path: root/chapters/textures.adoc

// Copyright 2015-2024 The Khronos Group Inc.
//
// SPDX-License-Identifier: CC-BY-4.0

[[textures]]
= Image Operations


== Image Operations Overview

Vulkan Image Operations are operations performed by those SPIR-V Image
Instructions which take an code:OpTypeImage (representing a
sname:VkImageView) or code:OpTypeSampledImage (representing a
(sname:VkImageView, sname:VkSampler) pair).
Read, write, and atomic operations also take texel coordinates as operands,
and return a value based on a neighborhood of texture elements (_texels_)
within the image.
Query operations return properties of the bound image or of the lookup
itself.
The "`Depth`" operand of code:OpTypeImage is ignored.

[NOTE]
.Note
====
Texel is a term which is a combination of the words texture and element.
Early interactive computer graphics supported texture operations on
textures, a small subset of the image operations on images described here.
The discrete samples remain essentially equivalent, however, so we retain
the historical term texel to refer to them.
====

Image Operations include the functionality of the following SPIR-V Image
Instructions:

  * code:OpImageSample* and code:OpImageSparseSample* read one or more
    neighboring texels of the image, and <<textures-texel-filtering,filter>>
    the texel values based on the state of the sampler.
  ** Instructions with code:ImplicitLod in the name
     <<textures-level-of-detail-operation,determine>> the LOD used in the
     sampling operation based on the coordinates used in neighboring
     fragments.
  ** Instructions with code:ExplicitLod in the name
     <<textures-level-of-detail-operation,determine>> the LOD used in the
     sampling operation based on additional coordinates.
  ** Instructions with code:Proj in the name apply homogeneous
     <<textures-projection,projection>> to the coordinates.
  * code:OpImageFetch and code:OpImageSparseFetch return a single texel of
    the image.
    No sampler is used.
  * code:OpImage*Gather and code:OpImageSparse*Gather read neighboring
    texels and <<textures-gather,return a single component>> of each.
  * code:OpImageRead (and code:OpImageSparseRead) and code:OpImageWrite read
    and write, respectively, a texel in the image.
    No sampler is used.
ifdef::VK_NV_shader_image_footprint[]
  * code:OpImageSampleFootprintNV identifies and returns information about
    the set of texels in the image that would be accessed by an equivalent
    code:OpImageSample* instruction.
endif::VK_NV_shader_image_footprint[]
  * code:OpImage*Dref* instructions apply
    <<textures-depth-compare-operation,depth comparison>> on the texel
    values.
  * code:OpImageSparse* instructions additionally return a
    <<textures-sparse-residency,sparse residency>> code.
  * code:OpImageQuerySize, code:OpImageQuerySizeLod,
    code:OpImageQueryLevels, and code:OpImageQuerySamples return properties
    of the image descriptor that would be accessed.
    The image itself is not accessed.
  * code:OpImageQueryLod returns the LOD parameters that would be used in a
    sample operation.
    The actual operation is not performed.
ifdef::VK_QCOM_image_processing[]
  * code:OpImageWeightedSampleQCOM reads a 2D neighborhood of texels and
    computes a weighted average using weight values from a separate weight
    texture.
  * code:opImageBlockMatchSADQCOM and code:opTextureBlockMatchSSD compare 2D
    neighborhoods of texels from two textures.
  * code:OpImageBoxFilterQCOM reads a 2D neighborhood of texels and computes
    a weighted average of the texels.
endif::VK_QCOM_image_processing[]
ifdef::VK_QCOM_image_processing2[]
  * code:opImageBlockMatchWindowSADQCOM and
    code:opImageBlockMatchWindowSSDQCOM compare 2D neighborhoods of texels
    from two textures with the comparison repeated across a window region in
    the target texture.
  * code:opImageBlockMatchGatherSADQCOM and
    code:opImageBlockMatchWindowSSDQCOM compares four 2D neighborhoods of
    texels from a target texture with a single 2D neighborhood in the
    reference texture.
    The R component of each comparison is gathered and returned in the
    output.
endif::VK_QCOM_image_processing2[]


[[textures-texel-coordinate-systems]]
=== Texel Coordinate Systems

Images are addressed by _texel coordinates_.
There are three _texel coordinate systems_:

  * normalized texel coordinates [eq]#[0.0, 1.0]#
  * unnormalized texel coordinates [eq]#[0.0, width / height / depth)#
  * integer texel coordinates [eq]#[0, width / height / depth)#

SPIR-V code:OpImageFetch, code:OpImageSparseFetch, code:OpImageRead,
code:OpImageSparseRead,
ifdef::VK_QCOM_image_processing[]
code:opImageBlockMatchSADQCOM, code:opImageBlockMatchSSDQCOM,
endif::VK_QCOM_image_processing[]
ifdef::VK_QCOM_image_processing2[]
code:opImageBlockMatchWindowSADQCOM, code:opImageBlockMatchWindowSSDQCOM,
endif::VK_QCOM_image_processing2[]
and code:OpImageWrite instructions use integer texel coordinates.

Other image instructions can: use either normalized or unnormalized texel
coordinates (selected by the pname:unnormalizedCoordinates state of the
sampler used in the instruction), but there are
<<samplers-unnormalizedCoordinates,limitations>> on what operations, image
state, and sampler state is supported.
Normalized coordinates are logically
<<textures-normalized-to-unnormalized,converted>> to unnormalized as part of
image operations, and <<textures-normalized-operations,certain steps>> are
only performed on normalized coordinates.
The array layer coordinate is always treated as unnormalized even when other
coordinates are normalized.

Normalized texel coordinates are referred to as [eq]#(s,t,r,q,a)#, with the
coordinates having the following meanings:

  * [eq]#s#: Coordinate in the first dimension of an image.
  * [eq]#t#: Coordinate in the second dimension of an image.
  * [eq]#r#: Coordinate in the third dimension of an image.
  ** [eq]#(s,t,r)# are interpreted as a direction vector for Cube images.
  * [eq]#q#: Fourth coordinate, for homogeneous (projective) coordinates.
  * [eq]#a#: Coordinate for array layer.

The coordinates are extracted from the SPIR-V operand based on the
dimensionality of the image variable and type of instruction.
For code:Proj instructions, the components are in order [eq]#(s, [t,] [r,]
q)#, with [eq]#t# and [eq]#r# being conditionally present based on the
code:Dim of the image.
For non-code:Proj instructions, the coordinates are [eq]#(s [,t] [,r]
[,a])#, with [eq]#t# and [eq]#r# being conditionally present based on the
code:Dim of the image and [eq]#a# being conditionally present based on the
code:Arrayed property of the image.
Projective image instructions are not supported on code:Arrayed images.

Unnormalized texel coordinates are referred to as [eq]#(u,v,w,a)#, with the
coordinates having the following meanings:

  * [eq]#u#: Coordinate in the first dimension of an image.
  * [eq]#v#: Coordinate in the second dimension of an image.
  * [eq]#w#: Coordinate in the third dimension of an image.
  * [eq]#a#: Coordinate for array layer.

Only the [eq]#u# and [eq]#v# coordinates are directly extracted from the
SPIR-V operand, because only 1D and 2D (non-code:Arrayed) dimensionalities
support unnormalized coordinates.
The components are in order [eq]#(u [,v])#, with [eq]#v# being conditionally
present when the dimensionality is 2D.
When normalized coordinates are converted to unnormalized coordinates, all
four coordinates are used.

Integer texel coordinates are referred to as [eq]#(i,j,k,l,n)#, with the
coordinates having the following meanings:

  * [eq]#i#: Coordinate in the first dimension of an image.
  * [eq]#j#: Coordinate in the second dimension of an image.
  * [eq]#k#: Coordinate in the third dimension of an image.
  * [eq]#l#: Coordinate for array layer.
  * [eq]#n#: Index of the sample within the texel.

They are extracted from the SPIR-V operand in order [eq]#(i [,j] [,k] [,l]
[,n])#, with [eq]#j# and [eq]#k# conditionally present based on the code:Dim
of the image, and [eq]#l# conditionally present based on the code:Arrayed
property of the image.
[eq]#n# is conditionally present and is taken from the code:Sample image
operand.

ifdef::VK_EXT_image_sliced_view_of_3d[]
If an accessed image was created from a view using
slink:VkImageViewSlicedCreateInfoEXT and accessed through a
ename:VK_DESCRIPTOR_TYPE_STORAGE_IMAGE descriptor, then the value of [eq]#k#
is incremented by slink:VkImageViewSlicedCreateInfoEXT::pname:sliceOffset,
giving [eq]#k <- sliceOffset {plus} k#.
The image's accessible range in the third dimension is [eq]#k < sliceOffset
+ sliceCount#.
If slink:VkImageViewSlicedCreateInfoEXT::pname:sliceCount is
ename:VK_REMAINING_3D_SLICES_EXT, the range is inherited from the image's
depth extent as specified by <<resources-image-mip-level-sizing, Image Mip
Level Sizing>>.
endif::VK_EXT_image_sliced_view_of_3d[]

For all coordinate types, unused coordinates are assigned a value of zero.

[[textures-texel-coordinate-systems-diagrams]]
image::{images}/vulkantexture0-ll.svg[align="center",title="Texel Coordinate Systems, Linear Filtering",opts="{imageopts}"]
The Texel Coordinate Systems - For the example shown of an 8{times}4 texel
two dimensional image.

  * Normalized texel coordinates:
  ** The [eq]#s# coordinate goes from 0.0 to 1.0.
  ** The [eq]#t# coordinate goes from 0.0 to 1.0.
  * Unnormalized texel coordinates:
  ** The [eq]#u# coordinate within the range 0.0 to 8.0 is within the image,
     otherwise it is outside the image.
  ** The [eq]#v# coordinate within the range 0.0 to 4.0 is within the image,
     otherwise it is outside the image.
  * Integer texel coordinates:
  ** The [eq]#i# coordinate within the range 0 to 7 addresses texels within
     the image, otherwise it is outside the image.
  ** The [eq]#j# coordinate within the range 0 to 3 addresses texels within
     the image, otherwise it is outside the image.
  * Also shown for linear filtering:
  ** Given the unnormalized coordinates [eq]#(u,v)#, the four texels
     selected are [eq]#i~0~j~0~#, [eq]#i~1~j~0~#, [eq]#i~0~j~1~#, and
     [eq]#i~1~j~1~#.
  ** The fractions [eq]#{alpha}# and [eq]#{beta}#.
  ** Given the offset [eq]#{DeltaUpper}~i~# and [eq]#{DeltaUpper}~j~#, the
     four texels selected by the offset are [eq]#i~0~j'~0~#,
     [eq]#i~1~j'~0~#, [eq]#i~0~j'~1~#, and [eq]#i~1~j'~1~#.

ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
[NOTE]
.Note
====
For formats with reduced-resolution components, [eq]#{DeltaUpper}~i~# and
[eq]#{DeltaUpper}~j~# are relative to the resolution of the
highest-resolution component, and therefore may be divided by two relative
to the unnormalized coordinate space of the lower-resolution components.
====
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]

image::{images}/vulkantexture1-ll.svg[align="center",title="Texel Coordinate Systems, Nearest Filtering",opts="{imageopts}"]

The Texel Coordinate Systems - For the example shown of an 8{times}4 texel
two dimensional image.

  * Texel coordinates as above.
    Also shown for nearest filtering:
  ** Given the unnormalized coordinates [eq]#(u,v)#, the texel selected is
     [eq]#ij#.
  ** Given the offset [eq]#{DeltaUpper}~i~# and [eq]#{DeltaUpper}~j~#, the
     texel selected by the offset is [eq]#ij'#.

ifdef::VK_NV_corner_sampled_image[]
For corner-sampled images, the texel samples are located at the grid
intersections instead of the texel centers.

image::{images}/vulkantexture0-corner-alternative-a-ll.svg[align="center",title="Texel Coordinate Systems, Corner Sampling",opts="{imageopts}"]

endif::VK_NV_corner_sampled_image[]


== Conversion Formulas

ifdef::editing-notes[]
[NOTE]
.editing-note
====
(Bill) These Conversion Formulas will likely move to Section 2.7 Fixed-Point
Data Conversions (RGB to sRGB and sRGB to RGB) and section 2.6 Numeric
Representation and Computation (RGB to Shared Exponent and Shared Exponent
to RGB)
====
endif::editing-notes[]


[[textures-RGB-sexp]]
=== RGB to Shared Exponent Conversion

An RGB color [eq]#(red, green, blue)# is transformed to a shared exponent
color [eq]#(red~shared~, green~shared~, blue~shared~, exp~shared~)# as
follows:

First, the components [eq]#(red, green, blue)# are clamped to
[eq]#(red~clamped~, green~clamped~, blue~clamped~)# as:

  {empty}:: [eq]#red~clamped~ = max(0, min(sharedexp~max~, red))#
  {empty}:: [eq]#green~clamped~ = max(0, min(sharedexp~max~, green))#
  {empty}:: [eq]#blue~clamped~ = max(0, min(sharedexp~max~, blue))#

where:

[latexmath]
+++++++++++++++++++
\begin{aligned}
N               & = 9  & \text{number of mantissa bits per component} \\
B               & = 15 & \text{exponent bias} \\
E_{max}         & = 31 & \text{maximum possible biased exponent value} \\
sharedexp_{max} & = \frac{(2^N-1)}{2^N} \times 2^{(E_{max}-B)}
\end{aligned}
+++++++++++++++++++

[NOTE]
.Note
====
// The trailing + is to avoid the asciidoc parser treating the custom role
// as a block attribute in some cases.
[eq]#NaN#, if supported, is handled as in +
<<ieee-754,IEEE 754-2008>> `minNum()` and `maxNum()`.
This results in any [eq]#NaN# being mapped to zero.
====

The largest clamped component, [eq]#max~clamped~# is determined:

  {empty}:: [eq]#max~clamped~ = max(red~clamped~, green~clamped~,
            blue~clamped~)#

A preliminary shared exponent [eq]#exp'# is computed:
[latexmath]
+++++++++++++++++++
\begin{aligned}
exp' =
  \begin{cases}
    \left \lfloor \log_2(max_{clamped}) \right \rfloor + (B+1)
      & \text{for}\  max_{clamped} > 2^{-(B+1)} \\
    0
      & \text{for}\  max_{clamped} \leq 2^{-(B+1)}
  \end{cases}
\end{aligned}
+++++++++++++++++++

The shared exponent [eq]#exp~shared~# is computed:

[latexmath]
+++++++++++++++++++
\begin{aligned}
max_{shared} =
    \left \lfloor
        { \frac{max_{clamped}}{2^{(exp'-B-N)}} + \frac{1}{2} }
    \right \rfloor
\end{aligned}
+++++++++++++++++++

[latexmath]
+++++++++++++++++++
\begin{aligned}
exp_{shared} =
  \begin{cases}
    exp'   & \text{for}\  0 \leq max_{shared} < 2^N \\
    exp'+1 & \text{for}\  max_{shared} = 2^N
  \end{cases}
\end{aligned}
+++++++++++++++++++

Finally, three integer values in the range [eq]#0# to [eq]#2^N^# are
computed:

[latexmath]
+++++++++++++++++++
\begin{aligned}
red_{shared} & =
    \left \lfloor
        { \frac{red_{clamped}}{2^{(exp_{shared}-B-N)}}+ \frac{1}{2} }
    \right \rfloor \\
green_{shared} & =
    \left \lfloor
        { \frac{green_{clamped}}{2^{(exp_{shared}-B-N)}}+ \frac{1}{2} }
    \right \rfloor \\
blue_{shared} & =
    \left \lfloor
        { \frac{blue_{clamped}}{2^{(exp_{shared}-B-N)}}+ \frac{1}{2} }
    \right \rfloor
\end{aligned}
+++++++++++++++++++


[[textures-sexp-RGB]]
=== Shared Exponent to RGB

A shared exponent color [eq]#(red~shared~, green~shared~, blue~shared~,
exp~shared~)# is transformed to an RGB color [eq]#(red, green, blue)# as
follows:

  {empty}:: latexmath:[red = red_{shared} \times {2^{(exp_{shared}-B-N)}}]
  {empty}:: latexmath:[green = green_{shared} \times
            {2^{(exp_{shared}-B-N)}}]
  {empty}:: latexmath:[blue = blue_{shared} \times {2^{(exp_{shared}-B-N)}}]

where:

  {empty}:: [eq]#N = 9# (number of mantissa bits per component)
  {empty}:: [eq]#B = 15# (exponent bias)


== Texel Input Operations

_Texel input instructions_ are SPIR-V image instructions that read from an
image.
_Texel input operations_ are a set of steps that are performed on state,
coordinates, and texel values while processing a texel input instruction,
and which are common to some or all texel input instructions.
They include the following steps, which are performed in the listed order:

  * <<textures-input-validation,Validation operations>>
  ** <<textures-operation-validation,Instruction/Sampler/Image validation>>
  ** <<textures-integer-coordinate-validation,Coordinate validation>>
  ** <<textures-sparse-validation,Sparse validation>>
ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
  ** <<textures-layout-validation,Layout validation>>
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
  * <<textures-format-conversion,Format conversion>>
  * <<textures-texel-replacement,Texel replacement>>
  * <<textures-depth-compare-operation,Depth comparison>>
  * <<textures-conversion-to-rgba,Conversion to RGBA>>
  * <<textures-component-swizzle,Component swizzle>>
ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
  * <<textures-chroma-reconstruction,Chroma reconstruction>>
  * <<textures-sampler-YCbCr-conversion,{YCbCr} conversion>>
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]

For texel input instructions involving multiple texels (for sampling or
gathering), these steps are applied for each texel that is used in the
instruction.
Depending on the type of image instruction, other steps are conditionally
performed between these steps or involving multiple coordinate or texel
values.

ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
If <<textures-chroma-reconstruction,Chroma Reconstruction>> is implicit,
<<textures-texel-filtering, Texel Filtering>> instead takes place during
chroma reconstruction, before <<textures-sampler-YCbCr-conversion,sampler
{YCbCr} conversion>> occurs.
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]

ifdef::VK_QCOM_image_processing[]
The operations described in <<textures-blockmatch,block matching>> and
<<textures-weightimage,weight image sampling>> are performed before
<<textures-conversion-to-rgba,Conversion to RGBA>> and
<<textures-component-swizzle,Component swizzle>>.
endif::VK_QCOM_image_processing[]


[[textures-input-validation]]
=== Texel Input Validation Operations

_Texel input validation operations_ inspect instruction/image/sampler state
or coordinates, and in certain circumstances cause the texel value to be
replaced or become undefined:.
There are a series of validations that the texel undergoes.


[[textures-operation-validation]]
==== Instruction/Sampler/Image View Validation

There are a number of cases where a SPIR-V instruction can: mismatch with
the sampler, the image view, or both, and a number of further cases where
the sampler can: mismatch with the image view.
In such cases the value of the texel returned is undefined:.

