Add experimental packed and in-place DLSS output
- Tag redirected surfaces as R10G10B10A2 ('R3') as well as RGBA16F ('R2'), and
let the native output kernel store the game's R10G10B10A2 texels itself.
- Import linear, exportable game textures (vkd3d-proton patch 0003) into HIP
so the output kernel writes the game's output texture with no game-side copy.
- Add ID3D12DXVKInteropDeviceD4R3 with GetVulkanLinearImageInfo.
- Gate both paths behind D4R_SHIM_OUTPUT_PACKED and D4R_SHIM_INPLACE, and use
them only for split frames of preset M whose output kernel honours the redirect.
- Convert on the host when a recorded packed frame finds the wrong kernel.
- Give the harness an R10G10B10A2 output, and add D4R_SHIM_SPIN_POLL so
frame timing is not quantised to the poll sleep.
- Document the measured removal of the output copies in performance.md.
This commit is contained in:
woutermans committed
1 parent
7c60d53cd0
commit
87365077e0
8 files changed
+469
-22
No files matched your search
@@ -0,0 +1,204 @@
|
||||
diff --git a/include/vkd3d_device_vkd3d_ext.idl b/include/vkd3d_device_vkd3d_ext.idl
|
||||
index 860640d5..fc91e283 100644
|
||||
--- a/include/vkd3d_device_vkd3d_ext.idl
|
||||
+++ b/include/vkd3d_device_vkd3d_ext.idl
|
||||
@@ -132,6 +132,21 @@ interface ID3D12DXVKInteropDeviceD4R2 : ID3D12DXVKInteropDeviceD4R
|
||||
HRESULT RetainExternalResources(ID3D12CommandList *list, IUnknown *resources);
|
||||
}
|
||||
|
||||
+[
|
||||
+ uuid(7b1e5d0a-93c4-4e8f-b6a2-51c08f2e4d73),
|
||||
+ object,
|
||||
+ local,
|
||||
+ pointer_default(unique)
|
||||
+]
|
||||
+interface ID3D12DXVKInteropDeviceD4R3 : ID3D12DXVKInteropDeviceD4R2
|
||||
+{
|
||||
+ /* Textures created linear and exportable (VKD3D_D4R_LINEAR_TEXTURES=1): the
|
||||
+ * dedicated VkDeviceMemory behind the image and the layout of its only
|
||||
+ * subresource, so an external API can read or write the texels in place. */
|
||||
+ HRESULT GetVulkanLinearImageInfo(ID3D12Resource *resource, UINT64 *vk_memory, UINT64 *memory_size,
|
||||
+ UINT64 *offset, UINT64 *row_pitch);
|
||||
+}
|
||||
+
|
||||
[
|
||||
uuid(f3112584-41f9-348d-a59b-00b7e1d285d6),
|
||||
object,
|
||||
diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c
|
||||
index fd753830..3ba6b1b0 100644
|
||||
--- a/libs/vkd3d/device.c
|
||||
+++ b/libs/vkd3d/device.c
|
||||
@@ -4659,7 +4659,8 @@ HRESULT STDMETHODCALLTYPE d3d12_device_QueryInterface(d3d12_device_iface *iface,
|
||||
|| IsEqualGUID(riid, &IID_ID3D12DXVKInteropDevice2)
|
||||
|| IsEqualGUID(riid, &IID_ID3D12DXVKInteropDevice3)
|
||||
|| IsEqualGUID(riid, &IID_ID3D12DXVKInteropDeviceD4R)
|
||||
- || IsEqualGUID(riid, &IID_ID3D12DXVKInteropDeviceD4R2))
|
||||
+ || IsEqualGUID(riid, &IID_ID3D12DXVKInteropDeviceD4R2)
|
||||
+ || IsEqualGUID(riid, &IID_ID3D12DXVKInteropDeviceD4R3))
|
||||
{
|
||||
d3d12_dxvk_interop_device_AddRef(&device->ID3D12DXVKInteropDevice_iface);
|
||||
*object = &device->ID3D12DXVKInteropDevice_iface;
