The assumption about the size of matrices is not correct: it is legitimate to compose a matrix 2x2 with a vector of length 4, in which case it appears that the result has the shape of the first (leftmost) operand. Even for matrices 1xN or Nx1, the result is not always a vector: in general it has the shape of the first operand again, which gets possibly decayed to a vector.
This algorithm is unfortunately not complete yet: sometimes, but not always, vectors of length 1 are decayed to scalars.
Signed-off-by: Giovanni Mascellani gmascellani@codeweavers.com --- libs/vkd3d-shader/hlsl.y | 13 +------------ 1 file changed, 1 insertion(+), 12 deletions(-)
diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index bfe2bbb9..c42e372e 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -971,18 +971,7 @@ static bool expr_common_shape(struct hlsl_ctx *ctx, struct hlsl_type *t1, struct } else { - /* Two vectors or a vector and a matrix (matrix must be 1xn or nx1) */ - unsigned int max_dim_1, max_dim_2; - - max_dim_1 = max(t1->dimx, t1->dimy); - max_dim_2 = max(t2->dimx, t2->dimy); - if (t1->dimx * t1->dimy == t2->dimx * t2->dimy) - { - *type = HLSL_CLASS_VECTOR; - *dimx = max(t1->dimx, t2->dimx); - *dimy = 1; - } - else if (max_dim_1 <= max_dim_2) + if (t1->dimx * t1->dimy <= t2->dimx * t2->dimy) { *type = t1->type; *dimx = t1->dimx;