Also update known goods to Vulkan 1.3 support
Also re-enable SPIR-V 1.6 tests with vulkan1.3 target
Also re-cache SPIRV 1.6 header which somehow regressed back to 1.5
Add command line support which enables the following behavior:
- Remap discard
Map discard to DemoteToHelperInvocation for HLSL shaders. Map to
OpTerminateInvocation for GLSL shaders.
- Decorate HelperInvocation with Volatile
- Use localSizeId for execution mode
WorkGroupSize is deprecated in spirv1.6
Also update known goods to SPIRV 1.6
When using this qualifier for a parameter, we make it as a pointer.
However, the function TranslateStorageClass() is therefore called and
the storage class should only be set to Function when it is invoked
to translate parameter types rather than actual argument types.
We previously use createOp() in SPV builder to create type declaration.
However, all type declarations should be placed in const-type-variable
declaration section. And duplicated type defintions ought to be avoided.
We now make a method in SPV builder to perform this operation with a
more general solution: makeGenericType().
Per SPIR-V spec, a string literal's UTF-8 octets are encoded packed into
words with little-endian convention. Explicitly perform that encoding
instead of assuming that the host system is little-endian.
Note that this change requires corresponding fixes in SPIRV-Tools.
Fixes#202
1. spirv_execution_mode_id and spirv_decorate_id could support
specialization constants. The original implementation always
assume only normal frontend constants are valid. It is not true.
2. spirv_type donesn't support type_specifier as an option of
spirv_type_parameter. At present, only constant_expression is the
valid option.
Full vector and fully specified vector swizzle stores are not affected by this change, only partial swizzles ie swizzles with fewer components than the vector being stored to.
Previously the vector being stored to loaded and any components not specified in the swizzle were used to create a full store to the vector.
While this change generates more SPIR-V instructions, it is necessary for correctness.
Fixes#2518.
Fix OpImageRead result type when compiling HLSL
- Per the Vulkan spec, OpImageRead must return a 4-component vector always. When compiling HLSL, loads from a RWTexture of a template type with < 4 components would incorrectly generate an OpImageRead with a < 4 component result, resulting in validation errors.
- This was previously fixed for OpImageFetch in commit 4425f24; this commit does the same thing for OpImageRead.
- Added associated tests and expanded existing image fetch tests to check all the different types of textures, in both float and int incarnations, for completeness.
- Update other HLSL tests involving OpImageRead
SPIR-V disallows bool in interface blocks, which is emulated with uint.
When loading a bool variable (through accessChainLoad()), it's converted
from uint to bool if it came from an interface block.
This was handled for bool and bvecN, but not for bool arrays. This
change implements the conversion for bool arrays.
Closes#2694
- Add support of SPIR-V execution mode qualifiers.
- Add support of SPIR-V storage class qualifier.
- Add support of SPIR-V decorate qualifiers.
- Add support of SPIR-V type specifier.
- Add support of SPIR-V intruction qualifiers.
- Add support of spirv_by_reference/spirv_literal parameter qualifier.
- Add shader stage macros introduced by this extension.
In a matrix constructor that takes a number of components, as many
components as necessary must be taken, with the rest discarded, as GLSL
allows more components than necessary to be specified. For example, the
following:
mat4 m4 = mat4(v4, v4.yzwx, v4.zwx, v4.zwxy, v4.wxyz);
is equivalent to:
mat4 m4 = mat4(v4, v4.yzwx, v4.zwx, v4.zwxy, v4.w);
glslang takes the components from the constructor and builds the single
components of the matrix in a 2D array before constructing the matrix
itself. It however did not check for extra parameters and was thus
writing OOB to said 2D array. This is fixed in this change
Signed-off-by: Shahbaz Youssefi <shabbyx@gmail.com>
Previous check was missing type difference between uniform array
actual arg with stride decoration and the formal arg without. Now
does logical or component-wise copy where needed.
Fixes#2637
Generate load of interpolant for first operand to GLSLstd450
InterpolateAt* SPIR-V ops. This allows the interpolants to
propagate from the input struct in the wrapper around main
into the shader during HLSL legalization. A new pass has been
added to legalization which will remove the load and replace
with the pointer of the load to create valid external
interpolate op.
Fixes#2584
Using const char* is not a good choice. We assume strings are always
from spv::E_SPV_XXX definitions. However, it is not the case. We can
store the name strings of extended intruction set in local variables.