These cases include:

  * The sampler pname:borderColor is an integer type and the image view
    pname:format is not one of the elink:VkFormat integer types or a stencil
    component of a depth/stencil format.
  * The sampler pname:borderColor is a float type and the image view
    pname:format is not one of the elink:VkFormat float types or a depth
    component of a depth/stencil format.
ifndef::VK_EXT_border_color_swizzle[]
  * The sampler pname:borderColor is one of the opaque black colors
    (ename:VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK or
    ename:VK_BORDER_COLOR_INT_OPAQUE_BLACK) and the image view
    elink:VkComponentSwizzle for any of the slink:VkComponentMapping
    components is not the <<resources-image-views-identity-mappings,identity
    swizzle>>.
endif::VK_EXT_border_color_swizzle[]
ifdef::VK_EXT_border_color_swizzle[]
  * The sampler pname:borderColor is one of the opaque black colors
    (ename:VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK or
    ename:VK_BORDER_COLOR_INT_OPAQUE_BLACK) and the image view
    elink:VkComponentSwizzle for any of the slink:VkComponentMapping
    components is not the <<resources-image-views-identity-mappings,identity
    swizzle>>, and
    slink:VkPhysicalDeviceBorderColorSwizzleFeaturesEXT::pname:borderColorSwizzleFromImage
    feature is not enabled, and
    slink:VkSamplerBorderColorComponentMappingCreateInfoEXT is not
    specified.
  * slink:VkSamplerBorderColorComponentMappingCreateInfoEXT::pname:components,
    if specified, has a component swizzle that does not match the component
    swizzle of the image view, and either component swizzle is not a form of
    identity swizzle.
  * slink:VkSamplerBorderColorComponentMappingCreateInfoEXT::pname:srgb, if
    specified, does not match the sRGB encoding of the image view.
endif::VK_EXT_border_color_swizzle[]
ifdef::VK_EXT_custom_border_color[]
  * The sampler pname:borderColor is a custom color
    (ename:VK_BORDER_COLOR_FLOAT_CUSTOM_EXT or
    ename:VK_BORDER_COLOR_INT_CUSTOM_EXT) and the supplied
    slink:VkSamplerCustomBorderColorCreateInfoEXT::pname:customBorderColor
    is outside the bounds of the values representable in the image view's
    pname:format.
ifndef::VK_EXT_border_color_swizzle[]
  * The sampler pname:borderColor is a custom color
    (ename:VK_BORDER_COLOR_FLOAT_CUSTOM_EXT or
    ename:VK_BORDER_COLOR_INT_CUSTOM_EXT) and the image view
    elink:VkComponentSwizzle for any of the slink:VkComponentMapping
    components is not the <<resources-image-views-identity-mappings,identity
    swizzle>>.
endif::VK_EXT_border_color_swizzle[]
ifdef::VK_EXT_border_color_swizzle[]
  * The sampler pname:borderColor is a custom color
    (ename:VK_BORDER_COLOR_FLOAT_CUSTOM_EXT or
    ename:VK_BORDER_COLOR_INT_CUSTOM_EXT) and the image view
    elink:VkComponentSwizzle for any of the slink:VkComponentMapping
    components is not the <<resources-image-views-identity-mappings,identity
    swizzle>>, and
    slink:VkPhysicalDeviceBorderColorSwizzleFeaturesEXT::pname:borderColorSwizzleFromImage
    feature is not enabled, and
    slink:VkSamplerBorderColorComponentMappingCreateInfoEXT is not
    specified.
endif::VK_EXT_border_color_swizzle[]
endif::VK_EXT_custom_border_color[]
  * The elink:VkImageLayout of any subresource in the image view does not
    match the slink:VkDescriptorImageInfo::pname:imageLayout used to write
    the image descriptor.
  * The SPIR-V Image Format is not <<spirvenv-image-formats,compatible>>
    with the image view's pname:format.
  * The sampler pname:unnormalizedCoordinates is ename:VK_TRUE and any of
    the <<samplers-unnormalizedCoordinates,limitations of unnormalized
    coordinates>> are violated.
ifdef::VK_EXT_fragment_density_map[]
  * The sampler was created with pname:flags containing
    ename:VK_SAMPLER_CREATE_SUBSAMPLED_BIT_EXT and the image was not created
    with pname:flags containing ename:VK_IMAGE_CREATE_SUBSAMPLED_BIT_EXT.
  * The sampler was not created with pname:flags containing
    ename:VK_SAMPLER_CREATE_SUBSAMPLED_BIT_EXT and the image was created
    with pname:flags containing ename:VK_IMAGE_CREATE_SUBSAMPLED_BIT_EXT.
  * The sampler was created with pname:flags containing
    ename:VK_SAMPLER_CREATE_SUBSAMPLED_BIT_EXT and is used with a function
    that is not code:OpImageSampleImplicitLod or
    code:OpImageSampleExplicitLod, or is used with operands code:Offset or
    code:ConstOffsets.
endif::VK_EXT_fragment_density_map[]
  * The SPIR-V instruction is one of the code:OpImage*Dref* instructions and
    the sampler pname:compareEnable is ename:VK_FALSE
  * The SPIR-V instruction is not one of the code:OpImage*Dref* instructions
    and the sampler pname:compareEnable is ename:VK_TRUE
ifndef::VK_VERSION_1_3,VK_KHR_format_feature_flags2[]
  * The SPIR-V instruction is one of the code:OpImage*Dref* instructions and
    the image view pname:format is not one of the depth/stencil formats with
    a depth component, or the image view aspect is not
    ename:VK_IMAGE_ASPECT_DEPTH_BIT.
endif::VK_VERSION_1_3,VK_KHR_format_feature_flags2[]
ifdef::VK_VERSION_1_3,VK_KHR_format_feature_flags2[]
  * The SPIR-V instruction is one of the code:OpImage*Dref* instructions,
    the image view pname:format is one of the depth/stencil formats, and the
    image view aspect is not ename:VK_IMAGE_ASPECT_DEPTH_BIT.
endif::VK_VERSION_1_3,VK_KHR_format_feature_flags2[]
  * The SPIR-V instruction's image variable's properties are not compatible
    with the image view:
  ** Rules for pname:viewType:
  *** ename:VK_IMAGE_VIEW_TYPE_1D must: have code:Dim = 1D, code:Arrayed =
      0, code:MS = 0.
  *** ename:VK_IMAGE_VIEW_TYPE_2D must: have code:Dim = 2D, code:Arrayed = 0.
  *** ename:VK_IMAGE_VIEW_TYPE_3D must: have code:Dim = 3D, code:Arrayed =
      0, code:MS = 0.
  *** ename:VK_IMAGE_VIEW_TYPE_CUBE must: have code:Dim = Cube, code:Arrayed
      = 0, code:MS = 0.
  *** ename:VK_IMAGE_VIEW_TYPE_1D_ARRAY must: have code:Dim = 1D,
      code:Arrayed = 1, code:MS = 0.
  *** ename:VK_IMAGE_VIEW_TYPE_2D_ARRAY must: have code:Dim = 2D,
      code:Arrayed = 1.
  *** ename:VK_IMAGE_VIEW_TYPE_CUBE_ARRAY must: have code:Dim = Cube,
      code:Arrayed = 1, code:MS = 0.
  ** If the image was created with slink:VkImageCreateInfo::pname:samples
     equal to ename:VK_SAMPLE_COUNT_1_BIT, the instruction must: have
     code:MS = 0.
  ** If the image was created with slink:VkImageCreateInfo::pname:samples
     not equal to ename:VK_SAMPLE_COUNT_1_BIT, the instruction must: have
     code:MS = 1.
  ** If the code:Sampled code:Type of the code:OpTypeImage does not match
     the <<spirv-type,SPIR-V Type>>.
  ** If the <<spirvenv-image-signedness,signedness of any read or sample
     operation>> does not match the signedness of the image's format.
ifdef::VK_NV_corner_sampled_image[]
  * If the image was created with slink:VkImageCreateInfo::pname:flags
    containing ename:VK_IMAGE_CREATE_CORNER_SAMPLED_BIT_NV, the sampler
    addressing modes must: only use a elink:VkSamplerAddressMode of
    ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE.
endif::VK_NV_corner_sampled_image[]
ifdef::VK_NV_shader_image_footprint[]
  * The SPIR-V instruction is code:OpImageSampleFootprintNV with code:Dim =
    2D and pname:addressModeU or pname:addressModeV in the sampler is not
    ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE.
  * The SPIR-V instruction is code:OpImageSampleFootprintNV with code:Dim =
    3D and pname:addressModeU, pname:addressModeV, or pname:addressModeW in
    the sampler is not ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE.
endif::VK_NV_shader_image_footprint[]
ifdef::VK_EXT_custom_border_color[]
  * The sampler was created with a specified
    slink:VkSamplerCustomBorderColorCreateInfoEXT::pname:format which does
    not match the elink:VkFormat of the image view(s) it is sampling.
  * The sampler is sampling an image view of
    ename:VK_FORMAT_B4G4R4A4_UNORM_PACK16,
    ename:VK_FORMAT_B5G6R5_UNORM_PACK16, or
    ename:VK_FORMAT_B5G5R5A1_UNORM_PACK16 format without a specified
    slink:VkSamplerCustomBorderColorCreateInfoEXT::pname:format.
endif::VK_EXT_custom_border_color[]

ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
Only code:OpImageSample* and code:OpImageSparseSample* can: be used with a
sampler or image view that enables <<samplers-YCbCr-conversion,sampler
{YCbCr} conversion>>.

code:OpImageFetch, code:OpImageSparseFetch, code:OpImage*Gather, and
code:OpImageSparse*Gather must: not be used with a sampler or image view
that enables <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>>.

The code:ConstOffset and code:Offset operands must: not be used with a
sampler or image view that enables <<samplers-YCbCr-conversion,sampler
{YCbCr} conversion>>.
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]


[[textures-integer-coordinate-validation]]
==== Integer Texel Coordinate Validation

Integer texel coordinates are validated against the size of the image level,
and the number of layers and number of samples in the image.
For SPIR-V instructions that use integer texel coordinates, this is
performed directly on the integer coordinates.
For instructions that use normalized or unnormalized texel coordinates, this
is performed on the coordinates that result after
<<textures-unnormalized-to-integer,conversion>> to integer texel
coordinates.

If the integer texel coordinates do not satisfy all of the conditions

  {empty}:: [eq]#0 {leq} i < w~s~#
  {empty}:: [eq]#0 {leq} j < h~s~#
  {empty}:: [eq]#0 {leq} k < d~s~#
  {empty}:: [eq]#0 {leq} l < layers#
  {empty}:: [eq]#0 {leq} n < samples#

where:

  {empty}:: [eq]#w~s~ =# width of the image level
  {empty}:: [eq]#h~s~ =# height of the image level
  {empty}:: [eq]#d~s~ =# depth of the image level
  {empty}:: [eq]#layers =# number of layers in the image
  {empty}:: [eq]#samples =# number of samples per texel in the image

then the texel fails integer texel coordinate validation.

There are four cases to consider:

  . Valid Texel Coordinates
+
  * If the texel coordinates pass validation (that is, the coordinates lie
    within the image),
+
then the texel value comes from the value in image memory.

  . Border Texel
+
  * If the texel coordinates fail validation, and
  * If the read is the result of an image sample instruction or image gather
    instruction, and
  * If the image is not a cube image,
ifdef::VK_EXT_non_seamless_cube_map[]
    or if a sampler created with
    ename:VK_SAMPLER_CREATE_NON_SEAMLESS_CUBE_MAP_BIT_EXT is used,
endif::VK_EXT_non_seamless_cube_map[]

+
then the texel is a border texel and <<textures-texel-replacement,texel
replacement>> is performed.

  . Invalid Texel
+
  * If the texel coordinates fail validation, and
  * If the read is the result of an image fetch instruction, image read
    instruction, or atomic instruction,
+
then the texel is an invalid texel and <<textures-texel-replacement,texel
replacement>> is performed.

  . Cube Map Edge or Corner
+
Otherwise the texel coordinates lie beyond the edges or corners of the
selected cube map face, and <<textures-cubemapedge, Cube map edge handling>>
is performed.


[[textures-cubemapedge]]
==== Cube Map Edge Handling

If the texel coordinates lie beyond the edges or corners of the selected
cube map face (as described in the prior section), the following steps are
performed.
Note that this does not occur when using ename:VK_FILTER_NEAREST filtering
within a mip level, since ename:VK_FILTER_NEAREST is treated as using
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE.

  * Cube Map Edge Texel
+
  ** If the texel lies beyond the selected cube map face in either only
     [eq]#i# or only [eq]#j#, then the coordinates [eq]#(i,j)# and the array
     layer [eq]#l# are transformed to select the adjacent texel from the
     appropriate neighboring face.

  * Cube Map Corner Texel
+
  ** If the texel lies beyond the selected cube map face in both [eq]#i# and
     [eq]#j#, then there is no unique neighboring face from which to read
     that texel.
     The texel should: be replaced by the average of the three values of the
     adjacent texels in each incident face.
     However, implementations may: replace the cube map corner texel by
     other methods.
ifndef::VK_EXT_filter_cubic[]
The methods are subject to the constraint that if the three available texels
have the same value, the resulting filtered texel must: have that value.
endif::VK_EXT_filter_cubic[]
ifdef::VK_EXT_filter_cubic[]
The methods are subject to the constraint that for linear filtering if the
three available texels have the same value, the resulting filtered texel
must: have that value, and for cubic filtering if the twelve available
samples have the same value, the resulting filtered texel must: have that
value.
endif::VK_EXT_filter_cubic[]


[[textures-sparse-validation]]
==== Sparse Validation

If the texel reads from an unbound region of a sparse image, the texel is a
_sparse unbound texel_, and processing continues with
<<textures-texel-replacement,texel replacement>>.


ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
[[textures-layout-validation]]
==== Layout Validation

If all planes of a _disjoint_ _multi-planar_ image are not in the same
<<resources-image-layouts,image layout>>, the image must: not be sampled
with <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>> enabled.

endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]


[[textures-format-conversion]]
=== Format Conversion

Texels undergo a format conversion from the elink:VkFormat of the image view
to a vector of either floating point or signed or unsigned integer
components, with the number of components based on the number of components
present in the format.

  * Color formats have one, two, three, or four components, according to the
    format.
  * Depth/stencil formats are one component.
    The depth or stencil component is selected by the pname:aspectMask of
    the image view.

Each component is converted based on its type and size (as defined in the
<<formats-definition,Format Definition>> section for each elink:VkFormat),
using the appropriate equations in <<fundamentals-fp16,16-Bit Floating-Point
Numbers>>, <<fundamentals-fp11,Unsigned 11-Bit Floating-Point Numbers>>,
<<fundamentals-fp10,Unsigned 10-Bit Floating-Point Numbers>>,
<<fundamentals-fixedconv,Fixed-Point Data Conversion>>, and
<<textures-sexp-RGB,Shared Exponent to RGB>>.
Signed integer components smaller than 32 bits are sign-extended.

If the image view format is sRGB, the color components are first converted
as if they are UNORM, and then sRGB to linear conversion is applied to the
R, G, and B components as described in the "`sRGB EOTF`" section of the
<<data-format,Khronos Data Format Specification>>.
The A component, if present, is unchanged.

ifdef::VK_QCOM_ycbcr_degamma[]
[[textures-ycbcr-degamma]]
If
slink:VkSamplerYcbcrConversionYcbcrDegammaCreateInfoQCOM::pname:enableYDegamma
is equal to ename:VK_TRUE, then sRGB to linear conversion is applied to the
G component as described in the "`sRGB EOTF`" section of the
<<data-format,Khronos Data Format Specification>>.
If
slink:VkSamplerYcbcrConversionYcbcrDegammaCreateInfoQCOM::pname:enableCbCrDegamma
is equal to ename:VK_TRUE, then sRGB to linear conversion is applied to the
R and B components as described in the "`sRGB EOTF`" section of the
<<data-format,Khronos Data Format Specification>>.
The A component, if present, is unchanged.
endif::VK_QCOM_ycbcr_degamma[]

If the image view format is block-compressed, then the texel value is first
decoded, then converted based on the type and number of components defined
by the compressed format.


[[textures-texel-replacement]]
=== Texel Replacement

A texel is replaced if it is one (and only one) of:

  * a border texel,
  * an invalid texel, or
  * a sparse unbound texel.

Border texels are replaced with a value based on the image format and the
pname:borderColor of the sampler.
The border color is:

[[textures-border-replacement-color]]
ifdef::VK_EXT_custom_border_color[]
.Border Color [eq]#B#, Custom Border Color slink:VkSamplerCustomBorderColorCreateInfoEXT::pname:customBorderColor [eq]#U#
endif::VK_EXT_custom_border_color[]
ifndef::VK_EXT_custom_border_color[]
.Border Color [eq]#B#
endif::VK_EXT_custom_border_color[]
[options="header",cols="60%,40%"]
|====
| Sampler pname:borderColor                     | Corresponding Border Color
| ename:VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK | [eq]#[B~r~, B~g~, B~b~, B~a~] = [0.0, 0.0, 0.0, 0.0]#
| ename:VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK      | [eq]#[B~r~, B~g~, B~b~, B~a~] = [0.0, 0.0, 0.0, 1.0]#
| ename:VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE      | [eq]#[B~r~, B~g~, B~b~, B~a~] = [1.0, 1.0, 1.0, 1.0]#
| ename:VK_BORDER_COLOR_INT_TRANSPARENT_BLACK   | [eq]#[B~r~, B~g~, B~b~, B~a~] = [0, 0, 0, 0]#
| ename:VK_BORDER_COLOR_INT_OPAQUE_BLACK        | [eq]#[B~r~, B~g~, B~b~, B~a~] = [0, 0, 0, 1]#
| ename:VK_BORDER_COLOR_INT_OPAQUE_WHITE        | [eq]#[B~r~, B~g~, B~b~, B~a~] = [1, 1, 1, 1]#
ifdef::VK_EXT_custom_border_color[]
| ename:VK_BORDER_COLOR_FLOAT_CUSTOM_EXT        | [eq]#[B~r~, B~g~, B~b~, B~a~] = [U~r~, U~g~, U~b~, U~a~]#
| ename:VK_BORDER_COLOR_INT_CUSTOM_EXT          | [eq]#[B~r~, B~g~, B~b~, B~a~] = [U~r~, U~g~, U~b~, U~a~]#
endif::VK_EXT_custom_border_color[]
|====

ifdef::VK_EXT_custom_border_color[]
The custom border color ([eq]#U#) may: be rounded by implementations prior
to texel replacement, but the error introduced by such a rounding must: not
exceed one ULP of the image's pname:format.
endif::VK_EXT_custom_border_color[]

[NOTE]
.Note
====
The names etext:VK_BORDER_COLOR_*\_TRANSPARENT_BLACK,
etext:VK_BORDER_COLOR_*\_OPAQUE_BLACK, and
etext:VK_BORDER_COLOR_*_OPAQUE_WHITE are meant to describe which components
are zeros and ones in the vocabulary of compositing, and are not meant to
imply that the numerical value of ename:VK_BORDER_COLOR_INT_OPAQUE_WHITE is
a saturating value for integers.
====

This is substituted for the texel value by replacing the number of
components in the image format

[[textures-border-replacement-table]]
.Border Texel Components After Replacement
[width="100%",options="header"]
|====
| Texel Aspect or Format      | Component Assignment
| Depth aspect                | [eq]#D                                     = B~r~#
ifdef::VK_EXT_custom_border_color[]
| Stencil aspect              | [eq]#S                                     = B~r~#{sym2}
endif::VK_EXT_custom_border_color[]
ifndef::VK_EXT_custom_border_color[]
| Stencil aspect              | [eq]#S                                     = B~r~#
endif::VK_EXT_custom_border_color[]
| One component color format  | [eq]#Color~r~                              = B~r~#
| Two component color format  | [eq]#[Color~r~,Color~g~]                   = [B~r~,B~g~]#
| Three component color format| [eq]#[Color~r~,Color~g~,Color~b~]          = [B~r~,B~g~,B~b~]#
| Four component color format | [eq]#[Color~r~,Color~g~,Color~b~,Color~a~] = [B~r~,B~g~,B~b~,B~a~]#
ifdef::VK_KHR_maintenance5[]
| Single component alpha format | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [0,0,0,B~a~]#
endif::VK_KHR_maintenance5[]
|====
ifdef::VK_EXT_custom_border_color[]
{sym2} [eq]#S = B~g~# may: be substituted as the replacement method by the
implementation when slink:VkSamplerCreateInfo::pname:borderColor is
ename:VK_BORDER_COLOR_INT_CUSTOM_EXT and
slink:VkSamplerCustomBorderColorCreateInfoEXT::pname:format is
ename:VK_FORMAT_UNDEFINED.
Implementations should: use [eq]#S = B~r~# as the replacement method.
endif::VK_EXT_custom_border_color[]

The value returned by a read of an invalid texel is undefined:, unless that
read operation is from a buffer resource and the pname:robustBufferAccess
feature is enabled.
In that case, an invalid texel is replaced as described by the
<<features-robustBufferAccess, pname:robustBufferAccess>> feature.
ifdef::VK_VERSION_1_3,VK_EXT_image_robustness,VK_EXT_robustness2[]
If the access is to an image resource and the x, y, z, or layer coordinate
validation fails and
ifdef::VK_VERSION_1_3,VK_EXT_image_robustness[]
the <<features-robustImageAccess, pname:robustImageAccess>> feature is
enabled, then zero must: be returned for the R, G, and B components, if
present.
Either zero or one must: be returned for the A component, if present.
ifdef::VK_EXT_robustness2[If]
endif::VK_VERSION_1_3,VK_EXT_image_robustness[]
ifdef::VK_EXT_robustness2[]
If the <<features-robustImageAccess2, pname:robustImageAccess2>> feature is
enabled, zero values must: be returned.
endif::VK_EXT_robustness2[]
If only the sample index was invalid, the values returned are undefined:.
endif::VK_VERSION_1_3,VK_EXT_image_robustness,VK_EXT_robustness2[]

ifdef::VK_VERSION_1_3,VK_EXT_image_robustness[]
Additionally, if the <<features-robustImageAccess, pname:robustImageAccess>>
feature is enabled,
ifdef::VK_EXT_robustness2[]
but the <<features-robustImageAccess2, pname:robustImageAccess2>> feature is
not,
endif::VK_EXT_robustness2[]
any invalid texels may: be expanded to four components prior to texel
replacement.
This means that components not present in the image format may be replaced
with 0 or may undergo <<textures-conversion-to-rgba,conversion to RGBA>> as
normal.
endif::VK_VERSION_1_3,VK_EXT_image_robustness[]

ifdef::VK_EXT_robustness2[]
Loads from a null descriptor return a four component color value of all
zeros.
However, for storage images and storage texel buffers using an explicit
SPIR-V Image Format, loads from a null descriptor may: return an alpha value
of 1 (float or integer, depending on format) if the format does not include
alpha.
endif::VK_EXT_robustness2[]

If the
slink:VkPhysicalDeviceSparseProperties::pname:residencyNonResidentStrict
property is ename:VK_TRUE, a sparse unbound texel is replaced with 0 or 0.0
values for integer and floating-point components of the image format,
respectively.

