Upgrade to LLVM 11 Stable

3rdparty/llvm.cmake

@@ -11,7 +11,6 @@ if(WITH_LLVM)
 		option(LLVM_INCLUDE_TESTS OFF)
 		option(LLVM_INCLUDE_TOOLS OFF)
 		option(LLVM_INCLUDE_UTILS OFF)
-		option(WITH_POLLY OFF)
 		option(LLVM_CCACHE_BUILD ON)
 
 		set(CXX_FLAGS_OLD ${CMAKE_CXX_FLAGS})

Utilities/JIT.cpp

@@ -285,6 +285,7 @@ asmjit::Runtime& asmjit::get_global_runtime()
 #endif
 #include "llvm/Support/TargetSelect.h"
 #include "llvm/Support/FormattedStream.h"
+#include "llvm/Support/Host.h"
 #include "llvm/ExecutionEngine/ExecutionEngine.h"
 #include "llvm/ExecutionEngine/RTDyldMemoryManager.h"
 #include "llvm/ExecutionEngine/ObjectCache.h"

Vulkan/glslang-build/glslang-build.vcxproj

@@ -38,7 +38,7 @@
   </ImportGroup>
   <PropertyGroup Label="UserMacros">
     <CmakeGenerator>"Visual Studio $(VisualStudioVersion.Substring(0,2))"</CmakeGenerator>
-    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DCMAKE_CXX_STANDARD=20 -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT ../glslang</CmakeCLI>
+    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT ../glslang</CmakeCLI>
   </PropertyGroup>
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
     <NMakeBuildCommandLine>$(CmakeCLI)

Vulkan/spirv-tools-build/spirv-tools-build.vcxproj

@@ -38,7 +38,7 @@
   </ImportGroup>
   <PropertyGroup Label="UserMacros">
     <CmakeGenerator>"Visual Studio $(VisualStudioVersion.Substring(0,2))"</CmakeGenerator>
-    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DCMAKE_CXX_STANDARD=20 -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT -DSPIRV-Headers_SOURCE_DIR=$([System.IO.Path]::GetFullPath('$(MSBuildThisFileDirectory)../spirv-headers')) ../spirv-tools</CmakeCLI>
+    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT -DSPIRV-Headers_SOURCE_DIR=$([System.IO.Path]::GetFullPath('$(MSBuildThisFileDirectory)../spirv-headers')) ../spirv-tools</CmakeCLI>
     <PropsAbsPath>$([System.IO.Path]::GetFullPath('$(MSBuildThisFileDirectory)..\..\common_default.props'))</PropsAbsPath>
   </PropertyGroup>
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">

llvm_build/UpdateProjectFiles.bat

@@ -1,7 +1,7 @@
 REM You need cmake and python to update the project files
 REM this script relies on CWD being the path that this script is in
 
-cmake -G "Visual Studio 15 2017 Win64" -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_BUILD_RUNTIME=OFF -DLLVM_BUILD_TOOLS=OFF -DLLVM_INCLUDE_DOCS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_TOOLS=OFF -DLLVM_INCLUDE_UTILS=OFF -DWITH_POLLY=OFF -DCMAKE_SYSTEM_VERSION=6.1 -DCMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION=8.1 ../llvm
+cmake -G "Visual Studio 16 2019 Win64" -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_DEFAULT_TARGET_TRIPLE:STRING=x86_64-pc-windows-msvc -DLLVM_HOST_TRIPLE:STRING=x86_64-pc-windows-msvc -DLLVM_BUILD_RUNTIME=OFF -DLLVM_BUILD_TOOLS=OFF -DLLVM_INCLUDE_DOCS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_TOOLS=OFF -DLLVM_INCLUDE_BENCHMARKS=OFF -DLLVM_INCLUDE_UTILS=OFF -DLLVM_USE_INTEL_JITEVENTS=ON -DLLVM_ENABLE_Z3_SOLVER=OFF -DCMAKE_SYSTEM_VERSION=6.1 -DCMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION=10.0 -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT ../llvm
 