If GL_EXT_scalar_block_layout is requested by the shader, set the
option to allow scalar blocks in the SPIR-V validator.
Fix the existing tests using scalar layout to not expect "Validation
failed".
Fixes#2400.
Issue #2483
According to GLSL spec the prototype is:
uint rayQueryGetIntersectionInstanceShaderBindingTableRecordOffsetEXT(rayQueryEXT q, bool committed);
but that was incorrectly getting translated to SPIRV as an `int`, and this was
causing SPIR-V validation errors when used.
Added explicit testing for the return types of all the builtin functions in GL_EXT_ray_query
* update spirv-headers and fix handling of gl_HitTEXT
Update spirv-headers known_good to f027d53ded7e230e008d37c8b47ede7cd308e19d
and update SPIRV/spirv.hpp to copy from that version as well.
In GLSL gl_HitTNV/gl_HitTEXT is defined as an alias of gl_RayTmaxNV/gl_RayTmaxEXT
SPV_NV_ray_tracing has a dedicated HitTNV which gl_HitTNV maps to.
For SPV_KHR_ray_tracing, gl_HitTEXT gets mapped to a RayTmaxKHR decoraged variable
to simplify the SPIRV consumer.
This change fixes the mapping for the GL_EXT_ray_tracing extension, and updates
the test results to match.
* update MissNV shader test to not use ObjectRay builtins
They shouldn't existing in the miss stage because there is no object intersected
* Add ray query capability if acceleration structure used
Fixes#2430
in non-ray tracing stages and the extension is enabled
* Add ray query capability if ray query declared
* Fix printing of TypeRayQueryKHR
It's no longer spelled with "Provisional"
* Fix traceRay/executeCallable to have id instead of constant.
Update to final (non-provisional) SPIR-V capabilities
(includes review feedback)
- Change visibilty of findLinkerObjects.
See merge request GLSL/glslang!78
* Add support for OpConvertUToAccelerationStructureKHR.
GLSL : https://gitlab.khronos.org/GLSL/GLSL/-/merge_requests/60
SPV : https://gitlab.khronos.org/spirv/spirv-extensions/-/merge_requests/182
See merge request GLSL/glslang!77
* Add volatile qualifier to certain builtins for ray tracing.
See merge request GLSL/glslang!81
* make gl_RayTmaxEXT volatile in intersection shader
Vulkan Issue #2268
* Add testing for layouts on SBT
vulkan/vulkan#2230
- no layout specified should be same as std430
- explicitly test std140, std430, scalar layouts
See merge request GLSL/glslang!86
* Support for new opcodes OpIgnoreIntersectionKHR and OpTerminateRayKHR
vulkan/vulkan#2374
Add support for ignoreIntersectionEXT and terminateRayEXT as block
terminator statements.
See merge request GLSL/glslang!87
* Fix code-generation issues with global ray query variables
See merge request GLSL/glslang!88
* update dependencies for spirv-headers and tools
And update mesh shader results
* Fix indeterminate argument ordering
Authored-by: David Neto <dneto@google.com>
Co-authored-by: Ashwin Lele (NVIDIA Corporation) <alele@nvidia.com>
Co-authored-by: Neslisah <Neslisah.Torosdagli@amd.com>
Use of gl_Layer and gl_ViewportIndex in tessellation and vertex
shaders should not trigger the addition of the Geometry capability.
Fixes#2461
Added tests for use of gl_Layer and gl_ViewportIndex in a tessellation
evaluation shader.
Several tests for NVIDIA features for tessellation, vertex, or mesh
shaders now lose the Geometry or MultiViewport capabilities.
This is ok because the functionality is already covered by
the ShaderViewportIndexLayerNV capability.
The spv.meshShaderPerViewBuiltins.mesh test now fails validation
because the validator does not know that PrimitiveId (and possibly
other) builtins are enabled by the MeshShadingNV capability.
I filed https://github.com/KhronosGroup/SPIRV-Headers/issues/179 to
fix the grammar upstream.
This implements a new nonunifom analysis suggested by @jbolz. This change
generates nonUniform decorations that were previously missing and avoids
generation of incorrect decorations. Most notably, it now generates
decorations for nonuniform functions and out params. It avoids generating
decorations for lvalues which themselves are not nonuniform.
* Implement GL_EXT_terminate_invocation.