If pname:residencyNonResidentStrict is ename:VK_FALSE, the value of the
sparse unbound texel is undefined:.


[[textures-depth-compare-operation]]
=== Depth Compare Operation

If the image view has a depth/stencil format, the depth component is
selected by the pname:aspectMask, and the operation is an code:OpImage*Dref*
instruction, a depth comparison is performed.
The result is [eq]#1.0# if the comparison evaluates to [eq]#true#, and
[eq]#0.0# otherwise.
This value replaces the depth component [eq]#D#.

The compare operation is selected by the elink:VkCompareOp value set by
slink:VkSamplerCreateInfo::pname:compareOp.
The reference value from the SPIR-V operand [eq]#D~ref~# and the texel depth
value [eq]#D~tex~# are used as the _reference_ and _test_ values,
respectively, in that operation.

If the image being sampled has an unsigned normalized fixed-point format,
then [eq]#D~ref~# is clamped to [eq]#[0,1]# before the compare operation.


[[textures-conversion-to-rgba]]
=== Conversion to RGBA

The texel is expanded from one, two, or three components to four components
based on the image base color:

[[textures-texel-color-rgba-conversion-table]]
.Texel Color After Conversion To RGBA
[width="100%", options="header", cols="<4,<6"]
|====
| Texel Aspect or Format      | RGBA Color
| Depth aspect                | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [D,0,0,one]#
| Stencil aspect              | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [S,0,0,one]#
| One component color format  | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [Color~r~,0,0,one]#
| Two component color format  | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [Color~r~,Color~g~,0,one]#
| Three component color format| [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [Color~r~,Color~g~,Color~b~,one]#
| Four component color format | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [Color~r~,Color~g~,Color~b~,Color~a~]#
ifdef::VK_KHR_maintenance5[]
| One alpha component color format | [eq]#[Color~r~,Color~g~,Color~b~, Color~a~] = [0,0,0,Color~a~]#
endif::VK_KHR_maintenance5[]
|====

where [eq]#one = 1.0f# for floating-point formats and depth aspects, and
[eq]#one = 1# for integer formats and stencil aspects.


[[textures-component-swizzle]]
=== Component Swizzle

ifndef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
All texel input instructions apply a _swizzle_ based on the
elink:VkComponentSwizzle enums in the pname:components member of the
slink:VkImageViewCreateInfo structure for the image being read.
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
All texel input instructions apply a _swizzle_ based on:

  * the elink:VkComponentSwizzle enums in the pname:components member of the
    slink:VkImageViewCreateInfo structure for the image being read if
    <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>> is not enabled,
    and
  * the elink:VkComponentSwizzle enums in the pname:components member of the
    slink:VkSamplerYcbcrConversionCreateInfo structure for the
    <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>> if sampler
    {YCbCr} conversion is enabled.

endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]

The swizzle can: rearrange the components of the texel, or substitute zero
or one for any components.
It is defined as follows for each color [eq]#component#:


[latexmath]
+++++++++++++++++++
\begin{aligned}
Color'_{component} & =
\begin{cases}
Color_r          & \text{for RED swizzle}   \\
Color_g          & \text{for GREEN swizzle} \\
Color_b          & \text{for BLUE swizzle}  \\
Color_a          & \text{for ALPHA swizzle} \\
0                & \text{for ZERO swizzle}  \\
one              & \text{for ONE swizzle} \\
identity         & \text{for IDENTITY swizzle}
\end{cases}
\end{aligned}
+++++++++++++++++++

where:

[latexmath]
+++++++++++++++++++
\begin{aligned}
one & =
\begin{cases}
& 1.0\text{f}  & \text{for floating point components} \\
& 1            & \text{for integer components} \\
\end{cases}
\\
identity & =
\begin{cases}
& Color_r          & \text{for}\ component = r \\
& Color_g          & \text{for}\ component = g \\
& Color_b          & \text{for}\ component = b \\
& Color_a          & \text{for}\ component = a \\
\end{cases}
\end{aligned}
+++++++++++++++++++

If the border color is one of the etext:VK_BORDER_COLOR_*_OPAQUE_BLACK enums
and the elink:VkComponentSwizzle is not the
<<resources-image-views-identity-mappings,identity swizzle>> for all
components, the value of the texel after swizzle is undefined:.

ifndef::VK_KHR_maintenance5[]
If the image view has a depth/stencil format and the
elink:VkComponentSwizzle is ename:VK_COMPONENT_SWIZZLE_ONE, the value of the
texel after swizzle is undefined:.
endif::VK_KHR_maintenance5[]
ifdef::VK_KHR_maintenance5[]
If the image view has a depth/stencil format and the
elink:VkComponentSwizzle is ename:VK_COMPONENT_SWIZZLE_ONE, and
sname:VkPhysicalDeviceMaintenance5PropertiesKHR::pname:depthStencilSwizzleOneSupport
is not set to ename:VK_TRUE, the value of the texel after swizzle is
undefined:.
endif::VK_KHR_maintenance5[]


[[textures-sparse-residency]]
=== Sparse Residency

code:OpImageSparse* instructions return a structure which includes a
_residency code_ indicating whether any texels accessed by the instruction
are sparse unbound texels.
This code can: be interpreted by the code:OpImageSparseTexelsResident
instruction which converts the residency code to a boolean value.


ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
[[textures-chroma-reconstruction]]
=== Chroma Reconstruction

In some color models, the color representation is defined in terms of
monochromatic light intensity (often called "`luma`") and color differences
relative to this intensity, often called "`chroma`".
It is common for color models other than RGB to represent the chroma
components at lower spatial resolution than the luma component.
This approach is used to take advantage of the eye's lower spatial
sensitivity to color compared with its sensitivity to brightness.
Less commonly, the same approach is used with additive color, since the
green component dominates the eye's sensitivity to light intensity and the
spatial sensitivity to color introduced by red and blue is lower.

Lower-resolution components are "`downsampled`" by resizing them to a lower
spatial resolution than the component representing luminance.
This process is also commonly known as "`chroma subsampling`".
There is one luminance sample in each texture texel, but each chrominance
sample may be shared among several texels in one or both texture dimensions.

  * "`etext:_444`" formats do not spatially downsample chroma values
    compared with luma: there are unique chroma samples for each texel.
  * "`etext:_422`" formats have downsampling in the x dimension
    (corresponding to _u_ or _s_ coordinates): they are sampled at half the
    resolution of luma in that dimension.
  * "`etext:_420`" formats have downsampling in the x dimension
    (corresponding to _u_ or _s_ coordinates) and the y dimension
    (corresponding to _v_ or _t_ coordinates): they are sampled at half the
    resolution of luma in both dimensions.

The process of reconstructing a full color value for texture access involves
accessing both chroma and luma values at the same location.
To generate the color accurately, the values of the lower-resolution
components at the location of the luma samples must be reconstructed from
the lower-resolution sample locations, an operation known here as "`chroma
reconstruction`" irrespective of the actual color model.

The location of the chroma samples relative to the luma coordinates is
determined by the pname:xChromaOffset and pname:yChromaOffset members of the
slink:VkSamplerYcbcrConversionCreateInfo structure used to create the
sampler {YCbCr} conversion.

The following diagrams show the relationship between unnormalized (_u_,_v_)
coordinates and (_i_,_j_) integer texel positions in the luma component
(shown in black, with circles showing integer sample positions) and the
texel coordinates of reduced-resolution chroma components, shown as crosses
in red.

[NOTE]
.Note
====
If the chroma values are reconstructed at the locations of the luma samples
by means of interpolation, chroma samples from outside the image bounds are
needed; these are determined according to <<textures-wrapping-operation>>.
These diagrams represent this by showing the bounds of the "`chroma texel`"
extending beyond the image bounds, and including additional chroma sample
positions where required for interpolation.
The limits of a sample for etext:NEAREST sampling is shown as a grid.
====

image::{images}/chromasamples_422_cosited.svg[align="center",title="422 downsampling, xChromaOffset=COSITED_EVEN",opts="{imageopts}"]

image::{images}/chromasamples_422_midpoint.svg[align="center",title="422 downsampling, xChromaOffset=MIDPOINT",opts="{imageopts}"]

image::{images}/chromasamples_420_xcosited_ycosited.svg[align="center",title="420 downsampling, xChromaOffset=COSITED_EVEN, yChromaOffset=COSITED_EVEN",opts="{imageopts}"]

image::{images}/chromasamples_420_xmidpoint_ycosited.svg[align="center",title="420 downsampling, xChromaOffset=MIDPOINT, yChromaOffset=COSITED_EVEN",opts="{imageopts}"]

image::{images}/chromasamples_420_xcosited_ymidpoint.svg[align="center",title="420 downsampling, xChromaOffset=COSITED_EVEN, yChromaOffset=MIDPOINT",opts="{imageopts}"]

image::{images}/chromasamples_420_xmidpoint_ymidpoint.svg[align="center",title="420 downsampling, xChromaOffset=MIDPOINT, yChromaOffset=MIDPOINT",opts="{imageopts}"]

Reconstruction is implemented in one of two ways:

If the format of the image that is to be sampled sets
ename:VK_FORMAT_FEATURE_SAMPLED_IMAGE_YCBCR_CONVERSION_CHROMA_RECONSTRUCTION_EXPLICIT_BIT,
or the slink:VkSamplerYcbcrConversionCreateInfo's
pname:forceExplicitReconstruction is set to ename:VK_TRUE, reconstruction is
performed as an explicit step independent of filtering, described in the
<<textures-explicit-reconstruction>> section.

If the format of the image that is to be sampled does not set
ename:VK_FORMAT_FEATURE_SAMPLED_IMAGE_YCBCR_CONVERSION_CHROMA_RECONSTRUCTION_EXPLICIT_BIT
and if the slink:VkSamplerYcbcrConversionCreateInfo's
pname:forceExplicitReconstruction is set to ename:VK_FALSE, reconstruction
is performed as an implicit part of filtering prior to color model
conversion, with no separate post-conversion texel filtering step, as
described in the <<textures-implict-reconstruction,Implicit Reconstruction>>
section.


[[textures-explicit-reconstruction]]
==== Explicit Reconstruction

  * If the pname:chromaFilter member of the
    slink:VkSamplerYcbcrConversionCreateInfo structure is
    ename:VK_FILTER_NEAREST:
  ** If the format's R and B components are reduced in resolution in just
     width by a factor of two relative to the G component (i.e. this is a
     "`etext:_422`" format), the latexmath:[\tau_{ijk}[level\]] values
     accessed by <<textures-texel-filtering,texel filtering>> are
     reconstructed as follows:
+
[latexmath]
++++++++++++++
\begin{aligned}
\tau_R'(i, j) & = \tau_R(\left\lfloor{i\times 0.5}\right\rfloor, j)[level] \\
\tau_B'(i, j) & = \tau_B(\left\lfloor{i\times 0.5}\right\rfloor, j)[level]
\end{aligned}
++++++++++++++

  ** If the format's R and B components are reduced in resolution in width
     and height by a factor of two relative to the G component (i.e. this is
     a "`etext:_420`" format), the latexmath:[\tau_{ijk}[level\]] values
     accessed by <<textures-texel-filtering,texel filtering>> are
     reconstructed as follows:
+
[latexmath]
++++++++++++++
\begin{aligned}
\tau_R'(i, j) & = \tau_R(\left\lfloor{i\times 0.5}\right\rfloor, \left\lfloor{j\times 0.5}\right\rfloor)[level] \\
\tau_B'(i, j) & = \tau_B(\left\lfloor{i\times 0.5}\right\rfloor, \left\lfloor{j\times 0.5}\right\rfloor)[level]
\end{aligned}
++++++++++++++
+
[NOTE]
.Note
====
pname:xChromaOffset and pname:yChromaOffset have no effect if
pname:chromaFilter is ename:VK_FILTER_NEAREST for explicit reconstruction.
====

  * If the pname:chromaFilter member of the
    slink:VkSamplerYcbcrConversionCreateInfo structure is
    ename:VK_FILTER_LINEAR:
  ** If the format's R and B components are reduced in resolution in just
     width by a factor of two relative to the G component (i.e. this is a
     "`etext:_422`" format):
  *** If pname:xChromaOffset is ename:VK_CHROMA_LOCATION_COSITED_EVEN:
+
[latexmath]
+++++
\tau_{RB}'(i,j) = \begin{cases}
\tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor,j)[level], & 0.5 \times i = \left\lfloor{0.5 \times i}\right\rfloor\\
0.5\times\tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor,j)[level] + \\
0.5\times\tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor + 1,j)[level], & 0.5 \times i \neq \left\lfloor{0.5 \times i}\right\rfloor
\end{cases}
+++++
+
  *** If pname:xChromaOffset is ename:VK_CHROMA_LOCATION_MIDPOINT:
+
[latexmath]
+++++
\tau_{RB}'(i,j) = \begin{cases}
0.25 \times \tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor - 1,j)[level] + \\
0.75 \times \tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor,j)[level], & 0.5 \times i = \left\lfloor{0.5 \times i}\right\rfloor\\
0.75 \times \tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor,j)[level] + \\
0.25 \times \tau_{RB}(\left\lfloor{i\times 0.5}\right\rfloor + 1,j)[level], & 0.5 \times i \neq \left\lfloor{0.5 \times i}\right\rfloor
\end{cases}
+++++

  ** If the format's R and B components are reduced in resolution in width
     and height by a factor of two relative to the G component (i.e. this is
     a "`etext:_420`" format), a similar relationship applies.
     Due to the number of options, these formulae are expressed more
     concisely as follows:
+
[latexmath]
+++++
\begin{aligned}
  i_{RB} & =
    \begin{cases}
      0.5 \times (i) & \textrm{xChromaOffset = COSITED}\_\textrm{EVEN} \\
      0.5 \times (i - 0.5) & \textrm{xChromaOffset = MIDPOINT}
    \end{cases}\\
  j_{RB} & =
    \begin{cases}
      0.5 \times (j) & \textrm{yChromaOffset = COSITED}\_\textrm{EVEN} \\
      0.5 \times (j - 0.5) & \textrm{yChromaOffset = MIDPOINT}
    \end{cases}\\
  \\
  i_{floor} & = \left\lfloor i_{RB} \right\rfloor \\
  j_{floor} & = \left\lfloor j_{RB} \right\rfloor \\
  \\
  i_{frac} & = i_{RB} - i_{floor} \\
  j_{frac} & = j_{RB} - j_{floor}
\end{aligned}
+++++
+
[latexmath]
+++++
\begin{aligned}
\tau_{RB}'(i,j) =
    & \tau_{RB}(     i_{floor},     j_{floor})[level]
        & \times & ( 1 - i_{frac} ) &
        & \times & ( 1 - j_{frac} ) & + \\
    & \tau_{RB}( 1 + i_{floor},     j_{floor})[level]
        & \times & (     i_{frac} ) &
        & \times & ( 1 - j_{frac} ) & + \\
    & \tau_{RB}(     i_{floor}, 1 + j_{floor})[level]
        & \times & ( 1 - i_{frac} ) &
        & \times & (     j_{frac} ) & + \\
    & \tau_{RB}( 1 + i_{floor}, 1 + j_{floor})[level]
        & \times & (     i_{frac} ) &
        & \times & (     j_{frac} ) &
\end{aligned}
+++++

[NOTE]
.Note
====
In the case where the texture itself is bilinearly interpolated as described
in <<textures-texel-filtering,Texel Filtering>>, thus requiring four
full-color samples for the filtering operation, and where the reconstruction
of these samples uses bilinear interpolation in the chroma components due to
pname:chromaFilter=ename:VK_FILTER_LINEAR, up to nine chroma samples may be
required, depending on the sample location.
====


[[textures-implict-reconstruction]]
==== Implicit Reconstruction

Implicit reconstruction takes place by the samples being interpolated, as
required by the filter settings of the sampler, except that
pname:chromaFilter takes precedence for the chroma samples.

If pname:chromaFilter is ename:VK_FILTER_NEAREST, an implementation may:
behave as if pname:xChromaOffset and pname:yChromaOffset were both
ename:VK_CHROMA_LOCATION_MIDPOINT, irrespective of the values set.

[NOTE]
.Note
====
This will not have any visible effect if the locations of the luma samples
coincide with the location of the samples used for rasterization.
====

The sample coordinates are adjusted by the downsample factor of the
component (such that, for example, the sample coordinates are divided by two
if the component has a downsample factor of two relative to the luma
component):

[latexmath]
++++++
\begin{aligned}
u_{RB}' (422/420) &=
  \begin{cases}
     0.5\times (u + 0.5), & \textrm{xChromaOffset = COSITED}\_\textrm{EVEN} \\
     0.5\times u, & \textrm{xChromaOffset = MIDPOINT}
  \end{cases} \\
v_{RB}' (420) &=
  \begin{cases}
     0.5\times (v + 0.5), & \textrm{yChromaOffset = COSITED}\_\textrm{EVEN} \\
     0.5\times v, & \textrm{yChromaOffset = MIDPOINT}
  \end{cases}
\end{aligned}
++++++


[[textures-sampler-YCbCr-conversion]]
=== Sampler {YCbCr} Conversion

Sampler {YCbCr} conversion performs the following operations, which an
implementation may: combine into a single mathematical operation:

  * <<textures-sampler-YCbCr-conversion-rangeexpand,Sampler {YCbCr} Range
    Expansion>>
  * <<textures-sampler-YCbCr-conversion-modelconversion,Sampler {YCbCr}
    Model Conversion>>


[[textures-sampler-YCbCr-conversion-rangeexpand]]
==== Sampler {YCbCr} Range Expansion

Sampler {YCbCr} range expansion is applied to color component values after
all texel input operations which are not specific to sampler {YCbCr}
conversion.
For example, the input values to this stage have been converted using the
normal <<textures-format-conversion,format conversion>> rules.

ifdef::VK_QCOM_ycbcr_degamma[]
The input values to this stage may have been converted using sRGB to linear
conversion if <<features-ycbcr-degamma,pname:ycbcrDegamma>> is enabled.
endif::VK_QCOM_ycbcr_degamma[]

Sampler {YCbCr} range expansion is not applied if pname:ycbcrModel is
ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_RGB_IDENTITY.
That is, the shader receives the vector C'~rgba~ as output by the Component
Swizzle stage without further modification.

For other values of pname:ycbcrModel, range expansion is applied to the
texel component values output by the <<textures-component-swizzle,Component
Swizzle>> defined by the pname:components member of
slink:VkSamplerYcbcrConversionCreateInfo.
Range expansion applies independently to each component of the image.
For the purposes of range expansion and {YCbCr} model conversion, the R and
B components contain color difference (chroma) values and the G component
contains luma.
The A component is not modified by sampler {YCbCr} range expansion.

The range expansion to be applied is defined by the pname:ycbcrRange member
of the slink:VkSamplerYcbcrConversionCreateInfo structure:

  * If pname:ycbcrRange is ename:VK_SAMPLER_YCBCR_RANGE_ITU_FULL, the
    following transformations are applied:
+
[latexmath]
+++++++++++++++++++
\begin{aligned}
Y' &= C'_{rgba}[G] \\
C_B &= C'_{rgba}[B] - {{2^{(n-1)}}\over{(2^n) - 1}} \\
C_R &= C'_{rgba}[R] - {{2^{(n-1)}}\over{(2^n) - 1}}
\end{aligned}
+++++++++++++++++++
+
[NOTE]
.Note
====
These formulae correspond to the "`full range`" encoding in the
"`Quantization schemes`" chapter of the <<data-format,Khronos Data Format
Specification>>.

