Re: [PATCH 0/4] MR2: tests: Test float4 UAV buffer load/store.

[Henri Verbeet (＠hverbeet)]

From patch 3/4:
> @@ -1391,6 +1391,45 @@ static void vkd3d_device_vk_heaps_descriptor_limits_init(struct vkd3d_device_des
>      limits->sampler_max_descriptors = min(limits->sampler_max_descriptors, VKD3D_MAX_DESCRIPTOR_SET_SAMPLERS);
>  }
>
> +static bool d3d12_device_supports_typed_uav_load_additional_formats(const struct d3d12_device *device)
> +{
> +    const struct vkd3d_vk_instance_procs *vk_procs = &device->vkd3d_instance->vk_procs;
> +    const struct vkd3d_format *format;
> +    VkFormatProperties properties;
> +    unsigned int i;
> +
> +    static const DXGI_FORMAT additional_formats[] =
> +    {
> +        DXGI_FORMAT_R32G32B32A32_FLOAT,
> +        DXGI_FORMAT_R32G32B32A32_UINT,
> +        DXGI_FORMAT_R32G32B32A32_SINT,
> +        DXGI_FORMAT_R16G16B16A16_FLOAT,
> +        DXGI_FORMAT_R16G16B16A16_UINT,
> +        DXGI_FORMAT_R16G16B16A16_SINT,
> +        DXGI_FORMAT_R8G8B8A8_UNORM,
> +        DXGI_FORMAT_R8G8B8A8_UINT,
> +        DXGI_FORMAT_R8G8B8A8_SINT,
> +        DXGI_FORMAT_R16_FLOAT,
> +        DXGI_FORMAT_R16_UINT,
> +        DXGI_FORMAT_R16_SINT,
> +        DXGI_FORMAT_R8_UNORM,
> +        DXGI_FORMAT_R8_UINT,
> +        DXGI_FORMAT_R8_SINT,
> +    };
> +
> +    for (i = 0; i < ARRAY_SIZE(additional_formats); ++i)
> +    {
> +        format = vkd3d_get_format(device, additional_formats[i], false);
> +        assert(format);
> +
> +        VK_CALL(vkGetPhysicalDeviceFormatProperties(device->vk_physical_device, format->vk_format, &properties));
> +        if (!((properties.linearTilingFeatures | properties.optimalTilingFeatures) & VK_FORMAT_FEATURE_STORAGE_IMAGE_BIT))
> +            return false;
> +    }
> +
> +    return true;
> +}
> +
>  static HRESULT vkd3d_init_device_caps(struct d3d12_device *device,
>          const struct vkd3d_device_create_info *create_info,
>          struct vkd3d_physical_device_info *physical_device_info,
> @@ -1455,7 +1494,7 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device,
>      else
>          device->feature_options.ResourceBindingTier = D3D12_RESOURCE_BINDING_TIER_3;
>
> -    device->feature_options.TypedUAVLoadAdditionalFormats = features->shaderStorageImageExtendedFormats;
> +    device->feature_options.TypedUAVLoadAdditionalFormats = d3d12_device_supports_typed_uav_load_additional_formats(device);
>      /* GL_INTEL_fragment_shader_ordering, no Vulkan equivalent. */
>      device->feature_options.ROVsSupported = FALSE;
>      /* GL_INTEL_conservative_rasterization, no Vulkan equivalent. */

I think we should also require `shaderStorageImageReadWithoutFormat` here; vkd3d-shader will be unable to properly use those additional formats unless that capability is available.

From patch 4/4:
> @@ -1310,6 +1310,7 @@ static void vkd3d_init_feature_level(struct vkd3d_vulkan_info *vk_info,
>      CHECK_FEATURE(shaderClipDistance);
>      CHECK_FEATURE(shaderCullDistance);
>      CHECK_FEATURE(shaderImageGatherExtended);
> +    CHECK_FEATURE(shaderStorageImageReadWithoutFormat);
>      CHECK_FEATURE(shaderStorageImageWriteWithoutFormat);
>      CHECK_FEATURE(tessellationShader);

I'm not quite sure where we left this discussion last time, but to be clear, I think this change shouldn't be included. Based on [[1]](https://docs.microsoft.com/en-us/windows/win32/direct3d12/hardware-feature-l...), I'd say `shaderStorageImageReadWithoutFormat` should only be a requirement for feature level 12_0 and up. We already include a check for `TypedUAVLoadAdditionalFormats` for feature level 12_0 and up, which should cover this.