[Build] Add Windows ARM64 support

premake5.lua

@@ -54,7 +54,7 @@ filter("configurations:Checked")
   defines({
     "DEBUG",
   })
-filter({"configurations:Checked", "platforms:Windows"})
+filter({"configurations:Checked", "platforms:Windows-*"})
   buildoptions({
     "/RTCsu",           -- Full Run-Time Checks.
   })
@@ -153,7 +153,7 @@ filter("platforms:Android-*")
     "log",
   })
 
-filter("platforms:Windows")
+filter("platforms:Windows-*")
   system("windows")
   toolset("msc")
   buildoptions({
@@ -179,8 +179,12 @@ filter("platforms:Windows")
     "_CRT_SECURE_NO_WARNINGS",
     "WIN32",
     "_WIN64=1",
-    "_AMD64=1",
   })
+  filter("architecture:x86_64")
+    defines({
+      "_AMD64=1",
+    })
+  filter({})
   linkoptions({
     "/ignore:4006",  -- Ignores complaints about empty obj files.
     "/ignore:4221",
@@ -198,7 +202,7 @@ filter("platforms:Windows")
   })
 
 -- Embed the manifest for things like dependencies and DPI awareness.
-filter({"platforms:Windows", "kind:ConsoleApp or WindowedApp"})
+filter({"platforms:Windows-*", "kind:ConsoleApp or WindowedApp"})
   files({
     "src/xenia/base/app_win32.manifest"
   })
@@ -228,7 +232,12 @@ workspace("xenia")
         ["ARCHS"] = "x86_64"
       })
     elseif os.istarget("windows") then
-      platforms({"Windows"})
+      platforms({"Windows-ARM64", "Windows-x86_64"})
+      filter("platforms:Windows-ARM64")
+        architecture("ARM64")
+      filter("platforms:Windows-x86_64")
+        architecture("x86_64")
+      filter({})
       -- 10.0.15063.0: ID3D12GraphicsCommandList1::SetSamplePositions.
       -- 10.0.19041.0: D3D12_HEAP_FLAG_CREATE_NOT_ZEROED.
       -- 10.0.22000.0: DWMWA_WINDOW_CORNER_PREFERENCE.
@@ -284,7 +293,11 @@ workspace("xenia")
   include("src/xenia/apu/nop")
   include("src/xenia/base")
   include("src/xenia/cpu")
-  include("src/xenia/cpu/backend/x64")
+
+  filter("architecture:x86_64")
+    include("src/xenia/cpu/backend/x64")
+  filter({})
+
   include("src/xenia/debug/ui")
   include("src/xenia/gpu")
   include("src/xenia/gpu/null")

src/xenia/app/premake5.lua

@@ -32,6 +32,7 @@ project("xenia-app")
     "libavcodec",
     "libavutil",
     "mspack",
+    "SDL2",
     "snappy",
     "xxhash",
   })
@@ -78,7 +79,7 @@ project("xenia-app")
       "xenia_main.cc",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     files({
       "main_resources.rc",
     })
@@ -104,7 +105,7 @@ project("xenia-app")
       "SDL2",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     links({
       "xenia-apu-xaudio2",
       "xenia-gpu-d3d12",
@@ -113,13 +114,13 @@ project("xenia-app")
       "xenia-ui-d3d12",
     })
 
-  filter({"platforms:Windows", SINGLE_LIBRARY_FILTER})
+  filter({"platforms:Windows-*", SINGLE_LIBRARY_FILTER})
     links({
       "xenia-gpu-d3d12-trace-viewer",
       "xenia-ui-window-d3d12-demo",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     -- Only create the .user file if it doesn't already exist.
     local user_file = project_root.."/build/xenia-app.vcxproj.user"
     if not os.isfile(user_file) then

src/xenia/base/exception_handler_win.cc

@@ -36,12 +36,22 @@ LONG CALLBACK ExceptionHandlerCallback(PEXCEPTION_POINTERS ex_info) {
   }
 
