Updated spirv-cross.

3rdparty/spirv-cross/main.cpp

@@ -558,6 +558,9 @@ struct CLIArguments
 	bool msl_enable_frag_stencil_ref_builtin = true;
 	uint32_t msl_enable_frag_output_mask = 0xffffffff;
 	bool msl_enable_clip_distance_user_varying = true;
+	bool msl_multi_patch_workgroup = false;
+	bool msl_vertex_for_tessellation = false;
+	uint32_t msl_additional_fixed_sample_mask = 0xffffffff;
 	bool glsl_emit_push_constant_as_ubo = false;
 	bool glsl_emit_ubo_as_plain_uniforms = false;
 	SmallVector<pair<uint32_t, uint32_t>> glsl_ext_framebuffer_fetch;
@@ -747,9 +750,17 @@ static void print_help_msl()
 	                "\t[--msl-enable-frag-output-mask <mask>]:\n\t\tOnly selectively enable fragment outputs. Useful if pipeline does not enable fragment output for certain locations, as pipeline creation might otherwise fail.\n"
 	                "\t[--msl-no-clip-distance-user-varying]:\n\t\tDo not emit user varyings to emulate gl_ClipDistance in fragment shaders.\n"
 	                "\t[--msl-shader-input <index> <format> <size>]:\n\t\tSpecify the format of the shader input at <index>.\n"
-	                "\t\t<format> can be 'u16', 'u8', or 'other', to indicate a 16-bit unsigned integer, 8-bit unsigned integer, "
+	                "\t\t<format> can be 'any32', 'any16', 'u16', 'u8', or 'other', to indicate a 32-bit opaque value, 16-bit opaque value, 16-bit unsigned integer, 8-bit unsigned integer, "
 	                "or other-typed variable. <size> is the vector length of the variable, which must be greater than or equal to that declared in the shader.\n"
-	                "\t\tUseful if shader stage interfaces don't match up, as pipeline creation might otherwise fail.\n");
+	                "\t\tUseful if shader stage interfaces don't match up, as pipeline creation might otherwise fail.\n"
+	                "\t[--msl-multi-patch-workgroup]:\n\t\tUse the new style of tessellation control processing, where multiple patches are processed per workgroup.\n"
+					"\t\tThis should increase throughput by ensuring all the GPU's SIMD lanes are occupied, but it is not compatible with the old style.\n"
+					"\t\tIn addition, this style also passes input variables in buffers directly instead of using vertex attribute processing.\n"
+					"\t\tIn a future version of SPIRV-Cross, this will become the default.\n"
+	                "\t[--msl-vertex-for-tessellation]:\n\t\tWhen handling a vertex shader, marks it as one that will be used with a new-style tessellation control shader.\n"
+					"\t\tThe vertex shader is output to MSL as a compute kernel which outputs vertices to the buffer in the order they are received, rather than in index order as with --msl-capture-output normally.\n"
+	                "\t[--msl-additional-fixed-sample-mask <mask>]:\n"
+	                "\t\tSet an additional fixed sample mask. If the shader outputs a sample mask, then the final sample mask will be a bitwise AND of the two.\n");
 	// clang-format on
 }
 
