Re: [PATCH 1/2] dxgi: Partially implement dxgi_adapter_CheckInterfaceSupport().

[Henri Verbeet]

On 12 October 2015 at 03:17, Józef Kucia  wrote:
> @@ -200,9 +200,39 @@ static HRESULT STDMETHODCALLTYPE dxgi_adapter_GetDesc(IDXGIAdapter1 *iface, DXGI
>  static HRESULT STDMETHODCALLTYPE dxgi_adapter_CheckInterfaceSupport(IDXGIAdapter1 *iface,
>          REFGUID guid, LARGE_INTEGER *umd_version)
>  {
> - FIXME("iface %p, guid %s, umd_version %p stub!\n", iface, debugstr_guid(guid), umd_version);
> - struct dxgi_adapter *adapter = impl_from_IDXGIAdapter1(iface);
> - struct wined3d_adapter_identifier adapter_id;
> - HRESULT hr;
> -
> - TRACE("iface %p, guid %s, umd_version %p.\n", iface, debugstr_guid(guid), umd_version);
> -
> - /* This method works only for D3D10 interfaces. */
> - if (!(IsEqualGUID(guid, &IID_ID3D10Device)
> -
>        || IsEqualGUID(guid, &IID_ID3D10Device1)))
> - {
> -
>    WARN("Returning DXGI_ERROR_UNSUPPORTED for %s.\n", debugstr_guid(guid));
> -
>    return DXGI_ERROR_UNSUPPORTED;
> - }
> -
> - FIXME("Determine actual capabilities of adapter.\n");

Do you know what this does for hardware that doesn't support
Direct3D10? Getting the capabilities isn't that hard, we already do
that in dxgi_device_init(). (The wined3d_get_device_caps() call in
particular.)
> #include "dxgi.h"
> +#include "d3d10_1.h"

If you include d3d10_1.h you should be able to drop dxgi.h.

[PATCH 1/2] dxgi: Partially implement dxgi_adapter_CheckInterfaceSupport().

[Józef Kucia]

On Mon, Oct 12, 2015 at 12:16 PM, Henri Verbeet  wrote:
> > - FIXME("Determine actual capabilities of adapter.\n");
>
> Do you know what this does for hardware that doesn't support
> Direct3D10? Getting the capabilities isn't that hard, we already do
> that in dxgi_device_init(). (The wined3d_get_device_caps() call in
> particular.)

A simple check based on the supported shader version is indeed easy.
However, implementing a proper feature level detection requires more
work. I expect that determining which D3D10_FEATURE_LEVEL_9_* we can
support on a given hardware will be more involved. Would you accept a
restrictive implementation which returns S_OK just for hardware that
supports D3D10?

[PATCH 1/2] dxgi: Partially implement dxgi_adapter_CheckInterfaceSupport().

[Henri Verbeet]

On 12 October 2015 at 17:37, Józef Kucia  wrote:
> On Mon, Oct 12, 2015 at 12:16 PM, Henri Verbeet  wrote:
> > > - FIXME("Determine actual capabilities of adapter.\n");
> >
> > Do you know what this does for hardware that doesn't support
> > Direct3D10? Getting the capabilities isn't that hard, we already do
> > that in dxgi_device_init(). (The wined3d_get_device_caps() call in
> > particular.)
>
> A simple check based on the supported shader version is indeed easy.
> However, implementing a proper feature level detection requires more
> work. I expect that determining which D3D10_FEATURE_LEVEL_9_* we can
> support on a given hardware will be more involved. Would you accept a
> restrictive implementation which returns S_OK just for hardware that
> supports D3D10?

Sure. At this point I think we mainly care that it's consistent with
dxgi_device_init().
As far as dxgi is concerned though, I think it's mostly a matter of
making the result of d3d_level_from_caps() from wined3d/directx.c
available somewhere. Perhaps as a separate call, perhaps just as part
of the wined3d_caps structure.