ComputeSharp VS range-v3

A couple other options than what has been suggested so far: - TerraFX.Interop.Windows. Raw, blittable, 1:1 bindings for all Win32, D2D/D3D11/D3D12 APIs (there's also a version with Vulkan bindings). As close to doing #include as you can get in C#. This is my personal favorite, I use it in my own ComputeSharp library, and transitively we use it in the Microsoft Store too 🙂 - Silk.NET another version of high-performance bindings, more opinionated than TerraFX and with some additional helpers to make it a bit easier to use.

A hello world .NET console app with NativeAOT is 1.08 MB. Goes down to 1.08 MB with some tweaking. Goes below 1MB with more tweaking. I have a fully self-contained NativeAOT sample app that renders some fancy animated shaders with DX12 that's about 2MB in size.

I dont think it's exactly what you're looking for, but I've found ComputeSharp to be a great C# GPU tool

Here's some: - MVVM Toolkit - PolySharp - ComputeSharp

If you don't mind being Windows only, I would also recommend trying out (shameless plug) ComputeSharp. You can find it on NuGet and it's available for UWP and WinUI 3 as well with ready to use XAML controls. The point is that it would allow you do your entire rendering via pixel/compute shaders, so it could basically run entirely on the GPU. That'd allow you to render even millions of particles with pretty good performance 🙂

Create a UWP or WinUI 3 applicaation and use the AnimatedComputeShaderPanel from ComputeSharp. The library lets you write shaders (ie. code to run on the GPU) entirely in C#, and handles all the logic to setup a DX12 animated swapchain, similar to Win2D. If you look at the repo you can also see a whole bunch of examples, which you can also try out for yourself. There's also several wiki pages with lots of info.

std::ranges is in c++20, but you can pull in the library it was based on if you use 17 (https://github.com/ericniebler/range-v3)

What you are describing seems to be std::ranges. If you’re interested in understanding how to implement it, I recommend checking out the original reference implementation, rangev3. Trying to implement your own ranges framework is really good practice for learning how to do efficient, advanced generic programming in C++. I highly recommend it as a hobby learning project. But it’s also really, really hard to do correctly, so please just use the stdlib and/or rangev3 in any real project.

In C++23, there will be std::views::cartesian_product. It is already available in the range-v3 library, the one that the standard is based on.

/** * @brief Forwards value equivalent to the std::forward. * * Using cast instead of std::forward to avoid template instantiation. Used by * Eric Niebler in range library. * * @see https://github.com/ericniebler/range-v3 */

It uses a custom zip_iterator (which isn't very good, and you should really use the one from boost or from range-v3).

It contains very little functionality compared to the Eric Niebler’s reference implementation for my liking. Especially views. This will undoubtedly change in the future. But the point is moot, because they are not really supported the is no other option for now other than https://github.com/ericniebler/range-v3.

range-v3 is a great library allowing you to bridge the gap: https://github.com/ericniebler/range-v3

[range-v3](https://github.com/ericniebler/range-v3) which std::ranges was based on has the `to>()` which as far as I know is expected to get into c++23 :)