* terminateInvocation: declare the SPV extension
* Update test results for spirv-tools and bison version bumps
Co-authored-by: John Kessenich <cepheus@frii.com>
* Code refine and adding missing features
1. Add new level for built in symbols.
2. Fix issues for structure members' qualifiers.
3. Global qualifier fix.
4. IO Mapper refine. Add support for checking with mangle names.
* Additional missing features
* Invariant member. (Only check non-interface).
* Split block nesting level and struct nesting level. To fix issues of checking 'invariant' qualifier.
Current grammar would check block/struct member without its parent class's information.
So we split nesting level, and 'invariant' would only be checked within a struct.
* Format anonymous block names. Refine codes for symbols from all kinds of resouces.
* Fix writeonly check.
* Use LValueBase to find operator.
* Fix random null ptr issue.
* invariant check, stage in io mapping, reference parameter should be used and remove wrong codes introduced with ordering vector.
* Remained: to be fixed with double check link.vk.multiblocksValid
* Fix version error.
invariant
* Revert loc modification.
* Add new SpirvToolsDisassemble API interface + Improve Doc on existing API interface (#2408)
* Add more flexible SpirvToolsDisassemble interface to allow specifying spv_target_env for disassembly output.
Improve documentation on existing SpirvToolsDisassemble interface.
* Update pre-processor check - following existing ENABLE_OPT checks.
* Fix not-found header paths for glslangValidator and glslangtests.
* Add spirv_tools/include path where there is an ENABLE_OPT=1 in the BUILD.gn configuration.
* When targeting SPIR-V 1.5, using gl_ViewportIndex will emit OpCapability ShaderViewportIndex and using gl_Layer will emit OpCapability CapabilityShaderLayer. OpCapability ShaderViewportIndexLayerEXT will only get emitted if the target < SPIR-V 1.5
* When using one of the viewport/layer arrays extensions, fallback to OpCapability ShaderViewportIndexLayerEXT, even when targeting SPIR-V 1.5
* Revert "When using one of the viewport/layer arrays extensions, fallback to OpCapability ShaderViewportIndexLayerEXT, even when targeting SPIR-V 1.5"
This reverts commit dccca82f40.
* Using gl_Layer and gl_ViewportIndex outside of the geometry shader stage still requires one of the viewport extensions even when targeting SPIR-V 1.5
(Fixes a problem introduced by 670536b663)
* Do not use PropagateLineInfoPass and RedundantLineInfoElimPass
Since spirv-opt will remove PropagateLineInfoPass and
RedundantLineInfoElimPass, glslang should not use it. spirv-opt will
propagate the line instructions and eliminate the redundant lines by
default in IR loading/emission.
* Update known_good.json for spirv-tool
Translate printf() to what GL_EXT_debug_printf has done. HLSL could
define non-constant string variable and we don't have such features
in SPIR-V, so just support constant string variable.
disassemble.cpp appears not to be using anything from SpvTools.h, but the inclusion of it prevents standalone building of the SPIR-V portion (for instance, when needed purely for generation and disassembly) without SPIRV-Tools dependency.
There's a statement that intends to generate a 32-bit hashcode, but due
to integer promotion, the intermediate values can trigger signed integer
overflow, which is undefined behavior.
To avoid this, cast at least one operand to unsigned int before
multiplying, which will cause the result to be promoted to unsigned int
instead of signed int.
With this patch, I'm able to build core for qemu-x64 with host_asan-ubsan.
Fixed: 60128
Change-Id: Idd644e534116bf29dca8013936ac39901bbe68fc
Reviewed-on: https://fuchsia-review.googlesource.com/c/third_party/glslang/+/428254
Reviewed-by: John Bauman <jbauman@google.com>
Co-authored-by: Drew Fisher <zarvox@google.com>
Add more builder functions to OpExecutionMode, OpExecutionModeId,
OpDecorateString, OpMemberDecorateString.
According to SPIR-V, OpExecutionMode and OpExecutionModeId could
take variable extra operands. Current implementation doesn't support
this and assumes at most 3 operands are extra operands. It is not
true. Similarly, OpDecorateString and OpMemberDecorateString could
support multiple strings either as literal strings or as string
operands. Further, OpDecorate and OpDecorateId have the same problem,
taking variable extra operands.
According to the extension SPV_GOOGLE_decorate_string,
OpDecorateString (or OpMemberDecorateString) ought to be capable of
supporting multiple literal strings. Each literal strings are padded
with null terminator to make word alignment. The layout is:
Inst | Target | Decoration | Literal String, Literal String, ...
By directly creating the OpImage instructions we were not propagating
the appropriate decorations.
