Vulkan-Docs/xml/registry.rnc

501 lines
18 KiB
Plaintext

# Copyright (c) 2013-2018 The Khronos Group Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Relax NG schema for Khronos Vulkan API Registry XML
#
# See https://www.khronos.org/vulkan/
#
# This definition is subject to change (mostly in the form of additions)
namespace xsd = "http://www.w3.org/2001/XMLSchema-datatypes"
# Toplevel is a <registry> tag.
# May be led by an optional <comment> tag containing e.g. copyrights.
start = element registry {
(
element comment { text } ? |
Platforms * |
Tags * |
Types * |
Enums * |
Commands * |
Feature * |
Extensions *
) *
}
# <platforms> defines a group of platform names
Platforms = element platforms {
Comment ? ,
Platform *
}
# <platform> defines a single platform name.
# name - string name of the platform, used as part of extension names
# protect - preprocessor symbol to include platform headers from <vulkan.h>
# comment - platform description
Platform = element platform {
attribute name { text } ,
attribute protect { text } ,
Comment
}
# <tags> defines a group of author tags
Tags = element tags {
Comment ? ,
Tag *
}
# <tag> defines a single author tag.
# name - name of the tag
# author - name of the author (usually a company or project name)
# contact - contact responsible for the tag (name and contact information)
Tag = element tag {
attribute name { text } ,
attribute author { text } ,
attribute contact { text }
}
# <types> defines a group of types
Types = element types {
Comment ? ,
(
Type |
element comment { text }
) *
}
# <type> defines a single type. It is usually a C typedef but
# may contain arbitrary C code.
# name - name of this type, if not present in the <name> tag
# api - matches a <feature> api attribute, if present
# requires - name of another type definition required by this one
# category - if present, 'enum' indicates a matching <enums>
# block to generate an enumerated type for, and 'struct'
# causes special interpretation of the contents of the type
# tag including ... TBD ...
# Other allowed values are 'include', 'define', 'handle' and 'bitmask',
# which don't change syntactic interpretation but allow organization in
# the generated header.
# comment - unused
# parent - only applicable if category is 'handle'. Notes another type with
# the 'handle' category that acts as a parent object for this type.
# returnedonly - only applicable if category is 'struct'. Notes that this
# struct is going to be filled in by the API, rather than an application
# filling it out and passing it to the API.
# structextends - only applicable if category is 'struct'. Lists parent
# structures which this structure may extend via the pNext chain
# of the parent.
# When present it suppresses generation of automatic validity for the
# pNext member of that structure, and instead the structure is added
# to pNext chain validity for the parent structures it extends.
# For types without a category, contents include
# <apientry /> - substitutes for an APIENTRY-style macro on output
# <name> - contains name of the type being defined
# <type> - contains name of types used to define this type. There
# may be multiple imbedded <type> tags
# For types with category 'enum', contents should be empty
# For types with category 'struct', contents should be one or more
# <member> - like <param> for a struct or union member
# len - if the member is an array, len may be one or more of the following
# things, separated by commas (one for each array indirection): another
# member of that struct, 'null-terminated' for a string, '1' to indicate it's
# just a pointer (used for nested pointers), or a latex equation (prefixed with
# 'latexmath:')
# altlen - if len has latexmath equations, this contains equivalent C99
# expressions separated by commas.
# externsync - denotes that the member should be externally synchronized
# when accessed by Vulkan
# optional - whether this value can be omitted by providing NULL (for
# pointers), VK_NULL_HANDLE (for handles) or 0 (for bitmasks/values)
# noautovalidity - tag stating that no automatic validity language should be generated
# values - comma-separated list of legal values, usually used only for sType enums
# <comment> - containing arbitrary text (unused)
#
# *** There's a problem here: I'm not sure how to represent the <type>
# syntax where it may contain arbitrarily interleaved text, <type>, and
# <enum> child tags. This allows only the syntax
# text <type>name</type> text <enum>name</enum> text
# where <type> and <enum> are both optional and occur in the specified
# order, which might eventually be a problem.
Type = element type {
attribute api { text } ? ,
attribute alias { text } ? ,
attribute requires { text } ? ,
attribute name { TypeName } ? ,
attribute category { text } ? ,
attribute parent { TypeName } ? ,
attribute returnedonly { text } ? ,
attribute structextends { text } ? ,
Comment ? ,
(
(
( text ,
element type { text } *
) * ,
element apientry { text } ? ,
( text ,
element type { text } *
) * ,
element name { TypeName } ? ,
( text ,
element type { text } *
) *
) |
(
element member {
attribute len { text } ? ,
attribute altlen { text } ? ,
attribute externsync { text } ? ,
attribute optional { text } ? ,
attribute noautovalidity { text } ? ,
attribute values { text } ? ,
mixed {
element type { TypeName } ? ,
element name { text } ,
element enum { EnumName } ? ,
element comment { text } ?