 RD /S /Q cmake
 RD /S /Q CMakeFiles

llvm_build/llvm_build.vcxproj

@@ -39,7 +39,7 @@
   </ImportGroup>
   <PropertyGroup Label="UserMacros">
     <CmakeGenerator>"Visual Studio $(VisualStudioVersion.Substring(0,2))"</CmakeGenerator>
-    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -Thost=x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_BUILD_RUNTIME=OFF -DLLVM_BUILD_TOOLS=OFF -DLLVM_INCLUDE_DOCS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_TOOLS=OFF -DLLVM_INCLUDE_UTILS=OFF -DWITH_POLLY=OFF -DCMAKE_SYSTEM_VERSION=6.1 -DLLVM_TEMPORARILY_ALLOW_OLD_TOOLCHAIN=ON -DCMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION=$(WindowsTargetPlatformVersion) -DCMAKE_CXX_STANDARD=20 -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT ../llvm</CmakeCLI>
+    <CmakeCLI>cmake -G $(CmakeGenerator) -A x64 -Thost=x64 -DCMAKE_CONFIGURATION_TYPES="Debug;Release" -DLLVM_TARGETS_TO_BUILD=X86 -DLLVM_BUILD_RUNTIME=OFF -DLLVM_BUILD_TOOLS=OFF -DLLVM_INCLUDE_DOCS=OFF -DLLVM_INCLUDE_EXAMPLES=OFF -DLLVM_INCLUDE_TESTS=OFF -DLLVM_INCLUDE_TOOLS=OFF -DLLVM_INCLUDE_UTILS=OFF -DCMAKE_SYSTEM_VERSION=6.1 -DLLVM_TEMPORARILY_ALLOW_OLD_TOOLCHAIN=ON -DCMAKE_VS_WINDOWS_TARGET_PLATFORM_VERSION=$(WindowsTargetPlatformVersion) -DLLVM_USE_CRT_DEBUG=MTd -DLLVM_USE_CRT_RELEASE=MT ../llvm</CmakeCLI>
   </PropertyGroup>
   <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">
     <NMakePreprocessorDefinitions>

rpcs3/Emu/CPU/CPUTranslator.cpp

@@ -74,9 +74,9 @@ llvm::Value* cpu_translator::bitcast(llvm::Value* val, llvm::Type* type)
 	uint s2 = val->getType()->getScalarSizeInBits();
 
 	if (type->isVectorTy())
-		s1 *= type->getVectorNumElements();
+		s1 *= llvm::cast<llvm::VectorType>(type)->getNumElements();
 	if (val->getType()->isVectorTy())
-		s2 *= val->getType()->getVectorNumElements();
+		s2 *= llvm::cast<llvm::VectorType>(val->getType())->getNumElements();
 
 	if (s1 != s2)
 	{
@@ -120,9 +120,9 @@ std::pair<bool, v128> cpu_translator::get_const_vector<v128>(llvm::Value* c, u32
 		fmt::throw_exception("[0x%x, %u] Not a vector" HERE, a, b);
 	}
 
-	if (uint sz = llvm::cast<llvm::VectorType>(t)->getBitWidth() - 128)
+	if (auto v = llvm::cast<llvm::VectorType>(t); v->getScalarSizeInBits() * v->getNumElements() != 128)
 	{
-		fmt::throw_exception("[0x%x, %u] Bad vector size: %u" HERE, a, b, sz + 128);
+		fmt::throw_exception("[0x%x, %u] Bad vector size: i%ux%u" HERE, a, b, v->getScalarSizeInBits(), v->getNumElements());
 	}
 
 	const auto cv = llvm::dyn_cast<llvm::ConstantDataVector>(c);
@@ -203,7 +203,8 @@ llvm::Constant* cpu_translator::make_const_vector<v128>(v128 v, llvm::Type* t)
 		return llvm::ConstantInt::get(t, llvm::APInt(128, llvm::makeArrayRef(reinterpret_cast<const u64*>(v._bytes), 2)));
 	}
 