|
||||
@@ -11381,7 +11382,7 @@ static void d3d12_device_replace_vtable(struct d3d12_device *device)
|
||||
}
|
||||
|
||||
extern CONST_VTBL struct ID3D12DeviceExt5Vtbl d3d12_device_vkd3d_ext_vtbl;
|
||||
-extern CONST_VTBL struct ID3D12DXVKInteropDeviceD4R2Vtbl d3d12_dxvk_interop_device_vtbl;
|
||||
+extern CONST_VTBL struct ID3D12DXVKInteropDeviceD4R3Vtbl d3d12_dxvk_interop_device_vtbl;
|
||||
extern CONST_VTBL struct ID3DLowLatencyDeviceVtbl d3d_low_latency_device_vtbl;
|
||||
extern CONST_VTBL struct IAmdExtAntiLagApiVtbl d3d_amd_ext_anti_lag_vtbl;
|
||||
|
||||
diff --git a/libs/vkd3d/device_vkd3d_ext.c b/libs/vkd3d/device_vkd3d_ext.c
|
||||
index fc5290da..f18e3c1a 100644
|
||||
--- a/libs/vkd3d/device_vkd3d_ext.c
|
||||
+++ b/libs/vkd3d/device_vkd3d_ext.c
|
||||
@@ -1145,6 +1145,35 @@ static HRESULT STDMETHODCALLTYPE d3d12_dxvk_interop_device_GetVulkanHeapInfo(d3d
|
||||
return S_OK;
|
||||
}
|
||||
|
||||
+static HRESULT STDMETHODCALLTYPE d3d12_dxvk_interop_device_GetVulkanLinearImageInfo(d3d12_dxvk_interop_device_iface *iface,
|
||||
+ ID3D12Resource *resource, UINT64 *vk_memory, UINT64 *memory_size, UINT64 *offset, UINT64 *row_pitch)
|
||||
+{
|
||||
+ struct d3d12_device *device = d3d12_device_from_ID3D12DXVKInteropDevice(iface);
|
||||
+ const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
|
||||
+ struct d3d12_resource *resource_impl;
|
||||
+ VkImageSubresource subresource;
|
||||
+ VkSubresourceLayout layout;
|
||||
+
|
||||
+ if (!resource || !vk_memory || !memory_size || !offset || !row_pitch)
|
||||
+ return E_INVALIDARG;
|
||||
+
|
||||
+ resource_impl = impl_from_ID3D12Resource(resource);
|
||||
+ if (!(resource_impl->flags & VKD3D_RESOURCE_D4R_LINEAR) ||
|
||||
+ resource_impl->mem.device_allocation.vk_memory == VK_NULL_HANDLE)
|
||||
+ return E_FAIL;
|
||||
+
|
||||
+ subresource.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT;
|
||||
+ subresource.mipLevel = 0;
|
||||
+ subresource.arrayLayer = 0;
|
||||
+ VK_CALL(vkGetImageSubresourceLayout(device->vk_device, resource_impl->res.vk_image, &subresource, &layout));
|
||||
+
|
||||
+ *vk_memory = (UINT64)resource_impl->mem.device_allocation.vk_memory;
|
||||
+ *memory_size = (UINT64)resource_impl->mem.device_allocation.size;
|
||||
+ *offset = (UINT64)resource_impl->mem.offset + layout.offset;
|
||||
+ *row_pitch = (UINT64)layout.rowPitch;
|
||||
+ return S_OK;
|
||||
+}
|
||||
+
|
||||
static HRESULT STDMETHODCALLTYPE d3d12_dxvk_interop_device_SplitCommandListForExternalWait(
|
||||
d3d12_dxvk_interop_device_iface *iface, ID3D12CommandList *list, UINT64 vk_semaphore, UINT64 value)
|
||||
{
|
||||
@@ -1166,7 +1195,7 @@ static HRESULT STDMETHODCALLTYPE d3d12_dxvk_interop_device_RetainExternalResourc
|
||||
return d3d12_command_list_retain_external_resources(cmd_list, resources);
|
||||
}
|
||||
|
||||
-CONST_VTBL struct ID3D12DXVKInteropDeviceD4R2Vtbl d3d12_dxvk_interop_device_vtbl =