Since this had a lot of cases I centralized the OpImage creation
a bit too.
The direct pointer argument to stores has to use
the NonUniform decoration but we were not using
qualifiers at all to decorate the NonUniform pointer.
(Test fixes by Greg Fischer <greg@lunarg.com>)
We need separate concepts for
- total set of extensions ever enabled, for the back end
- current state of extensions, for parsing
- the set of features currently enabled for building the AST
This change strips a few features similar to GLSLANG_WEB but doesn't
remove every detail like the latter. It also hardcodes profile/version
to core/450.
In particular, TBuiltIns::initialize is specialized to remove most of
what is not supported or won't be supported by ANGLE. The result of
this function is parsed with TParseContext::parseShaderStrings which is
a performance bottleneck.
This change shaves about 300KB off of ANGLE's binary size and reduces
the cost of SetupBuiltinSymbolTable to nearly a sixth.
Signed-off-by: Shahbaz Youssefi <ShabbyX@gmail.com>
iOS 8 does not support `thread_local`, which is still in use.
Another approach will have to be found.
This change is a revert of the following changes:
a3845240 - "Simplify PoolAlloc with use of thread_local."
abf92c80 - "Deprecate InitializeDll functions"
33585c87 - "Limit visibility of symbols for internal libraries"
Issue: #2346
Also remove `SPIRV/doc.cpp` from the `SPVRemapper` target as this
is part of `SPIRV`, causing ODR violations. Instead have
`SPVRemapper` link against `SPIRV`.
Fixes ODR violations.
This PR significantly reworks the way glslang is versioned.
Instead of committing changes to the `GLSLANG_MINOR_VERSION` define in
`glslang/Public/ShaderLang.h`, and using `make-revision` to generate
`GLSLANG_PATCH_LEVEL` in `glslang/Include/revision.h`, all version
information is now derived from the new `CHANGES.md` file.
`CHANGES.md` acts as the single source of truth for glslang version
information, along with a convenient place to put all release notes for
each notable change made.
`CHANGES.md` is parsed using the new `build_info.py` python script.
This script can read basic template files to produce new source files,
which it does to read the new `build_info.h.tmpl` to generate (at build
time) a glslang private header at
`<build-dir>/include/glslang/build_info.h`.
I've written generators for each of the CMake, Bazel, gn, and
`Android.mk` build scripts.
The new version code conforms to the Semantic Versioning 2.0 spec.
This new version is also used by the CMake rules to produce versioned
shared objects, including a major-versioned SONAME.
New APIs:
---------
* `glslang::GetVersion()` returns a `Version` struct with the version
major, minor, patch and flavor.
Breaking API changes:
---------------------
* The public defines `GLSLANG_MINOR_VERSION` and `GLSLANG_PATCH_LEVEL`
have been entirely removed.
* `glslang/Public/ShaderLang.h` and `glslang/Include/revision.h` have
been deleted.
* Instead, `<build-dir>/include/glslang/build_info.h` is created in
the build directory, and `<build-dir>/include` is a CMake `PUBLIC`
(dependee-inherited) include directory for the glslang targets.
* `<build-dir>/include/glslang/build_info.h` contains the following
new #defines:
`GLSLANG_VERSION_MAJOR`, `GLSLANG_VERSION_MINOR`,
`GLSLANG_VERSION_PATCH`, `GLSLANG_VERSION_FLAVOR`,
`GLSLANG_VERSION_GREATER_THAN(major, minor, patch)`,
`GLSLANG_VERSION_GREATER_OR_EQUAL_TO(major, minor, patch)`,
`GLSLANG_VERSION_LESS_THAN(major, minor, patch)`,
`GLSLANG_VERSION_LESS_OR_EQUAL_TO(major, minor, patch)`
* The CMake install output directory contains a copy of
`build_info.h` at: `include/glslang/build_info.h`
* Python3 is now always required to build glslang (likely always
required for transitive dependency builds).
This change strips a few features similar to GLSLANG_WEB but doesn't
remove every detail like the latter. It also hardcodes profile/version
to core/450.
In particular, TBuiltIns::initialize is specialized to remove most of
what is not supported or won't be supported by ANGLE. The result of
this function is parsed with TParseContext::parseShaderStrings which is
a performance bottleneck.
This change shaves about 300KB off of ANGLE's binary size and reduces
the cost of SetupBuiltinSymbolTable to nearly a sixth.
