Re: [PATCH 2/9] wined3d: Prepare wined3d_format_convert_from_float for > 32bpp formats.

[Henri Verbeet]

On Sun, 1 May 2022 at 20:39, Stefan Dösinger  wrote:
> @@ -3689,7 +3690,8 @@ static HRESULT process_vertices_strided(const struct wined3d_device *device, DWO
>                  diffuse_colour = material_diffuse;
>              }
>              wined3d_color_clamp(&diffuse_colour, &diffuse_colour, 0.0f, 1.0f);
> -
>        *((DWORD *)dest_ptr) = wined3d_format_convert_from_float(output_colour_format, &diffuse_colour);
> -
>        wined3d_format_convert_from_float(output_colour_format, &diffuse_colour, argb_color);
> -
>        *((DWORD *)dest_ptr) = argb_color[0];
>        dest_ptr += sizeof(DWORD);
>    }

Why not "wined3d_format_convert_from_float(output_colour_format,
&diffuse_colour, (uint32_t *)dest_ptr);"? This patch always writes 4
uint32_t's in wined3d_format_convert_from_float(), but that could
easily be avoided. And arguably, if we're going to write to an output
pointer in wined3d_format_convert_from_float(), that doesn't need to
be a pointer to a 32-bit value either. (I.e., compare the
upload/download/decompress operations from struct wined3d_format.)