status-im
/
Vulkan-Docs
mirror of https://github.com/status-im/Vulkan-Docs.git
2
0
Fork 0
Code Issues Projects Releases Wiki Activity
Vulkan-Docs/doc/specs/vulkan/chapters/resources.txt

1217 lines
54 KiB
Plaintext
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

// Copyright (c) 2015-2016 The Khronos Group Inc.
// Copyright notice at https://www.khronos.org/registry/speccopyright.html
[[resources]]
= Resource Creation
{apiname} supports two primary resource types: _buffers_ and _images_.
Resources are views of memory with associated formatting and dimensionality.
Buffers are essentially unformatted arrays of bytes whereas images contain
format information, can: be multidimensional and may: have associated
metadata.
[[resources-buffers]]
== Buffers
Buffers represent linear arrays of data which are used for various
purposes by binding them to a graphics or compute pipeline via descriptor
sets or via certain commands, or by directly specifying them as parameters
to certain commands.
Buffers are created by calling:
include::../protos/vkCreateBuffer.txt[]
* pname:device is the logical device that creates the buffer object.
* pname:pCreateInfo is a pointer to an instance of the
sname:VkBufferCreateInfo structure containing parameters affecting
creation of the buffer.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
* pname:pBuffer points to a sname:VkBuffer handle in which the resulting
buffer object is returned.
include::../validity/protos/vkCreateBuffer.txt[]
The definition of sname:VkBufferCreateInfo is:
include::../structs/VkBufferCreateInfo.txt[]
The members of sname:VkBufferCreateInfo have the following meanings:
* pname:sType is the type of this structure.
* pname:pNext is `NULL` or a pointer to an extension-specific structure.
* pname:flags is a bitfield describing additional parameters of the
buffer. See elink:VkBufferCreateFlagBits below for a description of the
supported bits.
* pname:size is the size in bytes of the buffer to be created.
* pname:usage is a bitfield describing the allowed usages of the buffer.
See elink:VkBufferUsageFlagBits below for a description of the supported
bits.
* pname:sharingMode is the sharing mode of the buffer when it will be
accessed by multiple queue families, see elink:VkSharingMode in the
<<resources-sharing,Resource Sharing>> section below for supported
values.
* pname:queueFamilyIndexCount is the number of entries in the
pname:pQueueFamilyIndices array.
* pname:pQueueFamilyIndices is a list of queue families that will
access this buffer (ignored if pname:sharingMode is not
ename:VK_SHARING_MODE_CONCURRENT).
include::../validity/structs/VkBufferCreateInfo.txt[]
Bits which may: be set in pname:usage are:
include::../enums/VkBufferUsageFlagBits.txt[]
* ename:VK_BUFFER_USAGE_TRANSFER_SRC_BIT indicates that the buffer can: be
used as the source of a _transfer command_ (see the definition of
<<synchronization-transfer,ename:VK_PIPELINE_STAGE_TRANSFER_BIT>>).
* ename:VK_BUFFER_USAGE_TRANSFER_DST_BIT indicates that the buffer
can: be used as the destination of a transfer command.
* ename:VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT indicates that the buffer
can: be used to create a sname:VkBufferView suitable for occupying a
sname:VkDescriptorSet slot of type
ename:VK_DESCRIPTOR_TYPE_UNIFORM_TEXEL_BUFFER.
* ename:VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT indicates that the buffer
can: be used to create a sname:VkBufferView suitable for occupying a
sname:VkDescriptorSet slot of type
ename:VK_DESCRIPTOR_TYPE_STORAGE_TEXEL_BUFFER.
* ename:VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT indicates that the buffer can:
be used in a sname:VkDescriptorBufferInfo suitable for occupying a
sname:VkDescriptorSet slot either of type
ename:VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER or
ename:VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC.
* ename:VK_BUFFER_USAGE_STORAGE_BUFFER_BIT indicates that the buffer can:
be used in a sname:VkDescriptorBufferInfo suitable for occupying a
sname:VkDescriptorSet slot either of type
ename:VK_DESCRIPTOR_TYPE_STORAGE_BUFFER or
ename:VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC.
* ename:VK_BUFFER_USAGE_INDEX_BUFFER_BIT indicates that the buffer is
suitable for passing as the pname:buffer parameter to
fname:vkCmdBindIndexBuffer.
* ename:VK_BUFFER_USAGE_VERTEX_BUFFER_BIT indicates that the buffer is
suitable for passing as an element of the pname:pBuffers array to
fname:vkCmdBindVertexBuffers.
* ename:VK_BUFFER_USAGE_INDIRECT_BUFFER_BIT indicates that the buffer is
suitable for passing as the pname:buffer parameter to
fname:vkCmdDrawIndirect, fname:vkCmdDrawIndexedIndirect, or
fname:vkCmdDispatchIndirect.
Any combination of bits can: be specified for pname:usage, but at least one
of the bits must: be set in order to create a valid buffer.
Bits which may: be set in pname:flags are:
include::../enums/VkBufferCreateFlagBits.txt[]
These bitfields have the following meanings:
* ename:VK_BUFFER_CREATE_SPARSE_BINDING_BIT indicates that the buffer will
be backed using sparse memory binding.
* ename:VK_BUFFER_CREATE_SPARSE_RESIDENCY_BIT indicates that the buffer
can: be partially backed using sparse memory binding.
* ename:VK_BUFFER_CREATE_SPARSE_ALIASED_BIT indicates that the buffer will
be backed using sparse memory binding with memory ranges that might also
simultaneously be backing another buffer (or another portion of the same
buffer).
See <<sparsememory-sparseresourcefeatures,Sparse Resource Features>> and
<<features-features,Physical Device Features>> for details of the sparse
memory features supported on a device.
To destroy a buffer, call:
include::../protos/vkDestroyBuffer.txt[]
* pname:device is the logical device that destroys the buffer.
* pname:buffer is the buffer to destroy.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
include::../validity/protos/vkDestroyBuffer.txt[]
[[resources-buffer-views]]
== Buffer Views
A _buffer view_ represents a contiguous range of a buffer and a specific
format to be used to interpret the data. Buffer views are used to enable
shaders to access buffer contents interpreted as formatted data. In order to
create a valid buffer view, the buffer must: have been created with at least
one of the following usage flags:
* ename:VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT
* ename:VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT
A buffer view is created by calling:
include::../protos/vkCreateBufferView.txt[]
* pname:device is the logical device that creates the buffer view.
