Yes, to add to the other comments, what many don't realize is that CUDA is an ecosystem, C, C++ and Fortran foremost, however NVidia quickly realized that supporting any programming language community to target PTX was a very good idea.
Their GPUs were re-designed to follow C++ memory model, and many NVidia engineers are seat at ISO C++, yet making CUDA the best way to run heterogenous C++. Something that Intel also realized, by acquiring CodePlay, key players in SYCL, and also employing ISO C++ contributors.
Then there are the Visual Studio and Eclipse plugins, and graphical debuggers that allow even to single step shaders if you so wish.
Their GPUs were re-designed to follow C++ memory model, and many NVidia engineers are seat at ISO C++, yet making CUDA the best way to run heterogenous C++. Something that Intel also realized, by acquiring CodePlay, key players in SYCL, and also employing ISO C++ contributors.
Then there are the Visual Studio and Eclipse plugins, and graphical debuggers that allow even to single step shaders if you so wish.