Common: Add GSVector

Mostly based on PCSX2.

CMakeModules/DuckStationUtils.cmake

@@ -57,6 +57,8 @@ function(detect_architecture)
     if("x86_64" IN_LIST CMAKE_OSX_ARCHITECTURES)
       message(STATUS "Building x86_64 MacOS binaries.")
       set(CPU_ARCH_X64 TRUE PARENT_SCOPE)
+      set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Xarch_x86_64 -msse4.1" PARENT_SCOPE)
+      set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Xarch_864_64 -msse4.1" PARENT_SCOPE)
     endif()
     if("arm64" IN_LIST CMAKE_OSX_ARCHITECTURES)
       message(STATUS "Building ARM64 MacOS binaries.")
@@ -67,6 +69,8 @@ function(detect_architecture)
          CMAKE_SIZEOF_VOID_P EQUAL 8)
     message(STATUS "Building x86_64 binaries.")
     set(CPU_ARCH_X64 TRUE PARENT_SCOPE)
+    set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse4.1" PARENT_SCOPE)
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse4.1" PARENT_SCOPE)
   elseif(("${CMAKE_SYSTEM_PROCESSOR}" STREQUAL "aarch64" OR "${CMAKE_SYSTEM_PROCESSOR}" STREQUAL "arm64") AND
          CMAKE_SIZEOF_VOID_P EQUAL 8) # Might have an A64 kernel, e.g. Raspbian.
     message(STATUS "Building ARM64 binaries.")

dep/msvc/vsprops/Base.props

@@ -30,6 +30,7 @@
       <PreprocessorDefinitions>_HAS_EXCEPTIONS=0;_CRT_INTERNAL_NONSTDC_NAMES;_CRT_NONSTDC_NO_DEPRECATE;_CRT_SECURE_NO_WARNINGS;WIN32;%(PreprocessorDefinitions)</PreprocessorDefinitions>
       <AdditionalIncludeDirectories>%(AdditionalIncludeDirectories);$(DepsIncludeDir)</AdditionalIncludeDirectories>
       <AdditionalOptions Condition="!$(Configuration.Contains(Clang))">/Zc:__cplusplus /Zo /utf-8 %(AdditionalOptions)</AdditionalOptions>
+      <AdditionalOptions Condition="$(Configuration.Contains(Clang)) And '$(Platform)'=='x64'"> -msse4.1 %(AdditionalOptions)</AdditionalOptions>
       <!-- Force ThinLTO for Release builds, MSVC doesn't seem to do it otherwise. -->
       <AdditionalOptions Condition="$(Configuration.Contains(Clang)) And $(Configuration.Contains(ReleaseLTCG))"> -flto=thin %(AdditionalOptions)</AdditionalOptions>
       <ExceptionHandling>false</ExceptionHandling>

src/common/CMakeLists.txt

@@ -21,6 +21,11 @@ add_library(common
   fifo_queue.h
   file_system.cpp
   file_system.h
+  gsvector.h
+  gsvector_formatter.h
+  gsvector_neon.h
+  gsvector_nosimd.h
+  gsvector_sse.h
   intrin.h
   hash_combine.h
   heap_array.h

src/common/common.vcxproj

@@ -16,6 +16,11 @@
     <ClInclude Include="fastjmp.h" />
     <ClInclude Include="fifo_queue.h" />
     <ClInclude Include="file_system.h" />
+    <ClInclude Include="gsvector.h" />
+    <ClInclude Include="gsvector_formatter.h" />
+    <ClInclude Include="gsvector_neon.h" />
+    <ClInclude Include="gsvector_nosimd.h" />
+    <ClInclude Include="gsvector_sse.h" />
     <ClInclude Include="hash_combine.h" />
     <ClInclude Include="heap_array.h" />
     <ClInclude Include="intrin.h" />
@@ -70,6 +75,7 @@
   </ItemGroup>
   <ItemGroup>
     <Natvis Include="bitfield.natvis" />
+    <Natvis Include="gsvector.natvis" />
     <Natvis Include="thirdparty\llvm.natvis" />
   </ItemGroup>
   <ItemGroup>

src/common/common.vcxproj.filters

@@ -46,6 +46,11 @@
     </ClInclude>
     <ClInclude Include="dynamic_library.h" />
     <ClInclude Include="binary_span_reader_writer.h" />
+    <ClInclude Include="gsvector_sse.h" />
+    <ClInclude Include="gsvector_neon.h" />
+    <ClInclude Include="gsvector.h" />
+    <ClInclude Include="gsvector_formatter.h" />
+    <ClInclude Include="gsvector_nosimd.h" />
   </ItemGroup>
   <ItemGroup>
     <ClCompile Include="small_string.cpp" />
@@ -80,6 +85,7 @@
     <Natvis Include="thirdparty\llvm.natvis">
       <Filter>thirdparty</Filter>
     </Natvis>
+    <Natvis Include="gsvector.natvis" />
   </ItemGroup>
   <ItemGroup>
     <Filter Include="thirdparty">