* pname:pCreateInfo is a pointer to an instance of the
sname:VkBufferViewCreateInfo structure containing parameters to be used
to create the buffer.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
* pname:pView points to a sname:VkBufferView handle in which the resulting
buffer view object is returned.
include::../validity/protos/vkCreateBufferView.txt[]
The definition of sname:VkBufferViewCreateInfo is:
include::../structs/VkBufferViewCreateInfo.txt[]
The members of sname:VkBufferViewCreateInfo have the following meanings:
* pname:sType is the type of this structure.
* pname:pNext is `NULL` or a pointer to an extension-specific structure.
* pname:flags is reserved for future use.
* pname:buffer is a sname:VkBuffer on which the view will be created.
* pname:format is a elink:VkFormat describing the format of the data
elements in the buffer.
* pname:offset is an offset in bytes from the base address of the buffer.
Accesses to the buffer view from shaders use addressing that is relative
to this starting offset.
* pname:range is a size in bytes of the buffer view. If pname:range is
equal to ename:VK_WHOLE_SIZE, the range from pname:offset to the end of
the buffer is used. If ename:VK_WHOLE_SIZE is used and the remaining
size of the buffer is not a multiple of the element size of
pname:format, then the nearest smaller multiple is used.
include::../validity/structs/VkBufferViewCreateInfo.txt[]
To destroy a buffer view, call:
include::../protos/vkDestroyBufferView.txt[]
* pname:device is the logical device that destroys the buffer view.
* pname:bufferView is the buffer view to destroy.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
include::../validity/protos/vkDestroyBufferView.txt[]
[[resources-images]]
== Images
Images represent multidimensional - up to 3 - arrays of data which can: be
used for various purposes (e.g. attachments, textures), by binding them to a
graphics or compute pipeline via descriptor sets, or by directly specifying
them as parameters to certain commands.
Images are created by calling:
include::../protos/vkCreateImage.txt[]
* pname:device is the logical device that creates the image.
* pname:pCreateInfo is a pointer to an instance of the
sname:VkImageCreateInfo structure containing parameters to be used to
create the image.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
* pname:pImage points to a sname:VkImage handle in which the resulting
image object is returned.
include::../validity/protos/vkCreateImage.txt[]
The definition of sname:VkImageCreateInfo is:
include::../structs/VkImageCreateInfo.txt[]
The members of sname:VkImageCreateInfo have the following meanings:
* pname:sType is the type of this structure.
* pname:pNext is `NULL` or a pointer to an extension-specific structure.
* pname:flags is a bitfield describing additional parameters of the image.
See elink:VkImageCreateFlagBits below for a description of the supported
bits.
* pname:imageType is the basic dimensionality of the image, and must: be
one of the values
+
--
include::../enums/VkImageType.txt[]
specifying one-, two-, or three-dimensionality, respectively. Layers in
array textures do not count as a dimension for the purposes of the image
type.
--
* pname:format is a elink:VkFormat describing the format and type of the
data elements that will be contained in the image.
* pname:extent is a slink:VkExtent3D describing the number of data
elements in each dimension of the base level.
* pname:mipLevels describes the number of levels of detail available for
minified sampling of the image.
* pname:arrayLayers is the number of layers in the image.
* pname:samples is the number of sub-data element samples in the image as
defined in elink:VkSampleCountFlagBits. See
<<primsrast-multisampling,Multisampling>>.
* pname:tiling is the tiling arrangement of the data elements in
memory, and must: have one of the values:
+
--
include::../enums/VkImageTiling.txt[]
ename:VK_IMAGE_TILING_OPTIMAL specifies optimal tiling (texels are laid out
in an implementation-dependent arrangement, for more optimal memory access),
and ename:VK_IMAGE_TILING_LINEAR specifies linear tiling (texels are laid
out in memory in row-major order, possibly with some padding on each row).
--
* pname:usage is a bitfield describing the intended usage of the image.
See elink:VkImageUsageFlagBits below for a description of the supported
bits.
* pname:sharingMode is the sharing mode of the image when it will be
accessed by multiple queue families, and must: be one of the values
described for elink:VkSharingMode in the <<resources-sharing,Resource
Sharing>> section below.
* pname:queueFamilyIndexCount is the number of entries in the
pname:pQueueFamilyIndices array.
* pname:pQueueFamilyIndices is a list of queue families that will
access this image (ignored if pname:sharingMode is not
ename:VK_SHARING_MODE_CONCURRENT).
* pname:initialLayout selects the initial elink:VkImageLayout state of all
subresources of the image. See <<resources-image-layouts,Image
Layouts>>. pname:initialLayout must: be ename:VK_IMAGE_LAYOUT_UNDEFINED
or ename:VK_IMAGE_LAYOUT_PREINITIALIZED.
include::../validity/structs/VkImageCreateInfo.txt[]
Valid limits for the image pname:extent, pname:mipLevels, pname:arrayLayers
and pname:samples members are queried with the
flink:vkGetPhysicalDeviceImageFormatProperties command.
Images created with pname:tiling equal to ename:VK_IMAGE_TILING_LINEAR have
further restrictions on their limits and capabilities compared to images
created with pname:tiling equal to ename:VK_IMAGE_TILING_OPTIMAL. Creation
of images with tiling ename:VK_IMAGE_TILING_LINEAR may: not be supported
unless other parameters meet all of the constraints:
* pname:imageType is ename:VK_IMAGE_TYPE_2D
* pname:format is not a depth/stencil format
* pname:mipLevels is 1
* pname:arrayLayers is 1
* pname:samples is ename:VK_SAMPLE_COUNT_1_BIT
* pname:usage only includes ename:VK_IMAGE_USAGE_TRANSFER_SRC_BIT
and/or ename:VK_IMAGE_USAGE_TRANSFER_DST_BIT
Implementations may: support additional limits and capabilities beyond those
listed above. To determine the specific capabilities of an implementation,
query the valid pname:usage bits by calling
flink:vkGetPhysicalDeviceFormatProperties and the valid limits for
pname:mipLevels and pname:arrayLayers by calling
flink:vkGetPhysicalDeviceImageFormatProperties.
