Re: [PATCH 2/5] wined3d: Clamp per-vertex point size in vertex shaders.

[Henri Verbeet]

On 2 June 2015 at 16:39, Matteo Bruni  wrote:
> @@ -6711,7 +6721,8 @@ static void set_glsl_shader_program(const struct wined3d_context *context, const
>          WORD map = vshader->reg_maps.input_registers;
>          struct wined3d_string_buffer *tmp_name = string_buffer_get(&priv->string_buffers);
> -
>    reorder_shader_id = generate_param_reorder_function(priv, vshader, pshader, gl_info);
> -
>    reorder_shader_id = generate_param_reorder_function(priv, vshader, pshader,
> -
>            state->gl_primitive_type == GL_POINTS && vshader->reg_maps.point_size, gl_info);

This works because "state->gl_primitive_type == GL_POINTS" is mirrored
in struct vs_compile_args, but it isn't necessarily very obvious or
pretty.

[PATCH 2/5] wined3d: Clamp per-vertex point size in vertex shaders.

[Matteo Bruni]

2015-06-03 12:46 GMT+02:00 Henri Verbeet :
> On 2 June 2015 at 16:39, Matteo Bruni  wrote:
> > @@ -6711,7 +6721,8 @@ static void set_glsl_shader_program(const struct wined3d_context *context, const
> >          WORD map = vshader->reg_maps.input_registers;
> >          struct wined3d_string_buffer *tmp_name = string_buffer_get(&priv->string_buffers);
> > -
> >    reorder_shader_id = generate_param_reorder_function(priv, vshader, pshader, gl_info);
> > -
> >    reorder_shader_id = generate_param_reorder_function(priv, vshader, pshader,
> > -
> >            state->gl_primitive_type == GL_POINTS && vshader->reg_maps.point_size, gl_info);
>
> This works because "state->gl_primitive_type == GL_POINTS" is mirrored
> in struct vs_compile_args, but it isn't necessarily very obvious or
> pretty.

Right. The problem is that vs_compile_args isn't necessarily around
(we might be reusing the previously selected vertex shader) and
recomputing it just for this seemed a bit overkill. Any idea?

[PATCH 2/5] wined3d: Clamp per-vertex point size in vertex shaders.

[Henri Verbeet]

On 3 June 2015 at 16:27, Matteo Bruni  wrote:
> Right. The problem is that vs_compile_args isn't necessarily around
> (we might be reusing the previously selected vertex shader) and
> recomputing it just for this seemed a bit overkill. Any idea?

No particular good one. We could perhaps store it in struct
glsl_vs_program, but it would probably be only marginally better.