Signed-off-by: Shahbaz Youssefi <ShabbyX@gmail.com>
Fix#2298.
The AST has two precisions, an operation precision and a result precision.
Actual use of GLSL with mediump samplers wants the result precision, so
pick that up instead of the operation precision.
This simplifies and enforces use of precision in many more places,
to help avoid accidental loss of RelaxedPrecision through intermediate
operations. Known fixes are:
- ?:
- function return values with mis-matched precision
- precision of function return values when a copy was needed to fix types
When arguments are copied to make space for a writable formal parameter,
and the formal parameter is relaxed precision, make the copy also
relaxed precision.
The normal IEEE not equal operation tests whether operands are unordered
or not equal (so comparison with a NaN returns true). This corresponds
to the SPIR-V OpFUnordNotEqual, so change to using that.
Analogous to gcc, -g0 would strip all debug info. This is done
regardless of whether optimizations are enabled.
Signed-off-by: Shahbaz Youssefi <ShabbyX@gmail.com>
If a uniform has an initializer it will now be given as the optional
initializer operand to the OpVariable instruction.
Fixes: https://github.com/KhronosGroup/glslang/issues/1259
Signed-off-by: Neil Roberts <nroberts@igalia.com> (the code)
Signed-off-by: Alejandro Piñeiro <apinheiro@igalia.com> (the tests)
Signed-off-by: Arcady Goldmints-Orlov <agoldmints@igalia.com>
Co-authored-by: Neil Roberts <nroberts@igalia.com>
* HLSL: Fix handling of uniforms in entry point parameters
* HLSL: fix handling of "uniform in"
* Tests: Update baseResults of hlsl.function.frag.out for #2254
* HLSL: fix uniforms in function parameters for opaque types
This CL updates the build scripts to only install to glslang/SPIRV
instead of also installing to the SPIRV/ folder. The deprecation notice
is also removed.
Note, this may cause downstream build issues if include directories have
not been updated
Fixes#1964#2216
Swizzled out operands were added in bbbd9a2a. This was sufficient
for most tests, but we ran into problems with umulExtended and
imulExtended, which have two.
This CL converts the tracking values to vectors so multiple operands
can be supported.
Test: KHR-GLES31.core.shader_bitfield_operation.*
Test: ctest
For GLSL function imageAtomicStore, it will be translated to
OpImageTexelPointer + OpAtomicStore. The result type of
OpImageTexelPointer must be the same as the sampled type of OpTypeImage.
On translation, the result type is mistakenly fetched from operand list
operands[2] while operands[2] corresponds to sampleNum whose type is
always uint. This leads to an error if the image type is iimageXXX that is
int image.
Root cause:
GlslangToSpv use symbol structure's ptr as a map key, but multiple shader object can build a new AST.
In the AST the the same symbol has different ptr point to their structure, so indext map faild.
solution:
Add a new map glslangTypeToIdMap to map ptr to symbol id, and use symbol id to index memberRemapper.
v2: Move addCapability and addExtension extension calls from
TGlslangToSpvTraverser::createBinaryOperation to
TGlslangToSpvTraverser::visitAggregate. Suggested by JohnK.
There is a flag to disable the SPVRemapper during the GLSLang build.
That flag is check in some, but not all spots so if you try to build
with SPVRemapper disabled you get CMake errors and compile errors.
This CL fixs up the build so building with -DENABLE_SPVREMAPPER=0 will
complete correclty.
The SPIR-V post-processing to discover capabilities and
extensions does not apply to WebGPU compilation. So don't include
that code.
This reclaims some of the code space added by #1943
More aggressively prune unreachable code as follows.
When no control flow edges reach a merge block or continue target:
- delete their contents so that:
- a merge block becomes OpLabel, then OpUnreachable
- a continue target becomes OpLabel, then an OpBranch back to the
loop header
- any basic block which is dominated by such a merge block or continue
target is removed as well.
- decorations targeting the removed instructions are removed.
Enables the SPIR-V builder post-processing step the GLSLANG_WEB case.
If the semantics that require the KHR form over the EXT form are seen
(OpBitcast between a vector and a pointer) promote the requested extension
from the EXT to the KHR.
Currently the SPIRV/ folder will get installed into the include
directory. This folder is part of GLSLang, so it makes more sense under
glslang/SPIRV.
Currently, GLSLang will install a SPIRV/ folder while spirv-headers will
install a spirv/ folder. This is confusing and will cause issues on a
case sensitive filesystem if both are installed at the same time.