Bits which may: be set in pname:usage are:
include::../enums/VkImageUsageFlagBits.txt[]
These bitfields have the following meanings:
* ename:VK_IMAGE_USAGE_TRANSFER_SRC_BIT indicates that the image can: be
used as the source of a transfer command.
* ename:VK_IMAGE_USAGE_TRANSFER_DST_BIT indicates that the image
can: be used as the destination of a transfer command.
* ename:VK_IMAGE_USAGE_SAMPLED_BIT indicates that the image can: be used
to create a sname:VkImageView suitable for occupying a
sname:VkDescriptorSet slot either of type
ename:VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE or
ename:VK_DESCRIPTOR_TYPE_COMBINED_IMAGE_SAMPLER, and be sampled by a
shader.
* ename:VK_IMAGE_USAGE_STORAGE_BIT indicates that the image can: be used
to create a sname:VkImageView suitable for occupying a
sname:VkDescriptorSet slot of type
ename:VK_DESCRIPTOR_TYPE_STORAGE_IMAGE.
* ename:VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT indicates that the image can:
be used to create a sname:VkImageView suitable for use as a color or
resolve attachment in a sname:VkFramebuffer.
* ename:VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT indicates that the
image can: be used to create a sname:VkImageView suitable for use as a
depth/stencil attachment in a sname:VkFramebuffer.
* ename:VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT indicates that the memory
bound to this image will have been allocated with the
ename:VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT (see <<memory>> for more
detail). If this is set, then bits other than
ename:VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT,
ename:VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT, and
ename:VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT mustnot: be set.
* ename:VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT indicates that the image can:
be used to create a sname:VkImageView suitable for occupying
sname:VkDescriptorSet slot of type
ename:VK_DESCRIPTOR_TYPE_INPUT_ATTACHMENT; be read from a shader as an
input attachment; and be used as an input attachment in a framebuffer.
Bits which may: be set in pname:flags are:
include::../enums/VkImageCreateFlagBits.txt[]
These bitfields have the following meanings:
* ename:VK_IMAGE_CREATE_SPARSE_BINDING_BIT indicates that the image will
be backed using sparse memory binding.
* ename:VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT indicates that the image can:
be partially backed using sparse memory binding.
* ename:VK_IMAGE_CREATE_SPARSE_ALIASED_BIT indicates that the image will
be backed using sparse memory binding with memory ranges that might also
simultaneously be backing another image (or another portion of the same
image). Sparse images created with this flag must: also be created with
the ename:VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT.
If any of these three bits are set,
ename:VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT mustnot: also be set.
See <<sparsememory-sparseresourcefeatures,Sparse Resource Features>> and
<<sparsememory-physicalfeatures,Sparse Physical Device Featuers>> for
more details.
* ename:VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT indicates that the image can:
be used to create a slink:VkImageView with a different format from the
image.
* ename:VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT indicates that the image can:
be used to create a slink:VkImageView of type
ename:VK_IMAGE_VIEW_TYPE_CUBE or ename:VK_IMAGE_VIEW_TYPE_CUBE_ARRAY.
The layout of a subresource (mipLevel/arrayLayer) of an image created with
linear tiling is queried by calling:
include::../protos/vkGetImageSubresourceLayout.txt[]
* pname:device is the logical device that owns the image.
* pname:image is the image whose layout is being queried.
* pname:pSubresource is a pointer to a slink:VkImageSubresource structure
selecting a specific image for the subresource.
* pname:pLayout points to a slink:VkSubresourceLayout structure in which
the layout is returned.
include::../validity/protos/vkGetImageSubresourceLayout.txt[]
The definition of the sname:VkImageSubresource structure is:
include::../structs/VkImageSubresource.txt[]
* pname:aspectMask is a elink:VkImageAspectFlags selecting the image
aspect.
* pname:mipLevel selects the mipmap level.
* pname:arrayLayer selects the array layer.
include::../validity/structs/VkImageSubresource.txt[]
Information about the layout of the subresource is returned in a
sname:VkSubresourceLayout structure:
include::../structs/VkSubresourceLayout.txt[]
* pname:offset is the byte offset from the start of the image where the
subresource begins.
* pname:size is the size in bytes of the subresource. pname:size includes
any extra memory that is required based on pname:rowPitch.
* pname:rowPitch describes the number of bytes between each row of texels
in an image.
* pname:arrayPitch describes the number of bytes between each array layer
of an image.
* pname:depthPitch describes the number of bytes between each slice of 3D
image.
include::../validity/structs/VkSubresourceLayout.txt[]
For images created with linear tiling, pname:rowPitch, pname:arrayPitch and
pname:depthPitch describe the layout of the subresource in linear memory.
For uncompressed formats, pname:rowPitch is the number of bytes between
texels with the same x coordinate in adjacent rows (y coordinates differ by
one). pname:arrayPitch is the number of bytes between texels with the same x
and y coordinate in adjacent array layers of the image (array layer values
differ by one). pname:depthPitch is the number of bytes between texels with
the same x and y coordinate in adjacent slices of a 3D image (z coordinates
differ by one). Expressed as an addressing formula, the starting byte of a
texel in the subresource has address:
[source,c]
---------------------------------------------------
// (x,y,z,layer) are in texel coordinates
address(x,y,z,layer) = layer*arrayPitch + z*depthPitch + y*rowPitch + x*texelSize + offset
---------------------------------------------------
For compressed formats, the pname:rowPitch is the number of bytes between
compressed texel blocks in adjacent rows. pname:arrayPitch is the number of
bytes between compressed texel blocks in adjacent array layers.
pname:depthPitch is the number of bytes between compressed texel blocks in
adjacent slices of a 3D image.
[source,c]
---------------------------------------------------
// (x,y,z,layer) are in compressed texel block coordinates
address(x,y,z,layer) = layer*arrayPitch + z*depthPitch + y*rowPitch + x*compressedTexelBlockByteSize + offset;
---------------------------------------------------
pname:arrayPitch is undefined for images that were not created as arrays.
pname:depthPitch is defined only for 3D images.
