[0035] Ambiguity in linalg matrix coordinates

[jenatali]

Which document does this relate to?
https://github.com/microsoft/hlsl-specs/blob/main/proposals/0035-linalg-matrix.md

Describe the issue you see with the spec
For Matrix::GetCoordinate, the wording is just "Converts a specified index into row and column coordinates." This is somewhat ambiguous. Since matrices are typically MxN (Rows by Columns), while a float2 return would have .xy components, one could interpret the return value as .xy = x and y coordinates, i.e. column then row, or one could interpret the return value as .xy = row then column.

See microsoft/DirectXShaderCompiler#8387 (comment)

[llvm-beanz]

My thinking when I wrote this was that this should effectively match how subscript ([]) on matrix<T,N,M> works. For a matrix m, m[x] returns a row vector and the second subscript in the form m[x][y] returns the yth element of the nth row vector.

So the x component of the matrix would be the row, and the y would be the column in that row.