Should any future amendments be made to the ITU specifications from which
these equations are derived, the formulae used by Vulkan may: also be
updated to maintain parity.
====
  * If pname:ycbcrRange is ename:VK_SAMPLER_YCBCR_RANGE_ITU_NARROW, the
    following transformations are applied:
+
[latexmath]
+++++++++++++++++++
\begin{aligned}
Y' &= {{C'_{rgba}[G] \times (2^n-1) - 16\times 2^{n-8}}\over{219\times 2^{n-8}}} \\
C_B &= {{C'_{rgba}[B] \times \left(2^n-1\right) - 128\times 2^{n-8}}\over{224\times 2^{n-8}}} \\
C_R &= {{C'_{rgba}[R] \times \left(2^n-1\right) - 128\times 2^{n-8}}\over{224\times 2^{n-8}}}
\end{aligned}
+++++++++++++++++++
+
[NOTE]
.Note
====
These formulae correspond to the "`narrow range`" encoding in the
"`Quantization schemes`" chapter of the <<data-format,Khronos Data Format
Specification>>.
====
  * _n_ is the bit-depth of the components in the format.

The precision of the operations performed during range expansion must: be at
least that of the source format.

An implementation may: clamp the results of these range expansion operations
such that Y{prime} falls in the range [0,1], and/or such that C~B~ and C~R~
fall in the range [-0.5,0.5].


[[textures-sampler-YCbCr-conversion-modelconversion]]
==== Sampler {YCbCr} Model Conversion

The range-expanded values are converted between color models, according to
the color model conversion specified in the pname:ycbcrModel member:

ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_RGB_IDENTITY::
    The color components are not modified by the color model conversion
    since they are assumed already to represent the desired color model in
    which the shader is operating; {YCbCr} range expansion is also ignored.
ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_IDENTITY::
    The color components are not modified by the color model conversion and
    are assumed to be treated as though in {YCbCr} form both in memory and
    in the shader; {YCbCr} range expansion is applied to the components as
    for other {YCbCr} models, with the vector (C~R~,Y{prime},C~B~,A)
    provided to the shader.
ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_709::
    The color components are transformed from a {YCbCr} representation to an
    {RGBprime} representation as described in the "`BT.709 {YCbCr}
    conversion`" section of the <<data-format,Khronos Data Format
    Specification>>.
ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_601::
    The color components are transformed from a {YCbCr} representation to an
    {RGBprime} representation as described in the "`BT.601 {YCbCr}
    conversion`" section of the <<data-format,Khronos Data Format
    Specification>>.
ename:VK_SAMPLER_YCBCR_MODEL_CONVERSION_YCBCR_2020::
    The color components are transformed from a {YCbCr} representation to an
    {RGBprime} representation as described in the "`BT.2020 {YCbCr}
    conversion`" section of the <<data-format,Khronos Data Format
    Specification>>.

In this operation, each output component is dependent on each input
component.

An implementation may: clamp the {RGBprime} results of these conversions to
the range [0,1].

The precision of the operations performed during model conversion must: be
at least that of the source format.

The alpha component is not modified by these model conversions.

[NOTE]
.Note
====
Sampling operations in a non-linear color space can introduce color and
intensity shifts at sharp transition boundaries.
To avoid this issue, the technically precise color correction sequence
described in the "`Introduction to Color Conversions`" chapter of the
<<data-format,Khronos Data Format Specification>> may be performed as
follows:

  * Calculate the <<textures-normalized-to-unnormalized,unnormalized texel
    coordinates>> corresponding to the desired sample position.
  * For a pname:minFilter or pname:magFilter of ename:VK_FILTER_NEAREST:
    . Calculate (_i_,_j_) for the sample location as described under the
      "`nearest filtering`" formulae in <<textures-unnormalized-to-integer>>
    . Calculate the normalized texel coordinates corresponding to these
      integer coordinates.
    . Sample using <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>>
      at this location.
  * For a pname:minFilter or pname:magFilter of ename:VK_FILTER_LINEAR:
    . Calculate (_i~[0,1]~_,_j~[0,1]~_) for the sample location as described
      under the "`linear filtering`" formulae in
      <<textures-unnormalized-to-integer>>
    . Calculate the normalized texel coordinates corresponding to these
      integer coordinates.
    . Sample using <<samplers-YCbCr-conversion,sampler {YCbCr} conversion>>
      at each of these locations.
    . Convert the non-linear A{prime}{RGBprime} outputs of the {YCbCr}
      conversions to linear ARGB values as described in the "`Transfer
      Functions`" chapter of the <<data-format,Khronos Data Format
      Specification>>.
    . Interpolate the linear ARGB values using the [eq]#{alpha}# and
      [eq]#{beta}# values described in the "`linear filtering`" section of
      <<textures-unnormalized-to-integer>> and the equations in
      <<textures-texel-filtering>>.

The additional calculations and, especially, additional number of sampling
operations in the ename:VK_FILTER_LINEAR case can be expected to have a
performance impact compared with using the outputs directly.
Since the variations from "`correct`" results are subtle for most content,
the application author should determine whether a more costly implementation
is strictly necessary.

If pname:chromaFilter, and pname:minFilter or pname:magFilter are both
ename:VK_FILTER_NEAREST, these operations are redundant and sampling using
<<samplers-YCbCr-conversion,sampler {YCbCr} conversion>> at the desired
sample coordinates will produce the "`correct`" results without further
processing.
====
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]


== Texel Output Operations

_Texel output instructions_ are SPIR-V image instructions that write to an
image.
_Texel output operations_ are a set of steps that are performed on state,
coordinates, and texel values while processing a texel output instruction,
and which are common to some or all texel output instructions.
They include the following steps, which are performed in the listed order:

  * <<textures-output-validation,Validation operations>>
  ** <<textures-format-validation,Format validation>>
  ** <<textures-type-validation,Type validation>>
  ** <<textures-output-coordinate-validation,Coordinate validation>>
  ** <<textures-output-sparse-validation,Sparse validation>>
  * <<textures-output-format-conversion,Texel output format conversion>>


[[textures-output-validation]]
=== Texel Output Validation Operations

_Texel output validation operations_ inspect instruction/image state or
coordinates, and in certain circumstances cause the write to have no effect.
There are a series of validations that the texel undergoes.


[[textures-format-validation]]
==== Texel Format Validation

If the image format of the code:OpTypeImage is not
<<spirvenv-image-formats,compatible>> with the sname:VkImageView's
pname:format, the write causes the contents of the image's memory to become
undefined:.


[[textures-type-validation]]
==== Texel Type Validation

If the code:Sampled code:Type of the code:OpTypeImage does not match the
<<spirv-type,SPIR-V Type>>, the write causes the value of the texel to
become undefined:.
For integer types, if the <<spirvenv-image-signedness,signedness of the
access>> does not match the signedness of the accessed resource, the write
causes the value of the texel to become undefined:.


[[textures-output-coordinate-validation]]
=== Integer Texel Coordinate Validation

The integer texel coordinates are validated according to the same rules as
for texel input <<textures-integer-coordinate-validation,coordinate
validation>>.

If the texel fails integer texel coordinate validation, then the write has
no effect.


[[textures-output-sparse-validation]]
=== Sparse Texel Operation

If the texel attempts to write to an unbound region of a sparse image, the
texel is a sparse unbound texel.
In such a case, if the
slink:VkPhysicalDeviceSparseProperties::pname:residencyNonResidentStrict
property is ename:VK_TRUE, the sparse unbound texel write has no effect.
If pname:residencyNonResidentStrict is ename:VK_FALSE, the write may: have a
side effect that becomes visible to other accesses to unbound texels in any
resource, but will not be visible to any device memory allocated by the
application.


[[textures-output-format-conversion]]
=== Texel Output Format Conversion

If the image format is sRGB, a linear to sRGB conversion is applied to the
R, G, and B components as described in the "`sRGB EOTF`" section of the
<<data-format,Khronos Data Format Specification>>.
The A component, if present, is unchanged.

Texels then undergo a format conversion from the floating point, signed, or
unsigned integer type of the texel data to the elink:VkFormat of the image
view.
If the number of components in the texel data is larger than the number of
components in the format, additional components are discarded.

Each component is converted based on its type and size (as defined in the
<<formats-definition,Format Definition>> section for each elink:VkFormat).
Floating-point outputs are converted as described in
<<fundamentals-fp-conversion,Floating-Point Format Conversions>> and
<<fundamentals-fixedconv,Fixed-Point Data Conversion>>.
Integer outputs are converted such that their value is preserved.
The converted value of any integer that cannot be represented in the target
format is undefined:.


[[textures-normalized-operations]]
== Normalized Texel Coordinate Operations

If the image sampler instruction provides normalized texel coordinates, some
of the following operations are performed.


[[textures-projection]]
=== Projection Operation

For code:Proj image operations, the normalized texel coordinates
[eq]#(s,t,r,q,a)# and (if present) the [eq]#D~ref~# coordinate are
transformed as follows:

[latexmath]
+++++++++++++++++++
\begin{aligned}
s       & = \frac{s}{q},       & \text{for 1D, 2D, or 3D image} \\
\\
t       & = \frac{t}{q},       & \text{for 2D or 3D image} \\
\\
r       & = \frac{r}{q},       & \text{for 3D image} \\
\\
D_{\textit{ref}} & = \frac{D_{\textit{ref}}}{q}, & \text{if provided}
\end{aligned}
+++++++++++++++++++


[[textures-derivative-image-operations]]
=== Derivative Image Operations

Derivatives are used for LOD selection.
These derivatives are either implicit (in an code:ImplicitLod image
instruction in a fragment shader) or explicit (provided explicitly by shader
to the image instruction in any shader).

For implicit derivatives image instructions, the derivatives of texel
coordinates are calculated in the same manner as
<<shaders-derivative-operations, derivative operations>>.
That is:

[latexmath]
+++++++++++++++++++
\begin{aligned}
\partial{s}/\partial{x} & = dPdx(s), & \partial{s}/\partial{y} & = dPdy(s), & \text{for 1D, 2D, Cube, or 3D image} \\
\partial{t}/\partial{x} & = dPdx(t), & \partial{t}/\partial{y} & = dPdy(t), & \text{for 2D, Cube, or 3D image} \\
\partial{r}/\partial{x} & = dPdx(r), & \partial{r}/\partial{y} & = dPdy(r), & \text{for Cube or 3D image}
\end{aligned}
+++++++++++++++++++

Partial derivatives not defined above for certain image dimensionalities are
set to zero.

For explicit LOD image instructions, if the optional: SPIR-V operand
code:Grad is provided, then the operand values are used for the derivatives.
The number of components present in each derivative for a given image
dimensionality matches the number of partial derivatives computed above.

If the optional: SPIR-V operand code:Lod is provided, then derivatives are
set to zero, the cube map derivative transformation is skipped, and the
scale factor operation is skipped.
Instead, the floating point scalar coordinate is directly assigned to
[eq]#{lambda}~base~# as described in <<textures-level-of-detail-operation,
LOD Operation>>.

ifdef::VK_VERSION_1_2,VK_EXT_descriptor_indexing[]
If the image or sampler object used by an implicit derivative image
instruction is not uniform across the quad and
<<limits-quadDivergentImplicitLod, pname:quadDivergentImplicitLod>> is not
supported, then the derivative and LOD values are undefined:.
Implicit derivatives are well-defined when the image and sampler and control
flow are uniform across the quad, even if they diverge between different
quads.

If <<limits-quadDivergentImplicitLod, pname:quadDivergentImplicitLod>> is
supported, then derivatives and implicit LOD values are well-defined even if
the image or sampler object are not uniform within a quad.
The derivatives are computed as specified above, and the implicit LOD
calculation proceeds for each shader invocation using its respective image
and sampler object.
endif::VK_VERSION_1_2,VK_EXT_descriptor_indexing[]


=== Cube Map Face Selection and Transformations

For cube map image instructions, the [eq]#(s,t,r)# coordinates are treated
as a direction vector [eq]#(r~x~,r~y~,r~z~)#.
The direction vector is used to select a cube map face.
The direction vector is transformed to a per-face texel coordinate system
[eq]#(s~face~,t~face~)#, The direction vector is also used to transform the
derivatives to per-face derivatives.


=== Cube Map Face Selection

The direction vector selects one of the cube map's faces based on the
largest magnitude coordinate direction (the major axis direction).
Since two or more coordinates can: have identical magnitude, the
implementation must: have rules to disambiguate this situation.

The rules should: have as the first rule that [eq]#r~z~# wins over
[eq]#r~y~# and [eq]#r~x~#, and the second rule that [eq]#r~y~# wins over
[eq]#r~x~#.
An implementation may: choose other rules, but the rules must: be
deterministic and depend only on [eq]#(r~x~,r~y~,r~z~)#.

The layer number (corresponding to a cube map face), the coordinate
selections for [eq]#s~c~#, [eq]#t~c~#, [eq]#r~c~#, and the selection of
derivatives, are determined by the major axis direction as specified in the
following two tables.

.Cube map face and coordinate selection
[width="75%",frame="all",options="header"]
|====
| Major Axis Direction | Layer Number | Cube Map Face | [eq]#s~c~#  | [eq]#t~c~#  | [eq]#r~c~#
| [eq]#+r~x~#          | [eq]#0#      | Positive X    | [eq]#-r~z~# | [eq]#-r~y~# | [eq]#r~x~#
| [eq]#-r~x~#          | [eq]#1#      | Negative X    | [eq]#+r~z~# | [eq]#-r~y~# | [eq]#r~x~#
| [eq]#+r~y~#          | [eq]#2#      | Positive Y    | [eq]#+r~x~# | [eq]#+r~z~# | [eq]#r~y~#
| [eq]#-r~y~#          | [eq]#3#      | Negative Y    | [eq]#+r~x~# | [eq]#-r~z~# | [eq]#r~y~#
| [eq]#+r~z~#          | [eq]#4#      | Positive Z    | [eq]#+r~x~# | [eq]#-r~y~# | [eq]#r~z~#
| [eq]#-r~z~#          | [eq]#5#      | Negative Z    | [eq]#-r~x~# | [eq]#-r~y~# | [eq]#r~z~#
|====


.Cube map derivative selection
[width="75%",frame="all",options="header"]
|====
| Major Axis Direction | [eq]#{partial}s~c~ / {partial}x# | [eq]#{partial}s~c~ / {partial}y# | [eq]#{partial}t~c~ / {partial}x# | [eq]#{partial}t~c~ / {partial}y# | [eq]#{partial}r~c~ / {partial}x# | [eq]#{partial}r~c~ / {partial}y#

| [eq]#+r~x~#
| [eq]#-{partial}r~z~ / {partial}x# | [eq]#-{partial}r~z~ / {partial}y#
| [eq]#-{partial}r~y~ / {partial}x# | [eq]#-{partial}r~y~ / {partial}y#
| [eq]#+{partial}r~x~ / {partial}x# | [eq]#+{partial}r~x~ / {partial}y#

| [eq]#-r~x~#
| [eq]#+{partial}r~z~ / {partial}x# | [eq]#+{partial}r~z~ / {partial}y#
| [eq]#-{partial}r~y~ / {partial}x# | [eq]#-{partial}r~y~ / {partial}y#
| [eq]#-{partial}r~x~ / {partial}x# | [eq]#-{partial}r~x~ / {partial}y#

| [eq]#+r~y~#
| [eq]#+{partial}r~x~ / {partial}x# | [eq]#+{partial}r~x~ / {partial}y#
| [eq]#+{partial}r~z~ / {partial}x# | [eq]#+{partial}r~z~ / {partial}y#
| [eq]#+{partial}r~y~ / {partial}x# | [eq]#+{partial}r~y~ / {partial}y#

| [eq]#-r~y~#
| [eq]#+{partial}r~x~ / {partial}x# | [eq]#+{partial}r~x~ / {partial}y#
| [eq]#-{partial}r~z~ / {partial}x# | [eq]#-{partial}r~z~ / {partial}y#
| [eq]#-{partial}r~y~ / {partial}x# | [eq]#-{partial}r~y~ / {partial}y#

| [eq]#+r~z~#
| [eq]#+{partial}r~x~ / {partial}x# | [eq]#+{partial}r~x~ / {partial}y#
| [eq]#-{partial}r~y~ / {partial}x# | [eq]#-{partial}r~y~ / {partial}y#
| [eq]#+{partial}r~z~ / {partial}x# | [eq]#+{partial}r~z~ / {partial}y#

| [eq]#-r~z~#
| [eq]#-{partial}r~x~ / {partial}x# | [eq]#-{partial}r~x~ / {partial}y#
| [eq]#-{partial}r~y~ / {partial}x# | [eq]#-{partial}r~y~ / {partial}y#
| [eq]#-{partial}r~z~ / {partial}x# | [eq]#-{partial}r~z~ / {partial}y#
|====


=== Cube Map Coordinate Transformation

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
s_{\textit{face}} & =
    \frac{1}{2} \times \frac{s_c}{|r_c|} + \frac{1}{2} \\
t_{\textit{face}} & =
    \frac{1}{2} \times \frac{t_c}{|r_c|} + \frac{1}{2} \\
\end{aligned}
++++++++++++++++++++++++


=== Cube Map Derivative Transformation

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\frac{\partial{s_{\textit{face}}}}{\partial{x}} &=
    \frac{\partial}{\partial{x}} \left ( \frac{1}{2} \times \frac{s_{c}}{|r_{c}|}
    + \frac{1}{2}\right ) \\
\frac{\partial{s_{\textit{face}}}}{\partial{x}} &=
    \frac{1}{2} \times \frac{\partial}{\partial{x}}
    \left ( \frac{s_{c}}{|r_{c}|}  \right ) \\
\frac{\partial{s_{\textit{face}}}}{\partial{x}} &=
    \frac{1}{2} \times
    \left (
    \frac{
      |r_{c}| \times \partial{s_c}/\partial{x}
      -s_c \times {\partial{r_{c}}}/{\partial{x}}}
    {\left ( r_{c} \right )^2}
    \right )
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\frac{\partial{s_{\textit{face}}}}{\partial{y}} &=
    \frac{1}{2} \times
    \left (
    \frac{
      |r_{c}| \times \partial{s_c}/\partial{y}
      -s_c \times {\partial{r_{c}}}/{\partial{y}}}
    {\left ( r_{c} \right )^2}
    \right )\\
\frac{\partial{t_{\textit{face}}}}{\partial{x}} &=
    \frac{1}{2} \times
    \left (
    \frac{
      |r_{c}| \times \partial{t_c}/\partial{x}
      -t_c \times {\partial{r_{c}}}/{\partial{x}}}
    {\left ( r_{c} \right )^2}
    \right ) \\
\frac{\partial{t_{\textit{face}}}}{\partial{y}} &=
    \frac{1}{2} \times
    \left (
    \frac{
       |r_{c}| \times \partial{t_c}/\partial{y}
      -t_c \times {\partial{r_{c}}}/{\partial{y}}}
    {\left ( r_{c} \right )^2}
    \right )
\end{aligned}
++++++++++++++++++++++++

ifdef::editing-notes[]
[NOTE]
.editing-note
====
(Bill) Note that we never revisited ARB_texture_cubemap after we introduced
dependent texture fetches (ARB_fragment_program and ARB_fragment_shader).

The derivatives of [eq]#s~face~# and [eq]#t~face~# are only valid for
non-dependent texture fetches (pre OpenGL 2.0).
====
endif::editing-notes[]


[[textures-lod-and-scale-factor]]
=== Scale Factor Operation, LOD Operation and Image Level(s) Selection

LOD selection can: be either explicit (provided explicitly by the image
instruction) or implicit (determined from a scale factor calculated from the
derivatives).
The LOD must: be computed with pname:mipmapPrecisionBits of accuracy.


[[textures-scale-factor]]
==== Scale Factor Operation

The magnitude of the derivatives are calculated by:

  {empty}:: [eq]#m~ux~ = {vert}{partial}s/{partial}x{vert} {times} w~base~#
  {empty}:: [eq]#m~vx~ = {vert}{partial}t/{partial}x{vert} {times} h~base~#
  {empty}:: [eq]#m~wx~ = {vert}{partial}r/{partial}x{vert} {times} d~base~#

  {empty}:: [eq]#m~uy~ = {vert}{partial}s/{partial}y{vert} {times} w~base~#
  {empty}:: [eq]#m~vy~ = {vert}{partial}t/{partial}y{vert} {times} h~base~#
  {empty}:: [eq]#m~wy~ = {vert}{partial}r/{partial}y{vert} {times} d~base~#


where:

  {empty}:: [eq]#{partial}t/{partial}x = {partial}t/{partial}y = 0# (for 1D
            images)
  {empty}:: [eq]#{partial}r/{partial}x = {partial}r/{partial}y = 0# (for 1D,
            2D or Cube images)

and:

  {empty}:: [eq]#w~base~ = image.w#
  {empty}:: [eq]#h~base~ = image.h#
  {empty}:: [eq]#d~base~ = image.d#

(for the pname:baseMipLevel, from the image descriptor).

ifdef::VK_NV_corner_sampled_image[]

For corner-sampled images, the [eq]#w~base~#, [eq]#h~base~#, and
[eq]#d~base~# are instead:

  {empty}:: [eq]#w~base~ = image.w - 1#
  {empty}:: [eq]#h~base~ = image.h - 1#
  {empty}:: [eq]#d~base~ = image.d - 1#

endif::VK_NV_corner_sampled_image[]

A point sampled in screen space has an elliptical footprint in texture
space.
The minimum and maximum scale factors [eq]#({rho}~min~, {rho}~max~)# should:
be the minor and major axes of this ellipse.

