Move the parameter verifictation to a centralized place where all the builtins
are verified for correctness.
Add verification for the new builtins with version and extension check
These builtins are supported on GLSL since version 130 and GLES since
version 300.
Adds the --no-link option which outputs the compiled shader binaries
without linking them. This is a first step towards allowing users to
create SPIR-v binary, non-executable libraries.
When using the --no-link option, all functions are decorated with the
Export linkage attribute.
float textureOffset(sampler2DArrayShadow sampler, vec4 P, ivec2 offset)
was incorrectly requiring the GL_EXT_texture_shadow_lod extension.
NOTE: Prior to GLSL 440, this prototype was defined as
float textureOffset(sampler2DArrayShadow sampler, vec4 P, vec2 offset)
i.e., the type of 'offset' was specified as 'vec2' rather than 'ivec2'.
This is believed to be a typo.
Fixes#3325.
Previously, when GL_EXT_spirv_intrinsics are enabled, we disable all
checks in layoutTypeCheck. This is too coarse because we can use nothing
in GL_EXT_spirv_intrinsics in a shader while the necessary processing is
skipped, such as addUsedLocation.
In this change, we apply fine check and more might be added if we
encounter new cases in the future.
Before this change, using gl_MeshPrimitivesEXT in mesh shader would
unconditionally create gl_MeshPrimitivesEXT.gl_PrimitiveShadingRateEXT
field and add PrimitiveShadingRateKHR capability to the output SPIRV
file, which would subsequently trigger validation errors when creating
the shader module unless the application requested primitive shading
rate feature.
What should happen instead is that unless GL_EXT_fragment_shading_rate
extension is enabled, we should not allow using
gl_PrimitiveShadingRateEXT and should not emit the associated fields
into the output.
This change fixes this by using existing filterMember mechanism that is
already used in a few other cases like this, and adjusting the required
extension on the field member which will generate an error when
gl_PrimitiveShadingRateEXT is used without enabling the extension.
The usage of GetGlobalLock/ReleaseGlobalLock/InitGlobalLock is replaced
by std::lock_guard which is available as of c++11, and the functions are
removed from the OSDependent ossource.cpp files.
The standalone glslang binary now explicitly depends on OSDependent, as
nothing in in the glslang library uses those functions anymore and they
are not implicitly picked up by the linker.
glslang is using C++ 11, which has first class support for variables of the `thread_local` storage class.
By dropping the use of the `OS_[GS]etTLSValue`, we can simplify the logic, and have it support a thread-local default allocator if none is provided.
Issue: #2346
The 'set' and 'setRT' variables were warning as maybe-uninitialized even
though in practice that case would never trigger (due to how the
function flow-controls).
The code blocks where these variables are actually read do not overlap,
so merge them into the same 'set' variable.
Simplify the control flow of the function with early-returns, which
drops indentation and simplifies the function.
Ensure we traverse the entire tree and upgrade all references to the
given symbol so it can be upgraded to push_constant. Without this change
only one instance was upgraded, and others were left as uniform buffers.
Previously, spirv_type doesn't accept type specifier as its parameter.
With this change, we can input non-array type specifier. This is because
some SPIR-V type definition intructions often need to reference other
SPIR-V types as its source operands. We add the support to facilitate
such usage.
Vertex shaders can have pipeline inputs that are arrays, but not structure
inputs. (GLSL 4.5 section 4.3.4)
But the error message for struct inputs says "cannot be a structure or array".
This PR removes the "or array" part.
Note: The array case is handled immediately after the check for
structure type.
Co-authored-by: Arcady Goldmints-Orlov <arcady@lunarg.com>
spirv_decorate_id will generate OpDecorateId. The parameter list should
accept variables as part of decorations. This is because OpDecorateId
allows this. The spec says:
All such <id> Extra Operands must be constant instructions or
OpVariable instructions.
This adds a new fullyExpanded flag that makes sure that macro arguments
only get expanded once. This can happen either in PrescanMacroArg, or, if
there is token pasting or a function-like macro name has been passed as
an argument and may need to be expanded when used as a function.
New versions of Clang warn:
```
glslang/MachineIndependent/Constant.cpp(216,114):
error: overflow in expression; result is -9223372036854775808 with type 'long long' [-Werror,-Winteger-overflow]
else if (rightUnionArray[i].getI64Const() == -1 && leftUnionArray[i].getI64Const() == (long long)-0x8000000000000000ll)
^
glslang/MachineIndependent/Constant.cpp(217,61):
error: overflow in expression; result is -9223372036854775808 with type 'long long' [-Werror,-Winteger-overflow]
newConstArray[i].setI64Const((long long)-0x8000000000000000ll);
^
2 errors generated.
```
Using LLONG_MIN instead avoids the problem. I think it's also more
clear, and the code for EOpMod further down already does this.
From TParseContext used only by GLSL, to TParseContextBase inherited by both GLSL and HLSL paths.
It caused compilations from HLSL to SPIR-V 1.3+ to use BufferBlock decoration which is no longer valid.
GLSL 1.20 and prior stated that "the attribute qualifier can be used
only with float, floating-point vectors, and matrices" and likewise
for varying.
Fixes: #3111
Add missing callgraph clean for bindless status flag.
Add test cases. Add support to check special extensions not be available for Vulkan when using GLSL.
ANGLE no longer links with glslang. This change reverts
1ef2e250fc which added a flag to strip
glslang to reduce its binary size. This flag is no longer needed.