-	verify(HERE), t->isVectorTy() && llvm::cast<llvm::VectorType>(t)->getBitWidth() == 128;
+	verify(HERE), t->isVectorTy();
+	verify(HERE), 128 == t->getScalarSizeInBits() * llvm::cast<llvm::VectorType>(t)->getNumElements();
 
 	const auto sct = t->getScalarType();
 

rpcs3/Emu/CPU/CPUTranslator.h

@@ -363,7 +363,7 @@ struct llvm_value_t<T[N]> : llvm_value_t<std::conditional_t<(std::extent_v<T> >
 		}
 		else if constexpr (N > 1)
 		{
-			return llvm::VectorType::get(base::get_type(context), N);
+			return llvm::VectorType::get(base::get_type(context), N, false);
 		}
 		else
 		{
@@ -2283,9 +2283,11 @@ struct llvm_splat
 
 		if (auto i = llvm::dyn_cast_or_null<llvm::ShuffleVectorInst>(value))
 		{
-			if (llvm::isa<llvm::ConstantAggregateZero>(i->getOperand(2)))
+			if (llvm::isa<llvm::ConstantAggregateZero>(i->getOperand(1)) || llvm::isa<llvm::UndefValue>(i->getOperand(1)))
 			{
-				if (auto j = llvm::dyn_cast<llvm::InsertElementInst>(i->getOperand(0)))
+				static constexpr int zero_array[llvm_value_t<U>::is_vector]{};
+
+				if (auto j = llvm::dyn_cast<llvm::InsertElementInst>(i->getOperand(0)); j && i->getShuffleMask().equals(zero_array))
 				{
 					if (llvm::cast<llvm::ConstantInt>(j->getOperand(2))->isZero())
 					{
@@ -2311,7 +2313,7 @@ struct llvm_zshuffle
 	using type = std::remove_extent_t<T>[N];
 
 	llvm_expr_t<A1> a1;
-	u32 index_array[N];
+	int index_array[N];
 
 	static_assert(llvm_value_t<T>::is_vector, "llvm_zshuffle<>: invalid type");
 
@@ -2334,7 +2336,7 @@ struct llvm_zshuffle
 
 			if (auto z = llvm::dyn_cast<llvm::ConstantAggregateZero>(i->getOperand(1)); z && z->getType() == v1->getType())
 			{
-				if (llvm::ConstantDataVector::get(value->getContext(), index_array) == i->getOperand(2))
+				if (i->getShuffleMask().equals(index_array))
 				{
 					if (auto r1 = a1.match(v1); v1)
 					{
@@ -2356,7 +2358,7 @@ struct llvm_shuffle2
 
 	llvm_expr_t<A1> a1;
 	llvm_expr_t<A2> a2;
-	u32 index_array[N];
+	int index_array[N];
 
 	static_assert(llvm_value_t<T>::is_vector, "llvm_shuffle2<>: invalid type");
 
@@ -2382,7 +2384,7 @@ struct llvm_shuffle2
 
 			if (v1->getType() == v2->getType() && v1->getType() == llvm_value_t<T>::get_type(v1->getContext()))
 			{
-				if (llvm::ConstantDataVector::get(value->getContext(), index_array) == i->getOperand(2))
+				if (i->getShuffleMask().equals(index_array))
 				{
 					if (auto r1 = a1.match(v1); v1)
 					{
@@ -2423,7 +2425,7 @@ class cpu_translator
 	// Allow FMA
 	bool m_use_fma = false;
 
-	// Allow Icelake tier AVX-512 
+	// Allow Icelake tier AVX-512
 	bool m_use_avx512_icl = false;
 
