Re: [PATCH 1/6] d3d11: Implement d3d11_device_CheckMultisampleQualityLevels().

[Henri Verbeet]

On 29 January 2016 at 00:51, Matteo Bruni  wrote:
> - if (sample_count > 32)
> -
>    return E_FAIL;

I think that's D3D11_MAX_MULTISAMPLE_SAMPLE_COUNT.
> - if (sample_count > 16)
> - {
> -
>    FIXME("sample_count %u not handled yet.\n", sample_count);
> -
>    return S_OK;
> - }

I think that belongs in wined3d_check_device_multisample_type().
> - hr = wined3d_device_check_multisample_quality_levels(device->wined3d_device,
> -
>        wined3dformat_from_dxgi_format(format), sample_count, quality_level_count);
> - if (hr == WINED3DERR_INVALIDCALL)
> -
>    return E_INVALIDARG;
> - if (hr == WINED3DERR_NOTAVAILABLE)
> -
>    return S_OK;

This is fine, but the d3d10+ API probably makes more sense in the long
run. I.e., eventually we'd want to handle the difference in d3d8/9
instead.
> +HRESULT CDECL wined3d_device_check_multisample_quality_levels(const struct wined3d_device *device,
> -
>    enum wined3d_format_id format_id, enum wined3d_multisample_type multisample_type,
> -
>    DWORD *quality_level_count)

I don't think we really want to introduce a new entry point for this.
I guess the issue is that we can't get at the wined3d object that
created the device from inside d3d11, but I think it would be
preferable to do something about that instead. It seems likely that
CheckFeatureSupport() and CheckFormatSupport() would run into this as
well.

[PATCH 1/6] d3d11: Implement d3d11_device_CheckMultisampleQualityLevels().

[Matteo Bruni]

2016-01-29 15:35 GMT+01:00 Henri Verbeet :
> On 29 January 2016 at 00:51, Matteo Bruni  wrote:
> > - if (sample_count > 16)
> > - {
> > -
> >    FIXME("sample_count %u not handled yet.\n", sample_count);
> > -
> >    return S_OK;
> > - }
>
> I think that belongs in wined3d_check_device_multisample_type().
> > - hr = wined3d_device_check_multisample_quality_levels(device->wined3d_device,
> > -
> >        wined3dformat_from_dxgi_format(format), sample_count, quality_level_count);
> > - if (hr == WINED3DERR_INVALIDCALL)
> > -
> >    return E_INVALIDARG;
> > - if (hr == WINED3DERR_NOTAVAILABLE)
> > -
> >    return S_OK;
>
> This is fine, but the d3d10+ API probably makes more sense in the long
> run. I.e., eventually we'd want to handle the difference in d3d8/9
> instead.

Yes, agreed, but handling the (d3d9-only) D3DMULTISAMPLE_NONMASKABLE
setting outside of wined3d seems to be a bit awkward. Other possible
options might be to keep the non-maskable option (and the enum
wined3d_multisample_type instead of the sample count) but otherwise
follow the d3d10+ API or simply return the multisample_types field
from struct wined3d_format and let the clients deal with it.

Mostly for curiosity I had a quick look at the d3d12 API on MSDN, at a
first glance it mostly resembles the d3d10 interface except that there
isn't a separate function for it i.e. it's been merged into
CheckFeatureSupport().

Anyway, for the time being I've kept the current API although I can
certainly change that.