   HostThreadContext thread_context;
+
+#if XE_ARCH_AMD64
   thread_context.rip = ex_info->ContextRecord->Rip;
   thread_context.eflags = ex_info->ContextRecord->EFlags;
   std::memcpy(thread_context.int_registers, &ex_info->ContextRecord->Rax,
               sizeof(thread_context.int_registers));
   std::memcpy(thread_context.xmm_registers, &ex_info->ContextRecord->Xmm0,
               sizeof(thread_context.xmm_registers));
+#elif XE_ARCH_ARM64
+  thread_context.pc = ex_info->ContextRecord->Pc;
+  thread_context.cpsr = ex_info->ContextRecord->Cpsr;
+  std::memcpy(thread_context.x, &ex_info->ContextRecord->X,
+              sizeof(thread_context.x));
+  std::memcpy(thread_context.v, &ex_info->ContextRecord->V,
+              sizeof(thread_context.v));
+#endif
 
   // https://msdn.microsoft.com/en-us/library/ms679331(v=vs.85).aspx
   // https://msdn.microsoft.com/en-us/library/aa363082(v=vs.85).aspx
@@ -78,6 +88,7 @@ LONG CALLBACK ExceptionHandlerCallback(PEXCEPTION_POINTERS ex_info) {
   for (size_t i = 0; i < xe::countof(handlers_) && handlers_[i].first; ++i) {
     if (handlers_[i].first(&ex, handlers_[i].second)) {
       // Exception handled.
+#if XE_ARCH_AMD64
       ex_info->ContextRecord->Rip = thread_context.rip;
       ex_info->ContextRecord->EFlags = thread_context.eflags;
       uint32_t modified_register_index;
@@ -98,6 +109,28 @@ LONG CALLBACK ExceptionHandlerCallback(PEXCEPTION_POINTERS ex_info) {
                     &thread_context.xmm_registers[modified_register_index],
                     sizeof(vec128_t));
       }
+#elif XE_ARCH_ARM64
+      ex_info->ContextRecord->Pc = thread_context.pc;
+      ex_info->ContextRecord->Cpsr = thread_context.cpsr;
+      uint32_t modified_register_index;
+      uint16_t modified_int_registers_remaining = ex.modified_x_registers();
+      while (xe::bit_scan_forward(modified_int_registers_remaining,
+                                  &modified_register_index)) {
+        modified_int_registers_remaining &=
+            ~(UINT16_C(1) << modified_register_index);
+        ex_info->ContextRecord->X[modified_register_index] =
+            thread_context.x[modified_register_index];
+      }
+      uint16_t modified_xmm_registers_remaining = ex.modified_v_registers();
+      while (xe::bit_scan_forward(modified_xmm_registers_remaining,
+                                  &modified_register_index)) {
+        modified_xmm_registers_remaining &=
+            ~(UINT16_C(1) << modified_register_index);
+        std::memcpy(&ex_info->ContextRecord->V + modified_register_index,
+                    &thread_context.v[modified_register_index],
+                    sizeof(vec128_t));
+      }
+#endif
       return EXCEPTION_CONTINUE_EXECUTION;
     }
   }

src/xenia/base/host_thread_context.cc

@@ -67,7 +67,7 @@ std::string HostThreadContext::GetStringFromValue(HostRegister reg,
     case Arm64Register::kPc:
       return hex ? string_util::to_hex_string(pc) : std::to_string(pc);
     case Arm64Register::kPstate:
-      return hex ? string_util::to_hex_string(pstate) : std::to_string(pstate);
+      return hex ? string_util::to_hex_string(cpsr) : std::to_string(cpsr);
     case Arm64Register::kFpsr:
       return hex ? string_util::to_hex_string(fpsr) : std::to_string(fpsr);
     case Arm64Register::kFpcr:

src/xenia/base/host_thread_context.h

@@ -202,7 +202,7 @@ class HostThreadContext {
   uint64_t x[31];
   uint64_t sp;
   uint64_t pc;
-  uint64_t pstate;
+  uint32_t cpsr;
   uint32_t fpsr;
   uint32_t fpcr;
   vec128_t v[32];

src/xenia/base/main_init_win.cc

@@ -11,6 +11,8 @@
 
 #include <cstdlib>
 
+#if XE_ARCH_AMD64
+
 // Includes Windows headers, so it goes after platform_win.h.
 #include "third_party/xbyak/xbyak/xbyak_util.h"
 