 	// IR builder
@@ -2665,13 +2667,13 @@ class cpu_translator
 	template <typename T, typename... Args, typename = std::enable_if_t<llvm_zshuffle<sizeof...(Args), T>::is_ok>>
 	static auto zshuffle(T&& v, Args... indices)
 	{
-		return llvm_zshuffle<sizeof...(Args), T>{std::forward<T>(v), {static_cast<u32>(indices)...}};
+		return llvm_zshuffle<sizeof...(Args), T>{std::forward<T>(v), {static_cast<int>(indices)...}};
 	}
 
 	template <typename T, typename U, typename... Args, typename = std::enable_if_t<llvm_shuffle2<sizeof...(Args), T, U>::is_ok>>
 	static auto shuffle2(T&& v1, U&& v2, Args... indices)
 	{
-		return llvm_shuffle2<sizeof...(Args), T, U>{std::forward<T>(v1), std::forward<U>(v2), {static_cast<u32>(indices)...}};
+		return llvm_shuffle2<sizeof...(Args), T, U>{std::forward<T>(v1), std::forward<U>(v2), {static_cast<int>(indices)...}};
 	}
 
 	// Average: (a + b + 1) >> 1
@@ -2685,7 +2687,7 @@ class cpu_translator
 		const auto cast_op = result.is_sint ? llvm::Instruction::SExt : llvm::Instruction::ZExt;
 		llvm::Type* cast_to = m_ir->getIntNTy(result.esize * 2);
 		if constexpr (result.is_vector != 0)
-			cast_to = llvm::VectorType::get(cast_to, result.is_vector);
+			cast_to = llvm::VectorType::get(cast_to, result.is_vector, false);
 
 		const auto axt = m_ir->CreateCast(cast_op, a.eval(m_ir), cast_to);
 		const auto bxt = m_ir->CreateCast(cast_op, b.eval(m_ir), cast_to);
@@ -2752,7 +2754,7 @@ class cpu_translator
 	{
 		value_t<typename T::type> result;
 		const auto av = a.eval(m_ir);
-		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.rcp.ps", av->getType(), av->getType()).getCallee(), {av});
+		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.rcp.ps", av->getType(), av->getType()), {av});
 		return result;
 	}
 
@@ -2761,7 +2763,7 @@ class cpu_translator
 	{
 		value_t<typename T::type> result;
 		const auto av = a.eval(m_ir);
-		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.rsqrt.ps", av->getType(), av->getType()).getCallee(), {av});
+		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.rsqrt.ps", av->getType(), av->getType()), {av});
 		return result;
 	}
 
@@ -2771,7 +2773,7 @@ class cpu_translator
 		value_t<typename T::type> result;
 		const auto av = a.eval(m_ir);
 		const auto bv = b.eval(m_ir);
-		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.max.ps", av->getType(), av->getType(), av->getType()).getCallee(), {av, bv});
+		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.max.ps", av->getType(), av->getType(), av->getType()), {av, bv});
 		return result;
 	}
 
@@ -2781,7 +2783,7 @@ class cpu_translator
 		value_t<typename T::type> result;
 		const auto av = a.eval(m_ir);
 		const auto bv = b.eval(m_ir);
-		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.min.ps", av->getType(), av->getType(), av->getType()).getCallee(), {av, bv});
+		result.value  = m_ir->CreateCall(m_module->getOrInsertFunction("llvm.x86.sse.min.ps", av->getType(), av->getType(), av->getType()), {av, bv});
 		return result;
 	}
 
@@ -2930,7 +2932,7 @@ struct fmt_unveil<llvm::TypeSize, void>
 template <>
 inline llvm::Type* cpu_translator::get_type<__m128i>()
 {
-	return llvm::VectorType::get(llvm::Type::getInt8Ty(m_context), 16);
+	return llvm::VectorType::get(llvm::Type::getInt8Ty(m_context), 16, false);
 }
 
 #ifndef _MSC_VER

rpcs3/Emu/Cell/PPUThread.cpp

@@ -30,6 +30,7 @@
 #endif
 #include "llvm/Support/FormattedStream.h"
 #include "llvm/Support/MemoryBuffer.h"
+#include "llvm/Support/Host.h"
 #include "llvm/Object/ObjectFile.h"
 #include "llvm/ADT/Triple.h"
 #include "llvm/IR/LLVMContext.h"