cuda like embed opencl

[MatejMagat305]

I know that is crazy idea, but I want share because I think that it would be interasting feature ..., so the situation seems like that it will be more and more computing on GPU ..., I have 2 sugestion which have advantages and dis. for human read and of course for implementation and they have sub variant:

1. use new extension .vcl (or .vl or .vocl) which would have syntax of v,
   a) in compilation it would be translated to string opencl by symbol @OCL("file") and then normal as currently vsl.vcl (I assume it is relative simplier to implementation)
   b) use this file as normal function (in editor) and in code use syntax https://github.com/vlang/v/blob/master/doc/docs.md#closures and call some_name[global][local](args)! - similar to cuda (I asume that it is harder for implementation)

2. use in .v special syntax instead pub fn name() import "example.of.new.cl" and use pub fn (cl)name()!, a) and b) are similar to 1.

of course it is idea after v1.0 and of course it will have backend depended on vsl.vcl or metal or ...

[MatejMagat305]

after 2 years I want add it would be like SYCL (similar)

[JalonSolov]

Have you checked the OpenCL bindings in https://github.com/vlang/vsl/tree/main/vcl ?