Merge pull request #60 from KhronosGroup/spec-constant

Add more complete specialization constant support

main.cpp

@@ -347,6 +347,16 @@ static void print_push_constant_resources(const Compiler &compiler, const vector
 	}
 }
 
+static void print_spec_constants(const Compiler &compiler)
+{
+	auto spec_constants = compiler.get_specialization_constants();
+	fprintf(stderr, "Specialization constants\n");
+	fprintf(stderr, "==================\n\n");
+	for (auto &c : spec_constants)
+		fprintf(stderr, "ID: %u, Spec ID: %u\n", c.id, c.constant_id);
+	fprintf(stderr, "==================\n\n");
+}
+
 struct PLSArg
 {
 	PlsFormat format;
@@ -650,6 +660,7 @@ int main(int argc, char *argv[])
 	{
 		print_resources(*compiler, res);
 		print_push_constant_resources(*compiler, res.push_constant_buffers);
+		print_spec_constants(*compiler);
 	}
 
 	if (combined_image_samplers)

reference/shaders/vulkan/frag/spec-constant.vk.frag

@@ -0,0 +1,59 @@
+#version 310 es
+precision mediump float;
+precision highp int;
+
+struct Foo
+{
+    float elems[(4 + 2)];
+};
+
+layout(location = 0) out vec4 FragColor;
+
+void main()
+{
+    float t0 = 1.0;
+    float t1 = 2.0;
+    mediump uint c0 = (uint(3) + 0u);
+    mediump int c1 = (-3);
+    mediump int c2 = (~3);
+    mediump int c3 = (3 + 4);
+    mediump int c4 = (3 - 4);
+    mediump int c5 = (3 * 4);
+    mediump int c6 = (3 / 4);
+    mediump uint c7 = (5u / 6u);
+    mediump int c8 = (3 % 4);
+    mediump uint c9 = (5u % 6u);
+    mediump int c10 = (3 >> 4);
+    mediump uint c11 = (5u >> 6u);
+    mediump int c12 = (3 << 4);
+    mediump int c13 = (3 | 4);
+    mediump int c14 = (3 ^ 4);
+    mediump int c15 = (3 & 4);
+    bool c16 = (false || true);
+    bool c17 = (false && true);
+    bool c18 = (!false);
+    bool c19 = (false == true);
+    bool c20 = (false != true);
+    bool c21 = (3 == 4);
+    bool c22 = (3 != 4);
+    bool c23 = (3 < 4);
+    bool c24 = (5u < 6u);
+    bool c25 = (3 > 4);
+    bool c26 = (5u > 6u);
+    bool c27 = (3 <= 4);
+    bool c28 = (5u <= 6u);
+    bool c29 = (3 >= 4);
+    bool c30 = (5u >= 6u);
+    mediump int c31 = (c8 + c3);
+    mediump int c32 = int(5u + 0u);
+    bool c33 = (3 != int(0u));
+    bool c34 = (5u != 0u);
+    mediump int c35 = int(false);
+    mediump uint c36 = uint(false);
+    float c37 = float(false);
+    float vec0[4][(3 + 3)];
+    float vec1[(3 + 2)][(4 + 5)];
+    Foo foo;
+    FragColor = (((vec4((t0 + t1)) + vec4(vec0[0][0])) + vec4(vec1[0][0])) + vec4(foo.elems[3]));
+}
+

reference/shaders/vulkan/frag/spec-constant.vk.frag.vk

@@ -0,0 +1,68 @@
+#version 310 es
+precision mediump float;
+precision highp int;
+
+layout(constant_id = 1) const float _9 = 1.0;
+layout(constant_id = 2) const float _11 = 2.0;
+layout(constant_id = 3) const int _16 = 3;
+layout(constant_id = 4) const int _25 = 4;
+layout(constant_id = 5) const uint _34 = 5u;
+layout(constant_id = 6) const uint _35 = 6u;
+layout(constant_id = 7) const bool _56 = false;
+layout(constant_id = 8) const bool _57 = true;
+
+struct Foo
+{
+    float elems[(_25 + 2)];
+};
+
+layout(location = 0) out vec4 FragColor;
+
+void main()
+{
+    float t0 = _9;
+    float t1 = _11;
+    mediump uint c0 = (uint(_16) + 0u);
+    mediump int c1 = (-_16);
+    mediump int c2 = (~_16);
+    mediump int c3 = (_16 + _25);
+    mediump int c4 = (_16 - _25);
+    mediump int c5 = (_16 * _25);
+    mediump int c6 = (_16 / _25);
+    mediump uint c7 = (_34 / _35);
+    mediump int c8 = (_16 % _25);
+    mediump uint c9 = (_34 % _35);
+    mediump int c10 = (_16 >> _25);
+    mediump uint c11 = (_34 >> _35);
+    mediump int c12 = (_16 << _25);
+    mediump int c13 = (_16 | _25);
+    mediump int c14 = (_16 ^ _25);
+    mediump int c15 = (_16 & _25);
+    bool c16 = (_56 || _57);
+    bool c17 = (_56 && _57);
+    bool c18 = (!_56);
+    bool c19 = (_56 == _57);
+    bool c20 = (_56 != _57);
+    bool c21 = (_16 == _25);
+    bool c22 = (_16 != _25);
+    bool c23 = (_16 < _25);
+    bool c24 = (_34 < _35);
+    bool c25 = (_16 > _25);
+    bool c26 = (_34 > _35);
+    bool c27 = (_16 <= _25);
+    bool c28 = (_34 <= _35);
+    bool c29 = (_16 >= _25);
+    bool c30 = (_34 >= _35);
+    mediump int c31 = (c8 + c3);
+    mediump int c32 = int(_34 + 0u);
+    bool c33 = (_16 != int(0u));
+    bool c34 = (_34 != 0u);
+    mediump int c35 = int(_56);
+    mediump uint c36 = uint(_56);
+    float c37 = float(_56);
+    float vec0[_25][(_16 + 3)];
+    float vec1[(_16 + 2)][(_25 + 5)];
+    Foo foo;
+    FragColor = (((vec4((t0 + t1)) + vec4(vec0[0][0])) + vec4(vec1[0][0])) + vec4(foo.elems[_16]));
+}
+

