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.
From the command line, the debug options "stack", with -gVS enabling all
of generateDebugInfo, emitNonSemanticShaderDebugInfo and
emitNonSemanticShaderDebugSource, however the programmatic interface
allows setting the latter options without the former. In this case, the
string corresponding to the source filename never gets emitted and some
debuginfo instructions end up with zero ID operands, resulting in
invalid SPIR-V.
Fixes#3240
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.
It is possible for the SPIR-V code generator to receive a conditional
where the two branches have the same shader language level type but
different SPIR-V types because of things like offset decorations. This
change modifies visitSelection() to handle this case by using either
multiTypeStore() or, if that is available, OpCopyLogical.
Fixes#3164
FixesKhronosGroup/glslang#2439
When decorating a struct for Vulkan, add [Buffer]Block decoration if the
struct has a RuntimeArray member, as required by VUID-...-04680.
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.
For SPIR-V 1.6 HelperInvocation accesses need to be volatile to avoid
undefined values when shaders execute 'demote'. Previously this was
always decorated on the gl_HelperInvocation variable, but this is not
valid when the Vulkan memory model is in use.
When the memory model is enabled, stop decorating the variable
declaration and apply the memory semantic to access chain loads instead.
Fixes#3042
Implement NonSemantic.Shader.DebugInfo.100 debug instruction generation.
These instructions will be generated under the -gV and -gVS command line options. These instructions enable source-level shader debugging with Renderdoc.
This is an alpha release of this capability. Additional improvements are forthcoming. Use and feedback are welcome.
Added following updates to GL_EXT_mesh_shader implementation:
1. Added SPIRV and GLSL test cases
2. Added checks to ensure NV and EXT mesh shader builtins cannot be used interchangeably.
3. Updated the language name by removing the postfix "NV" to MeshShader and TaskShader.
4. Added checks for grammar checking to comply with the spec.
5. Added gl_NumWorkGroups builtin to Mesh shader
6. Fixed data type of gl_PrimitiveLineIndicesEXT and gl_PrimitiveTriangleIndicesEXT
7. Added new constants to the resources table
8. Updates to handle new storage qualifier "taskPayloadSharedEXT"
9. Updated test cases by replacing "taskEXT" with storage qualifier "taskPayloadSharedEXT"
Addressed Review comments
1. Fixed instruction description used by glslang disassembly.
2. Updated OpEmitMeshTasksEXT as per spec update
3. Fixed implementation that errors out if there are more then one taskPayloadSharedEXT varjables.
4. Fixed miscellaneous error logs and removed unwanted code.
SPIRV 1.6 related build failure fixes
- Update SPIRV header to 1.6
- Fix conflict wiht SPIRV 1.6 change, where localSizeId is used for execution mode for mesh/task shaders
Enable SPIRV generated for EXT_mesh_shader to be version 1.4
GL_EXT_mesh_shader: Add checks for atomic support and corresponding test cases
This type was removed from spirv1.6. If samplerBuffer is specified in
GLSL, generate textureBuffer. If samplerBuffer type is constructed,
just return the buffer.
Fixes#2956
Fixes an issue where invalid SPIR-V was generated when
gl_ObjectToWorldEXT and gl_ObjectToWorld3x4EXT, or
gl_WorldToObjectEXT and gl_WorldToObject3x4EXT, were used in the same
shader. The SPIR-V specification requires that there be at most one
OpVariable decorated with a given BuiltIn value.
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().
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.
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
