ArrayFire VS sokol-samples

Loads of people have stated why easy GPU interfaces are difficult to create, but we solve many difficult things all the time.

Ultimately I think CPUs are just satisfactory for the vast vast majority of workloads. Servers rarely come with any GPUs to speak of. The ecosystem around GPUs is unattractive. CPUs have SIMD instructions that can help. There are so many reasons not to use GPUs. By the time anyone seriously considers using GPUs they're, in my imagination, typically seriously starved for performance, and looking to control as much of the execution details as possible. GPU programmers don't want an automagic solution.

So I think the demand for easy GPU interfaces is just very weak, and therefore no effort has taken off. The amount of work needed to make it as easy to use as CPUs is massive, and the only reason anyone would even attempt to take this on is to lock you in to expensive hardware (see CUDA).

For a practical suggestion, have you taken a look at https://arrayfire.com/ ? It can run on both CUDA and OpenCL, and it has C++, Rust and Python bindings.

What about Arrayfire. https://github.com/arrayfire/arrayfire

Low-overhead — not our goal, but Flashlight is on par with or outperforming most other ML/DL frameworks with its ArrayFire reference tensor implementation, especially on nonstandard setups where framework overhead matters

Looking for frameworks with Julia + OpenCL I found array fire. It seems quite good, bonus points for rust bindings. I will keep looking for more, Julia completely fell off my radar.

Your Problem may be the lazy evaluation, see this issue: https://github.com/arrayfire/arrayfire/issues/1709

Zig currently doesn't allow chained designators and also doesn't allow to partially initialize arrays and fill up the rest of the array with default values.

E.g. the closest Zig equivalent to this C99 code:

https://github.com/floooh/sokol-samples/blob/b3bc55c4411fa03...

...is this:

https://github.com/floooh/sokol-zig/blob/a4b3c287fadd153a504...

...note how part of the initialization had to be moved out into "code".

There's a ticket about this here, but it's currently not high-priority:

https://github.com/ziglang/zig/issues/6068

This is a simple Pong-like I've written in Raylib compiled to WASM - https://wasm-stuff.netlify.app/pong

Source code is here: https://github.com/rrampage/skitter/tree/master/pong-raylib

It is fairly straightforward to get Raylib running in the browser. I used @flohofwoe's HTML shell file ( https://github.com/floooh/sokol-samples/blob/master/webpage/... ).

PS: also if shader functions could be directly defined "inline" in regular CPU code (and behind a special function pointer type), there would be no need for string literal shenanigans like this:

https://github.com/floooh/sokol-samples/blob/3f10c1a0620cec9...

(and shader compilation would happen during the regular build and would also generate regular compiler errors - with current toolchains that's only possible with a lot of build system magic)

I think something along those lines (rarely used...) was also given by the C++ Committee as justification why so much of the C99 designated init feature set was left out of C++20. But in reality, all C99 designated-init 'sub-features' complement each other nicely once you start using them. It's a great way for building 'data-driven' APIs.

For instance check out this call to create a 3D-API pipeline-state-object in C99:

https://github.com/floooh/sokol-samples/blob/ad9047e228a8441...

I'm using anonymous nested structs extensively for grouping struct items, but I consider the extra field name a feature, not something that should be hidden:

https://github.com/floooh/sokol-samples/blob/bfb30ea00b5948f...

(also note the 'inplace initialization' which follows the state struct definition using C99's designated initialization)

It's an STB-style single-file library, which means the implementation is in a separate ifdef-block from the interface declarations, this allows to compile the implemenentation in a separate source file (which can be a C file), and be used in different source files (which can be C++ files).

For example (using stb_image.h):

https://github.com/floooh/sokol-samples/blob/master/libs/stb...

C99 has made all that a lot easier, you can setup, initialize a struct and call the function taking a pointer to that struct all in one call:

https://github.com/floooh/sokol-samples/blob/89f5825ab5d3690...

The Win32 window system functions and D3D11 are actually very convenient to work with in C99, better than C++ actually.