shaders/vulkan/frag/spec-constant.vk.frag

@@ -0,0 +1,77 @@
+#version 310 es
+precision mediump float;
+
+layout(location = 0) out vec4 FragColor;
+layout(constant_id = 1) const float a = 1.0;
+layout(constant_id = 2) const float b = 2.0;
+layout(constant_id = 3) const int c = 3;
+layout(constant_id = 4) const int d = 4;
+layout(constant_id = 5) const uint e = 5u;
+layout(constant_id = 6) const uint f = 6u;
+layout(constant_id = 7) const bool g = false;
+layout(constant_id = 8) const bool h = true;
+// glslang doesn't seem to support partial spec constants or composites yet, so only test the basics.
+
+struct Foo
+{
+	float elems[d + 2];
+};
+
+void main()
+{
+	float t0 = a;
+	float t1 = b;
+
+	uint c0 = uint(c); // OpIAdd with different types.
+	// FConvert, float-to-double.
+	int c1 = -c; // SNegate
+	int c2 = ~c; // OpNot
+	int c3 = c + d; // OpIAdd
+	int c4 = c - d; // OpISub
+	int c5 = c * d; // OpIMul
+	int c6 = c / d; // OpSDiv
+	uint c7 = e / f; // OpUDiv
+	int c8 = c % d; // OpSMod
+	uint c9 = e % f; // OpUMod
+	// TODO: OpSRem, any way to access this in GLSL?
+	int c10 = c >> d; // OpShiftRightArithmetic
+	uint c11 = e >> f; // OpShiftRightLogical
+	int c12 = c << d; // OpShiftLeftLogical
+	int c13 = c | d; // OpBitwiseOr
+	int c14 = c ^ d; // OpBitwiseXor
+	int c15 = c & d; // OpBitwiseAnd
+	// VectorShuffle, CompositeExtract, CompositeInsert, not testable atm.
+	bool c16 = g || h; // OpLogicalOr
+	bool c17 = g && h; // OpLogicalAnd
+	bool c18 = !g; // OpLogicalNot
+	bool c19 = g == h; // OpLogicalEqual
+	bool c20 = g != h; // OpLogicalNotEqual
+	// OpSelect not testable atm.
+	bool c21 = c == d; // OpIEqual
+	bool c22 = c != d; // OpINotEqual
+	bool c23 = c < d; // OpSLessThan
+	bool c24 = e < f; // OpULessThan
+	bool c25 = c > d; // OpSGreaterThan
+	bool c26 = e > f; // OpUGreaterThan
+	bool c27 = c <= d; // OpSLessThanEqual
+	bool c28 = e <= f; // OpULessThanEqual
+	bool c29 = c >= d; // OpSGreaterThanEqual
+	bool c30 = e >= f; // OpUGreaterThanEqual
+	// OpQuantizeToF16 not testable atm.
+
+	int c31 = c8 + c3;
+
+	int c32 = int(e); // OpIAdd with different types.
+	bool c33 = bool(c); // int -> bool
+	bool c34 = bool(e); // uint -> bool
+	int c35 = int(g); // bool -> int
+	uint c36 = uint(g); // bool -> uint
+	float c37 = float(g); // bool -> float
+
+	// Flexible sized arrays with spec constants and spec constant ops.
+	float vec0[d][c + 3];
+	float vec1[c + 2][d + 5];
+
+	Foo foo;
+	FragColor = vec4(t0 + t1) + vec0[0][0] + vec1[0][0] + foo.elems[c];
+}

spirv_common.hpp

@@ -134,6 +134,7 @@ enum Types
 	TypeBlock,
 	TypeExtension,
 	TypeExpression,
+	TypeConstantOp,
 	TypeUndef
 };
 
@@ -150,6 +151,25 @@ struct SPIRUndef : IVariant
 	uint32_t basetype;
 };
 
+struct SPIRConstantOp : IVariant
+{
+	enum
+	{
+		type = TypeConstantOp
+	};
+
+	SPIRConstantOp(uint32_t result_type, spv::Op op, const uint32_t *args, uint32_t length)
+	    : opcode(op)
+	    , arguments(args, args + length)
+	    , basetype(result_type)
+	{
+	}
+
+	spv::Op opcode;
+	std::vector<uint32_t> arguments;
+	uint32_t basetype;
+};
+
 struct SPIRType : IVariant
 {
 	enum
@@ -182,9 +202,16 @@ struct SPIRType : IVariant
 	uint32_t vecsize = 1;
 	uint32_t columns = 1;
 
-	// Arrays, suport array of arrays by having a vector of array sizes.
+	// Arrays, support array of arrays by having a vector of array sizes.
 	std::vector<uint32_t> array;
 
