Re: [PATCH 0/1] MR2839: gdiplus: Improve performance of GdipInvertMatrix.

[Esme Povirk (＠madewokherd)]

On Wed May 17 09:15:05 2023 +0000, Bartosz Kosiorek wrote:
> I have tried such approach, but unfortunately I don't know how to do not
> calculate `matrix_det(matrix)` twice.
> Maybe we could introduce function and provide calculated `det`, like:
> static inline BOOL check_det(GDIPCONST REAL det)
> {
>     return (fabs(det) >= 1e-5);
> }
>
> for
> if(matrix->matrix[1] == 0 && matrix->matrix[2] == 0)
>
> check we are already using that check for optimization. So with proposed
> optimization, we have limited that check to one.

I don't see a way to do this optimization while leaving the logic in GdipIsMatrixInvertible intact. There isn't much logic in there to begin with, so to me it seems fine to duplicate.