@@ -983,6 +994,9 @@ static string compile_iteration(const CLIArguments &args, std::vector<uint32_t>
 		msl_opts.enable_frag_stencil_ref_builtin = args.msl_enable_frag_stencil_ref_builtin;
 		msl_opts.enable_frag_output_mask = args.msl_enable_frag_output_mask;
 		msl_opts.enable_clip_distance_user_varying = args.msl_enable_clip_distance_user_varying;
+		msl_opts.multi_patch_workgroup = args.msl_multi_patch_workgroup;
+		msl_opts.vertex_for_tessellation = args.msl_vertex_for_tessellation;
+		msl_opts.additional_fixed_sample_mask = args.msl_additional_fixed_sample_mask;
 		msl_comp->set_msl_options(msl_opts);
 		for (auto &v : args.msl_discrete_descriptor_sets)
 			msl_comp->add_discrete_descriptor_set(v);
@@ -1381,15 +1395,23 @@ static int main_inner(int argc, char *argv[])
 		// Make sure next_uint() is called in-order.
 		input.location = parser.next_uint();
 		const char *format = parser.next_value_string("other");
-		if (strcmp(format, "u16") == 0)
-			input.format = MSL_VERTEX_FORMAT_UINT16;
+		if (strcmp(format, "any32") == 0)
+			input.format = MSL_SHADER_INPUT_FORMAT_ANY32;
+		else if (strcmp(format, "any16") == 0)
+			input.format = MSL_SHADER_INPUT_FORMAT_ANY16;
+		else if (strcmp(format, "u16") == 0)
+			input.format = MSL_SHADER_INPUT_FORMAT_UINT16;
 		else if (strcmp(format, "u8") == 0)
-			input.format = MSL_VERTEX_FORMAT_UINT8;
+			input.format = MSL_SHADER_INPUT_FORMAT_UINT8;
 		else
-			input.format = MSL_VERTEX_FORMAT_OTHER;
+			input.format = MSL_SHADER_INPUT_FORMAT_OTHER;
 		input.vecsize = parser.next_uint();
 		args.msl_shader_inputs.push_back(input);
 	});
+	cbs.add("--msl-multi-patch-workgroup", [&args](CLIParser &) { args.msl_multi_patch_workgroup = true; });
+	cbs.add("--msl-vertex-for-tessellation", [&args](CLIParser &) { args.msl_vertex_for_tessellation = true; });
+	cbs.add("--msl-additional-fixed-sample-mask",
+	        [&args](CLIParser &parser) { args.msl_additional_fixed_sample_mask = parser.next_hex_uint(); });
 	cbs.add("--extension", [&args](CLIParser &parser) { args.extensions.push_back(parser.next_string()); });
 	cbs.add("--rename-entry-point", [&args](CLIParser &parser) {
 		auto old_name = parser.next_string();

3rdparty/spirv-cross/spirv_common.hpp

@@ -1589,8 +1589,10 @@ enum ExtendedDecorations
 	// Marks a buffer block for using explicit offsets (GLSL/HLSL).
 	SPIRVCrossDecorationExplicitOffset,
 
-	// Apply to a variable in the Input storage class; marks it as holding the base group passed to vkCmdDispatchBase().
-	// In MSL, this is used to adjust the WorkgroupId and GlobalInvocationId variables.
+	// Apply to a variable in the Input storage class; marks it as holding the base group passed to vkCmdDispatchBase(),
+	// or the base vertex and instance indices passed to vkCmdDrawIndexed().
+	// In MSL, this is used to adjust the WorkgroupId and GlobalInvocationId variables in compute shaders,
+	// and to hold the BaseVertex and BaseInstance variables in vertex shaders.
 	SPIRVCrossDecorationBuiltInDispatchBase,
 
 	// Apply to a variable that is a function parameter; marks it as being a "dynamic"
@@ -1599,6 +1601,20 @@ enum ExtendedDecorations
 	// Y'CbCr conversion.
 	SPIRVCrossDecorationDynamicImageSampler,
 
+	// Apply to a variable in the Input storage class; marks it as holding the size of the stage
+	// input grid.
+	// In MSL, this is used to hold the vertex and instance counts in a tessellation pipeline
+	// vertex shader.
+	SPIRVCrossDecorationBuiltInStageInputSize,
+
+	// Apply to any access chain of a tessellation I/O variable; stores the type of the sub-object
+	// that was chained to, as recorded in the input variable itself. This is used in case the pointer
+	// is itself used as the base of an access chain, to calculate the original type of the sub-object
+	// chained to, in case a swizzle needs to be applied. This should not happen normally with valid
+	// SPIR-V, but the MSL backend can change the type of input variables, necessitating the
+	// addition of swizzles to keep the generated code compiling.
+	SPIRVCrossDecorationTessIOOriginalInputTypeID,
+
 	SPIRVCrossDecorationCount
 };
 