+	// Array elements can be either specialization constants or specialization ops.
+	// This array determines how to interpret the array size.
+	// If an element is true, the element is a literal,
+	// otherwise, it's an expression, which must be resolved on demand.
+	// The actual size is not really known until runtime.
+	std::vector<bool> array_size_literal;
+
 	// Pointers
 	bool pointer = false;
 	spv::StorageClass storage = spv::StorageClassGeneric;
@@ -826,6 +853,7 @@ struct Meta
 		uint32_t offset = 0;
 		uint32_t array_stride = 0;
 		uint32_t input_attachment = 0;
+		uint32_t spec_id = 0;
 		bool builtin = false;
 		bool per_instance = false;
 	};

spirv_cpp.cpp

@@ -410,8 +410,8 @@ string CompilerCPP::argument_decl(const SPIRFunction::Parameter &arg)
 	string variable_name = to_name(var.self);
 	remap_variable_type_name(type, variable_name, base);
 
-	for (auto &array : type.array)
+	for (uint32_t i = 0; i < type.array.size(); i++)
-		base = join("std::array<", base, ", ", array, ">");
+		base = join("std::array<", base, ", ", to_array_size(type, i), ">");
 
 	return join(constref ? "const " : "", base, " &", variable_name);
 }
@@ -421,16 +421,18 @@ string CompilerCPP::variable_decl(const SPIRType &type, const string &name)
 	string base = type_to_glsl(type);
 	remap_variable_type_name(type, name, base);
 	bool runtime = false;
-	for (auto &array : type.array)
+
+	for (uint32_t i = 0; i < type.array.size(); i++)
 	{
-		if (array)
+		auto &array = type.array[i];
-			base = join("std::array<", base, ", ", array, ">");
+		if (!array && type.array_size_literal[i])
-		else
 		{
 			// Avoid using runtime arrays with std::array since this is undefined.
 			// Runtime arrays cannot be passed around as values, so this is fine.
 			runtime = true;
 		}
+		else
+			base = join("std::array<", base, ", ", to_array_size(type, i), ">");
 	}
 	base += ' ';
 	return base + name + (runtime ? "[1]" : "");

spirv_cross.cpp

@@ -319,6 +319,9 @@ const SPIRType &Compiler::expression_type(uint32_t id) const
 	case TypeConstant:
 		return get<SPIRType>(get<SPIRConstant>(id).constant_type);
 
+	case TypeConstantOp:
+		return get<SPIRType>(get<SPIRConstantOp>(id).basetype);
+
 	case TypeUndef:
 		return get<SPIRType>(get<SPIRUndef>(id).basetype);
 
@@ -354,7 +357,8 @@ bool Compiler::is_immutable(uint32_t id) const
 	}
 	else if (ids[id].get_type() == TypeExpression)
 		return get<SPIRExpression>(id).immutable;
-	else if (ids[id].get_type() == TypeConstant || ids[id].get_type() == TypeUndef)
+	else if (ids[id].get_type() == TypeConstant || ids[id].get_type() == TypeConstantOp ||
+	         ids[id].get_type() == TypeUndef)
 		return true;
 	else
 		return false;
@@ -814,6 +818,10 @@ void Compiler::set_member_decoration(uint32_t id, uint32_t index, Decoration dec
 		dec.offset = argument;
 		break;
 
+	case DecorationSpecId:
+		dec.spec_id = argument;
+		break;
+
 	default:
 		break;
 	}
@@ -855,6 +863,8 @@ uint32_t Compiler::get_member_decoration(uint32_t id, uint32_t index, Decoration
 		return dec.location;
 	case DecorationOffset:
 		return dec.offset;
+	case DecorationSpecId:
+		return dec.spec_id;
 	default:
 		return 0;
 	}
@@ -892,6 +902,10 @@ void Compiler::unset_member_decoration(uint32_t id, uint32_t index, Decoration d
 		dec.offset = 0;
 		break;
 
+	case DecorationSpecId:
+		dec.spec_id = 0;
+		break;
+
 	default:
 		break;
 	}
@@ -933,6 +947,10 @@ void Compiler::set_decoration(uint32_t id, Decoration decoration, uint32_t argum
 		dec.input_attachment = argument;
 		break;
 
+	case DecorationSpecId:
+		dec.spec_id = argument;
+		break;
+
 	default:
 		break;
 	}
@@ -974,6 +992,8 @@ uint32_t Compiler::get_decoration(uint32_t id, Decoration decoration) const
 		return dec.set;
 	case DecorationInputAttachmentIndex:
 		return dec.input_attachment;
+	case DecorationSpecId:
+		return dec.spec_id;
 	default:
 		return 0;
 	}
@@ -1005,6 +1025,14 @@ void Compiler::unset_decoration(uint32_t id, Decoration decoration)
 		dec.set = 0;
 		break;
 
+	case DecorationInputAttachmentIndex:
+		dec.input_attachment = 0;
+		break;
+
+	case DecorationSpecId:
+		dec.spec_id = 0;
+		break;
+
 	default:
 		break;
 	}
@@ -1237,7 +1265,12 @@ void Compiler::parse(const Instruction &instruction)
 		auto &arraybase = set<SPIRType>(id);
 
 		arraybase = base;
-		arraybase.array.push_back(get<SPIRConstant>(ops[2]).scalar());
+
+		auto *c = maybe_get<SPIRConstant>(ops[2]);
+		bool literal = c && !c->specialization;
+
+		arraybase.array_size_literal.push_back(literal);
+		arraybase.array.push_back(literal ? c->scalar() : ops[2]);
 		// Do NOT set arraybase.self!
 		break;
 	}
@@ -1251,6 +1284,7 @@ void Compiler::parse(const Instruction &instruction)
 
 		arraybase = base;
 		arraybase.array.push_back(0);
+		arraybase.array_size_literal.push_back(true);
 		// Do NOT set arraybase.self!
 		break;
 	}
@@ -1729,6 +1763,19 @@ void Compiler::parse(const Instruction &instruction)
 		break;
 	}
 
