TensorView: Incorrect storage usage (prior to C++20)

[safocl]

A uint8_t array cannot be used as storage prior
to C++20 without some manipulation (the placement-new operator can be used).
as in struct TensorView

slang/prelude/slang-cuda-prelude.h

Line 4553 in 671b391

struct TensorView

Standart c++17
(https://timsong-cpp.github.io/cppwp/n4659/intro.object#1)

An object is created by a definition, by a new-expression, when implicitly changing the active member of a union, or when a temporary object is created ([conv.rval], [class.temporary]).

Standart c++20

An object is created by a definition, ... by an operation that implicitly creates objects (see below), ... .

https://eel.is/c++draft/intro.object#1.sentence-2
https://eel.is/c++draft/intro.object#3
https://eel.is/c++draft/intro.object#13
https://eel.is/c++draft/intro.object#14
https://eel.is/c++draft/intro.object#15
https://eel.is/c++draft/basic.lval#11

and the alignment of the uint8_t* data; member may differ from the required alignment for the type being converted to.
https://eel.is/c++draft/basic.align#1.sentence-4

https://www.en.cppreference.com/w/cpp/language/reinterpret_cast.html

[safocl]

also

slang/prelude/slang-cuda-prelude.h

Line 4654 in 671b391

template<typename T, unsigned int N>

slang/prelude/slang-cuda-prelude.h

Line 4713 in 671b391

template<typename T, unsigned int N>

template<typename T, unsigned int N>
void store(uint index[N], T val){}

(uint index[N] decay to uint* index and couldn't deduce template parameter 'N')

perhaps it was meant that it should be done like this:

template<typename T, unsigned int N>
void store(uint (&index)[N], T val){}

https://godbolt.org/z/3nWG9KxYT