For color formats, the pname:aspectMask member of sname:VkImageSubresource
must: be ename:VK_IMAGE_ASPECT_COLOR_BIT. For depth/stencil formats,
pname:aspect must: be either ename:VK_IMAGE_ASPECT_DEPTH_BIT or
ename:VK_IMAGE_ASPECT_STENCIL_BIT. On implementations that store depth and
stencil aspects separately, querying each of these subresource layouts will
return a different pname:offset and pname:size representing the region of
memory used for that aspect. On implementations that store depth and stencil
aspects interleaved, the same pname:offset and pname:size are returned and
represent the interleaved memory allocation.
To destroy an image, call:
include::../protos/vkDestroyImage.txt[]
* pname:device is the logical device that destroys the image.
* pname:image is the image to destroy.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
include::../validity/protos/vkDestroyImage.txt[]
[[resources-image-layouts]]
== Image Layouts
Images are stored in implementation-dependent opaque layouts in memory.
Implementations may: support several opaque layouts, and the layout used at
any given time is determined by the elink:VkImageLayout state of the
subresource. Each layout has limitations on what kinds of operations are
supported for subresources using the layout. Applications have control over
which layout each image subresource uses, and can: transition an image
subresource from one layout to another. Transitions can: happen with an
image memory barrier, included as part of a fname:vkCmdPipelineBarrier or a
fname:vkCmdWaitEvents command buffer command (see
<<synchronization-image-memory-barrier>>), or as part of a subpass
dependency within a render pass (see sname:VkSubpassDependency). The image
layout state is per-subresource, and separate subresources of the same image
can: be in different layouts at the same time with one exception - depth and
stencil aspects of a given subresource must: always be in the same layout.
[NOTE]
.Note
====
Each layout may: offer optimal performance for a specific usage of image
memory. For example, an image with a layout of
ename:VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL may: provide optimal
performance for use as a color attachment, but be unsupported for use in
transfer commands. Applications can: transition an image subresource from
one layout to another in order to achieve optimal performance when the
subresource is used for multiple kinds of operations. After initialization,
applications need not use any layout other than the general layout, though
this may: produce suboptimal performance on some implementations.
====
Upon creation, all subresources of an image are initially in the same
layout, where that layout is selected by the
sname:VkImageCreateInfo::pname:initialLayout member. The pname:initialLayout
must: be either ename:VK_IMAGE_LAYOUT_UNDEFINED or
ename:VK_IMAGE_LAYOUT_PREINITIALIZED. If it is
ename:VK_IMAGE_LAYOUT_PREINITIALIZED, then the image data can: be
pre-initialized by the host while using this layout, and the transition away
from this layout will preserve that data. If it is
ename:VK_IMAGE_LAYOUT_UNDEFINED, then the contents of the data are
considered to be undefined, and the transition away from this layout is not
guaranteed to preserve that data. For either of these initial layouts, any
subresources must: be transitioned to another layout before they are
accessed by the device.
Host access to image memory is only well-defined for images created with
ename:VK_IMAGE_TILING_LINEAR tiling and for subresources of those images
which are currently in either the ename:VK_IMAGE_LAYOUT_PREINITIALIZED or
ename:VK_IMAGE_LAYOUT_GENERAL layout.
The set of image layouts consists of:
include::../enums/VkImageLayout.txt[]
The type(s) of device access supported by each layout are:
* ename:VK_IMAGE_LAYOUT_UNDEFINED: Supports no device access. This layout
must: only be used as an pname:initialLayout or as the pname:oldLayout
in an image transition. When transitioning out of this layout, the
contents of the memory are not guaranteed to be preserved.
* ename:VK_IMAGE_LAYOUT_PREINITIALIZED: Supports no device access. This
layout must: only be used as an pname:initialLayout or as the
pname:oldLayout in an image transition. When transitioning out of this
layout, the contents of the memory are preserved. This
layout is intended to be used as the initial layout for an image whose
contents are written by the host, and hence the data can: be written to
memory immediately, without first executing a layout transition.
Currently, ename:VK_IMAGE_LAYOUT_PREINITIALIZED is only useful with
ename:VK_IMAGE_TILING_LINEAR images because there is not a standard
layout defined for ename:VK_IMAGE_TILING_OPTIMAL images.
* ename:VK_IMAGE_LAYOUT_GENERAL: Supports all types of device access.
* ename:VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL: must: only be used as a
color or resolve attachment in a sname:VkFramebuffer. This layout is
valid only for subresources of images created with the
ename:VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT usage bit enabled.
* ename:VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL: must: only be
used as a depth/stencil attachment in a sname:VkFramebuffer. This layout
is valid only for subresources of images created with the
ename:VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT usage bit enabled.
* ename:VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL: must: only be
used as a read-only depth/stencil attachment in a sname:VkFramebuffer
and/or as a read-only image in a shader (which can: be read as a sampled
image, combined image/sampler and/or input attachment). This layout is
valid only for subresources of images created with the
ename:VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT usage bit enabled.
* ename:VK_IMAGE_LAYOUT_SHADER_READ_ONLY_OPTIMAL: must: only be used as a
read-only image in a shader (which can: be read as a sampled image,
combined image/sampler and/or input attachment). This layout is valid
only for subresources of images created with the
ename:VK_IMAGE_USAGE_SAMPLED_BIT or
ename:VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT usage bit enabled.
* ename:VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL: must: only be used as a
source image of a transfer command (see the definition of
<<synchronization-transfer,ename:VK_PIPELINE_STAGE_TRANSFER_BIT>>).
This layout is valid only for subresources of images created with the
ename:VK_IMAGE_USAGE_TRANSFER_SRC_BIT usage bit enabled.
* ename:VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL: must: only be used as a
destination image of a transfer command. This layout is valid only for
subresources of images created with the
ename:VK_IMAGE_USAGE_TRANSFER_DST_BIT usage bit enabled.
For each mechanism of accessing an image in the API, there is a parameter or
structure member that controls the image layout used to access the image.
For transfer commands, this is a parameter to the command (see <<clears>>
and <<copies>>). For use as a framebuffer attachment, this is a member in
the substructures of the sname:VkRenderPassCreateInfo (see
<<renderpass,Render Pass>>). For use in a descriptor set, this is a member
in the sname:VkDescriptorImageInfo structure (see
<<descriptorsets-updates>>). At the time that any command buffer command
accessing an image executes on any queue, the layouts of the image
subresources that are accessed must: all match the layout specified via the
API controlling those accesses.