The _scale factors_ [eq]#{rho}~x~# and [eq]#{rho}~y~#, calculated from the
magnitude of the derivatives in x and y, are used to compute the minimum and
maximum scale factors.

[eq]#{rho}~x~# and [eq]#{rho}~y~# may: be approximated with functions
[eq]#f~x~# and [eq]#f~y~#, subject to the following constraints:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
& f_x \text{\ is\ continuous\ and\ monotonically\ increasing\ in\ each\ of\ }
    m_{ux},
    m_{vx}, \text{\ and\ }
    m_{wx} \\
& f_y \text{\ is\ continuous\ and\ monotonically\ increasing\ in\ each\ of\ }
    m_{uy},
    m_{vy}, \text{\ and\ }
    m_{wy}
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\max(|m_{ux}|, |m_{vx}|, |m_{wx}|) \leq f_{x}
\leq \sqrt{2} (|m_{ux}| + |m_{vx}| + |m_{wx}|) \\
\max(|m_{uy}|, |m_{vy}|, |m_{wy}|) \leq f_{y}
\leq \sqrt{2} (|m_{uy}| + |m_{vy}| + |m_{wy}|)
\end{aligned}
++++++++++++++++++++++++


ifdef::editing-notes[]
[NOTE]
.editing-note
====
(Bill) For reviewers only - anticipating questions.

We only support implicit derivatives for normalized texel coordinates.

So we are documenting the derivatives in s,t,r (normalized texel
coordinates) rather than u,v,w (unnormalized texel coordinates) as in OpenGL
and OpenGL ES specifications.
(I know, u,v,w is the way it has been documented since OpenGL V1.0.)

Also there is no reason to have conditional application of [eq]#w~base~,
h~base~, d~base~# for rectangle textures either, since they do not support
implicit derivatives.
====
endif::editing-notes[]


The minimum and maximum scale factors [eq]#({rho}~min~,{rho}~max~)# are
determined by:

  {empty}:: [eq]#{rho}~max~ = max({rho}~x~, {rho}~y~)#
  {empty}:: [eq]#{rho}~min~ = min({rho}~x~, {rho}~y~)#

The ratio of anisotropy is determined by:

  {empty}:: [eq]#{eta} = min({rho}~max~/{rho}~min~, max~Aniso~)#

where:

  {empty}:: [eq]#sampler.max~Aniso~ = pname:maxAnisotropy# (from sampler
            descriptor)
  {empty}:: [eq]#limits.max~Aniso~ = pname:maxSamplerAnisotropy# (from
            physical device limits)
  {empty}:: [eq]#max~Aniso~ = min(sampler.max~Aniso~, limits.max~Aniso~)#

If [eq]#{rho}~max~ = {rho}~min~ = 0#, then all the partial derivatives are
zero, the fragment's footprint in texel space is a point, and [eq]#{eta}#
should: be treated as 1.
If [eq]#{rho}~max~ {neq} 0# and [eq]#{rho}~min~ = 0# then all partial
derivatives along one axis are zero, the fragment's footprint in texel space
is a line segment, and [eq]#{eta}# should: be treated as [eq]#max~Aniso~#.
However, anytime the footprint is small in texel space the implementation
may: use a smaller value of [eq]#{eta}#, even when [eq]#{rho}~min~# is zero
or close to zero.
If either slink:VkPhysicalDeviceFeatures::pname:samplerAnisotropy or
slink:VkSamplerCreateInfo::pname:anisotropyEnable are ename:VK_FALSE,
[eq]#max~Aniso~# is set to 1.

If [eq]#{eta} = 1#, sampling is isotropic.
If [eq]#{eta} > 1#, sampling is anisotropic.

The sampling rate ([eq]#N#) is derived as:

  {empty}:: [eq]#N = {lceil}{eta}{rceil}#

An implementation may: round [eq]#N# up to the nearest supported sampling
rate.
An implementation may: use the value of [eq]#N# as an approximation of
[eq]#{eta}#.


[[textures-level-of-detail-operation]]
==== LOD Operation

The LOD parameter [eq]#{lambda}# is computed as follows:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\lambda_{base}(x,y) & =
  \begin{cases}
    shaderOp.Lod                                    & \text{(from optional SPIR-V operand)} \\
    \log_2 \left ( \frac{\rho_{max}}{\eta} \right ) & \text{otherwise}
  \end{cases} \\
\lambda'(x,y)       & = \lambda_{base} + \mathbin{clamp}(sampler.bias + shaderOp.bias,-maxSamplerLodBias,maxSamplerLodBias) \\
\lambda             & =
  \begin{cases}
    lod_{max}, & \lambda' > lod_{max} \\
    \lambda',  & lod_{min} \leq \lambda' \leq lod_{max} \\
    lod_{min}, & \lambda' < lod_{min} \\
    \textit{undefined}, & lod_{min} > lod_{max}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

where:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
sampler.bias       & = mipLodBias & \text{(from sampler descriptor)} \\
shaderOp.bias      & =
  \begin{cases}
    Bias & \text{(from optional SPIR-V operand)} \\
    0    & \text{otherwise}
  \end{cases} \\
sampler.lod_{min}  & = minLod & \text{(from sampler descriptor)} \\
shaderOp.lod_{min} & =
  \begin{cases}
    MinLod & \text{(from optional SPIR-V operand)} \\
    0      & \text{otherwise}
  \end{cases} \\
\\
lod_{min}          & = \max(sampler.lod_{min}, shaderOp.lod_{min}) \\
lod_{max}          & = maxLod & \text{(from sampler descriptor)}
\end{aligned}
++++++++++++++++++++++++

and [eq]#maxSamplerLodBias# is the value of the slink:VkPhysicalDeviceLimits
feature <<limits-maxSamplerLodBias, pname:maxSamplerLodBias>>.


[[textures-image-level-selection]]
==== Image Level(s) Selection

The image level(s) [eq]#d#, [eq]#d~hi~#, and [eq]#d~lo~# which texels are
read from are determined by an image-level parameter [eq]#d~l~#, which is
computed based on the LOD parameter, as follows:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
d_{l} =
  \begin{cases}
    nearest(d'),  & \text{mipmapMode is VK\_SAMPLER\_MIPMAP\_MODE\_NEAREST} \\
    d',           & \text{otherwise}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

where:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
ifdef::VK_EXT_image_view_min_lod[]
d' = max(level_{base} + \text{clamp}(\lambda, 0, q), minLod_{imageView})
endif::VK_EXT_image_view_min_lod[]
ifndef::VK_EXT_image_view_min_lod[]
d' = level_{base} + \text{clamp}(\lambda, 0, q)
endif::VK_EXT_image_view_min_lod[]

\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
nearest(d') & =
  \begin{cases}
    \left \lceil d' + 0.5\right \rceil - 1, &
        \text{preferred} \\
    \left \lfloor d' + 0.5\right \rfloor,   &
        \text{alternative}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

and:

ifdef::VK_EXT_image_view_min_lod[]
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
minLod_{imageView} & =
  \begin{cases}
    minLodFloat_{imageView}, & \text{preferred} \\
    minLodInteger_{imageView}, & \text{alternative}
  \end{cases} \\
level_{base}       & = baseMipLevel \\
q                  & = levelCount - 1
\end{aligned}
++++++++++++++++++++++++
endif::VK_EXT_image_view_min_lod[]
ifndef::VK_EXT_image_view_min_lod[]
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
level_{base}       & = baseMipLevel \\
q                  & = levelCount - 1
\end{aligned}
++++++++++++++++++++++++
endif::VK_EXT_image_view_min_lod[]

pname:baseMipLevel and pname:levelCount are taken from the
pname:subresourceRange of the image view.

ifdef::VK_EXT_image_view_min_lod[]
[eq]#minLod~imageView~# must: be less or equal to [eq]#level~base~ + q#.
endif::VK_EXT_image_view_min_lod[]

If the sampler's pname:mipmapMode is ename:VK_SAMPLER_MIPMAP_MODE_NEAREST,
then the level selected is [eq]#d = d~l~#.

If the sampler's pname:mipmapMode is ename:VK_SAMPLER_MIPMAP_MODE_LINEAR,
two neighboring levels are selected:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
d_{hi} & = \left\lfloor d_{l} \right\rfloor \\
d_{lo} & = min( d_{hi} + 1, level_{base} + q ) \\
\delta & = d_{l} - d_{hi}
\end{aligned}
++++++++++++++++++++++++

[eq]#{delta}# is the fractional value, quantized to the number of
<<limits-mipmapPrecisionBits, mipmap precision bits>>, used for
<<textures-texel-filtering, linear filtering>> between levels.


[[textures-normalized-to-unnormalized]]
=== (s,t,r,q,a) to (u,v,w,a) Transformation

The normalized texel coordinates are scaled by the image level dimensions
and the array layer is selected.

This transformation is performed once for each level used in
<<textures-texel-filtering,filtering>> (either [eq]#d#, or [eq]#d~hi~# and
[eq]#d~lo~#).

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
u(x,y) & = s(x,y) \times width_{scale} + \Delta_i\\
v(x,y) & =
  \begin{cases}
    0                         & \text{for 1D images} \\
    t(x,y) \times height_{scale} + \Delta_j & \text{otherwise}
  \end{cases} \\
w(x,y) & =
  \begin{cases}
    0                         & \text{for 2D or Cube images} \\
    r(x,y) \times depth_{scale}  + \Delta_k & \text{otherwise}
  \end{cases} \\
\\
a(x,y) & =
  \begin{cases}
    a(x,y)                    & \text{for array images} \\
    0                         & \text{otherwise}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

where:

  {empty}:: [eq]#width~scale~ = width~level~#
  {empty}:: [eq]#height~scale~ = height~level~#
  {empty}:: [eq]#depth~scale~ = depth~level~#

ifdef::VK_NV_corner_sampled_image[]
for conventional images, and:

  {empty}:: [eq]#width~scale~ = width~level~ - 1#
  {empty}:: [eq]#height~scale~ = height~level~ - 1#
  {empty}:: [eq]#depth~scale~ = depth~level~ - 1#

for corner-sampled images.
endif::VK_NV_corner_sampled_image[]

and where [eq]#({DeltaUpper}~i~, {DeltaUpper}~j~, {DeltaUpper}~k~)# are
taken from the image instruction if it includes a code:ConstOffset or
code:Offset operand, otherwise they are taken to be zero.


Operations then proceed to Unnormalized Texel Coordinate Operations.


== Unnormalized Texel Coordinate Operations


[[textures-unnormalized-to-integer]]
=== (u,v,w,a) to (i,j,k,l,n) Transformation and Array Layer Selection

The unnormalized texel coordinates are transformed to integer texel
coordinates relative to the selected mipmap level.

The layer index [eq]#l# is computed as:

  {empty}:: [eq]#l = clamp(RNE(a), 0, pname:layerCount - 1) {plus}
            pname:baseArrayLayer#

where pname:layerCount is the number of layers in the image subresource
range of the image view, pname:baseArrayLayer is the first layer from the
subresource range, and where:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\mathbin{RNE}(a) & =
  \begin{cases}
    \mathbin{roundTiesToEven}(a)                  & \text{preferred, from IEEE Std 754-2008 Floating-Point Arithmetic} \\
    \left \lfloor a + 0.5 \right \rfloor & \text{alternative}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

The sample index [eq]#n# is assigned the value 0.

Nearest filtering (ename:VK_FILTER_NEAREST) computes the integer texel
coordinates that the unnormalized coordinates lie within:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i &= \left\lfloor u + shift \right\rfloor \\
j &= \left\lfloor v + shift \right\rfloor \\
k &= \left\lfloor w + shift \right\rfloor
\end{aligned}
++++++++++++++++++++++++
where:

  {empty}:: [eq]#shift = 0.0#

ifdef::VK_NV_corner_sampled_image[]
for conventional images, and:

  {empty}:: [eq]#shift = 0.5#

for corner-sampled images.
endif::VK_NV_corner_sampled_image[]

Linear filtering (ename:VK_FILTER_LINEAR) computes a set of neighboring
coordinates which bound the unnormalized coordinates.
The integer texel coordinates are combinations of [eq]#i~0~# or [eq]#i~1~#,
[eq]#j~0~# or [eq]#j~1~#, [eq]#k~0~# or [eq]#k~1~#, as well as weights
[eq]#{alpha}, {beta}#, and [eq]#{gamma}#.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_0 &= \left\lfloor u - shift \right\rfloor \\
i_1 &= i_0 + 1 \\
j_0 &= \left\lfloor v - shift \right\rfloor \\
j_1 &= j_0 + 1 \\
k_0 &= \left\lfloor w - shift \right\rfloor \\
k_1 &= k_0 + 1
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\alpha &= \mathbin{frac}\left(u - shift\right)  \\[1em]
\beta &= \mathbin{frac}\left(v - shift\right)  \\[1em]
\gamma &= \mathbin{frac}\left(w - shift\right)
\end{aligned}
++++++++++++++++++++++++

where:

  {empty}:: [eq]#shift = 0.5#

ifdef::VK_NV_corner_sampled_image[]
for conventional images, and:

  {empty}:: [eq]#shift = 0.0#

for corner-sampled images,
endif::VK_NV_corner_sampled_image[]
and where:

[latexmath]
++++++++++++++++++++++++
\mathbin{frac}(x) = x -  \left\lfloor x \right\rfloor
++++++++++++++++++++++++
where the number of fraction bits retained is specified by
sname:VkPhysicalDeviceLimits::pname:subTexelPrecisionBits.

ifdef::VK_IMG_filter_cubic,VK_EXT_filter_cubic[]
Cubic filtering (ename:VK_FILTER_CUBIC_EXT) computes a set of neighboring
coordinates which bound the unnormalized coordinates.
The integer texel coordinates are combinations of [eq]#i~0~#, [eq]#i~1~#,
[eq]#i~2~# or [eq]#i~3~#, [eq]#j~0~#, [eq]#j~1~#, [eq]#j~2~# or [eq]#j~3~#,
ifndef::VK_EXT_filter_cubic[]
as well as weights [eq]#{alpha}# and [eq]#{beta}#.
endif::VK_EXT_filter_cubic[]
ifdef::VK_EXT_filter_cubic[]
[eq]#k~0~#, [eq]#k~1~#, [eq]#k~2~# or [eq]#k~3~#, as well as weights
[eq]#{alpha}#, [eq]#{beta}#, and [eq]#{gamma}#.
endif::VK_EXT_filter_cubic[]

ifndef::VK_EXT_filter_cubic[]
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_{0}  & = {\left \lfloor {u - \frac{3}{2}} \right \rfloor} & i_{1} & = i_{0} + 1 & i_{2} & = i_{1} + 1 & i_{3} & = i_{2} + 1 \\[1em]
j_{0}  & = {\left \lfloor {v - \frac{3}{2}} \right \rfloor} & j_{1} & = j_{0} + 1 & j_{2} & = j_{1} + 1 & j_{3} & = j_{2} + 1
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
alpha &= \mathbin{frac}\left(u - \frac{1}{2}\right)  \\[1em]
\beta &= \mathbin{frac}\left(v - \frac{1}{2}\right)
\end{aligned}
++++++++++++++++++++++++

endif::VK_EXT_filter_cubic[]

ifdef::VK_EXT_filter_cubic[]
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_{0}  & = {\left \lfloor {u - \frac{3}{2}} \right \rfloor} & i_{1} & = i_{0} + 1 & i_{2} & = i_{1} + 1 & i_{3} & = i_{2} + 1 \\[1em]
j_{0}  & = {\left \lfloor {v - \frac{3}{2}} \right \rfloor} & j_{1} & = j_{0} + 1 & j_{2} & = j_{1} + 1 & j_{3} & = j_{2} + 1 \\[1em]
k_{0}  & = {\left \lfloor {w - \frac{3}{2}} \right \rfloor} & k_{1} & = k_{0} + 1 & k_{2} & = k_{1} + 1 & k_{3} & = k_{2} + 1
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\alpha &= \mathbin{frac}\left(u - \frac{1}{2}\right)  \\[1em]
\beta &= \mathbin{frac}\left(v - \frac{1}{2}\right)  \\[1em]
\gamma &= \mathbin{frac}\left(w - \frac{1}{2}\right)
\end{aligned}
++++++++++++++++++++++++

endif::VK_EXT_filter_cubic[]

where:

[latexmath]
++++++++++++++++++++++++
\mathbin{frac}(x) = x -  \left\lfloor x \right\rfloor
++++++++++++++++++++++++

where the number of fraction bits retained is specified by
sname:VkPhysicalDeviceLimits::pname:subTexelPrecisionBits.
endif::VK_IMG_filter_cubic,VK_EXT_filter_cubic[]


[[textures-integer-coordinate-operations]]
== Integer Texel Coordinate Operations

ifdef::VK_AMD_shader_image_load_store_lod[]
Integer texel coordinate operations may: supply a LOD which texels are to be
read from or written to using the optional SPIR-V operand code:Lod.
endif::VK_AMD_shader_image_load_store_lod[]
ifndef::VK_AMD_shader_image_load_store_lod[]
The code:OpImageFetch and code:OpImageFetchSparse SPIR-V instructions may:
supply a LOD from which texels are to be fetched using the optional SPIR-V
operand code:Lod.
Other integer-coordinate operations must: not.
endif::VK_AMD_shader_image_load_store_lod[]
If the code:Lod is provided then it must: be an integer.

The image level selected is:
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
d & = level_{base} +
  \begin{cases}
    Lod & \text{(from optional SPIR-V operand)} \\
    0   & \text{otherwise}
  \end{cases} \\
\end{aligned}
++++++++++++++++++++++++

If [eq]#d# does not lie in the range [eq]#[pname:baseMipLevel,
pname:baseMipLevel {plus} pname:levelCount)#
ifdef::VK_EXT_image_view_min_lod[]
or [eq]#d# is less than minLodInteger~imageView~,
endif::VK_EXT_image_view_min_lod[]
then any values fetched are
ifdef::VK_EXT_robustness2[]
zero if the <<features-robustImageAccess2, pname:robustImageAccess2>>
feature is enabled, otherwise are
endif::VK_EXT_robustness2[]
undefined:, and any writes (if supported) are discarded.


[[textures-sample-operations]]
== Image Sample Operations


[[textures-wrapping-operation]]
=== Wrapping Operation

ifdef::VK_EXT_non_seamless_cube_map[]
If the used sampler was created without
ename:VK_SAMPLER_CREATE_NON_SEAMLESS_CUBE_MAP_BIT_EXT,
endif::VK_EXT_non_seamless_cube_map[]
code:Cube images ignore the wrap modes specified in the sampler.
Instead, if ename:VK_FILTER_NEAREST is used within a mip level then
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE is used, and if
ename:VK_FILTER_LINEAR is used within a mip level then sampling at the edges
is performed as described earlier in the <<textures-cubemapedge,Cube map
edge handling>> section.

The first integer texel coordinate i is transformed based on the
pname:addressModeU parameter of the sampler.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i &=
  \begin{cases}
    i \bmod size                                & \text{for repeat} \\
    (size - 1) - \mathbin{mirror}
        ((i \bmod (2 \times size)) - size)      & \text{for mirrored repeat} \\
    \mathbin{clamp}(i,0,size-1)                  & \text{for clamp to edge} \\
    \mathbin{clamp}(i,-1,size)                   & \text{for clamp to border} \\
    \mathbin{clamp}(\mathbin{mirror}(i),0,size-1) & \text{for mirror clamp to edge}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

where:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
& \mathbin{mirror}(n) =
  \begin{cases}
    n      & \text{for}\  n \geq 0 \\
    -(1+n) & \text{otherwise}
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

[eq]#j# (for 2D and Cube image) and [eq]#k# (for 3D image) are similarly
transformed based on the pname:addressModeV and pname:addressModeW
parameters of the sampler, respectively.