@@ -39,3 +41,5 @@ class StartupAvxCheck {
 #pragma warning(suppress : 4073)
 #pragma init_seg(lib)
 static StartupAvxCheck gStartupAvxCheck;
+
+#endif

src/xenia/base/math.h

@@ -133,10 +133,17 @@ constexpr inline uint32_t bit_count(T v) {
 }
 #else
 #if XE_COMPILER_MSVC || XE_COMPILER_INTEL
+#if XE_ARCH_AMD64
 inline uint32_t bit_count(uint32_t v) { return __popcnt(v); }
 inline uint32_t bit_count(uint64_t v) {
   return static_cast<uint32_t>(__popcnt64(v));
 }
+#elif XE_ARCH_ARM64
+inline uint32_t bit_count(uint32_t v) { return _CountOneBits(v); }
+inline uint32_t bit_count(uint64_t v) {
+  return static_cast<uint32_t>(_CountOneBits64(v));
+}
+#endif
 #elif XE_COMPILER_GCC || XE_COMPILER_CLANG
 static_assert(sizeof(unsigned int) == sizeof(uint32_t));
 static_assert(sizeof(unsigned long long) == sizeof(uint64_t));

src/xenia/cpu/backend/x64/premake5.lua

@@ -4,7 +4,11 @@ include(project_root.."/tools/build")
 group("src")
 project("xenia-cpu-backend-x64")
   uuid("7d8d5dce-4696-4197-952a-09506f725afe")
-  kind("StaticLib")
+  filter("architecture:x86_64")
+    kind("StaticLib")
+  filter("architecture:not x86_64")
+    kind("None")
+  filter({})
   language("C++")
   links({
     "capstone",

src/xenia/cpu/ppc/testing/premake5.lua

@@ -27,7 +27,7 @@ project("xenia-cpu-ppc-tests")
     links({
       "xenia-cpu-backend-x64",
     })
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     debugdir(project_root)
     debugargs({
       "2>&1",

src/xenia/cpu/stack_walker_win.cc

@@ -173,32 +173,62 @@ class Win32StackWalker : public StackWalker {
     } else {
       // Copy thread context local. We will be modifying it during stack
       // walking, so we don't want to mess with the incoming copy.
+#if XE_ARCH_AMD64
       thread_context.Rip = in_host_context->rip;
       thread_context.EFlags = in_host_context->eflags;
       std::memcpy(&thread_context.Rax, in_host_context->int_registers,
                   sizeof(in_host_context->int_registers));
       std::memcpy(&thread_context.Xmm0, in_host_context->xmm_registers,
                   sizeof(in_host_context->xmm_registers));
+#elif XE_ARCH_ARM64
+      thread_context.Pc = in_host_context->pc;
+      thread_context.Cpsr = in_host_context->cpsr;
+      std::memcpy(thread_context.X, in_host_context->x,
+                  sizeof(in_host_context->x));
+      std::memcpy(&thread_context.V, in_host_context->v,
+                  sizeof(in_host_context->v));
+#endif
     }
 
     if (out_host_context) {
       // Write out the captured thread context if the caller asked for it.
+#if XE_ARCH_AMD64
       out_host_context->rip = thread_context.Rip;
       out_host_context->eflags = thread_context.EFlags;
       std::memcpy(out_host_context->int_registers, &thread_context.Rax,
                   sizeof(out_host_context->int_registers));
       std::memcpy(out_host_context->xmm_registers, &thread_context.Xmm0,
                   sizeof(out_host_context->xmm_registers));
+#elif XE_ARCH_ARM64
+      out_host_context->pc = thread_context.Pc;
+      out_host_context->cpsr = thread_context.Cpsr;
+      std::memcpy(out_host_context->x, &thread_context.X,
+                  sizeof(out_host_context->x));
+      std::memcpy(out_host_context->v, &thread_context.V,
+                  sizeof(out_host_context->v));
+#endif
     }
 