|
||||
+CONST_VTBL struct ID3D12DXVKInteropDeviceD4R3Vtbl d3d12_dxvk_interop_device_vtbl =
|
||||
{
|
||||
/* IUnknown methods */
|
||||
d3d12_dxvk_interop_device_QueryInterface,
|
||||
@@ -1204,6 +1233,9 @@ CONST_VTBL struct ID3D12DXVKInteropDeviceD4R2Vtbl d3d12_dxvk_interop_device_vtbl
|
||||
|
||||
/* ID3D12DXVKInteropDeviceD4R2 methods */
|
||||
d3d12_dxvk_interop_device_RetainExternalResources,
|
||||
+
|
||||
+ /* ID3D12DXVKInteropDeviceD4R3 methods */
|
||||
+ d3d12_dxvk_interop_device_GetVulkanLinearImageInfo,
|
||||
};
|
||||
|
||||
static inline struct d3d12_device *d3d12_device_from_ID3DLowLatencyDevice(d3d_low_latency_device_iface *iface)
|
||||
diff --git a/libs/vkd3d/resource.c b/libs/vkd3d/resource.c
|
||||
index e521ec69..0683c75e 100644
|
||||
--- a/libs/vkd3d/resource.c
|
||||
+++ b/libs/vkd3d/resource.c
|
||||
@@ -701,6 +701,28 @@ static bool d3d12_device_supports_universal_color_ds_copy(struct d3d12_device *d
|
||||
return true;
|
||||
}
|
||||
|
||||
+/* d4r experiment (VKD3D_D4R_LINEAR_TEXTURES=1): committed single-level 2D textures in the formats the
|
||||
+ * d4r shim can write a DLSS result in are created with linear tiling on dedicated exportable memory. */
|
||||
+static bool d4r_linear_texture_candidate(const D3D12_HEAP_PROPERTIES *heap_properties,
|
||||
+ const D3D12_RESOURCE_DESC1 *desc, const struct d3d12_resource *resource, const struct vkd3d_format *format)
|
||||
+{
|
||||
+ /* Read from the live process environment until it is set: the d4r shim applies its d4r.ini after
|
||||
+ * this library has loaded. */
|
||||
+ static bool enabled;
|
||||
+ char value[8];
|
||||
+ if (!enabled && vkd3d_get_env_var("VKD3D_D4R_LINEAR_TEXTURES", value, sizeof(value)) && value[0] == '1')
|
||||
+ enabled = true;
|
||||
+ if (!enabled || !heap_properties || heap_properties->Type != D3D12_HEAP_TYPE_DEFAULT)
|
||||
+ return false;
|
||||
+ if (!(resource->flags & VKD3D_RESOURCE_COMMITTED) || (resource->flags & VKD3D_RESOURCE_LINEAR_STAGING_COPY))
|
||||
+ return false;
|
||||
+ if (desc->Dimension != D3D12_RESOURCE_DIMENSION_TEXTURE2D || desc->MipLevels != 1 ||
|
||||
+ desc->DepthOrArraySize != 1 || desc->SampleDesc.Count != 1 ||
|
||||
+ (desc->Flags & D3D12_RESOURCE_FLAG_ALLOW_DEPTH_STENCIL))
|
||||
+ return false;
|
||||
+ return format->vk_format == VK_FORMAT_R16G16B16A16_SFLOAT || format->vk_format == VK_FORMAT_A2B10G10R10_UNORM_PACK32;
|
||||
+}
|
||||
+
|
||||
static HRESULT vkd3d_get_image_create_info(struct d3d12_device *device,
|
||||
const D3D12_HEAP_PROPERTIES *heap_properties,
|
||||
const D3D12_RESOURCE_DESC1 *desc, struct d3d12_resource *resource,
|
||||
@@ -749,6 +771,14 @@ static HRESULT vkd3d_get_image_create_info(struct d3d12_device *device,
|
||||
external_info->handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT;
|
||||
vk_prepend_struct(image_info, external_info);
|
||||
}
|
||||
+ else if (resource && d4r_linear_texture_candidate(heap_properties, desc, resource, format))