The image layout of each image subresource must: be well-defined at each
point in the subresource's lifetime. This means that when performing a
layout transition on the subresource, the old layout value must: either
equal the current layout of the subresource (at the time the transition
executes), or else be ename:VK_IMAGE_LAYOUT_UNDEFINED (implying that the
contents of the subresource need not be preserved). The new layout used in a
transition mustnot: be ename:VK_IMAGE_LAYOUT_UNDEFINED or
ename:VK_IMAGE_LAYOUT_PREINITIALIZED.
[[resources-image-views]]
== Image Views
Image objects are not directly accessed by pipeline shaders for reading or
writing image data. Instead, _image views_ representing contiguous ranges of
the image subresources and containing additional metadata are used for that
purpose. Views must: be created on images of compatible types, and must:
represent a valid subset of image subresources.
The types of image views that can: be created are:
include::../enums/VkImageViewType.txt[]
The exact image view type is partially implicit, based on the image's type
and sample count, as well as the view creation parameters as described in
the <<resources-image-views-compatibility,table below>>. This table also
shows which SPIR-V OpTypeImage Dim and Arrayed parameters correspond to each
image view type.
To create an image view, call:
include::../protos/vkCreateImageView.txt[]
* pname:device is the logical device that creates the image view.
* pname:pCreateInfo is a pointer to an instance of the
sname:VkImageViewCreateInfo structure containing parameters to be used
to create the image view.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
* pname:pView points to a sname:VkImageView handle in which the resulting
image view object is returned.
Some of the image creation parameters are inherited by the view. The
remaining parameters are contained in the pname:pCreateInfo.
include::../validity/protos/vkCreateImageView.txt[]
The sname:VkImageViewCreateInfo structure is defined as:
include::../structs/VkImageViewCreateInfo.txt[]
* pname:sType is the type of this structure.
* pname:pNext is `NULL` or a pointer to an extension-specific structure.
* pname:flags is reserved for future use.
* pname:image is a sname:VkImage on which the view will be created.
* pname:viewType is the type of the image view.
* pname:format is a elink:VkFormat describing the format and type used to
interpret data elements in the image.
* pname:components specifies a remapping of color components (or of depth
or stencil components after they have been converted into color
components). See slink:VkComponentMapping.
* pname:subresourceRange selects the set of mipmap levels and array layers
to be accessible to the view.
include::../validity/structs/VkImageViewCreateInfo.txt[]
If pname:image was created with the ename:VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT
flag, pname:format can: be different from the image's format, but if they
are not equal they must: be _compatible_. Image format compatibility is
defined in the <<features-formats-compatibility-classes,Format Compatibility
Classes>> section.
[[resources-image-views-compatibility]]
.Image and image view parameter compatibility requirements
[cols="20%h,35%,45%",options="header"]
|========================================
| Dim, Arrayed, MS | Image parameters | View parameters
| 1D, 0, 0 |
imageType = IMAGE_TYPE_1D +
width >= 1 +
height = 1 +
depth = 1 +
arrayLayers >= 1 +
samples = 1 |
viewType = VIEW_TYPE_1D +
baseArrayLayer >= 0 +
arrayLayers = 1
| 1D, 1, 0 |
imageType = IMAGE_TYPE_1D +
width >= 1 +
height = 1 +
depth = 1 +
arrayLayers >= 1 +
samples = 1 |
viewType = VIEW_TYPE_1D_ARRAY +
baseArrayLayer >= 0 +
arrayLayers >= 1
| 2D, 0, 0 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height >= 1 +
depth = 1 +
arrayLayers >= 1 +
samples = 1 |
viewType = VIEW_TYPE_2D +
baseArrayLayer >= 0 +
arrayLayers = 1
| 2D, 1, 0 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height >= 1 +
depth = 1 +
arrayLayers >= 1 +
samples = 1 |
viewType = VIEW_TYPE_2D_ARRAY +
baseArrayLayer >= 0 +
arrayLayers >= 1
| 2D, 0, 1 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height >= 1 +
depth = 1 +
arrayLayers >= 1 +
samples > 1 |
viewType = VIEW_TYPE_2D +
baseArrayLayer >= 0 +
arrayLayers = 1
| 2D, 1, 1 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height >= 1 +
depth = 1 +
arrayLayers >= 1 +
samples > 1 |
viewType = VIEW_TYPE_2D_ARRAY +
baseArrayLayer >= 0 +
arrayLayers >= 1
| CUBE, 0, 0 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height = width +
depth = 1 +
arrayLayers >= 6 +
samples = 1 +
flags include ename:VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT |
viewType = VIEW_TYPE_CUBE +
baseArrayLayer >= 0 +
arrayLayers = 6
| CUBE, 1, 0 |
imageType = IMAGE_TYPE_2D +
width >= 1 +
height = width +
depth = 1 +
arrayLayers >= 6×N +
samples = 1 +
flags include ename:VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT |
viewType = VIEW_TYPE_CUBE_ARRAY +
baseArrayLayer >= 0 +
arrayLayers = 6×N
| 3D, 0, 0 |
imageType = IMAGE_TYPE_3D +
width >= 1 +
height >= 1 +
depth >= 1 +
arrayLayers = 1 +
samples = 1 |
viewType = VIEW_TYPE_3D +
baseArrayLayer = 0 +
arrayLayers = 1
|========================================
The pname:subresourceRange member is of type sname:VkImageSubresourceRange
and is defined as:
include::../structs/VkImageSubresourceRange.txt[]
* pname:aspectMask is a bitmask indicating which aspect(s) of the image
are included in the view. See elink:VkImageAspectFlagBits.
* pname:baseMipLevel is the first mipmap level accessible to the view.
* pname:levelCount is the number of mipmap levels (starting from
pname:baseMipLevel) accessible to the view.
* pname:baseArrayLayer is the first array layer accessible to the view.
* pname:layerCount is the number of array layers (starting from
pname:baseArrayLayer) accessible to the view.
include::../validity/structs/VkImageSubresourceRange.txt[]
The number of mip-map levels and array layers must: be a subset of the
subresources in the image. If an application wants to use all mip-levels or
layers in an image after the pname:baseMipLevel or pname:baseArrayLayer, it
can: set pname:levelCount and pname:layerCount to the special values
ename:VK_REMAINING_MIP_LEVELS and ename:VK_REMAINING_ARRAY_LAYERS without
knowing the exact number of mip-levels or layers.
