This repository contains machine-readable files for the SPIR-V Registry. This includes:
Headers are provided in the include directory, with up-to-date headers in the
unified1 subdirectory. Older headers are provided according to their version.
In contrast, the XML registry file has a linear history, so it is not tied to SPIR-V specification versions.
When a new version or revision of the SPIR-V specification is published, the SPIR-V Working Group will push new commits onto master, updating the files under include.
The SPIR-V XML registry file is updated by Khronos whenever a new enum range is allocated.
Pull requests can be made to
mkdir build cd build cmake .. # Linux cmake --build . --target install-headers # Windows cmake --build . --config Debug --target install-headers
Then, for example, you will have
If you want to install them somewhere else, then use
-DCMAKE_INSTALL_PREFIX=/other/path on the first
A CMake-based project can use the headers without installing, as follows:
add_subdirectorydirective to include this source tree.
#includedirectives that explicitly mention the
#include "spirv/unified1/GLSL.std.450.h" #include "spirv/unified1/OpenCL.std.h" #include "spirv/unified1/spirv.hpp"
See also the example subdirectory. But since that example is inside this repostory, it doesn't use and
This will generally be done by Khronos, for a change to the JSON grammar. However, the project for the tool to do this is included in this repository, and can be used to test a PR, or even to include the results in the PR. This is not required though.
The header-generation project is under the
tools/buildHeaders directory. Use CMake to build the project, in a
build subdirectory (under
tools/buildHeaders). There is then a bash script at
bin/makeHeaders that shows how to use the built header-generator binary to generate the headers from the JSON grammar. (Execute
bin/makeHeaders from the
How are different versions published?
The multiple versions of the headers have been simplified into a single
unified1 view. The JSON grammar has a “version” field saying what version things first showed up in.
How do you handle the evolution of extended instruction sets?
Extended instruction sets evolve asynchronously from the core spec. Right now there is only a single version of both the GLSL and OpenCL headers. So we don't yet have a problematic example to resolve.
Copyright (c) 2015-2018 The Khronos Group Inc. Permission is hereby granted, free of charge, to any person obtaining a copy of this software and/or associated documentation files (the "Materials"), to deal in the Materials without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Materials, and to permit persons to whom the Materials are furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Materials. MODIFICATIONS TO THIS FILE MAY MEAN IT NO LONGER ACCURATELY REFLECTS KHRONOS STANDARDS. THE UNMODIFIED, NORMATIVE VERSIONS OF KHRONOS SPECIFICATIONS AND HEADER INFORMATION ARE LOCATED AT https://www.khronos.org/registry/ THE MATERIALS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE MATERIALS OR THE USE OR OTHER DEALINGS IN THE MATERIALS.