3rdparty/spirv-cross/spirv_cross_c.cpp

@@ -635,6 +635,26 @@ spvc_result spvc_compiler_options_set_uint(spvc_compiler_options options, spvc_c
 	case SPVC_COMPILER_OPTION_MSL_ENABLE_CLIP_DISTANCE_USER_VARYING:
 		options->msl.enable_clip_distance_user_varying = value != 0;
 		break;
+
+	case SPVC_COMPILER_OPTION_MSL_MULTI_PATCH_WORKGROUP:
+		options->msl.multi_patch_workgroup = value != 0;
+		break;
+
+	case SPVC_COMPILER_OPTION_MSL_SHADER_INPUT_BUFFER_INDEX:
+		options->msl.shader_input_buffer_index = value;
+		break;
+
+	case SPVC_COMPILER_OPTION_MSL_SHADER_INDEX_BUFFER_INDEX:
+		options->msl.shader_index_buffer_index = value;
+		break;
+
+	case SPVC_COMPILER_OPTION_MSL_VERTEX_FOR_TESSELLATION:
+		options->msl.vertex_for_tessellation = value != 0;
+		break;
+
+	case SPVC_COMPILER_OPTION_MSL_VERTEX_INDEX_TYPE:
+		options->msl.vertex_index_type = static_cast<CompilerMSL::Options::IndexType>(value);
+		break;
 #endif
 
 	default:

3rdparty/spirv-cross/spirv_cross_c.h

@@ -33,7 +33,7 @@ extern "C" {
 /* Bumped if ABI or API breaks backwards compatibility. */
 #define SPVC_C_API_VERSION_MAJOR 0
 /* Bumped if APIs or enumerations are added in a backwards compatible way. */
-#define SPVC_C_API_VERSION_MINOR 35
+#define SPVC_C_API_VERSION_MINOR 36
 /* Bumped if internal implementation details change. */
 #define SPVC_C_API_VERSION_PATCH 0
 
@@ -258,12 +258,23 @@ typedef enum spvc_msl_platform
 	SPVC_MSL_PLATFORM_MAX_INT = 0x7fffffff
 } spvc_msl_platform;
 
+/* Maps to C++ API. */
+typedef enum spvc_msl_index_type
+{
+	SPVC_MSL_INDEX_TYPE_NONE = 0,
+	SPVC_MSL_INDEX_TYPE_UINT16 = 1,
+	SPVC_MSL_INDEX_TYPE_UINT32 = 2,
+	SPVC_MSL_INDEX_TYPE_MAX_INT = 0x7fffffff
+} spvc_msl_index_type;
+
 /* Maps to C++ API. */
 typedef enum spvc_msl_shader_input_format
 {
 	SPVC_MSL_SHADER_INPUT_FORMAT_OTHER = 0,
 	SPVC_MSL_SHADER_INPUT_FORMAT_UINT8 = 1,
 	SPVC_MSL_SHADER_INPUT_FORMAT_UINT16 = 2,
+	SPVC_MSL_SHADER_INPUT_FORMAT_ANY16 = 3,
+	SPVC_MSL_SHADER_INPUT_FORMAT_ANY32 = 4,
 
 	/* Deprecated names. */
 	SPVC_MSL_VERTEX_FORMAT_OTHER = SPVC_MSL_SHADER_INPUT_FORMAT_OTHER,
@@ -617,6 +628,12 @@ typedef enum spvc_compiler_option
 
 	SPVC_COMPILER_OPTION_HLSL_ENABLE_16BIT_TYPES = 60 | SPVC_COMPILER_OPTION_HLSL_BIT,
 
+	SPVC_COMPILER_OPTION_MSL_MULTI_PATCH_WORKGROUP = 61 | SPVC_COMPILER_OPTION_MSL_BIT,
+	SPVC_COMPILER_OPTION_MSL_SHADER_INPUT_BUFFER_INDEX = 62 | SPVC_COMPILER_OPTION_MSL_BIT,
+	SPVC_COMPILER_OPTION_MSL_SHADER_INDEX_BUFFER_INDEX = 63 | SPVC_COMPILER_OPTION_MSL_BIT,
+	SPVC_COMPILER_OPTION_MSL_VERTEX_FOR_TESSELLATION = 64 | SPVC_COMPILER_OPTION_MSL_BIT,
+	SPVC_COMPILER_OPTION_MSL_VERTEX_INDEX_TYPE = 65 | SPVC_COMPILER_OPTION_MSL_BIT,
+
 	SPVC_COMPILER_OPTION_INT_MAX = 0x7fffffff
 } spvc_compiler_option;
 

3rdparty/spirv-cross/spirv_msl.hpp

@@ -35,6 +35,8 @@ enum MSLShaderInputFormat
 	MSL_SHADER_INPUT_FORMAT_OTHER = 0,
 	MSL_SHADER_INPUT_FORMAT_UINT8 = 1,
 	MSL_SHADER_INPUT_FORMAT_UINT16 = 2,
+	MSL_SHADER_INPUT_FORMAT_ANY16 = 3,
+	MSL_SHADER_INPUT_FORMAT_ANY32 = 4,
 