}
} |
element comment { text }
) *
)
}
# <enums> defines a group of enumerants
# name - identifies a type name associated with this group. Should
# match a <type> name to trigger generation of the type.
# start, end - beginning and end of a numeric range
# vendor - owner of the numeric range
# type - 'enum' or 'bitmask', if present
# comment - unused
Enums = element enums {
attribute name { text } ? ,
attribute type { text } ? ,
attribute start { Integer } ? ,
attribute end { Integer } ? ,
Vendor ? ,
Comment ? ,
(
Enum |
Unused |
element comment { text}
) *
}
# <enum> defines or references a single enumerant. There are two places it
# can be used: in an <enums> block, providing a global definition which
# may later be required by a feature or extension; or in a feature or
# extension, defining an enumerant specific to that feature. The second
# form has more possible attributes. Some combinations of attributes are
# nonsensical in on or the other place, but these are not detected by the
# validator.
#
# Ways to specify the enumerant value:
# value - integer (including hex) value of the enumerant
# bitpos - integer bit position of the enumerant in a bitmask
# [extnumber], offset, [dir] - integer extension number specifying a
# base block value (inherited from surrounding <extension> if
# not specified); integer offset in that block; and direction
# of offset ('-' for negative, positive if not specified).
# alias - name of another enum this is an alias of
#
# value and bitpos allow, and extnumber/offset/dir require:
# extends - type name of the enumerant being extended
#
# Other attributes:
# api - matches a <feature> api attribute, if present
# type - 'u' (unsigned), 'ull' (uint64), or integer if not present
# name - enumerant name
# alias - another enumerant this is semantically identical to
# comment - unused
Enum = element enum {
(
(
(
attribute value { Integer } &
attribute extends { TypeName } ?
) |
(
attribute bitpos { Integer } &
attribute extends { TypeName } ?
) |
(
attribute extnumber { Integer } ? &
attribute offset { Integer } &
attribute dir { text } ? &
attribute extends { TypeName }
) |
(
attribute extends { TypeName } ? &
attribute alias { TypeName }
)
) ? &
attribute api { text } ? &
attribute type { TypeSuffix } ? &
attribute name { text } &
Comment ?
)
}
# <unused> defines a range of enumerants not currently being used
# start, end - beginning and end of an unused numeric range
# vendor - unused
# comment - unused
Unused = element unused {
attribute start { Integer } ,
attribute end { Integer } ? ,
Vendor ? ,
Comment ?
}
# <commands> defines a group of commands
Commands = element commands {
Comment ? ,
Command *
}
# <command> defines a single command
#
# There are two forms of the tag.
#
# The first only has 'name' and 'alias' attributes, and no contents.
# It defines a command alias.
#
# The second fully defines a command, and has the following structure:
# The possible attributes are not described in this comment block yet, but
# are in readme.pdf.
#
# <proto> is the C function prototype, including the return type
# <param> are function parameters, in order
# len - if the member is an array, len may be one or more of the following
# things, separated by commas (one for each array indirection): another
# member of that struct, 'null-terminated' for a string, '1' to indicate it's
# just a pointer (used for nested pointers), or a latex equation (prefixed with
# 'latexmath:')
# altlen - if len has latexmath equations, this contains equivalent C99
# expressions separated by commas.
# externsync - denotes that the member should be externally synchronized
# when accessed by Vulkan
# optional - whether this value can be omitted by providing NULL (for
# pointers), VK_NULL_HANDLE (for handles) or 0 (for bitmasks/values)
# noautovalidity - tag stating that no automatic validity language should be generated
# <type> is a <type> name, if present
# <name> is the function / parameter name
# The textual contents of <proto> and <param> should be legal C
# for those parts of a function declaration.
# <alias> - denotes function aliasing, if present
# name - name of aliased function
# <description> - unused text
# <implicitexternsyncparams> are spec-language descriptions of
# objects that are not parameters of the command, but
# are related to them and also require external synchronization.
Command = element command {
( attribute name { text } ,
attribute alias { text } ) |
(
attribute queues { text } ? ,
attribute successcodes { text } ? ,
attribute errorcodes { text } ? ,
attribute renderpass { text } ? ,
attribute cmdbufferlevel { text } ? ,
attribute pipeline { text } ? ,
Comment ? ,
element proto {
mixed {
element type { TypeName } ? ,
element name { text }
}
} ,
element param {
attribute len { text } ? ,
attribute altlen { text } ? ,
attribute externsync { text } ? ,
attribute optional { text } ? ,
attribute noautovalidity { text } ? ,
mixed {
element type { TypeName } ? ,
element name { text }
}
} * ,
(
element alias {
Name
} ? &
element description {
text
} ? &
element implicitexternsyncparams {
element param { text } *
} ?