|
||||
+ {
|
||||
+ /* d4r: linear, exportable texels an external API can address in place. */
|
||||
+ resource->flags |= VKD3D_RESOURCE_D4R_LINEAR;
|
||||
+ external_info->sType = VK_STRUCTURE_TYPE_EXTERNAL_MEMORY_IMAGE_CREATE_INFO;
|
||||
+ external_info->handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT;
|
||||
+ vk_prepend_struct(image_info, external_info);
|
||||
+ }
|
||||
|
||||
disable_compression = false;
|
||||
|
||||
@@ -905,6 +935,8 @@ static HRESULT vkd3d_get_image_create_info(struct d3d12_device *device,
|
||||
image_info->mipLevels = min(desc->MipLevels, max_miplevel_count(desc));
|
||||
image_info->samples = vk_samples_from_dxgi_sample_desc(&desc->SampleDesc);
|
||||
image_info->tiling = format->vk_image_tiling;
|
||||
+ if (resource && (resource->flags & VKD3D_RESOURCE_D4R_LINEAR))
|
||||
+ image_info->tiling = VK_IMAGE_TILING_LINEAR;
|
||||
image_info->initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
|
||||
|
||||
if (resource && (resource->flags & VKD3D_RESOURCE_COMMITTED) &&
|
||||
@@ -4430,6 +4462,15 @@ HRESULT d3d12_resource_create_committed(struct d3d12_device *device, const D3D12
|
||||
#endif
|
||||
}
|
||||
|
||||
+ if (object->flags & VKD3D_RESOURCE_D4R_LINEAR)
|
||||
+ {
|
||||
+ dedicated_requirements.prefersDedicatedAllocation = VK_TRUE;
|
||||
+ export_info.sType = VK_STRUCTURE_TYPE_EXPORT_MEMORY_ALLOCATE_INFO;
|
||||
+ export_info.pNext = allocate_info.pNext;
|
||||
+ export_info.handleTypes = VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_BIT;
|
||||
+ allocate_info.pNext = &export_info;
|
||||
+ }
|
||||
+
|
||||
/* Requires implies that prefers is also set by spec. */
|
||||
if (dedicated_requirements.prefersDedicatedAllocation)
|
||||
{
|
||||
diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h
|
||||
index b2f66434..c60ffe00 100644
|
||||
--- a/libs/vkd3d/vkd3d_private.h
|
||||
+++ b/libs/vkd3d/vkd3d_private.h
|
||||
@@ -1081,7 +1081,8 @@ enum vkd3d_resource_flag
|
||||
VKD3D_RESOURCE_ZERO_INITIALIZED = (1u << 8),
|
||||
VKD3D_RESOURCE_RETAINED_GPU_REFERENCE = (1u << 9),
|
||||
VKD3D_RESOURCE_COPY_QUEUE_COMPATIBLE = (1u << 10),
|
||||
- VKD3D_RESOURCE_INPUT_ATTACHMENT = (1u << 11)
|
||||
+ VKD3D_RESOURCE_INPUT_ATTACHMENT = (1u << 11),
|
||||
+ VKD3D_RESOURCE_D4R_LINEAR = (1u << 12)
|
||||
};
|
||||
|
||||
#define VKD3D_INVALID_TILE_INDEX (~0u)
|
||||
@@ -5480,7 +5481,7 @@ HRESULT d3d12_command_list_split_for_external_wait(struct d3d12_command_list *li
|
||||
|
||||
HRESULT d3d12_command_list_retain_external_resources(struct d3d12_command_list *list, IUnknown *resources);
|
||||
|
||||
-typedef ID3D12DXVKInteropDeviceD4R2 d3d12_dxvk_interop_device_iface;
|
||||
+typedef ID3D12DXVKInteropDeviceD4R3 d3d12_dxvk_interop_device_iface;
|
||||
|
||||
/* ID3DLowLatencyDevice */
|
||||
typedef ID3DLowLatencyDevice d3d_low_latency_device_iface;
|
||||
Reference in new issue
Block a user