 	// Deprecated aliases.
 	MSL_VERTEX_FORMAT_OTHER = MSL_SHADER_INPUT_FORMAT_OTHER,
@@ -271,9 +273,15 @@ public:
 		uint32_t buffer_size_buffer_index = 25;
 		uint32_t view_mask_buffer_index = 24;
 		uint32_t dynamic_offsets_buffer_index = 23;
+		uint32_t shader_input_buffer_index = 22;
+		uint32_t shader_index_buffer_index = 21;
 		uint32_t shader_input_wg_index = 0;
 		uint32_t device_index = 0;
 		uint32_t enable_frag_output_mask = 0xffffffff;
+		// Metal doesn't allow setting a fixed sample mask directly in the pipeline.
+		// We can evade this restriction by ANDing the internal sample_mask output
+		// of the shader with the additional fixed sample mask.
+		uint32_t additional_fixed_sample_mask = 0xffffffff;
 		bool enable_point_size_builtin = true;
 		bool enable_frag_depth_builtin = true;
 		bool enable_frag_stencil_ref_builtin = true;
@@ -329,6 +337,31 @@ public:
 		// can be read in subsequent stages.
 		bool enable_clip_distance_user_varying = true;
 
+		// In a tessellation control shader, assume that more than one patch can be processed in a
+		// single workgroup. This requires changes to the way the InvocationId and PrimitiveId
+		// builtins are processed, but should result in more efficient usage of the GPU.
+		bool multi_patch_workgroup = false;
+
+		// If set, a vertex shader will be compiled as part of a tessellation pipeline.
+		// It will be translated as a compute kernel, so it can use the global invocation ID
+		// to index the output buffer.
+		bool vertex_for_tessellation = false;
+
+		enum class IndexType
+		{
+			None = 0,
+			UInt16 = 1,
+			UInt32 = 2
+		};
+
+		// The type of index in the index buffer, if present. For a compute shader, Metal
+		// requires specifying the indexing at pipeline creation, rather than at draw time
+		// as with graphics pipelines. This means we must create three different pipelines,
+		// for no indexing, 16-bit indices, and 32-bit indices. Each requires different
+		// handling for the gl_VertexIndex builtin. We may as well, then, create three
+		// different shaders for these three scenarios.
+		IndexType vertex_index_type = IndexType::None;
+
 		bool is_ios() const
 		{
 			return platform == iOS;
@@ -431,7 +464,7 @@ public:
 	// input is a shader input description used to fix up shader input variables.
 	// If shader inputs are provided, is_msl_shader_input_used() will return true after
 	// calling ::compile() if the location was used by the MSL code.
-	void add_msl_shader_input(const MSLShaderInput &attr);
+	void add_msl_shader_input(const MSLShaderInput &input);
 
 	// resource is a resource binding to indicate the MSL buffer,
 	// texture or sampler index to use for a particular SPIR-V description set
@@ -692,7 +725,7 @@ protected:
 
 	void fix_up_interface_member_indices(spv::StorageClass storage, uint32_t ib_type_id);
 
-	void mark_location_as_used_by_shader(uint32_t location, spv::StorageClass storage);
+	void mark_location_as_used_by_shader(uint32_t location, const SPIRType &type, spv::StorageClass storage);
 	uint32_t ensure_correct_builtin_type(uint32_t type_id, spv::BuiltIn builtin);
 	uint32_t ensure_correct_input_type(uint32_t type_id, uint32_t location, uint32_t num_components = 0);
 
@@ -716,6 +749,7 @@ protected:
 	std::string to_sampler_expression(uint32_t id);
 	std::string to_swizzle_expression(uint32_t id);
 	std::string to_buffer_size_expression(uint32_t id);
+	bool is_direct_input_builtin(spv::BuiltIn builtin);
 	std::string builtin_qualifier(spv::BuiltIn builtin);
 	std::string builtin_type_decl(spv::BuiltIn builtin, uint32_t id = 0);
 	std::string built_in_func_arg(spv::BuiltIn builtin, bool prefix_comma);
@@ -738,7 +772,13 @@ protected:
 	uint32_t get_declared_struct_member_matrix_stride_msl(const SPIRType &struct_type, uint32_t index) const;
 	uint32_t get_declared_struct_member_alignment_msl(const SPIRType &struct_type, uint32_t index) const;
 