     // Setup the frame for walking.
     STACKFRAME64 stack_frame = {0};
     stack_frame.AddrPC.Mode = AddrModeFlat;
+#if XE_ARCH_AMD64
     stack_frame.AddrPC.Offset = thread_context.Rip;
+#elif XE_ARCH_ARM64
+    stack_frame.AddrPC.Offset = thread_context.Pc;
+#endif
     stack_frame.AddrFrame.Mode = AddrModeFlat;
+#if XE_ARCH_AMD64
     stack_frame.AddrFrame.Offset = thread_context.Rbp;
+#elif XE_ARCH_ARM64
+    stack_frame.AddrFrame.Offset = thread_context.Fp;
+#endif
     stack_frame.AddrStack.Mode = AddrModeFlat;
+#if XE_ARCH_AMD64
     stack_frame.AddrStack.Offset = thread_context.Rsp;
+#elif XE_ARCH_ARM64
+    stack_frame.AddrStack.Offset = thread_context.Sp;
+#endif
 
     // Walk the stack.
     // Note that StackWalk64 is thread safe, though other dbghelp functions are

src/xenia/debug/ui/debug_window.cc

@@ -956,6 +956,7 @@ void DebugWindow::DrawRegistersPane() {
     } break;
     case RegisterGroup::kHostGeneral: {
       ImGui::BeginChild("##host_general");
+#if XE_ARCH_AMD64
       for (int i = 0; i < 18; ++i) {
         auto reg = static_cast<X64Register>(i);
         ImGui::BeginGroup();
@@ -993,6 +994,9 @@ void DebugWindow::DrawRegistersPane() {
             i, thread_info->host_context.xmm_registers[i].f32);
         ImGui::EndGroup();
       }
+#elif XE_ARCH_ARM64
+      // TODO(wunkolo): print ARM64 registers
+#endif
       ImGui::EndChild();
     }
   }

src/xenia/emulator.cc

@@ -172,11 +172,19 @@ X_STATUS Emulator::Setup(
   if (cvars::cpu == "x64") {
     backend.reset(new xe::cpu::backend::x64::X64Backend());
   }
+#elif XE_ARCH_ARM64
+  // TODO(wunkolo): Arm64 backend
+  if (cvars::cpu == "a64") {
+    backend.reset(new xe::cpu::backend::NullBackend());
+  }
 #endif  // XE_ARCH
   if (cvars::cpu == "any") {
     if (!backend) {
 #if XE_ARCH_AMD64
       backend.reset(new xe::cpu::backend::x64::X64Backend());
+#elif XE_ARCH_ARM64
+      // TODO(wunkolo): Arm64 backend
+      backend.reset(new xe::cpu::backend::NullBackend());
 #endif  // XE_ARCH
     }
   }

src/xenia/gpu/premake5.lua

@@ -43,7 +43,7 @@ project("xenia-gpu-shader-compiler")
     "../base/console_app_main_"..platform_suffix..".cc",
   })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     -- Only create the .user file if it doesn't already exist.
     local user_file = project_root.."/build/xenia-gpu-shader-compiler.vcxproj.user"
     if not os.isfile(user_file) then

src/xenia/gpu/vulkan/premake5.lua

@@ -75,7 +75,7 @@ project("xenia-gpu-vulkan-trace-viewer")
       "X11-xcb",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     -- Only create the .user file if it doesn't already exist.
     local user_file = project_root.."/build/xenia-gpu-vulkan-trace-viewer.vcxproj.user"
     if not os.isfile(user_file) then
@@ -138,7 +138,7 @@ project("xenia-gpu-vulkan-trace-dump")
       "X11-xcb",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     -- Only create the .user file if it doesn't already exist.
     local user_file = project_root.."/build/xenia-gpu-vulkan-trace-dump.vcxproj.user"
     if not os.isfile(user_file) then

src/xenia/hid/premake5.lua

@@ -53,7 +53,7 @@ project("xenia-hid-demo")
       "X11-xcb",
     })
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     links({
       "xenia-hid-winkey",
       "xenia-hid-xinput",

src/xenia/ui/premake5.lua

@@ -19,7 +19,7 @@ project("xenia-ui")
     -- Exports JNI functions.
     wholelib("On")
 
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     links({
       "dwmapi",
       "dxgi",

third_party/SDL2.lua

@@ -26,7 +26,7 @@ end
 -- Call this function in project scope to include the SDL2 headers.
 --
 function sdl2_include()
-  filter("platforms:Windows")
+  filter("platforms:Windows-*")
     includedirs({
       path.getrelative(".", third_party_path) .. "/SDL2/include",
     })