[[textures-gather]]
=== Texel Gathering

SPIR-V instructions with code:Gather in the name return a vector derived
from 4 texels in the base level of the image view.
The rules for the ename:VK_FILTER_LINEAR minification filter are applied to
identify the four selected texels.
Each texel is then converted to an RGBA value according to
<<textures-conversion-to-rgba,conversion to RGBA>> and then
<<textures-component-swizzle,swizzled>>.
A four-component vector is then assembled by taking the component indicated
by the code:Component value in the instruction from the swizzled color value
of the four texels.
If the operation does not use the code:ConstOffsets image operand then the
four texels form the 2 {times} 2 rectangle used for texture filtering:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau[R] &= \tau_{i0j1}[level_{base}][comp] \\
\tau[G] &= \tau_{i1j1}[level_{base}][comp] \\
\tau[B] &= \tau_{i1j0}[level_{base}][comp] \\
\tau[A] &= \tau_{i0j0}[level_{base}][comp]
\end{aligned}
++++++++++++++++++++++++

If the operation does use the code:ConstOffsets image operand then the
offsets allow a custom filter to be defined:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau[R] &= \tau_{i0j0 + \Delta_0}[level_{base}][comp] \\
\tau[G] &= \tau_{i0j0 + \Delta_1}[level_{base}][comp] \\
\tau[B] &= \tau_{i0j0 + \Delta_2}[level_{base}][comp] \\
\tau[A] &= \tau_{i0j0 + \Delta_3}[level_{base}][comp]
\end{aligned}
++++++++++++++++++++++++

where:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau[level_{base}][comp] &=
  \begin{cases}
    \tau[level_{base}][R], & \text{for}\  comp = 0 \\
    \tau[level_{base}][G], & \text{for}\  comp = 1 \\
    \tau[level_{base}][B], & \text{for}\  comp = 2 \\
    \tau[level_{base}][A], & \text{for}\  comp = 3
  \end{cases}\\
comp & \,\text{from SPIR-V operand Component}
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]
code:OpImage*Gather must: not be used on a sampled image with
<<samplers-YCbCr-conversion,sampler {YCbCr} conversion>> enabled.
endif::VK_VERSION_1_1,VK_KHR_sampler_ycbcr_conversion[]

ifdef::VK_EXT_image_view_min_lod[]
If [eq]#level~base~ < minLodInteger~imageView~#, then any values fetched are
ifdef::VK_EXT_robustness2[]
zero if <<features-robustImageAccess2, pname:robustImageAccess2>> is
enabled.
Otherwise values are
endif::VK_EXT_robustness2[]
undefined:.
endif::VK_EXT_image_view_min_lod[]


[[textures-texel-filtering]]
=== Texel Filtering

Texel filtering is first performed for each level (either [eq]#d# or
[eq]#d~hi~# and [eq]#d~lo~#).

If [eq]#{lambda}# is less than or equal to zero, the texture is said to be
_magnified_, and the filter mode within a mip level is selected by the
pname:magFilter in the sampler.
If [eq]#{lambda}# is greater than zero, the texture is said to be
_minified_, and the filter mode within a mip level is selected by the
pname:minFilter in the sampler.


[[textures-texel-nearest-filtering]]
==== Texel Nearest Filtering

Within a mip level, ename:VK_FILTER_NEAREST filtering selects a single value
using the [eq]#(i, j, k)# texel coordinates, with all texels taken from
layer l.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau[level] &=
  \begin{cases}
     \tau_{ijk}[level], & \text{for 3D image} \\
     \tau_{ij}[level],  & \text{for 2D or Cube image} \\
     \tau_{i}[level],   & \text{for 1D image}
   \end{cases}
\end{aligned}
++++++++++++++++++++++++


[[textures-texel-linear-filtering]]
==== Texel Linear Filtering

Within a mip level, ename:VK_FILTER_LINEAR filtering combines 8 (for 3D), 4
(for 2D or Cube), or 2 (for 1D) texel values, together with their linear
weights.
The linear weights are derived from the fractions computed earlier:

[latexmath]

++++++++++++++++++++++++
\begin{aligned}
w_{i_0} &= (1-\alpha) \\
w_{i_1} &= (\alpha)   \\
w_{j_0} &= (1-\beta)  \\
w_{j_1} &= (\beta)    \\
w_{k_0} &= (1-\gamma) \\
w_{k_1} &= (\gamma)
\end{aligned}
++++++++++++++++++++++++

ifndef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
The values of multiple texels, together with their weights, are combined
using a weighted average to produce a filtered value:
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]

ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
The values of multiple texels, together with their weights, are combined to
produce a filtered value.

The slink:VkSamplerReductionModeCreateInfo::pname:reductionMode can: control
the process by which multiple texels, together with their weights, are
combined to produce a filtered texture value.

When the pname:reductionMode is set (explicitly or implicitly) to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, a weighted average is
computed:
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{3D} &= \sum_{k=k_0}^{k_1}\sum_{j=j_0}^{j_1}\sum_{i=i_0}^{i_1}(w_{i})(w_{j})(w_{k})\tau_{ijk} \\
\tau_{2D} &= \sum_{j=j_0}^{j_1}\sum_{i=i_0}^{i_1}(w_{i})(w_{j})\tau_{ij} \\
\tau_{1D} &= \sum_{i=i_0}^{i_1}(w_{i})\tau_{i}
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above set
of multiple texels, together with their weights, computing a component-wise
minimum or maximum, respectively, of the components of the set of texels
with non-zero weights.
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]


ifdef::VK_IMG_filter_cubic,VK_EXT_filter_cubic[]
[[textures-texel-cubic-filtering]]
==== Texel Cubic Filtering

Within a mip level, ename:VK_FILTER_CUBIC_EXT, filtering computes a weighted
average of
ifdef::VK_EXT_filter_cubic[]
64 (for 3D),
endif::VK_EXT_filter_cubic[]
16 (for 2D), or 4 (for 1D) texel values, together with their
ifndef::VK_QCOM_filter_cubic_weights[]
Catmull-Rom weights.
endif::VK_QCOM_filter_cubic_weights[]
ifdef::VK_QCOM_filter_cubic_weights[]
Catmull-Rom, Zero Tangent Cardinal, B-Spline, or Mitchell-Netravali weights
as specified by slink:VkSamplerCubicWeightsCreateInfoQCOM.
endif::VK_QCOM_filter_cubic_weights[]


Catmull-Rom weights
ifdef::VK_QCOM_filter_cubic_weights[]
specified by ename:VK_CUBIC_FILTER_WEIGHTS_CATMULL_ROM_QCOM
endif::VK_QCOM_filter_cubic_weights[]
are derived from the fractions computed earlier.

ifndef::VK_EXT_filter_cubic[]
[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\begin{bmatrix}
w_{i_0}\phantom{,} w_{i_1}\phantom{,} w_{i_2}\phantom{,} w_{i_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \alpha & \alpha^2 & \alpha^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-1 &  \phantom{-}0 &  \phantom{-}1 &  \phantom{-}0 \\
\phantom{-}2 & -5 &  \phantom{-}4 &  -1 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{j_0}\phantom{,} w_{j_1}\phantom{,} w_{j_2}\phantom{,} w_{j_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \beta & \beta^2 & \beta^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-1 &  \phantom{-}0 &  \phantom{-}1 &  \phantom{-}0 \\
\phantom{-}2 & -5 &  \phantom{-}4 &  -1 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\end{aligned}
++++++++++++++++++++++++

The values of multiple texels, together with their weights, are combined
using a weighted average to produce a filtered value:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{2D} &= \sum_{j=j_0}^{j_3}\sum_{i=i_0}^{i_3}(w_{i})(w_{j})\tau_{ij} \\
\tau_{1D} &= \sum_{i=i_0}^{i_3}(w_{i})\tau_{i}
\end{aligned}
++++++++++++++++++++++++
endif::VK_EXT_filter_cubic[]

ifdef::VK_EXT_filter_cubic[]

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\begin{bmatrix}
w_{i_0}\phantom{,} w_{i_1}\phantom{,} w_{i_2}\phantom{,} w_{i_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \alpha & \alpha^2 & \alpha^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-1 &  \phantom{-}0 &  \phantom{-}1 &  \phantom{-}0 \\
\phantom{-}2 & -5 &  \phantom{-}4 &  -1 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{j_0}\phantom{,} w_{j_1}\phantom{,} w_{j_2}\phantom{,} w_{j_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \beta & \beta^2 & \beta^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-1 &  \phantom{-}0 &  \phantom{-}1 &  \phantom{-}0 \\
\phantom{-}2 & -5 &  \phantom{-}4 &  -1 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{k_0}\phantom{,} w_{k_1}\phantom{,} w_{k_2}\phantom{,} w_{k_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \gamma & \gamma^2 & \gamma^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-1 &  \phantom{-}0 &  \phantom{-}1 &  \phantom{-}0 \\
\phantom{-}2 & -5 &  \phantom{-}4 &  -1 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_QCOM_filter_cubic_weights[]
Zero Tangent Cardinal weights specified by
ename:VK_CUBIC_FILTER_WEIGHTS_ZERO_TANGENT_CARDINAL_QCOM are derived from
the fractions computed earlier.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\begin{bmatrix}
w_{i_0}\phantom{,} w_{i_1}\phantom{,} w_{i_2}\phantom{,} w_{i_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \alpha & \alpha^2 & \alpha^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-2 &  \phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 \\
\phantom{-}4 & -4 &  \phantom{-}2 &  -2 \\
-2 &  \phantom{-}2 & -2 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{j_0}\phantom{,} w_{j_1}\phantom{,} w_{j_2}\phantom{,} w_{j_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \beta & \beta^2 & \beta^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-2 &  \phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 \\
\phantom{-}4 & -4 &  \phantom{-}2 &  -2 \\
-2 &  \phantom{-}2 & -2 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{k_0}\phantom{,} w_{k_1}\phantom{,} w_{k_2}\phantom{,} w_{k_3}
\end{bmatrix}
= \frac{1}{2}
\begin{bmatrix}
1 & \gamma & \gamma^2 & \gamma^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 &  \phantom{-}0 \\
-2 &  \phantom{-}0 &  \phantom{-}2 &  \phantom{-}0 \\
\phantom{-}4 & -4 &  \phantom{-}2 &  -2 \\
-2 &  \phantom{-}2 & -2 &  \phantom{-}1
\end{bmatrix}
\end{aligned}
++++++++++++++++++++++++

B-Spline weights specified by ename:VK_CUBIC_FILTER_WEIGHTS_B_SPLINE_QCOM
are derived from the fractions computed earlier.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\begin{bmatrix}
w_{i_0}\phantom{,} w_{i_1}\phantom{,} w_{i_2}\phantom{,} w_{i_3}
\end{bmatrix}
= \frac{1}{6}
\begin{bmatrix}
1 & \alpha & \alpha^2 & \alpha^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}4 &  \phantom{-}1 &  \phantom{-}0 \\
-3 &  \phantom{-}0 &  \phantom{-}3 &  \phantom{-}0 \\
\phantom{-}3 & -6 &  \phantom{-}3 &  \phantom{-}0 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{j_0}\phantom{,} w_{j_1}\phantom{,} w_{j_2}\phantom{,} w_{j_3}
\end{bmatrix}
= \frac{1}{6}
\begin{bmatrix}
1 & \beta & \beta^2 & \beta^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}4 &  \phantom{-}1 &  \phantom{-}0 \\
-3 &  \phantom{-}0 &  \phantom{-}3 &  \phantom{-}0 \\
\phantom{-}3 & -6 &  \phantom{-}3 &  \phantom{-}0 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\\
\begin{bmatrix}
w_{k_0}\phantom{,} w_{k_1}\phantom{,} w_{k_2}\phantom{,} w_{k_3}
\end{bmatrix}
= \frac{1}{6}
\begin{bmatrix}
1 & \gamma & \gamma^2 & \gamma^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}4 &  \phantom{-}1 &  \phantom{-}0 \\
-3 &  \phantom{-}0 &  \phantom{-}3 &  \phantom{-}0 \\
\phantom{-}3 & -6 &  \phantom{-}3 &  \phantom{-}0 \\
-1 &  \phantom{-}3 & -3 &  \phantom{-}1
\end{bmatrix}
\end{aligned}
++++++++++++++++++++++++

Mitchell-Netravali weights specified by
ename:VK_CUBIC_FILTER_WEIGHTS_MITCHELL_NETRAVALI_QCOM are derived from the
fractions computed earlier.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\begin{bmatrix}
w_{i_0}\phantom{,} w_{i_1}\phantom{,} w_{i_2}\phantom{,} w_{i_3}
\end{bmatrix}
= \frac{1}{18}
\begin{bmatrix}
1 & \alpha & \alpha^2 & \alpha^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}16 &  \phantom{-}1 &  \phantom{-}0 \\
-9 &  \phantom{-}0 &  \phantom{-}9 &  \phantom{-}0 \\
\phantom{-}15 & -36 &  \phantom{-}27 &  -6 \\
-7 &  \phantom{-}21 & -21 &  \phantom{-}7
\end{bmatrix}
\\
\begin{bmatrix}
w_{j_0}\phantom{,} w_{j_1}\phantom{,} w_{j_2}\phantom{,} w_{j_3}
\end{bmatrix}
= \frac{1}{18}
\begin{bmatrix}
1 & \beta & \beta^2 & \beta^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}16 &  \phantom{-}1 &  \phantom{-}0 \\
-9 &  \phantom{-}0 &  \phantom{-}9 &  \phantom{-}0 \\
\phantom{-}15 & -36 &  \phantom{-}27 &  -6 \\
-7 &  \phantom{-}21 & -21 &  \phantom{-}7
\end{bmatrix}
\\
\begin{bmatrix}
w_{k_0}\phantom{,} w_{k_1}\phantom{,} w_{k_2}\phantom{,} w_{k_3}
\end{bmatrix}
= \frac{1}{18}
\begin{bmatrix}
1 & \gamma & \gamma^2 & \gamma^3
\end{bmatrix}
\begin{bmatrix}
\phantom{-}1 &  \phantom{-}16 &  \phantom{-}1 &  \phantom{-}0 \\
-9 &  \phantom{-}0 &  \phantom{-}9 &  \phantom{-}0 \\
\phantom{-}15 & -36 &  \phantom{-}27 &  -6 \\
-7 &  \phantom{-}21 & -21 &  \phantom{-}7
\end{bmatrix}
\end{aligned}
++++++++++++++++++++++++

endif::VK_QCOM_filter_cubic_weights[]


The values of multiple texels, together with their weights, are combined to
produce a filtered value.

The slink:VkSamplerReductionModeCreateInfo::pname:reductionMode can: control
the process by which multiple texels, together with their weights, are
combined to produce a filtered texture value.

When the pname:reductionMode is set (explicitly or implicitly) to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE
ifdef::VK_QCOM_filter_cubic_clamp[]
or ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_RANGECLAMP_QCOM
endif::VK_QCOM_filter_cubic_clamp[]
, a weighted average is computed:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{3D} &= \sum_{k=j_0}^{k_3}\sum_{j=j_0}^{j_3}\sum_{i=i_0}^{i_3}(w_{i})(w_{j})(w_{k})\tau_{ijk} \\
\tau_{2D} &= \sum_{j=j_0}^{j_3}\sum_{i=i_0}^{i_3}(w_{i})(w_{j})\tau_{ij} \\
\tau_{1D} &= \sum_{i=i_0}^{i_3}(w_{i})\tau_{i}
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above set
of multiple texels, together with their weights, computing a component-wise
minimum or maximum, respectively, of the components of the set of texels
with non-zero weights.
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]


ifdef::VK_QCOM_filter_cubic_clamp[]
[[textures-texel-range-clamp]]
==== Texel Range Clamp
When the pname:reductionMode is set to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_RANGECLAMP_QCOM, the
weighted average is clamped to be within the component-wise minimum and
maximum of the set of texels with non-zero weights.
endif::VK_QCOM_filter_cubic_clamp[]

endif::VK_EXT_filter_cubic[]
endif::VK_IMG_filter_cubic,VK_EXT_filter_cubic[]


[[textures-texel-mipmap-filtering]]
==== Texel Mipmap Filtering

ename:VK_SAMPLER_MIPMAP_MODE_NEAREST filtering returns the value of a single
mipmap level,

[eq]#{tau} = {tau}[d]#.

ename:VK_SAMPLER_MIPMAP_MODE_LINEAR filtering combines the values of
multiple mipmap levels ({tau}[hi] and {tau}[lo]), together with their linear
weights.

The linear weights are derived from the fraction computed earlier:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
w_{hi} &= (1-\delta) \\
w_{lo} &= (\delta)   \\
\end{aligned}
++++++++++++++++++++++++

ifndef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
The values of multiple mipmap levels together with their linear weights, are
combined using a weighted average to produce a final filtered value:
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
The values of multiple mipmap levels, together with their weights, are
combined to produce a final filtered value.

The slink:VkSamplerReductionModeCreateInfo::pname:reductionMode can: control
the process by which multiple texels, together with their weights, are
combined to produce a filtered texture value.

When the pname:reductionMode is set (explicitly or implicitly) to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, a weighted average is
computed:
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau &= (w_{hi})\tau[hi]+(w_{lo})\tau[lo]
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above
values, together with their weights, computing a component-wise minimum or
maximum, respectively, of the components of the values with non-zero
weights.
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]


[[textures-texel-anisotropic-filtering]]
==== Texel Anisotropic Filtering

Anisotropic filtering is enabled by the pname:anisotropyEnable in the
sampler.
When enabled, the image filtering scheme accounts for a degree of
anisotropy.

The particular scheme for anisotropic texture filtering is
implementation-dependent.
Implementations should: consider the pname:magFilter, pname:minFilter and
pname:mipmapMode of the sampler to control the specifics of the anisotropic
filtering scheme used.
In addition, implementations should: consider pname:minLod and pname:maxLod
of the sampler.

[NOTE]
.Note
====
For historical reasons, vendor implementations of anisotropic filtering
interpret these sampler parameters in different ways, particularly in corner
cases such as pname:magFilter, pname:minFilter of ename:NEAREST or
pname:maxAnisotropy equal to 1.0.
Applications should not expect consistent behavior in such cases, and should
use anisotropic filtering only with parameters which are expected to give a
quality improvement relative to etext:LINEAR filtering.

The following describes one particular approach to implementing anisotropic
filtering for the 2D Image case; implementations may: choose other methods:

Given a pname:magFilter, pname:minFilter of ename:VK_FILTER_LINEAR and a
pname:mipmapMode of ename:VK_SAMPLER_MIPMAP_MODE_NEAREST:

Instead of a single isotropic sample, N isotropic samples are sampled within
the image footprint of the image level [eq]#d# to approximate an anisotropic
filter.
The sum [eq]#{tau}~2Daniso~# is defined using the single isotropic
[eq]#{tau}~2D~(u,v)# at level [eq]#d#.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{2Daniso} & =
     \frac{1}{N}\sum_{i=1}^{N}
     {\tau_{2D}\left (
       u \left ( x - \frac{1}{2} + \frac{i}{N+1} , y \right ),
       v \left (x-\frac{1}{2}+\frac{i}{N+1}, y \right )
     \right )},
     & \text{when}\  \rho_{x} > \rho_{y} \\
\tau_{2Daniso} &=
     \frac{1}{N}\sum_{i=1}^{N}
     {\tau_{2D}\left (
        u \left ( x, y - \frac{1}{2} + \frac{i}{N+1} \right ),
        v \left (x,y-\frac{1}{2}+\frac{i}{N+1} \right )
     \right )},
     & \text{when}\  \rho_{y} \geq \rho_{x}
\end{aligned}
++++++++++++++++++++++++

ifdef::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
When slink:VkSamplerReductionModeCreateInfo::pname:reductionMode is set to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, the above summation is
used.
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above
values, together with their weights, computing a component-wise minimum or
maximum, respectively, of the components of the values with non-zero
weights.
endif::VK_VERSION_1_2,VK_EXT_sampler_filter_minmax[]
====


ifdef::VK_NV_shader_image_footprint[]
[[textures-footprint]]
== Texel Footprint Evaluation

The SPIR-V instruction code:OpImageSampleFootprintNV evaluates the set of
texels from a single mip level that would be accessed during a
<<textures-texel-filtering, texel filtering>> operation.
In addition to the inputs that would be accepted by an equivalent
code:OpImageSample* instruction, code:OpImageSampleFootprintNV accepts two
additional inputs.
The code:Granularity input is an integer identifying the size of texel
groups used to evaluate the footprint.
Each bit in the returned footprint mask corresponds to an aligned block of
texels whose size is given by the following table:

.Texel footprint granularity values
[width="50%",options="header"]
|====
| code:Granularity  | code:Dim = 2D |  code:Dim = 3D
|         0         |  unsupported  |  unsupported
|         1         |      2x2      |     2x2x2
|         2         |      4x2      |  unsupported
|         3         |      4x4      |     4x4x2
|         4         |      8x4      |  unsupported
|         5         |      8x8      |  unsupported
|         6         |     16x8      |  unsupported
|         7         |     16x16     |  unsupported
|         8         |  unsupported  |  unsupported
|         9         |  unsupported  |  unsupported
|         10        |  unsupported  |    16x16x16
|         11        |     64x64     |    32x16x16
|         12        |    128x64     |    32x32x16
|         13        |    128x128    |    32x32x32
|         14        |    256x128    |    64x32x32
|         15        |    256x256    |  unsupported
|====

The code:Coarse input is used to select between the two mip levels that may:
be accessed during texel filtering when using a pname:mipmapMode of
ename:VK_SAMPLER_MIPMAP_MODE_LINEAR.
When filtering between two mip levels, a code:Coarse value of code:true
requests the footprint in the lower-resolution mip level (higher level
number), while code:false requests the footprint in the higher-resolution
mip level.
If texel filtering would access only a single mip level, the footprint in
that level would be returned when code:Coarse is set to code:false; an empty
footprint would be returned when code:Coarse is set to code:true.

