XM_CALLCONV

[matt77hias]

MSDN:

In addition to the type aliases, you must also use the XM_CALLCONV annotation to make sure the function uses the appropriate calling convention (__fastcall versus __vectorcall) based on your compiler and architecture. Because of limitations with __vectorcall, we recommend that you not use XM_CALLCONV for C++ constructors.

MSDN:

The __vectorcall calling convention specifies that arguments to functions are to be passed in registers, when possible. __vectorcall uses more registers for arguments than __fastcall or the default x64 calling convention use. The __vectorcall calling convention is only supported in native code on x86 and x64 processors that include Streaming SIMD Extensions 2 (SSE2) and above. Use __vectorcall to speed functions that pass several floating-point or SIMD vector arguments and perform operations that take advantage of the arguments loaded in registers.

Given these two documentation snippets, I wonder why XMLoadX (e.g. XMLoadFloat2, etc.) use XM_CALLCONV, since no XMVECTOR or XMMATRIX (or some type aliases) are passed as arguments for these functions?

[walbourn]

The load functions are annotated with XM_CALLCONV to take advantage of the __vectorcall register semantics for return values. This blog post talks a bit about the convention including this comment:

The return value of type HVA/HFA is returned via XMM0(/YMM0): XMM3(/YMM3), one register per element.

The MSDN __vectorcall also states:

Results of __vectorcall functions are returned by value in registers when possible. Results of integer type, including structs or unions of 8 bytes or less, are returned by value in RAX. Vector type results are returned by value in XMM0 or YMM0, depending on size. HVA results have each data element returned by value in registers XMM0:XMM3 or YMM0:YMM3, depending on element size. Result types that don't fit in the corresponding registers are returned by reference to memory allocated by the caller.

TL;DR: Normally in the x64 __fastcall the XMM0L is only used to return float or double values, so vector return types are returned on the stack with RAX pointing to that address. With __fastcall it can return XMVECTOR in XMM0 and XMMATRIX in XMM0:XMM3. For x86, XMM0 was already allowed for 128-bit SIMD return values with __fastcall, but __vectorcall let's it return XMMATRIX in XMM0:XMM3

[matt77hias]

Thanks for the clarification.