For cube and cube array image views, the layers of the image view starting
at pname:baseArrayLayer correspond to faces in the order +X, -X, +Y, -Y, +Z,
-Z. For cube arrays, each set of six sequential layers is a single cube, so
the number of cube maps in a cube map array view is _pname:layerCount / 6_,
and image array layer _pname:baseArrayLayer + i_ is face index _i mod 6_ of
cube _i / 6_. If the number of layers in the view, whether set explicitly in
pname:layerCount or implied by ename:VK_REMAINING_ARRAY_LAYERS, is not a
multiple of 6, behavior when indexing the last cube is undefined.
pname:aspectMask is a bitmask indicating the format being used. Bits which
may: be set include:
include::../enums/VkImageAspectFlagBits.txt[]
The mask must: be only ename:VK_IMAGE_ASPECT_COLOR_BIT,
ename:VK_IMAGE_ASPECT_DEPTH_BIT or ename:VK_IMAGE_ASPECT_STENCIL_BIT if
pname:format is a color, depth-only or stencil-only format, respectively. If
using a depth/stencil format with both depth and stencil components,
pname:aspectMask must: include at least one of
ename:VK_IMAGE_ASPECT_DEPTH_BIT and ename:VK_IMAGE_ASPECT_STENCIL_BIT, and
can: include both.
When using an imageView of a depth/stencil image to populate a descriptor
set (e.g. for sampling in the shader, or for use as an input attachment),
the pname:aspectMask must: only include one bit and selects whether the
imageView is used for depth reads (i.e. using a floating-point sampler or
input attachment in the shader) or stencil reads (i.e. using an unsigned
integer sampler or input attachment in the shader). When an imageView of a
depth/stencil image is used as a depth/stencil framebuffer attachment, the
pname:aspectMask is ignored and both depth and stencil subresources are
used.
The pname:components member is defined as follows:
include::../structs/VkComponentMapping.txt[]
and describes a remapping from components of the image to components of the
vector returned by shader image instructions. This remapping must: be
identity for storage image descriptors, input attachment descriptors, and
framebuffer attachments. The pname:r, pname:g, pname:b, and pname:a members
of pname:components are the values placed in the corresponding components of
the output vector:
include::../enums/VkComponentSwizzle.txt[]
* ename:VK_COMPONENT_SWIZZLE_IDENTITY: the component is set to the
identity swizzle.
* ename:VK_COMPONENT_SWIZZLE_ZERO: the component is set to zero.
* ename:VK_COMPONENT_SWIZZLE_ONE: the component is set to either 1 or 1.0
depending on whether the type of the image view format is integer or
floating-point respectively, as determined by the
<<features-formats-definition,Format Definition>> section for each
elink:VkFormat.
* ename:VK_COMPONENT_SWIZZLE_R: the component is set to the value
of the R component of the image.
* ename:VK_COMPONENT_SWIZZLE_G: the component is set to the value
of the G component of the image.
* ename:VK_COMPONENT_SWIZZLE_B: the component is set to the value
of the B component of the image.
* ename:VK_COMPONENT_SWIZZLE_A: the component is set to the value
of the A component of the image.
include::../validity/structs/VkComponentMapping.txt[]
Setting the identity swizzle on a component is equivalent to setting the
identity mapping on that component. That is:
[[resources-image-views-identity-mappings]]
.Component Mappings Equivalent To ename:VK_COMPONENT_SWIZZLE_IDENTITY
[options="header"]
|====
| Component | Identity Mapping
| pname:components.r | ename:VK_COMPONENT_SWIZZLE_R
| pname:components.g | ename:VK_COMPONENT_SWIZZLE_G
| pname:components.b | ename:VK_COMPONENT_SWIZZLE_B
| pname:components.a | ename:VK_COMPONENT_SWIZZLE_A
|====
To destroy an image view, call:
include::../protos/vkDestroyImageView.txt[]
* pname:device is the logical device that destroys the image view.
* pname:imageView is the image view to destroy.
* pname:pAllocator controls host memory allocation as described in the
<<memory-allocation, Memory Allocation>> chapter.
include::../validity/protos/vkDestroyImageView.txt[]
[[resources-association]]
== Resource Memory Association
Resources are initially created as _virtual allocations_ with no backing
memory. Device memory is allocated separately (see <<memory-device>>) and
then associated with the resource. This association is done differently for
sparse and non-sparse resources.
Resources created with any of the sparse creation flags are considered
sparse resources. Resources created without these flags are non-sparse. The
details on resource memory association for sparse resources is described in
<<sparsememory>>.
Non-sparse resources must: be bound completely and contiguously to a single
slink:VkDeviceMemory object before the resource is passed as a parameter to
any of the following operations:
* creating image or buffer views
* updating descriptor sets
* recording commands in a command buffer
Once bound, the memory binding is immutable for the lifetime of the
resource.
To determine the memory requirements for a non-sparse buffer resource, call:
include::../protos/vkGetBufferMemoryRequirements.txt[]
* pname:device is the logical device that owns the buffer.
* pname:buffer is the buffer to query.
* pname:pMemoryRequirements points to an instance of the
slink:VkMemoryRequirements structure in which the memory requirements of
the buffer object are returned.
include::../validity/protos/vkGetBufferMemoryRequirements.txt[]
To determine the memory requirements for a non-sparse image resource, call:
include::../protos/vkGetImageMemoryRequirements.txt[]
* pname:device is the logical device that owns the image.
* pname:image is the image to query.
* pname:pMemoryRequirements points to an instance of the
slink:VkMemoryRequirements structure in which the memory requirements of
the image object are returned.
include::../validity/protos/vkGetImageMemoryRequirements.txt[]
The sname:VkMemoryRequirements structure returned by
flink:vkGetBufferMemoryRequirements and flink:vkGetImageMemoryRequirements
is defined as follows:
include::../structs/VkMemoryRequirements.txt[]
* pname:size is the size, in bytes, of the memory allocation required: for
the resource.
* pname:alignment is the alignment, in bytes, of the offset within the
allocation required: for the resource.
* pname:memoryTypeBits is a bitfield and contains one bit set for every
supported memory type for the resource. Bit `i` is set if and only if
the memory type `i` in the sname:VkPhysicalDeviceMemoryProperties
structure for the physical device is supported for the resource.
include::../validity/structs/VkMemoryRequirements.txt[]
The implementation guarantees certain properties about the memory
requirements returned by flink:vkGetBufferMemoryRequirements and
flink:vkGetImageMemoryRequirements:
* The pname:memoryTypeBits member always contains at least one bit set.