src/common/gsvector.h

@@ -0,0 +1,65 @@
+// SPDX-FileCopyrightText: 2019-2023 Connor McLaughlin <stenzek@gmail.com>
+// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
+
+#pragma once
+
+#include "common/intrin.h"
+
+#include <cstring>
+
+enum Align_Mode
+{
+  Align_Outside,
+  Align_Inside,
+  Align_NegInf,
+  Align_PosInf
+};
+
+enum Round_Mode
+{
+  Round_NearestInt = 8,
+  Round_NegInf = 9,
+  Round_PosInf = 10,
+  Round_Truncate = 11
+};
+
+template<class T>
+class GSVector2T
+{
+public:
+  union
+  {
+    struct
+    {
+      T x, y;
+    };
+    struct
+    {
+      T r, g;
+    };
+    struct
+    {
+      T v[2];
+    };
+  };
+
+  GSVector2T() = default;
+
+  ALWAYS_INLINE constexpr GSVector2T(T x) : x(x), y(x) {}
+  ALWAYS_INLINE constexpr GSVector2T(T x, T y) : x(x), y(y) {}
+  ALWAYS_INLINE constexpr bool operator==(const GSVector2T& v) const { return std::memcmp(this, &v, sizeof(*this)) == 0; }
+  ALWAYS_INLINE constexpr bool operator!=(const GSVector2T& v) const { return std::memcmp(this, &v, sizeof(*this)) != 0; }
+  ALWAYS_INLINE constexpr GSVector2T operator*(const GSVector2T& v) const { return {x * v.x, y * v.y}; }
+  ALWAYS_INLINE constexpr GSVector2T operator/(const GSVector2T& v) const { return {x / v.x, y / v.y}; }
+};
+
+using GSVector2 = GSVector2T<float>;
+using GSVector2i = GSVector2T<s32>;
+
+#if defined(CPU_ARCH_SSE)
+#include "common/gsvector_sse.h"
+#elif defined(CPU_ARCH_NEON)
+#include "common/gsvector_neon.h"
+#else
+#include "common/gsvector_nosimd.h"
+#endif

src/common/gsvector.natvis

@@ -0,0 +1,10 @@
+<?xml version="1.0" encoding="utf-8"?>
+<AutoVisualizer xmlns="http://schemas.microsoft.com/vstudio/debugger/natvis/2010">
+	<Type Name="GSVector2T&lt;*&gt;">
+		<DisplayString>{{ {x}, {y} }}</DisplayString>
+	</Type>
+
+	<Type Name="GSVector4i">
+		<DisplayString>{{ {I32[0]}, {I32[1]}, {I32[2]}, {I32[3]} }}</DisplayString>
+	</Type>
+</AutoVisualizer>

src/common/gsvector_formatter.h

@@ -0,0 +1,21 @@
+// SPDX-FileCopyrightText: 2024 Connor McLaughlin <stenzek@gmail.com>
+// SPDX-License-Identifier: (GPL-3.0 OR CC-BY-NC-ND-4.0)
+
+#pragma once
+
+#include "gsvector.h"
+#include "small_string.h"
+
+#include "fmt/format.h"
+
+template<>
+struct fmt::formatter<GSVector4i> : formatter<std::string_view>
+{
+  auto format(const GSVector4i& rc, format_context& ctx) const
+  {
+    const TinyString str =
+      TinyString::from_format("{},{} => {},{} ({}x{})", rc.left, rc.top, rc.right, rc.bottom, rc.width(), rc.height());
+
+    return fmt::formatter<std::string_view>::format(str.view(), ctx);
+  }
+};

src/common/intrin.h

@@ -13,7 +13,21 @@
 #if defined(CPU_ARCH_X86) || defined(CPU_ARCH_X64)
 #define CPU_ARCH_SSE 1
 #include <emmintrin.h>
-#elif defined(CPU_ARCH_ARM64)
+#include <tmmintrin.h>
+#include <smmintrin.h>
+#include <immintrin.h>
+
+#if defined(__AVX2__)
+#define CPU_ARCH_AVX 1
+#define CPU_ARCH_AVX2 1
+#define CPU_ARCH_SSE41 1
+#elif defined(__AVX__)
+#define CPU_ARCH_AVX 1
+#define CPU_ARCH_SSE41 1
+#elif defined(__SSE4_1__) || defined(_MSC_VER)
+#define CPU_ARCH_SSE41 1
+#endif
+#elif defined(CPU_ARCH_ARM32) || defined(CPU_ARCH_ARM64)
 #define CPU_ARCH_NEON 1
 #if defined(_MSC_VER) && !defined(__clang__)
 #include <arm64_neon.h>