+	case OpSpecConstantOp:
+	{
+		if (length < 3)
+			throw CompilerError("OpSpecConstantOp not enough arguments.");
+
+		uint32_t result_type = ops[0];
+		uint32_t id = ops[1];
+		auto spec_op = static_cast<Op>(ops[2]);
+
+		set<SPIRConstantOp>(id, result_type, spec_op, ops + 3, length - 3);
+		break;
+	}
+
 	// Actual opcodes.
 	default:
 	{
@@ -2616,3 +2663,30 @@ void Compiler::build_combined_image_samplers()
 	CombinedImageSamplerHandler handler(*this);
 	traverse_all_reachable_opcodes(get<SPIRFunction>(entry_point), handler);
 }
+
+vector<SpecializationConstant> Compiler::get_specialization_constants() const
+{
+	vector<SpecializationConstant> spec_consts;
+	for (auto &id : ids)
+	{
+		if (id.get_type() == TypeConstant)
+		{
+			auto &c = id.get<SPIRConstant>();
+			if (c.specialization)
+			{
+				spec_consts.push_back({ c.self, get_decoration(c.self, DecorationSpecId) });
+			}
+		}
+	}
+	return spec_consts;
+}
+
+SPIRConstant &Compiler::get_constant(uint32_t id)
+{
+	return get<SPIRConstant>(id);
+}
+
+const SPIRConstant &Compiler::get_constant(uint32_t id) const
+{
+	return get<SPIRConstant>(id);
+}

spirv_cross.hpp

@@ -91,6 +91,14 @@ struct CombinedImageSampler
 	uint32_t sampler_id;
 };
 
+struct SpecializationConstant
+{
+	// The ID of the specialization constant.
+	uint32_t id;
+	// The constant ID of the constant, used in Vulkan during pipeline creation.
+	uint32_t constant_id;
+};
+
 struct BufferRange
 {
 	unsigned index;
@@ -288,6 +296,16 @@ public:
 		variable_remap_callback = std::move(cb);
 	}
 
+	// API for querying which specialization constants exist.
+	// To modify a specialization constant before compile(), use get_constant(constant.id),
+	// then update constants directly in the SPIRConstant data structure.
+	// For composite types, the subconstants can be iterated over and modified.
+	// constant_type is the SPIRType for the specialization constant,
+	// which can be queried to determine which fields in the unions should be poked at.
+	std::vector<SpecializationConstant> get_specialization_constants() const;
+	SPIRConstant &get_constant(uint32_t id);
+	const SPIRConstant &get_constant(uint32_t id) const;
+
 protected:
 	const uint32_t *stream(const Instruction &instr) const
 	{

spirv_glsl.cpp

@@ -23,6 +23,27 @@ using namespace spv;
 using namespace spirv_cross;
 using namespace std;
 
+// Returns true if an arithmetic operation does not change behavior depending on signedness.
+static bool opcode_is_sign_invariant(Op opcode)
+{
+	switch (opcode)
+	{
+	case OpIEqual:
+	case OpINotEqual:
+	case OpISub:
+	case OpIAdd:
+	case OpIMul:
+	case OpShiftLeftLogical:
+	case OpBitwiseOr:
+	case OpBitwiseXor:
+	case OpBitwiseAnd:
+		return true;
+
+	default:
+		return false;
+	}
+}
+
 static const char *to_pls_layout(PlsFormat format)
 {
 	switch (format)
@@ -708,6 +729,7 @@ uint32_t CompilerGLSL::type_to_std430_array_stride(const SPIRType &type, uint64_
 	// Array stride is equal to aligned size of the underlying type.
 	SPIRType tmp = type;
 	tmp.array.pop_back();
+	tmp.array_size_literal.pop_back();
 	uint32_t size = type_to_std430_size(tmp, flags);
 	uint32_t alignment = type_to_std430_alignment(tmp, flags);
 	return (size + alignment - 1) & ~(alignment - 1);
@@ -716,7 +738,7 @@ uint32_t CompilerGLSL::type_to_std430_array_stride(const SPIRType &type, uint64_
 uint32_t CompilerGLSL::type_to_std430_size(const SPIRType &type, uint64_t flags)
 {
 	if (!type.array.empty())
-		return type.array.back() * type_to_std430_array_stride(type, flags);
+		return to_array_size_literal(type, type.array.size() - 1) * type_to_std430_array_stride(type, flags);
 
 	const uint32_t base_alignment = type_to_std430_base_size(type);
 	uint32_t size = 0;
@@ -1046,6 +1068,15 @@ void CompilerGLSL::emit_uniform(const SPIRVariable &var)
 	statement(layout_for_variable(var), "uniform ", variable_decl(var), ";");
 }
 
+void CompilerGLSL::emit_specialization_constant(const SPIRConstant &constant)
+{
+	auto &type = get<SPIRType>(constant.constant_type);
+	auto name = to_name(constant.self);
+
+	statement("layout(constant_id = ", get_decoration(constant.self, DecorationSpecId), ") const ",
+	          variable_decl(type, name), " = ", constant_expression(constant), ";");
+}
+
 void CompilerGLSL::replace_illegal_names()
 {
 	for (auto &id : ids)
@@ -1185,6 +1216,34 @@ void CompilerGLSL::emit_resources()
 	if (!pls_inputs.empty() || !pls_outputs.empty())
 		emit_pls();
 