* If pname:buffer is a sname:VkBuffer, or if pname:image is a
sname:VkImage that was created with a ename:VK_IMAGE_TILING_LINEAR value
in the pname:tiling member of the sname:VkImageCreateInfo structure
passed to fname:vkCreateImage, then the pname:memoryTypeBits member
always contains at least one bit set corresponding to a
sname:VkMemoryType with a pname:propertyFlags that has both the
ename:VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT bit and the
ename:VK_MEMORY_PROPERTY_HOST_COHERENT_BIT bit set. In other words,
mappable coherent memory can: always be attached to these objects.
* The pname:memoryTypeBits member is identical for all sname:VkBuffer
objects created with the same value for the pname:flags and pname:usage
members in the sname:VkBufferCreateInfo structure passed to
fname:vkCreateBuffer. Further, if code:usage1 and code:usage2 of type
elink:VkBufferUsageFlags are such that the bits set in code:usage2 are a
subset of the bits set in code:usage1, and they have the same
pname:flags, then the bits set in pname:memoryTypeBits returned for
code:usage1 must: be a subset of the bits set in pname:memoryTypeBits
returned for code:usage2, for all values of pname:flags.
* The pname:alignment member is identical for all sname:VkBuffer objects
created with the same combination of values for the pname:usage and
pname:flags members in the sname:VkBufferCreateInfo structure passed to
fname:vkCreateBuffer.
* The pname:memoryTypeBits member is identical for all sname:VkImage
objects created with the same combination of values for the pname:tiling
member and the ename:VK_IMAGE_CREATE_SPARSE_BINDING_BIT bit of the
pname:flags member and the ename:VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT
of the pname:usage member in the sname:VkImageCreateInfo structure
passed to fname:vkCreateImage.
* The pname:memoryTypeBits member mustnot: refer to a sname:VkMemoryType
with a pname:propertyFlags that has the
ename:VK_MEMORY_PROPERTY_LAZILY_ALLOCATED_BIT bit set if the
sname:VkImage does not have
ename:VK_IMAGE_USAGE_TRANSIENT_ATTACHMENT_BIT bit set in the pname:usage
member of the sname:VkImageCreateInfo structure passed to
fname:vkCreateImage.
To attach memory to a buffer object, call:
include::../protos/vkBindBufferMemory.txt[]
* pname:device is the logical device that owns the buffer and memory.
* pname:buffer is the buffer.
* pname:memory is a sname:VkDeviceMemory object describing the device
memory to attach.
* pname:memoryOffset is the start offset of the region of pname:memory
which is to be bound to the buffer. The number of bytes returned in the
sname:VkMemoryRequirements::pname:size member in pname:memory, starting
from pname:memoryOffset bytes, will be bound to the specified buffer.
include::../validity/protos/vkBindBufferMemory.txt[]
To attach memory to a image object, call:
include::../protos/vkBindImageMemory.txt[]
* pname:device is the logical device that owns the image and memory.
* pname:image is the image.
* pname:memory is the a sname:VkDeviceMemory object describing the device
memory to attach.
* pname:memoryOffset is the start offset of the region of pname:memory
which is to be bound to the image. The number of bytes returned in the
sname:VkMemoryRequirements::pname:size member in pname:memory, starting
from pname:memoryOffset bytes, will be bound to the specified image.
include::../validity/protos/vkBindImageMemory.txt[]
[[resources-bufferimagegranularity,Buffer-Image Granularity]]
.Buffer-Image Granularity
There is an implementation-dependent limit, pname:bufferImageGranularity,
which specifies a page-like granularity at which buffer, linear image and
optimal image resources must: be placed in adjacent memory locations to
avoid aliasing. Two resources which do not satisfy this granularity
requirement are said to <<resources-memory-aliasing,alias>>. Linear image
resource are images created with ename:VK_IMAGE_TILING_LINEAR and optimal
image resources are those created with ename:VK_IMAGE_TILING_OPTIMAL.
pname:bufferImageGranularity is specified in bytes, and must: be a power of
two. Implementations which do not require such an additional granularity
may: report a value of one.
[NOTE]
.Note
====
pname:bufferImageGranularity is really a granularity between "linear"
resources, including buffers and images with linear tiling, vs. "optimal"
resources, i.e. images with optimal tiling. It would have been better named
"linearOptimalGranularity".
====
Given resourceA at the lower memory offset and resourceB at the higher
memory offset in the same sname:VkDeviceMemory object, where one of the
resources is a buffer or a linear image and the other is an optimal image,
and the following:
resourceA.end = resourceA.memoryOffset + resourceA.size - 1
resourceA.endPage = resourceA.end & ~(bufferImageGranularity-1)
resourceB.start = resourceB.memoryOffset
resourceB.startPage = resourceB.start & ~(bufferImageGranularity-1)
The following property must: hold:
resourceA.endPage < resourceB.startPage
That is, the end of the first resource (A) and the beginning of the second
resource (B) must: be on separate ``pages'' of size
pname:bufferImageGranularity. pname:bufferImageGranularity may: be
different than the physical page size of the memory heap. This
restriction is only needed when a buffer or a linear image is at adjacent
memory location with an optimal image and both will be used simultaneously.
Adjacent buffers' or adjacent images'
memory ranges can: be closer than pname:bufferImageGranularity, provided
they meet the pname:alignment requirement for the objects in question.
Sparse block size in bytes and sparse image and buffer memory alignments
must: all be multiples of the pname:bufferImageGranularity. Therefore,
memory bound to sparse resources naturally satisfies the
pname:bufferImageGranularity.
[[resources-sharing]]
== Resource Sharing Mode
Buffer and image objects are created with a _sharing mode_ controlling how
they can: be accessed from queues. The supported sharing modes are:
include::../enums/VkSharingMode.txt[]
* ename:VK_SHARING_MODE_EXCLUSIVE specifies that access to any range or
subresource of the object will be exclusive to a single queue family at
a time.
* ename:VK_SHARING_MODE_CONCURRENT specifies that concurrent access to any
range or subresource of the object from multiple queue families is
supported.