The footprint for code:OpImageSampleFootprintNV is returned in a structure
with six members:

  * The first member is a boolean value that is true if the texel filtering
    operation would access only a single mip level.
  * The second member is a two- or three-component integer vector holding
    the footprint anchor location.
    For two-dimensional images, the returned components are in units of
    eight texel groups.
    For three-dimensional images, the returned components are in units of
    four texel groups.
  * The third member is a two- or three-component integer vector holding a
    footprint offset relative to the anchor.
    All returned components are in units of texel groups.
  * The fourth member is a two-component integer vector mask, which holds a
    bitfield identifying the set of texel groups in an 8x8 or 4x4x4
    neighborhood relative to the anchor and offset.
  * The fifth member is an integer identifying the mip level containing the
    footprint identified by the anchor, offset, and mask.
  * The sixth member is an integer identifying the granularity of the
    returned footprint.

For footprints in two-dimensional images (code:Dim2D), the mask returned by
code:OpImageSampleFootprintNV indicates whether each texel group in a 8x8
local neighborhood of texel groups would have one or more texels accessed
during texel filtering.
In the mask, the texel group with local group coordinates
latexmath:[(lgx,lgy)] is considered covered if and only if

[latexmath]
+++++++++++++++++++
\begin{aligned}
0 \neq ((mask.x + (mask.y << 32)) \text{ \& } (1 << (lgy \times 8 + lgx)))
\end{aligned}
+++++++++++++++++++

where:

  * latexmath:[0 \leq lgx < 8] and latexmath:[0 \leq lgy < 8]; and
  * latexmath:[mask] is the returned two-component mask.

The local group with coordinates latexmath:[(lgx,lgy)] in the mask is
considered covered if and only if the texel filtering operation would access
one or more texels latexmath:[\tau_{ij}] in the returned mip level where:
[latexmath]
+++++++++++++++++++
\begin{aligned}
i0 & =
  \begin{cases}
    gran.x \times (8 \times anchor.x + lgx), & \text{if } lgx + offset.x < 8 \\
    gran.x \times (8 \times (anchor.x - 1) + lgx), & \text{otherwise}
  \end{cases} \\
i1 & = i0 + gran.x - 1 \\
j0 & =
  \begin{cases}
    gran.y \times (8 \times anchor.y + lgy), & \text{if } lgy + offset.y < 8 \\
    gran.y \times (8 \times (anchor.y - 1) + lgy), & otherwise
  \end{cases} \\
j1 & = j0 + gran.y - 1
\end{aligned}
+++++++++++++++++++
and

  * latexmath:[i0 \leq i \leq i1] and latexmath:[j0 \leq j \leq j1];
  * latexmath:[gran] is a two-component vector holding the width and height
    of the texel group identified by the granularity;
  * latexmath:[anchor] is the returned two-component anchor vector; and
  * latexmath:[offset] is the returned two-component offset vector.

For footprints in three-dimensional images (code:Dim3D), the mask returned
by code:OpImageSampleFootprintNV indicates whether each texel group in a
4x4x4 local neighborhood of texel groups would have one or more texels
accessed during texel filtering.
In the mask, the texel group with local group coordinates
latexmath:[(lgx,lgy,lgz)], is considered covered if and only if:
[latexmath]
+++++++++++++++++++
\begin{aligned}
0 \neq ((mask.x + (mask.y << 32)) \text{ \& } (1 << (lgz \times 16 + lgy \times 4 + lgx)))
\end{aligned}
+++++++++++++++++++
where:

  * latexmath:[0 \leq lgx < 4], latexmath:[0 \leq lgy < 4], and latexmath:[0
    \leq lgz < 4]; and
  * latexmath:[mask] is the returned two-component mask.

The local group with coordinates latexmath:[(lgx,lgy,lgz)] in the mask is
considered covered if and only if the texel filtering operation would access
one or more texels latexmath:[\tau_{ijk}] in the returned mip level where:
[latexmath]
+++++++++++++++++++
\begin{aligned}
i0 & =
  \begin{cases}
    gran.x \times (4 \times anchor.x + lgx), & \text{if } lgx + offset.x < 4 \\
    gran.x \times (4 \times (anchor.x - 1) + lgx), & \text{otherwise}
  \end{cases} \\
i1 & = i0 + gran.x - 1 \\
j0 & =
  \begin{cases}
    gran.y \times (4 \times anchor.y + lgy), & \text{if } lgy + offset.y < 4 \\
    gran.y \times (4 \times (anchor.y - 1) + lgy), & otherwise
  \end{cases} \\
j1 & = j0 + gran.y - 1 \\
k0 & =
  \begin{cases}
    gran.z \times (4 \times anchor.z + lgz), & \text{if } lgz + offset.z < 4 \\
    gran.z \times (4 \times (anchor.z - 1) + lgz), & otherwise
  \end{cases} \\
k1 & = k0 + gran.z - 1
\end{aligned}
+++++++++++++++++++
and

  * latexmath:[i0 \leq i \leq i1], latexmath:[j0 \leq j \leq j1],
    latexmath:[k0 \leq k \leq k1];
  * latexmath:[gran] is a three-component vector holding the width, height,
    and depth of the texel group identified by the granularity;
  * latexmath:[anchor] is the returned three-component anchor vector; and
  * latexmath:[offset] is the returned three-component offset vector.

If the sampler used by code:OpImageSampleFootprintNV enables anisotropic
texel filtering via pname:anisotropyEnable, it is possible that the set of
texel groups accessed in a mip level may be too large to be expressed using
an 8x8 or 4x4x4 mask using the granularity requested in the instruction.
In this case, the implementation uses a texel group larger than the
requested granularity.
When a larger texel group size is used, code:OpImageSampleFootprintNV
returns an integer granularity value that can: be interpreted in the same
manner as the granularity value provided to the instruction to determine the
texel group size used.
If anisotropic texel filtering is disabled in the sampler, or if an
anisotropic footprint can be represented as an 8x8 or 4x4x4 mask with the
requested granularity, code:OpImageSampleFootprintNV will use the requested
granularity as-is and return a granularity value of zero.

code:OpImageSampleFootprintNV supports only two- and three-dimensional image
accesses (code:Dim2D and code:Dim3D), and the footprint returned is
undefined: if a sampler uses an addressing mode other than
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE.

endif::VK_NV_shader_image_footprint[]


ifdef::VK_QCOM_image_processing[]
[[textures-weightimage]]
== Weight Image Sampling

The SPIR-V instruction code:OpImageWeightedSampleQCOM specifies a texture
sampling operation involving two images: the _sampled image_ and the _weight
image_.
It is similar to bilinear filtering except more than 2x2 texels may
participate in the filter and the filter weights are user-specified rather
than computed by fixed-function hardware.
The weight image view defines a 2D kernel weights used during sampling.

The code:OpImageWeightedSampleQCOM support normalized or unnormalized texel
coordinates.
In addition to the inputs that would be accepted by an equivalent
code:OpImageSample* instruction, code:OpImageWeightedSampleQCOM accepts a
code:weight input that specifies the view of a sample weight image

The input code:weight must: be a view of a 2D or 1D image with
code:miplevels equal to `1`, code:samples equal to
ename:VK_SAMPLE_COUNT_1_BIT, created with an identity swizzle, and created
with code:usage that includes ename:VK_IMAGE_USAGE_SAMPLE_WEIGHT_BIT_QCOM.
The slink:VkImageViewSampleWeightCreateInfoQCOM specifies additional
parameters of the view: pname:filterCenter, pname:filterSize, and
pname:numPhases.
described in more detail below.

The code:weight input must: be bound using a
<<descriptorsets-weightimage,sample weight image>> descriptor type.
The code:weight view defines a filtering kernel that is a region of view's
subresource range.
The kernel spans a region from integer texel coordinate [eq]#(0,0)# to
[eq]#(pname:filterSize.x-1, pname:filterSize.y-1)#.
It is valid for the view's subresource to have dimensions larger than the
kernel but the texels with integer coordinates greater than
[eq]#(pname:filterSize.width-1, pname:filterSize.height-1)# are ignored by
weight sampling.
The value returned by queries code:OpImageQuerySize,
code:OpImageQuerySizeLod, code:OpImageQueryLevels, and
code:OpImageQuerySamples return for a weight image is undefined:.

pname:filterCenter designates an integer texel coordinate within the filter
kernel as being the 'center' of the kernel.
The center must: be in the range [eq]#(0,0)# to [eq]#(pname:filterSize.x-1,
pname:filterSize.y-1)#.
pname:numPhases describes the number of filter phases used to provide
sub-pixel filtering.
Both are described in more detail below.


[[textures-weightimage-layout]]
=== Weight Image Layout

The weight image specifies filtering kernel weight values.
A 2D image view can be used to specify a 2D matrix of filter weights.
For separable filers, a 1D image view can be used to specity the horizontal
and vertical weights.


==== 2D Non-Separable Weight Filters

A 2D image view defined with slink:VkImageViewSampleWeightCreateInfoQCOM
describes a 2D matrix [eq]#(pname:filterSize.width {times}
pname:filterSize.height)# of weight elements with filter's center point at
pname:filterCenter.
Note that pname:filterSize can be smaller than the view's subresource, but
the filter will always be located starting at integer texel coordinate
[eq]#(0,0)#.

The following figure illustrates a 2D convolution filter having
pname:filterSize of [eq]#(4,3)# and pname:filterCenter at [eq]#(1, 1)#.

image::{images}/weight_filter_2d.svg[align="center",title="2D Convolution Filter",opts="{imageopts}"]

For a 2D weight filter, the phases are stored as layers of a 2D array image.
The width and height of the view's subresource range must: be less than or
equal to
slink:VkPhysicalDeviceImageProcessingPropertiesQCOM::pname:maxWeightFilterDimension.
The layers are stored in horizontal phase major order.
Expressed as a formula, the layer index for a each filter phase is computed
as:

[source,c]
----
layerIndex(horizPhase,vertPhase,horizPhaseCount) = (vertPhase * horizPhaseCount) + horizPhase
----


==== 1D Separable Weight Filters

A separable weight filter is a 2D filter that can be specified by two 1D
filters in the [eq]#x# and [eq]#y# directions such that their product yields
the 2D filter.
The following example shows a 2D filter and its associated separable 1D
horizontal and vertical filters.

image::{images}/weight_filter_1d_separable.svg[align="center",title="Separable 2D Convolution Filter",opts="{imageopts}"]

A 1D array image view defined with
slink:VkImageViewSampleWeightCreateInfoQCOM and with pname:layerCount equal
to '2' describes a separable weight filter.
The horizontal weights are specified in slice '0' and the vertical weights
in slice '1'.
The pname:filterSize and pname:filterCenter specify the size and origin of
the of the horizontal and vertical filters.
For many use cases, 1D separable filters can offer a performance advantage
over 2D filters.

For a 1D separable weight filter, the phases are arranged into a 1D array
image with two layers.
The horizontal weights are stored in layer 0 and the vertical weights in
layer 1.
Within each layer of the 1D array image, the weights are arranged into
groups of 4, and then arranged by phase.
Expressed as a formula, the 1D texel offset for each weight within each
layer is computed as:

[source,c]
----
// Let horizontal weights have a weightIndex of [0, filterSize.width - 1]
// Let vertical weights have a weightIndex of [0, filterSize.height - 1]
// Let phaseCount be the number of phases in either the vertical or horizontal direction.

texelOffset(phaseIndex,weightIndex,phaseCount) = (phaseCount * 4 * (weightIndex / 4)) + (phaseIndex * 4) + (weightIndex % 4)
----


[[textures-weightimage-filterphases]]
=== Weight Sampling Phases

When using weight image sampling, the texture coordinates may not align with
a texel center in the sampled image.
In this case, the filter weights can be adjusted based on the subpixel
location.
This is termed "`subpixel filtering`" to indicate that the origin of the
filter lies at a subpixel location other than the texel center.
Conceptually, this means that the weight filter is positioned such that
filter taps do not align with sampled texels exactly.
In such a case, modified filter weights may be needed to adjust for the
off-center filter taps.
Unlike bilinear filtering where the subpixel weights are computed by the
implementation, subpixel weight image sampling requires that the per-phase
filter weights are pre-computed by the application and stored in an array
where each slice of the array is a "`filter phase`".
The array is indexed by the implementation based on subpixel positioning.
Rather than a single 2D kernel of filter weights, the application provides
an array of kernels, one set of filter weights per phase.

The number of phases are restricted by following requirements, which apply
to both separable and non-separable filters:

  * The number of phases in the vertical direction, [eq]#phaseCount~vert~#,
    must: be a power of two (i.e., 1, 2, 4, etc.).
  * The number of phases in the horizontal direction
    [eq]#phaseCount~horiz~#, must: equal [eq]#phaseCount~vert~#.
  * The total number of phases, [eq]#phaseCount~vert~ {times}
    phaseCount~horiz~#, must: be less than or equal to
    slink:VkPhysicalDeviceImageProcessingPropertiesQCOM::pname:maxWeightFilterPhases.


[[textures-weightimage-sampler]]
=== Weight Sampler Parameters

Weight sampling requires sname:VkSamplerCreateInfo pname:addressModeU and
pname:addressModeV must: be set to
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE or
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER.
If ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER is used, then the border
color must: be set to transparent black.


[[textures-weightimage-filteroperation]]
=== Weight Sampling Operation

The 2D unnormalized texel coordinates latexmath:[(u,v)] are transformed by
latexmath:[filterCenter] to specify coordinates latexmath:[i_{0}, j_{0}].

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_{0} &= \left\lfloor u - filterCenter_{x} \right\rfloor \\[1em]
j_{0} &= \left\lfloor v - filterCenter_{y} \right\rfloor
\end{aligned}
++++++++++++++++++++++++
where latexmath:[filterCenter] is specified by
slink:VkImageViewSampleWeightCreateInfoQCOM::pname:filterCenter.


Two sets of neighboring integer 2D texel coordinates are generated.
The first set is used for selecting texels from the sampled image
latexmath:[\tau] and the second set used for selecting texels from the
weight image latexmath:[w].
The first set of neighboring coordinates are combinations of
latexmath:[i_{0}] to latexmath:[i_{filterWidth-1}] and latexmath:[j_{0}] to
latexmath:[j_{filterHeight-1}].
The second set of neighboring coordinates are combinations of
latexmath:[k_{0}] to latexmath:[k_{filterWidth-1}] and latexmath:[l_{0}] to
latexmath:[l_{filterHeight-1}].
The first and second sets each contain latexmath:[(filterWidth \times
filterHeight)] of pairs of latexmath:[(i,j)] and latexmath:[(k,l)]
coordinates respectively.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\{i_q\}_{q=0}^{q=filterWidth-1} \quad &= i_{0} + q  \\[1em]
\{j_q\}_{q=0}^{q=filterHeight-1} \quad &= j_{0} + q \\[1em]
\{k_q\}_{q=0}^{q=filterWidth-1} \quad &= q  \\[1em]
\{l_q\}_{q=0}^{q=filterHeight-1} \quad &= q
\end{aligned}
++++++++++++++++++++++++

where latexmath:[filterWidth] and latexmath:[filterHeight] are specified by
slink:VkImageViewSampleWeightCreateInfoQCOM::pname:filterSize.

Each of the generated integer coordinates latexmath:[({i_q}, {j_q})] is
transformed by <<textures-wrapping-operation, texture wrapping operation>>,
followed by <<textures-integer-coordinate-validation,integer texel
coordinate validation>>, If any coordinate fails coordinate validation, it
is a Border Texel and <<textures-texel-replacement,texel replacement>> is
performed.


The phase index latexmath:[\psi] is computed from the fraction bits of the
unnormalized 2D texel coordinates:

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
phaseCount_{h} = phaseCount_{v} &= \sqrt{numPhases}  \\[1em]
hPhase &= \left\lfloor\mathbin{frac}\left( u \right) \times phaseCount_{h} \right\rfloor  \\[1em]
vPhase &= \left\lfloor\mathbin{frac}\left( v \right) \times phaseCount_{v} \right\rfloor  \\[1em]
\psi &= \left(vPhase \times phaseCount_{h}\right) + hPhase
\end{aligned}
++++++++++++++++++++++++

where the number of fraction bits retained is
latexmath:[\mathbin{log2}\left( numPhases \right)] specified by
slink:VkImageViewSampleWeightCreateInfoQCOM::pname:numPhases

Each pair of texel coordinates latexmath:[(i,j)] in the first set selects a
single texel value latexmath:[\tau_{ij}] from the sampled image.
Each pair of texel coordinates latexmath:[(k,l)] in the second set, combined
with phaseIndex latexmath:[\psi], selects a single weight from the weight
image latexmath:[w(k,l,\psi)] .


[latexmath]
++++++++++++++++++++++++
\begin{aligned}
w(k,l,\psi) &=
  \begin{cases}
    w_{kl}[\psi]\quad\text{(}\psi\text{ as layer index)}   & \text{for 2D array view (non-separable filter) } \\
    weight_{h} \times weight_{v}                           & \text{for 1D array view (separable filter) } \\
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

If latexmath:[w] is a 2D array view, then non-separable filtering is
specified, and integer coordinates latexmath:[(k,l)] are used to select
texels from layer latexmath:[\psi] of latexmath:[(w)].
If latexmath:[w] is a 1D array view, then separable filtering is specified
and integer coordinates latexmath:[(k,l)] are transformed to
latexmath:[(k_{packed},l_{packed})], and used to select horizontal weight
latexmath:[(weight_{h})] and vertical weight latexmath:[(weight_{v})] texels
from layer 0 and layer 1 of latexmath:[(w)] respectively.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
k_{packed} &= \left(phaseCount_{h} \times 4 \times \left\lfloor k / 4 \right\rfloor\right) + \left(hPhase \times 4\right) + \left(k \mathbin{\%} 4\right)  \\[1em]
l_{packed}& = \left(phaseCount_{v} \times 4 \times \left\lfloor l / 4 \right\rfloor\right) + \left(vPhase  \times 4\right) + \left(l \mathbin{\%} 4\right) \\[1em]
weight_{h} &= w_{k_{packed}}[0]  & \text{(horizontal weights packed in layer 0)}  \\[1em]
weight_{v}  &= w_{l_{packed}}[1]  & \text{(vertical weights packed in layer 1)}

\end{aligned}
++++++++++++++++++++++++

Where latexmath:[\mathbin{\%}] refers to the integer modulo operator.

The values of multiple texels, together with their weights, are combined to
produce a filtered value.


[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{weightSampling} &= \sum_{{j=j_0} \atop {l=l_0}}^{j_{blockHeight-1} \atop {l_{blockHeight-1}}}\quad \sum_{{i=i_0}\atop {k=k_0}}^{i_{blockWidth-1} \atop {k_{blockWidth-1}}}w(k,l,\psi)\tau_{ij} \\
\end{aligned}
++++++++++++++++++++++++

When slink:VkSamplerReductionModeCreateInfo::pname:reductionMode is set to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, the above summation is
used.
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above
values, computing a component-wise minimum or maximum of the texels with
non-zero weights.
If the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, each latexmath:[w(k,l,\psi)] weight
must: be equal to 0.0 or 1.0, otherwise the undefined: values are returned.

Finally, the operations described in
<<textures-conversion-to-rgba,Conversion to RGBA>> and
<<textures-component-swizzle,Component swizzle>> are performed and the final
result is returned to the shader.