+	bool emitted = false;
+
+	// If emitted Vulkan GLSL,
+	// emit specialization constants as actual floats,
+	// spec op expressions will redirect to the constant name.
+	//
+	// TODO: If we have the fringe case that we create a spec constant which depends on a struct type,
+	// we'll have to deal with that, but there's currently no known way to express that.
+	if (options.vulkan_semantics)
+	{
+		for (auto &id : ids)
+		{
+			if (id.get_type() == TypeConstant)
+			{
+				auto &c = id.get<SPIRConstant>();
+				if (!c.specialization)
+					continue;
+
+				emit_specialization_constant(c);
+				emitted = true;
+			}
+		}
+	}
+
+	if (emitted)
+		statement("");
+	emitted = false;
+
 	// Output all basic struct types which are not Block or BufferBlock as these are declared inplace
 	// when such variables are instantiated.
 	for (auto &id : ids)
@@ -1233,8 +1292,6 @@ void CompilerGLSL::emit_resources()
 		}
 	}
 
-	bool emitted = false;
-
 	bool skip_separate_image_sampler = !combined_image_samplers.empty() || !options.vulkan_semantics;
 
 	// Output Uniform Constants (values, samplers, images, etc).
@@ -1366,7 +1423,16 @@ string CompilerGLSL::to_expression(uint32_t id)
 	}
 
 	case TypeConstant:
-		return constant_expression(get<SPIRConstant>(id));
+	{
+		auto &c = get<SPIRConstant>(id);
+		if (c.specialization && options.vulkan_semantics)
+			return to_name(id);
+		else
+			return constant_expression(c);
+	}
+
+	case TypeConstantOp:
+		return constant_op_expression(get<SPIRConstantOp>(id));
 
 	case TypeVariable:
 	{
@@ -1393,6 +1459,145 @@ string CompilerGLSL::to_expression(uint32_t id)
 	}
 }
 
+string CompilerGLSL::constant_op_expression(const SPIRConstantOp &cop)
+{
+	auto &type = get<SPIRType>(cop.basetype);
+	bool binary = false;
+	bool unary = false;
+	string op;
+
+	// TODO: Find a clean way to reuse emit_instruction.
+	switch (cop.opcode)
+	{
+	case OpSConvert:
+	case OpUConvert:
+	case OpFConvert:
+		op = type_to_glsl_constructor(type);
+		break;
+
+#define BOP(opname, x) \
+	case Op##opname:   \
+		binary = true; \
+		op = x;        \
+		break
+
+#define UOP(opname, x) \
+	case Op##opname:   \
+		unary = true;  \
+		op = x;        \
+		break
+
+		UOP(SNegate, "-");
+		UOP(Not, "~");
+		BOP(IAdd, "+");
+		BOP(ISub, "-");
+		BOP(IMul, "*");
+		BOP(SDiv, "/");
+		BOP(UDiv, "/");
+		BOP(UMod, "%");
+		BOP(SMod, "%");
+		BOP(ShiftRightLogical, ">>");
+		BOP(ShiftRightArithmetic, ">>");
+		BOP(ShiftLeftLogical, "<<");
+		BOP(BitwiseOr, "|");
+		BOP(BitwiseXor, "^");
+		BOP(BitwiseAnd, "&");
+		BOP(LogicalOr, "||");
+		BOP(LogicalAnd, "&&");
+		UOP(LogicalNot, "!");
+		BOP(LogicalEqual, "==");
+		BOP(LogicalNotEqual, "!=");
+		BOP(IEqual, "==");
+		BOP(INotEqual, "!=");
+		BOP(ULessThan, "<");
+		BOP(SLessThan, "<");
+		BOP(ULessThanEqual, "<=");
+		BOP(SLessThanEqual, "<=");
+		BOP(UGreaterThan, ">");
+		BOP(SGreaterThan, ">");
+		BOP(UGreaterThanEqual, ">=");
+		BOP(SGreaterThanEqual, ">=");
+
+	case OpSelect:
+	{
+		if (cop.arguments.size() < 3)
+			throw CompilerError("Not enough arguments to OpSpecConstantOp.");
+
+		// This one is pretty annoying. It's triggered from
+		// uint(bool), int(bool) from spec constants.
+		// In order to preserve its compile-time constness in Vulkan GLSL,
+		// we need to reduce the OpSelect expression back to this simplified model.
+		// If we cannot, fail.
+		if (!to_trivial_mix_op(type, op, cop.arguments[2], cop.arguments[1], cop.arguments[0]))
+		{
+			throw CompilerError(
+			    "Cannot implement specialization constant op OpSelect. "
+			    "Need trivial select implementation which can be resolved to a simple cast from boolean.");
+		}
+		break;
+	}
+
+	default:
+		// Some opcodes are unimplemented here, these are currently not possible to test from glslang.
+		throw CompilerError("Unimplemented spec constant op.");
+	}
+
+	SPIRType::BaseType input_type;
+	bool skip_cast_if_equal_type = opcode_is_sign_invariant(cop.opcode);
+
+	switch (cop.opcode)
+	{
+	case OpIEqual:
+	case OpINotEqual:
+		input_type = SPIRType::Int;
+		break;
+
+	default:
+		input_type = type.basetype;
+		break;
+	}
+
+#undef BOP
+#undef UOP
+	if (binary)
+	{
+		if (cop.arguments.size() < 2)
+			throw CompilerError("Not enough arguments to OpSpecConstantOp.");
+
+		string cast_op0;
+		string cast_op1;
+		auto expected_type = binary_op_bitcast_helper(cast_op0, cast_op1, input_type, cop.arguments[0],
+		                                              cop.arguments[1], skip_cast_if_equal_type);
+
+		if (type.basetype != input_type && type.basetype != SPIRType::Boolean)
+		{
+			expected_type.basetype = input_type;
+			auto expr = bitcast_glsl_op(type, expected_type);
+			expr += '(';
+			expr += join(cast_op0, " ", op, " ", cast_op1);
+			expr += ')';
+			return expr;
+		}
+		else
+			return join("(", cast_op0, " ", op, " ", cast_op1, ")");
+	}
+	else if (unary)
+	{
+		if (cop.arguments.size() < 1)
+			throw CompilerError("Not enough arguments to OpSpecConstantOp.");
+
+		// Auto-bitcast to result type as needed.
+		// Works around various casting scenarios in glslang as there is no OpBitcast for specialization constants.
+		return join("(", op, bitcast_glsl(type, cop.arguments[0]), ")");
+	}
+	else
+	{
+		if (cop.arguments.size() < 1)
+			throw CompilerError("Not enough arguments to OpSpecConstantOp.");
+		return join(op, "(", to_expression(cop.arguments[0]), ")");
+	}
+}
+
 string CompilerGLSL::constant_expression(const SPIRConstant &c)
 {
 	if (!c.subconstants.empty())
@@ -1406,7 +1611,12 @@ string CompilerGLSL::constant_expression(const SPIRConstant &c)
 