+	uint32_t get_declared_input_size_msl(const SPIRType &struct_type, uint32_t index) const;
+	uint32_t get_declared_input_array_stride_msl(const SPIRType &struct_type, uint32_t index) const;
+	uint32_t get_declared_input_matrix_stride_msl(const SPIRType &struct_type, uint32_t index) const;
+	uint32_t get_declared_input_alignment_msl(const SPIRType &struct_type, uint32_t index) const;
+
 	const SPIRType &get_physical_member_type(const SPIRType &struct_type, uint32_t index) const;
+	SPIRType get_presumed_input_type(const SPIRType &struct_type, uint32_t index) const;
 
 	uint32_t get_declared_struct_size_msl(const SPIRType &struct_type, bool ignore_alignment = false,
 	                                      bool ignore_padding = false) const;
@@ -757,6 +797,8 @@ protected:
 	SPIRType &get_patch_stage_in_struct_type();
 	SPIRType &get_patch_stage_out_struct_type();
 	std::string get_tess_factor_struct_name();
+	SPIRType &get_uint_type();
+	uint32_t get_uint_type_id();
 	void emit_atomic_func_op(uint32_t result_type, uint32_t result_id, const char *op, uint32_t mem_order_1,
 	                         uint32_t mem_order_2, bool has_mem_order_2, uint32_t op0, uint32_t op1 = 0,
 	                         bool op1_is_pointer = false, bool op1_is_literal = false, uint32_t op2 = 0);
@@ -771,6 +813,7 @@ protected:
 	void emit_entry_point_declarations() override;
 	uint32_t builtin_frag_coord_id = 0;
 	uint32_t builtin_sample_id_id = 0;
+	uint32_t builtin_sample_mask_id = 0;
 	uint32_t builtin_vertex_idx_id = 0;
 	uint32_t builtin_base_vertex_id = 0;
 	uint32_t builtin_instance_idx_id = 0;
@@ -782,10 +825,14 @@ protected:
 	uint32_t builtin_subgroup_invocation_id_id = 0;
 	uint32_t builtin_subgroup_size_id = 0;
 	uint32_t builtin_dispatch_base_id = 0;
+	uint32_t builtin_stage_input_size_id = 0;
 	uint32_t swizzle_buffer_id = 0;
 	uint32_t buffer_size_buffer_id = 0;
 	uint32_t view_mask_buffer_id = 0;
 	uint32_t dynamic_offsets_buffer_id = 0;
+	uint32_t uint_type_id = 0;
+
+	bool does_shader_write_sample_mask = false;
 
 	void bitcast_to_builtin_store(uint32_t target_id, std::string &expr, const SPIRType &expr_type) override;
 	void bitcast_from_builtin_load(uint32_t source_id, std::string &expr, const SPIRType &expr_type) override;
@@ -807,7 +854,8 @@ protected:
 
 	Options msl_options;
 	std::set<SPVFuncImpl> spv_function_implementations;
-	std::unordered_map<uint32_t, MSLShaderInput> inputs_by_location;
+	// Must be ordered to ensure declarations are in a specific order.
+	std::map<uint32_t, MSLShaderInput> inputs_by_location;
 	std::unordered_map<uint32_t, MSLShaderInput> inputs_by_builtin;
 	std::unordered_set<uint32_t> inputs_in_use;
 	std::unordered_map<uint32_t, uint32_t> fragment_output_components;
@@ -862,9 +910,11 @@ protected:
 	std::string buffer_size_name_suffix = "BufferSize";
 	std::string plane_name_suffix = "Plane";
 	std::string input_wg_var_name = "gl_in";
+	std::string input_buffer_var_name = "spvIn";
 	std::string output_buffer_var_name = "spvOut";
 	std::string patch_output_buffer_var_name = "spvPatchOut";
 	std::string tess_factor_buffer_var_name = "spvTessLevel";
+	std::string index_buffer_var_name = "spvIndices";
 	spv::Op previous_instruction_opcode = spv::OpNop;
 
 	// Must be ordered since declaration is in a specific order.