[[textures-blockmatch]]
== Block Matching

The SPIR-V instruction code:opImageBlockMatchSAD and
code:opImageBlockMatchSSD specify texture block matching operations where a
block or region of texels within a _target image_ is compared with a
same-sized region a _reference image_.
The instructions make use of two image views: the _target view_ and the
_reference view_.
The target view and reference view can be the same view, allowing block
matching of two blocks within a single image.

Similar to an equivalent code:OpImageFetch instruction,
code:opImageBlockMatchSAD and code:opImageBlockMatchSAD specify a code:image
and an integer texel code:coordinate which which describes the bottom-left
texel of the target block.
There are three additional inputs.
The code:reference and code:refCoodinate specifies bottom-left texel of the
reference block.
The code:blockSize specifies the integer width and height of the target and
reference blocks to be compared, and must: not be greater than
slink:VkPhysicalDeviceImageProcessingPropertiesQCOM.code:maxBlockMatchRegion.

ifdef::VK_QCOM_image_processing2[]
code:opImageBlockMatchWindowSAD and code:opImageBlockMatchWindowSAD take the
same input parameters as the corresponding non-window instructions.
The block matching comparison is performed for all pixel values within a 2D
window whose dimensions are specified in the sampler.
endif::VK_QCOM_image_processing2[]


[[textures-blockmatch-sampler]]
=== Block Matching Sampler Parameters

For code:opImageBlockMatchSAD and code:opImageBlockMatchSSD, the input
code:sampler must: be created with code:addressModeU and code:addressModeV,
equal to ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, or
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER with
ename:VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK.
The input code:sampler must: be created with code:unnormalizedCoordinates
equal to ename:VK_TRUE.
The input code:sampler must: be created with pname:components equal to
ename:VK_COMPONENT_SWIZZLE_IDENTITY.


ifdef::VK_QCOM_image_processing2[]
For code:opImageBlockMatchWindowSAD and code:opImageBlockMatchWindowSSD
instructions, the code:target sampler must: have been created with
slink:VkSamplerBlockMatchWindowCreateInfoQCOM in the code:pNext chain.

For code:opImageBlockMatchWindowSAD, code:opImageBlockMatchWindowSSD,
code:opImageBlockMatchGatherSAD, or
code:opImageBlockMatchGatherSSDinstructions, the input code:sampler must: be
created with code:addressModeU and code:addressModeV, equal to
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER with
ename:VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK.
endif::VK_QCOM_image_processing2[]

Other sampler states are ignored.


[[textures-blockmatch-filteroperation]]
=== Block Matching Operation

Block matching SPIR-V instructions code:opImageBlockMatchSAD and
code:opImageBlockMatchSSD specify two sets of 2D integer texel coordinates:
target coordinates latexmath:[(u,v)] and reference coordinates
latexmath:[(s,t)].

The coordinates define the bottom-left texel of the target block
latexmath:[(i_{0}, j_{0})] and the reference block latexmath:[(k_{0},
l_{0})].

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_{0} &= u  \\[1em]
j_{0} &= v  \\[1em]
k_{0} &= s  \\[1em]
l_{0} &= t
\end{aligned}
++++++++++++++++++++++++

For the target block, a set of neighboring integer texel coordinates are
generated.
The neighboring coordinates are combinations of latexmath:[i_{0}] to
latexmath:[i_{blockWidth-1}] and latexmath:[j_{0}] to
latexmath:[j_{blockHeight-1}].
The set is of size latexmath:[blockWidth \times blockHeight].

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\{i_q\}_{q=0}^{q=blockWidth-1} \quad &= i_{0} + q  \\[1em]
\{j_q\}_{q=0}^{q=blockHeight-1} \quad &= j_{0} + q
\end{aligned}
++++++++++++++++++++++++

where latexmath:[blockWidth] and latexmath:[blockHeight] is specified by the
code:blockSize operand.

If any target integer texel coordinate latexmath:[(i,j)] in the set fails
<<textures-integer-coordinate-validation,integer texel coordinate
validation>>, then the texel is an invalid texel and
<<textures-texel-replacement,texel replacement>> is performed.

Similarly for the reference block, a set of neighboring integer texel
coordinates are generated.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\{k_q\}_{q=0}^{q=blockWidth-1} \quad &= k_{0} + q  \\[1em]
\{l_q\}_{q=0}^{q=blockHeight-1} \quad &= l_{0} + q
\end{aligned}
++++++++++++++++++++++++

Each reference texel coordinate latexmath:[(k,l)] in the set must: not fail
<<textures-integer-coordinate-validation,integer texel coordinate
validation>>.
To avoid undefined: behavior, application shader should guarantee that the
reference block is fully within the bounds of the reference image.

Each pair of texel coordinates latexmath:[(i,j)] in the set selects a single
texel value from the target image latexmath:[\tau_{ij}].
Each pair of texel coordinates latexmath:[(k,l)] in the set selects a single
texel value from the reference image latexmath:[\upsilon_{kl}].

The difference between target and reference texel values is summed to
compute a difference metric.
The code:opTextureBlockMatchSAD computes the sum of absolute differences.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{SAD} &= \sum_{{j=j_0} \atop {l=l_0}}^{{j_{blockHeight-1}} \atop {l_{blockHeight-1}}} \quad\sum_{{i=i_0} \atop {k=k_0}}^{{i_{blockWidth-1}} \atop {k_{blockWidth-1}}}|\upsilon_{kl}-\tau_{ij}| \\
\end{aligned}
++++++++++++++++++++++++

The code:opImageBlockMatchSSD computes the sum of the squared differences.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{SSD} &= \sum_{{j=j_0} \atop {l=l_0}}^{{j_{blockHeight-1}} \atop {l_{blockHeight-1}}} \quad\sum_{{i=i_0} \atop {k=k_0}}^{{i_{blockWidth-1}} \atop {k_{blockWidth-1}}}|\upsilon_{kl}-\tau_{ij}|^2 \\
\end{aligned}
++++++++++++++++++++++++

When slink:VkSamplerReductionModeCreateInfo::pname:reductionMode is set to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, the above summation is
used.
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above
values, computing a component-wise minimum or maximum of
latexmath:[|\upsilon_{kl}-\tau_{ij}|], respectively.
For latexmath:[\tau_{SAD}], the minimum or maximum difference is computed
and for latexmath:[\tau_{SSD}], the square of the minimum or maximum is
computed.

Finally, the operations described in
<<textures-conversion-to-rgba,Conversion to RGBA>> and
<<textures-component-swizzle,Component swizzle>> are performed and the final
result is returned to the shader.
The component swizzle is specified by the _target image_ descriptor; any
swizzle specified by the _reference image_ descriptor is ignored.


ifdef::VK_QCOM_image_processing2[]
[[textures-blockmatchwindow-filteroperation]]
==== Block Matching Window Operation

Window block matching SPIR-V instructions code:opImageBlockMatchWindowSAD
and code:opImageBlockMatchWindowSSD specify two sets of 2D integer texel
coordinates: target coordinates latexmath:[(u,v)] and reference coordinates
latexmath:[(s,t)].
The <<textures-blockmatch-filteroperation,block matching operation>> is
performed repeatedly, for multiple sets of target integer coordinates within
the specified window.
These instructions effectively search a region or "`window`" within the
target texture and identify the window coordinates where the minimum or
maximum error metric is found.
These instructions only support single component image formats.

The target coordinates are combinations of coordinates from
latexmath:[(u,v)] to latexmath:[(u + windowWidth - 1, v + windowHeight - 1)]
where latexmath:[windowHeight] and latexmath:[windowWidth] are specified by
slink:VkSamplerBlockMatchWindowCreateInfoQCOM::pname:windowExtent.
At each each target coordinate, a
<<textures-blockmatch-filteroperation,block matching operation>> is
performed, resulting in a difference metric.
The reference coordinate latexmath:[(s,t)] is fixed.
The block matching operation is repeated latexmath:[windowWidth \times
windowHeight] times.

The resulting minimum or maximum error is returned in the R component of the
output.
The integer window coordinates latexmath:[(x,y)] are returned in the G and B
components of the output.
The A component is 0.
The minimum or maximum behavior is selected by
slink:VkSamplerBlockMatchWindowCreateInfoQCOM::pname:windowCompareMode.

The following pseudocode describes the operation
code:opImageBlockMatchWindowSAD.
The pseudocode for code:opImageBlockMatchWindowSSD follows an identical
pattern.

[source,c]
----
vec4 opImageBlockMatchGatherSAD( sampler2D target,
                                 uvec2 targetCoord,
                                 samler2D reference,
                                 uvec2 refCoord,
                                 uvec2 blocksize) {
    // Two parameters are sourced from the VkSampler associated with
    // `target`:
    //    compareMode  (which can be either `MIN` or `MAX`)
    //    uvec2 window (which defines the search window)

    minSAD = INF;
    maxSAD = -INF;
    uvec2 minCoord;
    uvec2 maxCoord;

    for (uint x=0, x<window.width; x++) {
        for (uint y=0; y<window.height; y++) {
            float SAD = textureBlockMatchSAD(target,
                                             targetCoord + uvec2(x, y),
                                             reference,
                                             refCoord,
                                             blocksize).x;
            if (SAD < minSAD) {
                minSAD = SAD;
                minCoord = uvec2(x,y);
            }
            if (SAD > maxSAD) {
                maxSAD = SAD;
                maxCoord = uvec2(x,y);
            }
        }
    }
    if (compareMode==MIN) {
        return vec4(minSAD, minCoord.x, minCoord.y, 0.0);
    } else {
        return vec4(maxSAD, maxCoord.x, maxCoord.y, 0.0);
    }
}
----


[[textures-blockmatchgather-filteroperation]]
==== Block Matching Gather Operation

Block matching Gather SPIR-V instructions code:opImageBlockMatchGatherSAD
and code:opImageBlockMatchGatherSSD specify two sets of 2D integer texel
coordinates: target coordinates latexmath:[(u,v)] and reference coordinates
latexmath:[(s,t)].

These instructions perform the <<textures-blockmatch-filteroperation,block
matching operation>> 4 times, using integer target coordinates
latexmath:[(u,v)], latexmath:[(u+1,v)], latexmath:[(u+2,v)], and
latexmath:[(u+3,v)].
The R component from each of those 4 operations is gathered and returned in
the R, G, B, and A components of the output respectively.
For each block match operation, the reference coordinate is
latexmath:[(s,t)].
For each block match operation, only the R component of the target and
reference images are compared.
The following pseudocode describes the operation opImageBlockMatchGatherSAD.
The pseudocode for opImageBlockMatchGatherSSD follows an identical pattern.

[source,c]
----
vec4 opImageBlockMatchGatherSAD(sampler2D target,
                                uvec2 targetCoord,
                                samler2D reference,
                                uvec2 refCoord,
                                uvec2 blocksize) {
    vec4 out;
    for (uint x=0, x<4; x++) {
            float SAD = textureBlockMatchSAD(target,
                                             targetCoord + uvec2(x, 0),
                                             reference,
                                             refCoord,
                                             blocksize).x;
            if (x == 0) {
                out.x = SAD;
            }
            if (x == 1) {
                out.y = SAD;
            }
            if (x == 2) {
                out.z = SAD;
            }
            if (x == 3) {
                out.w = SAD;
            }
    }
    return out;
}
----
endif::VK_QCOM_image_processing2[]


[[textures-boxfilter]]
== Box Filter Sampling

The SPIR-V instruction code:OpImageBoxFilterQCOM specifies texture box
filtering operation where a weighted average of a region of texels is
computed, with the weights proportional to the coverage of each of the
texels.

In addition to the inputs that would be accepted by an equivalent
code:OpImageSample* instruction, code:OpImageBoxFilterQCOM accepts one
additional input, code:boxSize which specifies the width and height in
texels of the region to be averaged.

The figure below shows an example of using code:OpImageBoxFilterQCOM to
sample from a [eq]#8 {times} 4# texel two-dimensional image, with
unnormalized texture coordinates [eq]#(4.125, 2.625)# and code:boxSize of
[eq]#(2.75, 2.25)#.
The filter will read 12 texel values and compute a weights based portion of
of each texel covered by the box.

[[textures-box-filter-diagrams]]
image::{images}/vulkantexture_boxFilter.svg[align="center",title="Box Filter Sampling Example",opts="{imageopts}"]

If code:boxSize has height and width both equal to 1.0, then this
instruction will behave as traditional bilinear filtering.
The code:boxSize parameter must: be greater than or equal to 1.0 and must:
not be greater than
slink:VkPhysicalDeviceImageProcessingPropertiesQCOM.code:maxBoxFilterBlockSize.


[[textures-boxfilter-sampler]]
=== Box Filter Sampler Parameters

The input code:sampler must: be created with code:addressModeU and
code:addressModeV, equal to ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE, or
ename:VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER with
ename:VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK.


[[textures-boxfilter-filteroperation]]
=== Box Filter Operation

The 2D unnormalized texel coordinates latexmath:[(u,v)] are transformed by
latexmath:[boxSize] to specify integer texel coordinates latexmath:[(i_{0},
j_{0})] of the bottom left texel for the filter.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
i_{0} &= \left\lfloor u - \frac{boxWidth}{2} \right\rfloor \\[1em]
j_{0} &= \left\lfloor v - \frac{boxHeight}{2} \right\rfloor
\end{aligned}
++++++++++++++++++++++++

where latexmath:[boxWidth] and latexmath:[boxHeight] are specified by the
code:(x,y) components of the code:boxSize operand.

The filter dimensions latexmath:[(filterWidth \times filterHeight)] are
computed from the fractional portion of the latexmath:[(u,v)] coordinates
and the latexmath:[boxSize].

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
startFracU &= \mathbin{frac}\left(u - \frac{boxWidth}{2} \right) \\[1em]
startFracV &= \mathbin{frac}\left(v - \frac{boxHeight}{2} \right) \\[1em]
endFracU &= \mathbin{frac}\left( startFracU + boxWidth \right) \\[1em]
endFracV &= \mathbin{frac}\left( startFracV + boxHeight \right) \\[1em]
filterWidth &= \left\lceil startFracU + boxWidth \right\rceil  \\[1em]
filterHeight &= \left\lceil startFracV + boxHeight \right\rceil
\end{aligned}
++++++++++++++++++++++++

where the number of fraction bits retained by latexmath:[frac()] is
specified by sname:VkPhysicalDeviceLimits::pname:subTexelPrecisionBits.

A set of neighboring integer texel coordinates are generated.
The neighboring coordinates are combinations of latexmath:[i_{0}] to
latexmath:[i_{filterWidth-1}] and latexmath:[j_{0}] to
latexmath:[j_{filterHeight-1}], with latexmath:[i_{0}, j_{0}] being the
top-left coordinate of this set.
The set is of size latexmath:[(filterWidth \times filterHeight)].

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\{i_q\}_{q=0}^{q=filterWidth-1} \quad &= i_{0} + q  \\[1em]
\{j_q\}_{q=0}^{q=filterHeight-1} \quad &= j_{0} + q
\end{aligned}
++++++++++++++++++++++++

Each of the generated integer coordinates latexmath:[({i_q}, {j_q})] is
transformed by <<textures-wrapping-operation, texture wrapping operation>>,
followed by <<textures-integer-coordinate-validation,integer texel
coordinate validation>>, If any coordinate fails coordinate validation, it
is a Border Texel and <<textures-texel-replacement,texel replacement>> is
performed.

Horizontal weights latexmath:[horizWeight_{0}] to
latexmath:[horizWeight_{boxWidth-1}] and vertical weights
latexmath:[vertWeight_{0}] to latexmath:[vertWeight_{boxHeight-1}] are
computed.
Texels that are fully covered by the box will have a horizontal and vertical
weight of 1.
Texels partially covered by the box will have will have a reduced weights
proportional to the coverage.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
horizWeight_{i} &=
  \begin{cases}
    \left(1-startFracU \right),         & \text{for }  (i == 0) \\
    \left(endFracU \right),             & \text{for }  (i == filterWidth-1) \text{ and } (endFracU != 0) \\
    \left(1\right),                     & \text{otherwise} \\
  \end{cases}
\end{aligned}
++++++++++++++++++++++++

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
vertWeight_{j} &=
  \begin{cases}
    \left(1-startFracV \right),         & \text{for } (j == 0)            \\
    \left(endFracV \right),             & \text{for } (j == filterHeight-1) \text{ and } (endFracV !=0)   \\
     \left(1\right),                    & \text{otherwise}  \\
  \end{cases}
\end{aligned}
++++++++++++++++++++++++


The values of multiple texels, together with their horizontal and vertical
weights, are combined to produce a box filtered value.

[latexmath]
++++++++++++++++++++++++
\begin{aligned}
\tau_{boxFilter} &= \frac{1}{boxHeight \times boxWidth} \sum_{j=j_0}^{j_{filterHeight-1}}\quad\sum_{i=i_0}^{i_{filterWidth-1}}(horizWeight_i)(vertWeight_j)\tau_{ij} \\
\end{aligned}
++++++++++++++++++++++++

When slink:VkSamplerReductionModeCreateInfo::pname:reductionMode is set to
ename:VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE, the above summation is
used.
However, if the reduction mode is ename:VK_SAMPLER_REDUCTION_MODE_MIN or
ename:VK_SAMPLER_REDUCTION_MODE_MAX, the process operates on the above
values, computing a component-wise minimum or maximum of the texels.

endif::VK_QCOM_image_processing[]


[[textures-instructions]]
== Image Operation Steps

Each step described in this chapter is performed by a subset of the image
instructions:

  * Texel Input Validation Operations, Format Conversion, Texel Replacement,
    Conversion to RGBA, and Component Swizzle: Performed by all instructions
    except code:OpImageWrite.
  * Depth Comparison: Performed by code:OpImage*Dref instructions.
  * All Texel output operations: Performed by code:OpImageWrite.
  * Projection: Performed by all code:OpImage*Proj instructions.
  * Derivative Image Operations, Cube Map Operations, Scale Factor
    Operation, LOD Operation and Image Level(s) Selection, and Texel
    Anisotropic Filtering: Performed by all code:OpImageSample* and
    code:OpImageSparseSample* instructions.
  * (s,t,r,q,a) to (u,v,w,a) Transformation, Wrapping, and (u,v,w,a) to
    (i,j,k,l,n) Transformation And Array Layer Selection: Performed by all
    code:OpImageSample, code:OpImageSparseSample, and code:OpImage*Gather
    instructions.
  * Texel Gathering: Performed by code:OpImage*Gather instructions.
ifdef::VK_NV_shader_image_footprint[]
  * Texel Footprint Evaluation: Performed by code:OpImageSampleFootprint
    instructions.
endif::VK_NV_shader_image_footprint[]
  * Texel Filtering: Performed by all code:OpImageSample* and
    code:OpImageSparseSample* instructions.
  * Sparse Residency: Performed by all code:OpImageSparse* instructions.
ifdef::VK_QCOM_image_processing[]
  * (s,t,r,q,a) to (u,v,w,a) Transformation, Wrapping, and Weight Image
    Sampling: Performed by code:OpImageWeightedSample* instructions.
  * (s,t,r,q,a) to (u,v,w,a) Transformation, Wrapping, and Block Matching:
    Performed by code:opImageBlockMatch* instructions.
  * (s,t,r,q,a) to (u,v,w,a) Transformation, Wrapping, and Box Filter
    Sampling: Performed by code:OpImageBoxFilter* instructions.
endif::VK_QCOM_image_processing[]


[[textures-queries]]
== Image Query Instructions


=== Image Property Queries

code:OpImageQuerySize, code:OpImageQuerySizeLod, code:OpImageQueryLevels,
and code:OpImageQuerySamples query properties of the image descriptor that
would be accessed by a shader image operation.
ifdef::VK_EXT_robustness2[]
They return 0 if the bound descriptor is a null descriptor.
endif::VK_EXT_robustness2[]

code:OpImageQuerySizeLod returns the size of the image level identified by
the code:Level code:of code:Detail operand.
If that level does not exist in the image,
ifdef::VK_EXT_robustness2[and the descriptor is not null,]
then the value returned is undefined:.


=== Lod Query

code:OpImageQueryLod returns the Lod parameters that would be used in an
image operation with the given image and coordinates.
ifdef::VK_EXT_robustness2[]
If the descriptor that would be accessed is a null descriptor then
[eq]#(0,0)# is returned.
endif::VK_EXT_robustness2[]
ifdef::VK_EXT_robustness2[Otherwise, the]
ifndef::VK_EXT_robustness2[The]
steps described in this chapter are performed as if for
code:OpImageSampleImplicitLod, up to <<textures-lod-and-scale-factor>>.
The return value is the vector [eq]#({lambda}', d~l~)#.
These values may: be subject to implementation-specific maxima and minima
for very large, out-of-range values.