 		for (auto &elem : c.subconstants)
 		{
-			res += constant_expression(get<SPIRConstant>(elem));
+			auto &subc = get<SPIRConstant>(elem);
+			if (subc.specialization && options.vulkan_semantics)
+				res += to_name(elem);
+			else
+				res += constant_expression(get<SPIRConstant>(elem));
+
 			if (&elem != &c.subconstants.back())
 				res += ", ";
 		}
@@ -1742,9 +1952,7 @@ void CompilerGLSL::emit_binary_op_cast(uint32_t result_type, uint32_t result_id,
 		extra_parens = false;
 	}
 	else
-	{
 		expr += join(cast_op0, " ", op, " ", cast_op1);
-	}
 
 	emit_op(result_type, result_id, expr, should_forward(op0) && should_forward(op1), extra_parens);
 }
@@ -1870,17 +2078,77 @@ string CompilerGLSL::legacy_tex_op(const std::string &op, const SPIRType &imgtyp
 		throw CompilerError(join("Unsupported legacy texture op: ", op));
 }
 
+bool CompilerGLSL::to_trivial_mix_op(const SPIRType &type, string &op, uint32_t left, uint32_t right, uint32_t lerp)
+{
+	auto *cleft = maybe_get<SPIRConstant>(left);
+	auto *cright = maybe_get<SPIRConstant>(right);
+	auto &lerptype = expression_type(lerp);
+
+	// If our targets aren't constants, we cannot use construction.
+	if (!cleft || !cright)
+		return false;
+
+	// If our targets are spec constants, we cannot use construction.
+	if (cleft->specialization || cright->specialization)
+		return false;
+
+	// We can only use trivial construction if we have a scalar
+	// (should be possible to do it for vectors as well, but that is overkill for now).
+	if (lerptype.basetype != SPIRType::Boolean || lerptype.vecsize > 1)
+		return false;
+
+	// If our bool selects between 0 and 1, we can cast from bool instead, making our trivial constructor.
+	bool ret = false;
+	switch (type.basetype)
+	{
+	case SPIRType::Int:
+	case SPIRType::UInt:
+		ret = cleft->scalar() == 0 && cright->scalar() == 1;
+		break;
+
+	case SPIRType::Float:
+		ret = cleft->scalar_f32() == 0.0f && cright->scalar_f32() == 1.0f;
+		break;
+
+	case SPIRType::Double:
+		ret = cleft->scalar_f64() == 0.0 && cright->scalar_f64() == 1.0;
+		break;
+
+	case SPIRType::Int64:
+	case SPIRType::UInt64:
+		ret = cleft->scalar_u64() == 0 && cright->scalar_u64() == 1;
+		break;
+
+	default:
+		break;
+	}
+
+	if (ret)
+		op = type_to_glsl_constructor(type);
+	return ret;
+}
+
 void CompilerGLSL::emit_mix_op(uint32_t result_type, uint32_t id, uint32_t left, uint32_t right, uint32_t lerp)
 {
 	auto &lerptype = expression_type(lerp);
 	auto &restype = get<SPIRType>(result_type);
 
+	string mix_op;
 	bool has_boolean_mix = (options.es && options.version >= 310) || (!options.es && options.version >= 450);
+	bool trivial_mix = to_trivial_mix_op(restype, mix_op, left, right, lerp);
 
-	// Boolean mix not supported on desktop without extension.
+	// If we can reduce the mix to a simple cast, do so.
-	// Was added in OpenGL 4.5 with ES 3.1 compat.
+	// This helps for cases like int(bool), uint(bool) which is implemented with
-	if (!has_boolean_mix && lerptype.basetype == SPIRType::Boolean)
+	// OpSelect bool 1 0.
+	if (trivial_mix)
 	{
+		emit_unary_func_op(result_type, id, lerp, mix_op.c_str());
+	}
+	else if (!has_boolean_mix && lerptype.basetype == SPIRType::Boolean)
+	{
+		// Boolean mix not supported on desktop without extension.
+		// Was added in OpenGL 4.5 with ES 3.1 compat.
+		//
 		// Could use GL_EXT_shader_integer_mix on desktop at least,
 		// but Apple doesn't support it. :(
 		// Just implement it as ternary expressions.
@@ -3013,12 +3281,14 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	uint32_t length = instruction.length;
 
