to showcase efficient C# code, using an infamous performance hog (LINQ) doesn't accomplish that.
@akiraveliara my goal is to create readable & efficient code for this particular problem. i know linq can be expensive in some cases, but it does not bother me as it really does not matter in this context
This particular problem does not require readable code, it requires efficient code. You are welcome to create a separate pull request if you believe otherwise.
@akiraveliara my goal is to create readable & efficient code for this particular problem. i know linq can be expensive in some cases, but it does not bother me as it really does not matter in this context
if your goal is to write both readable and efficient code you failed in both departments
LINQ is simply not as intuitive - especially to people new to c# - as just using a plain old array, and due to the abstractions that are hidden behind the scenes when using IEnumerable
you take a hit in performance. You wont get more efficient than just using an array. Even if it is just a few nanoseconds, this is supposed to be a benchmark, so even that magnitude matters.
Significantly improves .NET's code, closes https://github.com/pkolaczk/async-runtimes-benchmarks/pull/4
ImplicitUsings
indotnet.csproj
Task[]
) over a list (List<Task>
) to prevent constant resizing of the internally backed array.Task.Run
and passes theTask
itself in the for loop. This prevents the unnecessary creation of an additional async contexts.Additionally
ValueTask
s are typically used in hotpaths overTask
's due to their newer implementation and don't suffer the scars of .NET framework.