)
)
}
# Each <feature> defines the interface of an API version (e.g. OpenGL 1.2)
# api - API tag (e.g. 'gl', 'gles2', etc. - used internally, not
# necessarily an actual API name
# name - version name (C preprocessor name, e.g. GL_VERSION_4_2)
# number - version number, e.g. 4.2
# protect - additional #ifdef symbol to place around the feature
# <require> / <remove> contains features to require or remove in
# this version
# profile - only require/remove when generated profile matches
# comment - unused
Feature = element feature {
attribute api { text } ,
Name ,
attribute number { xsd:float } ,
attribute protect { text } ? ,
Comment ? ,
(
element require {
ProfileName ? ,
ExtensionName ? ,
Comment ? ,
(
InterfaceElement |
element comment { text }
) *
} |
element remove {
ProfileName ? ,
Comment ? ,
(
InterfaceElement |
element comment { text }
) *
}
) *
}
Extensions = element extensions {
Comment ? ,
Extension *
}
# Defines the interface of an API <extension>. Like a <feature>
# tag, but with slightly different attributes:
# api - regexp pattern matching one or more API tags, indicating
# which APIs the extension is known to work with. The only
# syntax supported is <name>{|<name>}* and each name must
# exactly match an API being generated (implicit ^$ surrounding).
# name - extension name string
# number - extension number (positive integer, should be unique)
# protect - C preprocessor symbol to conditionally define the interface
# platform - should be one of the platform names defined in the
# <platform> tag. Currently unused.
# author - name of the author (usually a company or project name)
# contact - contact responsible for the tag (name and contact information)
# type - 'device' or 'instance', if present
# requires - commas-separated list of extension names required by this
# extension
# requiresCore - core version of Vulkan required by the extension, e.g.
# "1.1". Defaults to "1.0".
# supported - profile name(s) supporting this extension, e.g. 'vulkan'
# or 'disabled' to never generate output.
# promotedto - Vulkan version or a name of an extension that this
# extension was promoted to; e.g. 'VK_VERSION_1_1', or
# 'VK_KHR_draw_indirect_county'
# deprecatedby - Vulkan version or a name of an extension that deprecates
# this extension. It may be empty string.
# e.g. 'VK_VERSION_1_1', or 'VK_EXT_debug_utils', or ''
# obsoletedby - Vulkan version or a name of an extension that obsoletes
# this extension. It may be empty string.
# e.g. 'VK_VERSION_1_1', or 'VK_EXT_debug_utils', or ''
# In addition, <require> / <remove> tags also support an api attribute:
# api - only require/remove these features for the matching API.
# Not a regular expression.
Extension = element extension {
Name ,
attribute number { Integer } ? ,
attribute protect { text } ? ,
attribute platform { text } ? ,
attribute author { text } ? ,
attribute contact { text } ? ,
attribute type { text } ? ,
attribute requires { text } ? ,
attribute requiresCore { text } ? ,
attribute supported { StringGroup } ? ,
attribute promotedto { text } ? ,
attribute deprecatedby { text } ? ,
attribute obsoletedby { text } ? ,
Comment ? ,
(
element require {
attribute api { text } ? ,
ProfileName ? ,
ExtensionName ? ,
FeatureName ? ,
Comment ? ,
(
InterfaceElement |
element comment { text }
) *
} |
element remove {
attribute api { text } ? ,
ProfileName ? ,
Comment ? ,
(
InterfaceElement |
element comment { text }
) *
}
) *
}
# Contents of a <require> / <remove> tag, defining a group
# of features to require or remove.
# <type> / <enum> / <command> all have attributes
# name - feature name which must match
InterfaceElement =
element type {
Name ,
Comment ?
} |
Enum |
element command {
Name ,
Comment ?
}
# Integers are allowed to be either decimal or C-hex (0x[0-9A-F]+), but
# XML Schema types don't seem to support hex notation, so we use this
# as a placeholder.
Integer = text
# EnumName is an compile-time constant name
EnumName = text
# TypeName is an argument/return value C type name
TypeName = text
# TypeSuffix is a C numeric type suffix, e.g. 'u' or 'ull'
TypeSuffix = text
# StringGroup is a regular expression with an implicit
# '^(' and ')$' bracketing it.
StringGroup = text
# Repeatedly used attributes
ProfileName = attribute profile { text }
ExtensionName = attribute extension { text }
FeatureName = attribute feature { text }
Vendor = attribute vendor { text }
Comment = attribute comment { text }
Name = attribute name { text }