Re: wined3d: Use shader and fragment capabilities to select d3d level

[Henri Verbeet]

On 23 June 2014 21:10, Andrei Slăvoiu  wrote:
> +static UINT d3d_level_from_gl_info(const struct wined3d_gl_info *gl_info)
> +{
> - struct shader_caps shader_caps;
> - struct fragment_caps fragment_caps;
> - const struct wined3d_shader_backend_ops *shader_backend;
> - const struct fragment_pipeline *fragment_pipeline;
> -
> - shader_backend = select_shader_backend(gl_info);
> - shader_backend->shader_get_caps(gl_info, &shader_caps);
> -
> - if (shader_caps.vs_version >= 4)
> -
>  return 10;
> - //wine can not use SM 4 on mesa drivers as the necessary functionality is not exposed on compatibility contexts, but still set the pci id accordingly

Please avoid C99 features like // comments.
> - if (shader_caps.vs_version == 3 && gl_info->glsl_version >= MAKEDWORD_VERSION(1, 30))
> -
>  return 10;
> - if (shader_caps.vs_version >= 2)
> -
>  return 9;
> - if (shader_caps.vs_version == 1)
> -
>  return 8;
> -
> - fragment_pipeline = select_fragment_implementation(gl_info, shader_backend);
> - fragment_pipeline->get_caps(gl_info, &fragment_caps);
> -
> - if (fragment_caps.TextureOpCaps & WINED3DTEXOPCAPS_DOTPRODUCT3)
> -
>  return 7;
> - if (gl_info->limits.textures > 1)
> -
>  return 6;

That's "fragment_caps.MaxSimultaneousTextures".

It probably makes sense to reorganize things a bit. For example, the
gl_info parameter to the select_card() callbacks in struct
gl_vendor_selection is really only used by select_card_nvidia_binary()
and select_card_amd_binary() to call d3d_level_from_gl_info(), but I
don't think anything of value would be lost if those used a plain
table like e.g. select_card_amd_mesa(). (At which point you wouldn't
need a callback there anymore either.) d3d_level_from_gl_info() should
probably get the shader and fragment caps passed as parameters, and be
renamed. It probably makes sense to make it return some kind of enum
instead of a plain number, so you can e.g. distinguish between D3D9
SM2 and D3D9 SM3. That would also allow the fallback card selection
code to use plain tables instead of callbacks.

wined3d: Use shader and fragment capabilities to select d3d level

[Andrei Slăvoiu]

În ziua de Mie 25 Iun 2014, la 10:50:49, Henri Verbeet a scris:
> Please avoid C99 features like // comments.

So I suppose C11 features are out of the question? Do you have any suggestion 
for an alternative to _Static_assert?
> > - if (gl_info->limits.textures > 1)
> > -
> >  return 6;
>
> That's "fragment_caps.MaxSimultaneousTextures".

Right, that makes more sense.
> It probably makes sense to reorganize things a bit. For example, the
> gl_info parameter to the select_card() callbacks in struct
> gl_vendor_selection is really only used by select_card_nvidia_binary()
> and select_card_amd_binary() to call d3d_level_from_gl_info(), but I
> don't think anything of value would be lost if those used a plain
> table like e.g. select_card_amd_mesa(). (At which point you wouldn't
> need a callback there anymore either.) d3d_level_from_gl_info() should
> probably get the shader and fragment caps passed as parameters, and be
> renamed. It probably makes sense to make it return some kind of enum
> instead of a plain number, so you can e.g. distinguish between D3D9
> SM2 and D3D9 SM3. That would also allow the fallback card selection
> code to use plain tables instead of callbacks.

Ok, I think I understand what you mean, I'll give it a try.

wined3d: Use shader and fragment capabilities to select d3d level

[Henri Verbeet]

On 25 June 2014 21:25, Andrei Slăvoiu  wrote:
> În ziua de Mie 25 Iun 2014, la 10:50:49, Henri Verbeet a scris:
> > Please avoid C99 features like // comments.
>
> So I suppose C11 features are out of the question? Do you have any suggestion
> for an alternative to _Static_assert?

There's C_ASSERT in winnt.h.