 #define BOP(op) emit_binary_op(ops[0], ops[1], ops[2], ops[3], #op)
-#define BOP_CAST(op, type, skip_cast) emit_binary_op_cast(ops[0], ops[1], ops[2], ops[3], #op, type, skip_cast)
+#define BOP_CAST(op, type) \
+	emit_binary_op_cast(ops[0], ops[1], ops[2], ops[3], #op, type, opcode_is_sign_invariant(opcode))
 #define UOP(op) emit_unary_op(ops[0], ops[1], ops[2], #op)
 #define QFOP(op) emit_quaternary_func_op(ops[0], ops[1], ops[2], ops[3], ops[4], ops[5], #op)
 #define TFOP(op) emit_trinary_func_op(ops[0], ops[1], ops[2], ops[3], ops[4], #op)
 #define BFOP(op) emit_binary_func_op(ops[0], ops[1], ops[2], ops[3], #op)
-#define BFOP_CAST(op, type, skip_cast) emit_binary_func_op_cast(ops[0], ops[1], ops[2], ops[3], #op, type, skip_cast)
+#define BFOP_CAST(op, type) \
+	emit_binary_func_op_cast(ops[0], ops[1], ops[2], ops[3], #op, type, opcode_is_sign_invariant(opcode))
 #define BFOP(op) emit_binary_func_op(ops[0], ops[1], ops[2], ops[3], #op)
 #define UFOP(op) emit_unary_func_op(ops[0], ops[1], ops[2], #op)
 
@@ -3433,7 +3703,7 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	{
 		// For simple arith ops, prefer the output type if there's a mismatch to avoid extra bitcasts.
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(+, type, true);
+		BOP_CAST(+, type);
 		break;
 	}
 
@@ -3444,7 +3714,7 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	case OpISub:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(-, type, true);
+		BOP_CAST(-, type);
 		break;
 	}
 
@@ -3455,7 +3725,7 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	case OpIMul:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(*, type, true);
+		BOP_CAST(*, type);
 		break;
 	}
 
@@ -3481,11 +3751,11 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 		break;
 
 	case OpSDiv:
-		BOP_CAST(/, SPIRType::Int, false);
+		BOP_CAST(/, SPIRType::Int);
 		break;
 
 	case OpUDiv:
-		BOP_CAST(/, SPIRType::UInt, false);
+		BOP_CAST(/, SPIRType::UInt);
 		break;
 
 	case OpFDiv:
@@ -3493,38 +3763,38 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 		break;
 
 	case OpShiftRightLogical:
-		BOP_CAST(>>, SPIRType::UInt, false);
+		BOP_CAST(>>, SPIRType::UInt);
 		break;
 
 	case OpShiftRightArithmetic:
-		BOP_CAST(>>, SPIRType::Int, false);
+		BOP_CAST(>>, SPIRType::Int);
 		break;
 
 	case OpShiftLeftLogical:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(<<, type, true);
+		BOP_CAST(<<, type);
 		break;
 	}
 
 	case OpBitwiseOr:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(|, type, true);
+		BOP_CAST(|, type);
 		break;
 	}
 
 	case OpBitwiseXor:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST (^, type, true);
+		BOP_CAST (^, type);
 		break;
 	}
 
 	case OpBitwiseAnd:
 	{
 		auto type = get<SPIRType>(ops[0]).basetype;
-		BOP_CAST(&, type, true);
+		BOP_CAST(&, type);
 		break;
 	}
 
@@ -3533,11 +3803,11 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 		break;
 
 	case OpUMod:
-		BOP_CAST(%, SPIRType::UInt, false);
+		BOP_CAST(%, SPIRType::UInt);
 		break;
 
 	case OpSMod:
-		BOP_CAST(%, SPIRType::Int, false);
+		BOP_CAST(%, SPIRType::Int);
 		break;
 
 	case OpFMod:
@@ -3572,9 +3842,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	case OpIEqual:
 	{
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(equal, SPIRType::Int, true);
+			BFOP_CAST(equal, SPIRType::Int);
 		else
-			BOP_CAST(==, SPIRType::Int, true);
+			BOP_CAST(==, SPIRType::Int);
 		break;
 	}
 
@@ -3591,9 +3861,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	case OpINotEqual:
 	{
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(notEqual, SPIRType::Int, true);
+			BFOP_CAST(notEqual, SPIRType::Int);
 		else
-			BOP_CAST(!=, SPIRType::Int, true);
+			BOP_CAST(!=, SPIRType::Int);
 		break;
 	}
 
@@ -3612,9 +3882,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	{
 		auto type = opcode == OpUGreaterThan ? SPIRType::UInt : SPIRType::Int;
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(greaterThan, type, false);
+			BFOP_CAST(greaterThan, type);
 		else
-			BOP_CAST(>, type, false);
+			BOP_CAST(>, type);
 		break;
 	}
 
@@ -3632,9 +3902,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	{
 		auto type = opcode == OpUGreaterThanEqual ? SPIRType::UInt : SPIRType::Int;
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(greaterThanEqual, type, false);
+			BFOP_CAST(greaterThanEqual, type);
 		else
-			BOP_CAST(>=, type, false);
+			BOP_CAST(>=, type);
 		break;
 	}
 
@@ -3652,9 +3922,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	{
 		auto type = opcode == OpULessThan ? SPIRType::UInt : SPIRType::Int;
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(lessThan, type, false);
+			BFOP_CAST(lessThan, type);
 		else
-			BOP_CAST(<, type, false);
+			BOP_CAST(<, type);
 		break;
 	}
 
