Re: [PATCH 4/6] wined3d: Use d3d_level_from_gl_info in match_dx10_capable

[Henri Verbeet]

On 14 May 2013 23:46, Stefan Dösinger  wrote:
> @@ -613,14 +635,7 @@ static BOOL match_apple_nonr500ati(const struct wined3d_gl_info *gl_info, const
>  static BOOL match_dx10_capable(const struct wined3d_gl_info *gl_info, const char *gl_renderer,
>          enum wined3d_gl_vendor gl_vendor, enum wined3d_pci_vendor card_vendor, enum wined3d_pci_device device)
>  {
> - /* DX9 cards support 40 single float varyings in hardware, most drivers report 32. ATI misreports
> -
> * 44 varyings. So assume that if we have more than 44 varyings we have a dx10 card.
> -
> * This detection is for the gl_ClipPos varying quirk. If a d3d9 card really supports more than 44
> -
> * varyings and we subtract one in dx9 shaders its not going to hurt us because the dx9 limit is
> -
> * hardcoded
> -
> *
> -
> * dx10 cards usually have 64 varyings */
> - return gl_info->limits.glsl_varyings > 44;
> - return d3d_level_from_gl_info(gl_info) >= 10;
>
> }

I'm not sure d3d_level_from_gl_info() is quite reliable enough for
this. For example, Mesa doesn't implement EXT_gpu_shader4, and
possibly never will. Using an actual hardware limit seems more
reliable.