[NOTE]
.Note
====
ename:VK_SHARING_MODE_CONCURRENT may: result in lower performance access to
the buffer or image than ename:VK_SHARING_MODE_EXCLUSIVE.
====
Ranges of buffers and subresources of image objects created using
ename:VK_SHARING_MODE_EXCLUSIVE must: only be accessed by queues in the same
queue family at any given time. In order for a different queue family to be
able to interpret the memory contents of a range or subresource, the
application must: transfer exclusive ownership of the range or subresource
between the source and destination queue families with the following
sequence of operations:
1. Release exclusive ownership from the source queue family to the
destination queue family.
2. Use semaphores to ensure proper execution control for the ownership
transfer.
3. Acquire exclusive ownership for the destination queue family from the
source queue family.
To release exclusive ownership of a range of a buffer or subresource of an
image object, the application must: execute a buffer or image memory
barrier, respectively (see slink:VkBufferMemoryBarrier and
slink:VkImageMemoryBarrier) on a queue from the source queue family. The
pname:srcQueueFamilyIndex parameter of the barrier must: be set to the
source queue family index, and the pname:dstQueueFamilyIndex parameter to
the destination queue family index.
To acquire exclusive ownership, the application must: execute the same
buffer or image memory barrier on a queue from the destination queue family.
Upon creation, resources using ename:VK_SHARING_MODE_EXCLUSIVE are not owned
by any queue family. A buffer or image memory barrier is not required to
acquire ownership when no queue family owns the resource - it is implicitly
acquired upon first use within a queue. However, images still require a
<<resources-image-layouts,layout transition>> from
ename:VK_IMAGE_LAYOUT_UNDEFINED or ename:VK_IMAGE_LAYOUT_PREINITIALIZED
before being used on the first queue. This layout transition can: either be
accomplished by an image memory barrier or by use in a render pass instance.
Once a queue family has used a range or subresource of an
ename:VK_SHARING_MODE_EXCLUSIVE resource, its contents are undefined to
other queue families unless ownership is transferred. The contents may: also
become undefined for other reasons, e.g. as a result of writes to an image
subresource that aliases the same memory. A queue family can: take ownership
of a range or subresource without an ownership transfer in the same way as
for a resource that was just created, however doing so means any contents
written by other queue families or via incompatible aliases are undefined.
[[resources-memory-aliasing]]
== Memory Aliasing
A range of a sname:VkDeviceMemory allocation is _aliased_ if it is bound to
multiple resources simultaneously, via flink:vkBindImageMemory,
flink:vkBindBufferMemory, or via <<sparsememory-resource-binding,sparse
memory bindings>>. A memory range aliased between two images or two buffers
is defined to be the intersection of the memory ranges bound to the two
resources. A memory range aliased between an image and a buffer is defined
to be the intersection of the memory ranges bound to the two resources,
where each range is first bloated to be aligned to the
pname:bufferImageGranularity. Applications can: alias memory, but use of
multiple aliases is subject to several constraints.
[NOTE]
.Note
====
Memory aliasing can: be useful to reduce the total device memory footprint
of an application, if some large resources are used for disjoint periods of
time.
====
When an opaque, non-ename:VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT image is
bound to an aliased range, all subresources of the image _overlap_ the
range. When a linear image is bound to an aliased range, the subresources
that (according to the image's advertised layout) include bytes from the
aliased range overlap the range. When a
ename:VK_IMAGE_CREATE_SPARSE_RESIDENCY_BIT image has sparse image blocks bound
to an aliased range, only subresources including those sparse image blocks
overlap the range, and when the memory bound to the image's miptail overlaps
an aliased range all subresources in the miptail overlap the range.
Buffers, and linear image subresources in either the
ename:VK_IMAGE_LAYOUT_PREINITIALIZED or ename:VK_IMAGE_LAYOUT_GENERAL
layouts, are _host-accessible subresources_. That is, the host has a
well-defined addressing scheme to interpret the contents, and thus the
layout of the data in memory can be consistently interpreted across aliases
if each of those aliases is a host-accessible subresource. Opaque images and
linear image subresources in other layouts are not host-accessible.
If two aliases are both host-accessible, then they interpret the contents of
the memory in consistent ways, and data written to one alias can: be read by
the other alias.
If either of two aliases is not host-accessible, then the aliases interpret
the contents of the memory differently, and writes via one alias make the
contents of memory partially or completely undefined to the other alias. If
the first alias is a host-accessible subresource, then the bytes affected
are those written by the memory operations according to its addressing
scheme. If the first alias is not host-accessible, then the bytes affected
are those overlapped by the image subresources that were written. If the
second alias is a host-accessible subresource, the affected bytes become
undefined. If the second alias is a not host-accessible, all sparse image
blocks (for sparse partially-resident images) or all subresources (for
non-sparse image and fully resident sparse images) that overlap the affected
bytes become undefined.
If any subresources are made undefined due to writes to an alias, then each
of those subresources must: have its layout transitioned from
ename:VK_IMAGE_LAYOUT_UNDEFINED to a valid layout before it is used, or from
ename:VK_IMAGE_LAYOUT_PREINITIALIZED if the memory has been written by the
host. If any sparse blocks of a sparse image have been made undefined, then
only the subresources containing them must: be transitioned.
Use of an overlapping range by two aliases must: be separated by a memory
dependency using the appropriate access types if at least one of those uses
performs writes, whether the aliases interpret memory consistently or not.
If buffer or image memory barriers are used, the scope of the barrier must:
contain the entire range and/or set of subresources that overlap.
If two aliasing image views are used in the same framebuffer, then the
render pass must: declare the attachments using the
<<renderpass-aliasing,ename:VK_ATTACHMENT_DESCRIPTION_MAY_ALIAS_BIT>>, and
follow the other rules listed in that section.
Access to resources which alias memory from shaders using variables
decorated with code:Coherent are not automatically coherent with each other.
[NOTE]
.Note
====
Memory recycled via an application suballocator (i.e. without freeing and
reallocating the memory objects) is not substantially different from memory
aliasing. However, a suballocator usually waits on a fence before recycling
a region of memory, and signalling a fence involves enough
<<synchronization-implicit-ordering,implicit ordering>> that the above
requirements are all satisfied.
====
Reference in New Issue View Git Blame Copy Permalink