@@ -3672,9 +3942,9 @@ void CompilerGLSL::emit_instruction(const Instruction &instruction)
 	{
 		auto type = opcode == OpULessThanEqual ? SPIRType::UInt : SPIRType::Int;
 		if (expression_type(ops[2]).vecsize > 1)
-			BFOP_CAST(lessThanEqual, type, false);
+			BFOP_CAST(lessThanEqual, type);
 		else
-			BOP_CAST(<=, type, false);
+			BOP_CAST(<=, type);
 		break;
 	}
 
@@ -4435,6 +4705,39 @@ string CompilerGLSL::pls_decl(const PlsRemap &var)
 	            to_name(variable.self));
 }
 
+uint32_t CompilerGLSL::to_array_size_literal(const SPIRType &type, uint32_t index) const
+{
+	assert(type.array.size() == type.array_size_literal.size());
+
+	if (!type.array_size_literal[index])
+		throw CompilerError("The array size is not a literal, but a specialization constant or spec constant op.");
+
+	return type.array[index];
+}
+
+string CompilerGLSL::to_array_size(const SPIRType &type, uint32_t index)
+{
+	assert(type.array.size() == type.array_size_literal.size());
+
+	auto &size = type.array[index];
+	if (!type.array_size_literal[index])
+		return to_expression(size);
+	else if (size)
+		return convert_to_string(size);
+	else if (!backend.flexible_member_array_supported)
+	{
+		// For runtime-sized arrays, we can work around
+		// lack of standard support for this by simply having
+		// a single element array.
+		//
+		// Runtime length arrays must always be the last element
+		// in an interface block.
+		return "1";
+	}
+	else
+		return "";
+}
+
 string CompilerGLSL::type_to_array_glsl(const SPIRType &type)
 {
 	if (type.array.empty())
@@ -4443,23 +4746,8 @@ string CompilerGLSL::type_to_array_glsl(const SPIRType &type)
 	string res;
 	for (size_t i = type.array.size(); i; i--)
 	{
-		auto &size = type.array[i - 1];
-
 		res += "[";
-		if (size)
+		res += to_array_size(type, i - 1);
-		{
-			res += convert_to_string(size);
-		}
-		else if (!backend.flexible_member_array_supported)
-		{
-			// For runtime-sized arrays, we can work around
-			// lack of standard support for this by simply having
-			// a single element array.
-			//
-			// Runtime length arrays must always be the last element
-			// in an interface block.
-			res += '1';
-		}
 		res += "]";
 	}
 	return res;

spirv_glsl.hpp

@@ -144,6 +144,7 @@ protected:
 	virtual std::string member_decl(const SPIRType &type, const SPIRType &member_type, uint32_t member);
 	virtual std::string image_type_glsl(const SPIRType &type);
 	virtual std::string constant_expression(const SPIRConstant &c);
+	std::string constant_op_expression(const SPIRConstantOp &cop);
 	virtual std::string constant_expression_vector(const SPIRConstant &c, uint32_t vector);
 	virtual void emit_fixup();
 	virtual std::string variable_decl(const SPIRType &type, const std::string &name);
@@ -203,6 +204,8 @@ protected:
 	Options options;
 
 	std::string type_to_array_glsl(const SPIRType &type);
+	std::string to_array_size(const SPIRType &type, uint32_t index);
+	uint32_t to_array_size_literal(const SPIRType &type, uint32_t index) const;
 	std::string variable_decl(const SPIRVariable &variable);
 
 	void add_local_variable_name(uint32_t id);
@@ -240,6 +243,7 @@ protected:
 	void emit_push_constant_block_glsl(const SPIRVariable &var);
 	void emit_interface_block(const SPIRVariable &type);
 	void emit_block_chain(SPIRBlock &block);
+	void emit_specialization_constant(const SPIRConstant &constant);
 	std::string emit_continue_block(uint32_t continue_block);
 	bool attempt_emit_loop_header(SPIRBlock &block, SPIRBlock::Method method);
 	void emit_uniform(const SPIRVariable &var);
@@ -254,6 +258,7 @@ protected:
 
 	bool should_forward(uint32_t id);
 	void emit_mix_op(uint32_t result_type, uint32_t id, uint32_t left, uint32_t right, uint32_t lerp);
+	bool to_trivial_mix_op(const SPIRType &type, std::string &op, uint32_t left, uint32_t right, uint32_t lerp);
 	void emit_glsl_op(uint32_t result_type, uint32_t result_id, uint32_t op, const uint32_t *args, uint32_t count);
 	void emit_quaternary_func_op(uint32_t result_type, uint32_t result_id, uint32_t op0, uint32_t op1, uint32_t op2,
 	                             uint32_t op3, const char *op);

spirv_msl.cpp

@@ -1260,6 +1260,7 @@ SPIRType &CompilerMSL::get_pad_type(uint32_t pad_len)
 	ib_type.basetype = SPIRType::Char;
 	ib_type.width = 8;
 	ib_type.array.push_back(pad_len);
+	ib_type.array_size_literal.push_back(true);
 	set_decoration(ib_type.self, DecorationArrayStride, pad_len);
 
 	pad_type_ids_by_pad_len[pad_len] = pad_type_id;
@@ -1616,7 +1617,7 @@ size_t CompilerMSL::get_declared_type_size(const SPIRType &type, uint64_t dec_ma
 			// ArrayStride is part of the array type not OpMemberDecorate.
 			auto &dec = meta[type.self].decoration;
 			if (dec.decoration_flags & (1ull << DecorationArrayStride))
-				return dec.array_stride * type.array.back();
+				return dec.array_stride * to_array_size_literal(type, type.array.size() - 1);
 			else
 				throw CompilerError("Type does not have ArrayStride set.");
 		}