[Mesa-dev] [PATCH 3/3] glsl: Reassociate multiplication of mat*mat*vec.

[Jason Ekstrand]

On Fri, Mar 27, 2015 at 9:22 PM, Matt Turner <mattst88 at gmail.com> wrote:
> The typical case of mat4*mat4*vec4 is 80 scalar multiplications, but
> mat4*(mat4*vec4) is only 32.
>
> On HSW (with vec4 vertex shaders):
> instructions in affected programs:     4420 -> 3194 (-27.74%)
>
> On BDW (with scalar vertex shaders):
> instructions in affected programs:     12756 -> 6726 (-47.27%)

Do you have any actual benchmark results with something that's
affected?  That would be nice to know.
--Jason

> Implementing a general matrix chain ordering is harder (or at least
> tedious) because of having to walk the GLSL IR to create a list of
> multiplicands. I'm guessing that this patch handles 90+% of cases, but
> of course to tell definitively you'd have to implement the general
> thing.
> ---
>  src/glsl/opt_algebraic.cpp | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
>
> diff --git a/src/glsl/opt_algebraic.cpp b/src/glsl/opt_algebraic.cpp
> index 98c852a..a940d2f 100644
> --- a/src/glsl/opt_algebraic.cpp
> +++ b/src/glsl/opt_algebraic.cpp
> @@ -290,6 +290,20 @@ ir_algebraic_visitor::handle_expression(ir_expression *ir)
>     ir_expression *op_expr[4] = {NULL, NULL, NULL, NULL};
>     unsigned int i;
>
> +   if (ir->operation == ir_binop_mul &&
> +       ir->operands[0]->type->is_matrix() &&
> +       ir->operands[1]->type->is_vector()) {
> +      ir_expression *matrix_mul = ir->operands[0]->as_expression();
> +
> +      if (matrix_mul && matrix_mul->operation == ir_binop_mul &&
> +         matrix_mul->operands[0]->type->is_matrix() &&
> +         matrix_mul->operands[1]->type->is_matrix()) {
> +
> +         return mul(matrix_mul->operands[0],
> +                    mul(matrix_mul->operands[1], ir->operands[1]));
> +      }
> +   }
> +
>     assert(ir->get_num_operands() <= 4);
>     for (i = 0; i < ir->get_num_operands(); i++) {
>        if (ir->operands[i]->type->is_matrix())
> --
> 2.0.5
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev