diff --git a/src/alloy/backend/x64/x64_assembler.cc b/src/alloy/backend/x64/x64_assembler.cc
index 5a7028e11..d70afe909 100644
--- a/src/alloy/backend/x64/x64_assembler.cc
+++ b/src/alloy/backend/x64/x64_assembler.cc
@@ -66,6 +66,8 @@ int X64Assembler::Assemble(
     FunctionInfo* symbol_info, HIRBuilder* builder,
     uint32_t debug_info_flags, DebugInfo* debug_info,
     Function** out_function) {
+  SCOPE_profile_cpu_f("alloy");
+
   int result = 0;
 
   // Lower HIR -> x64.
diff --git a/src/alloy/backend/x64/x64_code_cache.cc b/src/alloy/backend/x64/x64_code_cache.cc
index 7282c2e23..9d1c2ce60 100644
--- a/src/alloy/backend/x64/x64_code_cache.cc
+++ b/src/alloy/backend/x64/x64_code_cache.cc
@@ -75,6 +75,8 @@ int X64CodeCache::Initialize() {
 
 void* X64CodeCache::PlaceCode(void* machine_code, size_t code_size,
                               size_t stack_size) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Add unwind info into the allocation size. Keep things 16b aligned.
   code_size += XEROUNDUP(X64CodeChunk::UNWIND_INFO_SIZE, 16);
 
diff --git a/src/alloy/backend/x64/x64_emitter.cc b/src/alloy/backend/x64/x64_emitter.cc
index 58ac912d9..6616b0d52 100644
--- a/src/alloy/backend/x64/x64_emitter.cc
+++ b/src/alloy/backend/x64/x64_emitter.cc
@@ -77,6 +77,8 @@ int X64Emitter::Emit(
     HIRBuilder* builder,
     uint32_t debug_info_flags, runtime::DebugInfo* debug_info,
     void*& out_code_address, size_t& out_code_size) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Reset.
   if (debug_info_flags & DEBUG_INFO_SOURCE_MAP) {
     source_map_count_ = 0;
diff --git a/src/alloy/compiler/compiler.cc b/src/alloy/compiler/compiler.cc
index a28f6b48b..62c6e5a4b 100644
--- a/src/alloy/compiler/compiler.cc
+++ b/src/alloy/compiler/compiler.cc
@@ -49,6 +49,8 @@ void Compiler::Reset() {
 }
 
 int Compiler::Compile(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // TODO(benvanik): sophisticated stuff. Run passes in parallel, run until they
   //                 stop changing things, etc.
   for (auto it = passes_.begin(); it != passes_.end(); ++it) {
diff --git a/src/alloy/compiler/passes/constant_propagation_pass.cc b/src/alloy/compiler/passes/constant_propagation_pass.cc
index 5804ed218..f8430c509 100644
--- a/src/alloy/compiler/passes/constant_propagation_pass.cc
+++ b/src/alloy/compiler/passes/constant_propagation_pass.cc
@@ -23,6 +23,8 @@ ConstantPropagationPass::~ConstantPropagationPass() {
 }
 
 int ConstantPropagationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Once ContextPromotion has run there will likely be a whole slew of
   // constants that can be pushed through the function.
   // Example:
diff --git a/src/alloy/compiler/passes/context_promotion_pass.cc b/src/alloy/compiler/passes/context_promotion_pass.cc
index c880c4f0e..dc225aea6 100644
--- a/src/alloy/compiler/passes/context_promotion_pass.cc
+++ b/src/alloy/compiler/passes/context_promotion_pass.cc
@@ -51,6 +51,8 @@ int ContextPromotionPass::Initialize(Compiler* compiler) {
 }
 
 int ContextPromotionPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Like mem2reg, but because context memory is unaliasable it's easier to
   // check and convert LoadContext/StoreContext into value operations.
   // Example of load->value promotion:
diff --git a/src/alloy/compiler/passes/control_flow_analysis_pass.cc b/src/alloy/compiler/passes/control_flow_analysis_pass.cc
index 9c1abf118..5cf6ea6a6 100644
--- a/src/alloy/compiler/passes/control_flow_analysis_pass.cc
+++ b/src/alloy/compiler/passes/control_flow_analysis_pass.cc
@@ -30,6 +30,8 @@ ControlFlowAnalysisPass::~ControlFlowAnalysisPass() {
 }
 
 int ControlFlowAnalysisPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // TODO(benvanik): reset edges for all blocks? Needed to be re-runnable.
 
   // Add edges.
diff --git a/src/alloy/compiler/passes/data_flow_analysis_pass.cc b/src/alloy/compiler/passes/data_flow_analysis_pass.cc
index 2a44f076d..209410016 100644
--- a/src/alloy/compiler/passes/data_flow_analysis_pass.cc
+++ b/src/alloy/compiler/passes/data_flow_analysis_pass.cc
@@ -36,6 +36,8 @@ DataFlowAnalysisPass::~DataFlowAnalysisPass() {
 }
 
 int DataFlowAnalysisPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Linearize blocks so that we can detect cycles and propagate dependencies.
   uint32_t block_count = LinearizeBlocks(builder);
 
diff --git a/src/alloy/compiler/passes/dead_code_elimination_pass.cc b/src/alloy/compiler/passes/dead_code_elimination_pass.cc
index d295cebec..afb8d87b2 100644
--- a/src/alloy/compiler/passes/dead_code_elimination_pass.cc
+++ b/src/alloy/compiler/passes/dead_code_elimination_pass.cc
@@ -23,6 +23,8 @@ DeadCodeEliminationPass::~DeadCodeEliminationPass() {
 }
 
 int DeadCodeEliminationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // ContextPromotion/DSE will likely leave around a lot of dead statements.
   // Code generated for comparison/testing produces many unused statements and
   // with proper use analysis it should be possible to remove most of them:
diff --git a/src/alloy/compiler/passes/finalization_pass.cc b/src/alloy/compiler/passes/finalization_pass.cc
index 7f827da15..e6358f242 100644
--- a/src/alloy/compiler/passes/finalization_pass.cc
+++ b/src/alloy/compiler/passes/finalization_pass.cc
@@ -30,6 +30,8 @@ FinalizationPass::~FinalizationPass() {
 }
 
 int FinalizationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Process the HIR and prepare it for lowering.
   // After this is done the HIR should be ready for emitting.
 
diff --git a/src/alloy/compiler/passes/register_allocation_pass.cc b/src/alloy/compiler/passes/register_allocation_pass.cc
index a89e1415c..7c3a0a7a9 100644
--- a/src/alloy/compiler/passes/register_allocation_pass.cc
+++ b/src/alloy/compiler/passes/register_allocation_pass.cc
@@ -59,6 +59,8 @@ RegisterAllocationPass::~RegisterAllocationPass() {
 }
 
 int RegisterAllocationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Simple per-block allocator that operates on SSA form.
   // Registers do not move across blocks, though this could be
   // optimized with some intra-block analysis (dominators/etc).
diff --git a/src/alloy/compiler/passes/simplification_pass.cc b/src/alloy/compiler/passes/simplification_pass.cc
index 14cea8681..7fc53c940 100644
--- a/src/alloy/compiler/passes/simplification_pass.cc
+++ b/src/alloy/compiler/passes/simplification_pass.cc
@@ -23,6 +23,8 @@ SimplificationPass::~SimplificationPass() {
 }
 
 int SimplificationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   EliminateConversions(builder);
   SimplifyAssignments(builder);
   return 0;
diff --git a/src/alloy/compiler/passes/validation_pass.cc b/src/alloy/compiler/passes/validation_pass.cc
index bc77ab482..265c82fe9 100644
--- a/src/alloy/compiler/passes/validation_pass.cc
+++ b/src/alloy/compiler/passes/validation_pass.cc
@@ -30,6 +30,8 @@ ValidationPass::~ValidationPass() {
 }
 
 int ValidationPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   StringBuffer str;
   builder->Dump(&str);
   printf(str.GetString());
diff --git a/src/alloy/compiler/passes/value_reduction_pass.cc b/src/alloy/compiler/passes/value_reduction_pass.cc
index 4eb61a38b..94453e294 100644
--- a/src/alloy/compiler/passes/value_reduction_pass.cc
+++ b/src/alloy/compiler/passes/value_reduction_pass.cc
@@ -53,6 +53,8 @@ void ValueReductionPass::ComputeLastUse(Value* value) {
 }
 
 int ValueReductionPass::Run(HIRBuilder* builder) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Walk each block and reuse variable ordinals as much as possible.
 
   llvm::BitVector ordinals(builder->max_value_ordinal());
diff --git a/src/alloy/frontend/ppc/ppc_hir_builder.cc b/src/alloy/frontend/ppc/ppc_hir_builder.cc
index 1b254ea4e..a8bec8435 100644
--- a/src/alloy/frontend/ppc/ppc_hir_builder.cc
+++ b/src/alloy/frontend/ppc/ppc_hir_builder.cc
@@ -44,6 +44,8 @@ void PPCHIRBuilder::Reset() {
 }
 
 int PPCHIRBuilder::Emit(FunctionInfo* symbol_info, bool with_debug_info) {
+  SCOPE_profile_cpu_f("alloy");
+
   Memory* memory = frontend_->memory();
   const uint8_t* p = memory->membase();
 
diff --git a/src/alloy/frontend/ppc/ppc_scanner.cc b/src/alloy/frontend/ppc/ppc_scanner.cc
index f75229b9e..9658bd595 100644
--- a/src/alloy/frontend/ppc/ppc_scanner.cc
+++ b/src/alloy/frontend/ppc/ppc_scanner.cc
@@ -38,6 +38,8 @@ bool PPCScanner::IsRestGprLr(uint64_t address) {
 }
 
 int PPCScanner::FindExtents(FunctionInfo* symbol_info) {
+  SCOPE_profile_cpu_f("alloy");
+
   // This is a simple basic block analyizer. It walks the start address to the
   // end address looking for branches. Each span of instructions between
   // branches is considered a basic block. When the last blr (that has no
@@ -286,6 +288,8 @@ int PPCScanner::FindExtents(FunctionInfo* symbol_info) {
 }
 
 std::vector<BlockInfo> PPCScanner::FindBlocks(FunctionInfo* symbol_info) {
+  SCOPE_profile_cpu_f("alloy");
+
   Memory* memory = frontend_->memory();
   const uint8_t* p = memory->membase();
 
diff --git a/src/alloy/frontend/ppc/ppc_translator.cc b/src/alloy/frontend/ppc/ppc_translator.cc
index 61617db33..4f879336c 100644
--- a/src/alloy/frontend/ppc/ppc_translator.cc
+++ b/src/alloy/frontend/ppc/ppc_translator.cc
@@ -86,6 +86,8 @@ int PPCTranslator::Translate(
     FunctionInfo* symbol_info,
     uint32_t debug_info_flags,
     Function** out_function) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Scan the function to find its extents. We only need to do this if we
   // haven't already been provided with them from some other source.
   if (!symbol_info->has_end_address()) {
diff --git a/src/alloy/hir/hir_builder.cc b/src/alloy/hir/hir_builder.cc
index f93a310e8..158e08224 100644
--- a/src/alloy/hir/hir_builder.cc
+++ b/src/alloy/hir/hir_builder.cc
@@ -51,6 +51,8 @@ void HIRBuilder::Reset() {
 }
 
 int HIRBuilder::Finalize() {
+  SCOPE_profile_cpu_f("alloy");
+
   // Scan blocks in order and add fallthrough branches. These are needed for
   // analysis passes to work. We may have also added blocks out of order and
   // need to ensure they fall through in the right order.
@@ -141,6 +143,8 @@ void HIRBuilder::DumpOp(
 }
 
 void HIRBuilder::Dump(StringBuffer* str) {
+  SCOPE_profile_cpu_f("alloy");
+
   if (attributes_) {
     str->Append("; attributes = %.8X\n", attributes_);
   }
diff --git a/src/alloy/runtime/entry_table.cc b/src/alloy/runtime/entry_table.cc
index cf6da5d70..ebec56ea4 100644
--- a/src/alloy/runtime/entry_table.cc
+++ b/src/alloy/runtime/entry_table.cc
@@ -75,6 +75,8 @@ Entry::Status EntryTable::GetOrCreate(uint64_t address, Entry** out_entry) {
 }
 
 std::vector<Function*> EntryTable::FindWithAddress(uint64_t address) {
+  SCOPE_profile_cpu_f("alloy");
+
   std::vector<Function*> fns;
   LockMutex(lock_);
   for (auto it = map_.begin(); it != map_.end(); ++it) {
diff --git a/src/alloy/runtime/function.cc b/src/alloy/runtime/function.cc
index 853808d53..2dd0ddce5 100644
--- a/src/alloy/runtime/function.cc
+++ b/src/alloy/runtime/function.cc
@@ -74,6 +74,8 @@ Breakpoint* Function::FindBreakpoint(uint64_t address) {
 }
 
 int Function::Call(ThreadState* thread_state, uint64_t return_address) {
+  SCOPE_profile_cpu_f("alloy");
+
   ThreadState* original_thread_state = ThreadState::Get();
   if (original_thread_state != thread_state) {
     ThreadState::Bind(thread_state);
diff --git a/src/alloy/runtime/module.cc b/src/alloy/runtime/module.cc
index ea056e0dd..5e38c3902 100644
--- a/src/alloy/runtime/module.cc
+++ b/src/alloy/runtime/module.cc
@@ -161,6 +161,8 @@ SymbolInfo::Status Module::DefineVariable(VariableInfo* symbol_info) {
 }
 
 void Module::ForEachFunction(std::function<void (FunctionInfo*)> callback) {
+  SCOPE_profile_cpu_f("alloy");
+
   LockMutex(lock_);
   for (auto it = list_.begin(); it != list_.end(); ++it) {
     SymbolInfo* symbol_info = *it;
@@ -174,6 +176,8 @@ void Module::ForEachFunction(std::function<void (FunctionInfo*)> callback) {
 
 void Module::ForEachFunction(size_t since, size_t& version,
                              std::function<void (FunctionInfo*)> callback) {
+  SCOPE_profile_cpu_f("alloy");
+
   LockMutex(lock_);
   size_t count = list_.size();
   version = count;
diff --git a/src/alloy/runtime/runtime.cc b/src/alloy/runtime/runtime.cc
index 3fc45a447..1aff92e04 100644
--- a/src/alloy/runtime/runtime.cc
+++ b/src/alloy/runtime/runtime.cc
@@ -159,6 +159,8 @@ std::vector<Function*> Runtime::FindFunctionsWithAddress(uint64_t address) {
 }
 
 int Runtime::ResolveFunction(uint64_t address, Function** out_function) {
+  SCOPE_profile_cpu_f("alloy");
+
   *out_function = NULL;
   Entry* entry;
   Entry::Status status = entry_table_.GetOrCreate(address, &entry);
@@ -192,6 +194,8 @@ int Runtime::ResolveFunction(uint64_t address, Function** out_function) {
 
 int Runtime::LookupFunctionInfo(
     uint64_t address, FunctionInfo** out_symbol_info) {
+  SCOPE_profile_cpu_f("alloy");
+
   *out_symbol_info = NULL;
 
   // TODO(benvanik): fast reject invalid addresses/log errors.
@@ -220,6 +224,8 @@ int Runtime::LookupFunctionInfo(
 
 int Runtime::LookupFunctionInfo(Module* module, uint64_t address,
                                 FunctionInfo** out_symbol_info) {
+  SCOPE_profile_cpu_f("alloy");
+
   // Atomic create/lookup symbol in module.
   // If we get back the NEW flag we must declare it now.
   FunctionInfo* symbol_info = NULL;
@@ -241,6 +247,8 @@ int Runtime::LookupFunctionInfo(Module* module, uint64_t address,
 
 int Runtime::DemandFunction(
     FunctionInfo* symbol_info, Function** out_function) {
+  SCOPE_profile_cpu_f("alloy");
+
   *out_function = NULL;
 
   // Lock function for generation. If it's already being generated
diff --git a/src/xenia/apu/audio_system.cc b/src/xenia/apu/audio_system.cc
index 1793fc92d..144d6bb15 100644
--- a/src/xenia/apu/audio_system.cc
+++ b/src/xenia/apu/audio_system.cc
@@ -82,21 +82,26 @@ void AudioSystem::ThreadStart() {
     if (result == WAIT_FAILED) {
       DWORD err = GetLastError();
       XEASSERTALWAYS();
+      break;
     }
+
     size_t pumped = 0;
-    if (result >= WAIT_OBJECT_0 && result <= WAIT_OBJECT_0 + (maximum_client_count_ - 1)) {
-      size_t index = result - WAIT_OBJECT_0;
-      do {
-        xe_mutex_lock(lock_);
-        uint32_t client_callback = clients_[index].callback;
-        uint32_t client_callback_arg = clients_[index].wrapped_callback_arg;
-        xe_mutex_unlock(lock_);
-        if (client_callback) {
-          processor->Execute(thread_state_, client_callback, client_callback_arg, 0);
-        }
-        pumped++;
-        index++;
-      } while (index < maximum_client_count_ && WaitForSingleObject(client_wait_handles_[index], 0) == WAIT_OBJECT_0);
+    {
+      SCOPE_profile_cpu_i("apu", "Pump");
+      if (result >= WAIT_OBJECT_0 && result <= WAIT_OBJECT_0 + (maximum_client_count_ - 1)) {
+        size_t index = result - WAIT_OBJECT_0;
+        do {
+          xe_mutex_lock(lock_);
+          uint32_t client_callback = clients_[index].callback;
+          uint32_t client_callback_arg = clients_[index].wrapped_callback_arg;
+          xe_mutex_unlock(lock_);
+          if (client_callback) {
+            processor->Execute(thread_state_, client_callback, client_callback_arg, 0);
+          }
+          pumped++;
+          index++;
+        } while (index < maximum_client_count_ && WaitForSingleObject(client_wait_handles_[index], 0) == WAIT_OBJECT_0);
+      }
     }
 
     if (!running_) {
@@ -104,6 +109,7 @@ void AudioSystem::ThreadStart() {
     }
 
     if (!pumped) {
+      SCOPE_profile_cpu_i("apu", "Sleep");
       Sleep(500);
     }
   }
@@ -126,6 +132,8 @@ void AudioSystem::Shutdown() {
 
 X_STATUS AudioSystem::RegisterClient(
     uint32_t callback, uint32_t callback_arg, size_t* out_index) {
+  SCOPE_profile_cpu_f("apu");
+
   XEASSERTTRUE(unused_clients_.size());
   xe_mutex_lock(lock_);
 
@@ -157,6 +165,8 @@ X_STATUS AudioSystem::RegisterClient(
 }
 
 void AudioSystem::SubmitFrame(size_t index, uint32_t samples_ptr) {
+  SCOPE_profile_cpu_f("apu");
+
   xe_mutex_lock(lock_);
   XEASSERTTRUE(index < maximum_client_count_);
   XEASSERTTRUE(clients_[index].driver != NULL);
@@ -166,6 +176,8 @@ void AudioSystem::SubmitFrame(size_t index, uint32_t samples_ptr) {
 }
 
 void AudioSystem::UnregisterClient(size_t index) {
+  SCOPE_profile_cpu_f("apu");
+
   xe_mutex_lock(lock_);
   XEASSERTTRUE(index < maximum_client_count_);
   DestroyDriver(clients_[index].driver);
diff --git a/src/xenia/apu/xaudio2/xaudio2_audio_driver.cc b/src/xenia/apu/xaudio2/xaudio2_audio_driver.cc
index 0155753c4..6f2cd6659 100644
--- a/src/xenia/apu/xaudio2/xaudio2_audio_driver.cc
+++ b/src/xenia/apu/xaudio2/xaudio2_audio_driver.cc
@@ -121,6 +121,8 @@ void XAudio2AudioDriver::Initialize() {
 }
 
 void XAudio2AudioDriver::SubmitFrame(uint32_t frame_ptr) {
+  SCOPE_profile_cpu_f("apu");
+
   // Process samples! They are big-endian floats.
   HRESULT hr;
 
diff --git a/src/xenia/cpu/processor.cc b/src/xenia/cpu/processor.cc
index b77664482..0c780ce22 100644
--- a/src/xenia/cpu/processor.cc
+++ b/src/xenia/cpu/processor.cc
@@ -147,6 +147,8 @@ void Processor::AddRegisterAccessCallbacks(
 }
 
 int Processor::Execute(XenonThreadState* thread_state, uint64_t address) {
+  SCOPE_profile_cpu_f("cpu");
+
   // Attempt to get the function.
   Function* fn;
   if (runtime_->ResolveFunction(address, &fn)) {
@@ -171,6 +173,8 @@ int Processor::Execute(XenonThreadState* thread_state, uint64_t address) {
 
 uint64_t Processor::Execute(
     XenonThreadState* thread_state, uint64_t address, uint64_t arg0) {
+  SCOPE_profile_cpu_f("cpu");
+
   PPCContext* context = thread_state->context();
   context->r[3] = arg0;
   if (Execute(thread_state, address)) {
@@ -182,6 +186,8 @@ uint64_t Processor::Execute(
 uint64_t Processor::Execute(
     XenonThreadState* thread_state, uint64_t address, uint64_t arg0,
     uint64_t arg1) {
+  SCOPE_profile_cpu_f("cpu");
+
   PPCContext* context = thread_state->context();
   context->r[3] = arg0;
   context->r[4] = arg1;
diff --git a/src/xenia/gpu/d3d11/d3d11_geometry_shader.cc b/src/xenia/gpu/d3d11/d3d11_geometry_shader.cc
index 5984631fe..ba677f7a0 100644
--- a/src/xenia/gpu/d3d11/d3d11_geometry_shader.cc
+++ b/src/xenia/gpu/d3d11/d3d11_geometry_shader.cc
@@ -34,6 +34,8 @@ D3D11GeometryShader::~D3D11GeometryShader() {
 }
 
 int D3D11GeometryShader::Prepare(D3D11VertexShader* vertex_shader) {
+  SCOPE_profile_cpu_f("gpu");
+
   if (handle_) {
     return 0;
   }
@@ -74,6 +76,8 @@ int D3D11GeometryShader::Prepare(D3D11VertexShader* vertex_shader) {
 }
 
 ID3D10Blob* D3D11GeometryShader::Compile(const char* shader_source) {
+  SCOPE_profile_cpu_f("gpu");
+
   // TODO(benvanik): pick shared runtime mode defines.
   D3D10_SHADER_MACRO defines[] = {
     "TEST_DEFINE", "1",
@@ -161,6 +165,7 @@ D3D11PointSpriteGeometryShader::~D3D11PointSpriteGeometryShader() {
 
 int D3D11PointSpriteGeometryShader::Generate(D3D11VertexShader* vertex_shader,
                                              alloy::StringBuffer* output) {
+  SCOPE_profile_cpu_f("gpu");
   if (D3D11GeometryShader::Generate(vertex_shader, output)) {
     return 1;
   }
@@ -215,6 +220,7 @@ D3D11RectListGeometryShader::~D3D11RectListGeometryShader() {
 
 int D3D11RectListGeometryShader::Generate(D3D11VertexShader* vertex_shader,
                                           alloy::StringBuffer* output) {
+  SCOPE_profile_cpu_f("gpu");
   if (D3D11GeometryShader::Generate(vertex_shader, output)) {
     return 1;
   }
@@ -259,6 +265,7 @@ D3D11QuadListGeometryShader::~D3D11QuadListGeometryShader() {
 
 int D3D11QuadListGeometryShader::Generate(D3D11VertexShader* vertex_shader,
                                           alloy::StringBuffer* output) {
+  SCOPE_profile_cpu_f("gpu");
   if (D3D11GeometryShader::Generate(vertex_shader, output)) {
     return 1;
   }
diff --git a/src/xenia/gpu/d3d11/d3d11_graphics_driver.cc b/src/xenia/gpu/d3d11/d3d11_graphics_driver.cc
index 25410bf6f..11518f71c 100644
--- a/src/xenia/gpu/d3d11/d3d11_graphics_driver.cc
+++ b/src/xenia/gpu/d3d11/d3d11_graphics_driver.cc
@@ -190,6 +190,8 @@ void D3D11GraphicsDriver::SetShader(
 }
 
 int D3D11GraphicsDriver::SetupDraw(XE_GPU_PRIMITIVE_TYPE prim_type) {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   // Ignore copies.
@@ -296,6 +298,8 @@ void D3D11GraphicsDriver::DrawIndexBuffer(
     XE_GPU_PRIMITIVE_TYPE prim_type,
     bool index_32bit, uint32_t index_count,
     uint32_t index_base, uint32_t index_size, uint32_t endianness) {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   XETRACED3D("D3D11: draw indexed %d (%d indicies) from %.8X",
@@ -321,6 +325,8 @@ void D3D11GraphicsDriver::DrawIndexBuffer(
 void D3D11GraphicsDriver::DrawIndexAuto(
     XE_GPU_PRIMITIVE_TYPE prim_type,
     uint32_t index_count) {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   XETRACED3D("D3D11: draw indexed %d (%d indicies)",
@@ -346,6 +352,8 @@ int D3D11GraphicsDriver::RebuildRenderTargets(
     return 0;
   }
 
+  SCOPE_profile_cpu_f("gpu");
+
   // Remove old versions.
   for (int n = 0; n < XECOUNT(render_targets_.color_buffers); n++) {
     auto& cb = render_targets_.color_buffers[n];
@@ -426,6 +434,8 @@ int D3D11GraphicsDriver::RebuildRenderTargets(
 }
 
 int D3D11GraphicsDriver::UpdateState(uint32_t state_overrides) {
+  SCOPE_profile_cpu_f("gpu");
+
   // Most information comes from here:
   // https://chromium.googlesource.com/chromiumos/third_party/mesa/+/6173cc19c45d92ef0b7bc6aa008aa89bb29abbda/src/gallium/drivers/freedreno/freedreno_zsa.c
   // http://cgit.freedesktop.org/mesa/mesa/diff/?id=aac7f06ad843eaa696363e8e9c7781ca30cb4914
@@ -768,6 +778,8 @@ int D3D11GraphicsDriver::UpdateState(uint32_t state_overrides) {
 }
 
 int D3D11GraphicsDriver::UpdateConstantBuffers() {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   D3D11_MAPPED_SUBRESOURCE res;
@@ -799,6 +811,8 @@ int D3D11GraphicsDriver::UpdateConstantBuffers() {
 }
 
 int D3D11GraphicsDriver::BindShaders() {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
   xe_gpu_program_cntl_t program_cntl;
   program_cntl.dword_0 = rf.values[XE_GPU_REG_SQ_PROGRAM_CNTL].u32;
@@ -892,6 +906,8 @@ int D3D11GraphicsDriver::BindShaders() {
 }
 
 int D3D11GraphicsDriver::PrepareFetchers() {
+  SCOPE_profile_cpu_f("gpu");
+
   // Input assembly.
   XEASSERTNOTNULL(state_.vertex_shader);
   auto vtx_inputs = state_.vertex_shader->GetVertexBufferInputs();
@@ -934,6 +950,8 @@ int D3D11GraphicsDriver::PrepareFetchers() {
 }
 
 int D3D11GraphicsDriver::PrepareVertexBuffer(Shader::vtx_buffer_desc_t& desc) {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
   int r = XE_GPU_REG_SHADER_CONSTANT_FETCH_00_0 + (desc.fetch_slot / 3) * 6;
   xe_gpu_fetch_group_t* group = (xe_gpu_fetch_group_t*)&rf.values[r];
@@ -1009,6 +1027,8 @@ int D3D11GraphicsDriver::PrepareVertexBuffer(Shader::vtx_buffer_desc_t& desc) {
 }
 
 int D3D11GraphicsDriver::PrepareTextureFetchers() {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   for (int n = 0; n < XECOUNT(state_.texture_fetchers); n++) {
@@ -1275,6 +1295,8 @@ int D3D11GraphicsDriver::FetchTexture1D(
     xe_gpu_texture_fetch_t& fetch,
     TextureInfo& info,
     ID3D11Resource** out_texture) {
+  SCOPE_profile_cpu_f("gpu");
+
   uint32_t address = (fetch.address << 12) + address_translation_;
 
   uint32_t width = 1 + fetch.size_1d.width;
@@ -1299,6 +1321,8 @@ int D3D11GraphicsDriver::FetchTexture1D(
 }
 
 XEFORCEINLINE void TextureSwap(uint8_t* dest, const uint8_t* src, uint32_t pitch, XE_GPU_ENDIAN endianness) {
+  SCOPE_profile_cpu_f("gpu");
+
   switch (endianness) {
     case XE_GPU_ENDIAN_8IN16:
       for (uint32_t i = 0; i < pitch; i += 2, src += 2, dest += 2) {
@@ -1344,6 +1368,8 @@ int D3D11GraphicsDriver::FetchTexture2D(
     xe_gpu_texture_fetch_t& fetch,
     TextureInfo& info,
     ID3D11Resource** out_texture) {
+  SCOPE_profile_cpu_f("gpu");
+
   XEASSERTTRUE(fetch.dimension == 1);
 
   uint32_t address = (fetch.address << 12) + address_translation_;
@@ -1448,6 +1474,8 @@ int D3D11GraphicsDriver::FetchTexture3D(
     xe_gpu_texture_fetch_t& fetch,
     TextureInfo& info,
     ID3D11Resource** out_texture) {
+  SCOPE_profile_cpu_f("gpu");
+
   XELOGE("D3D11: FetchTexture2D not yet implemented");
   XEASSERTALWAYS();
   return 1;
@@ -1470,6 +1498,8 @@ int D3D11GraphicsDriver::FetchTextureCube(
     xe_gpu_texture_fetch_t& fetch,
     TextureInfo& info,
     ID3D11Resource** out_texture) {
+  SCOPE_profile_cpu_f("gpu");
+
   XELOGE("D3D11: FetchTextureCube not yet implemented");
   XEASSERTALWAYS();
   return 1;
@@ -1477,6 +1507,7 @@ int D3D11GraphicsDriver::FetchTextureCube(
 
 int D3D11GraphicsDriver::PrepareTextureSampler(
     xenos::XE_GPU_SHADER_TYPE shader_type, Shader::tex_buffer_desc_t& desc) {
+  SCOPE_profile_cpu_f("gpu");
 
   auto& fetcher = state_.texture_fetchers[desc.fetch_slot];
   auto& info = fetcher.info;
@@ -1588,6 +1619,8 @@ int D3D11GraphicsDriver::PrepareTextureSampler(
 int D3D11GraphicsDriver::PrepareIndexBuffer(
     bool index_32bit, uint32_t index_count,
     uint32_t index_base, uint32_t index_size, uint32_t endianness) {
+  SCOPE_profile_cpu_f("gpu");
+
   RegisterFile& rf = register_file_;
 
   uint32_t address = index_base + address_translation_;
@@ -1634,6 +1667,8 @@ int D3D11GraphicsDriver::PrepareIndexBuffer(
 }
 
 int D3D11GraphicsDriver::Resolve() {
+  SCOPE_profile_cpu_f("gpu");
+
   // No clue how this is supposed to work yet.
   ID3D11Texture2D* back_buffer = 0;
   swap_chain_->GetBuffer(0, __uuidof(ID3D11Texture2D),
diff --git a/src/xenia/gpu/d3d11/d3d11_graphics_system.cc b/src/xenia/gpu/d3d11/d3d11_graphics_system.cc
index 03c91038c..825e9f1ff 100644
--- a/src/xenia/gpu/d3d11/d3d11_graphics_system.cc
+++ b/src/xenia/gpu/d3d11/d3d11_graphics_system.cc
@@ -29,6 +29,7 @@ void __stdcall D3D11GraphicsSystemVsyncCallback(
     thread_name_set = true;
     Profiler::ThreadEnter("VsyncTimer");
   }
+  SCOPE_profile_cpu_f("gpu");
 
   gs->MarkVblank();
   gs->DispatchInterruptCallback(0);
@@ -151,6 +152,8 @@ void D3D11GraphicsSystem::Initialize() {
 }
 
 void D3D11GraphicsSystem::Pump() {
+  SCOPE_profile_cpu_f("gpu");
+
   if (swap_pending_) {
     swap_pending_ = false;
 
diff --git a/src/xenia/gpu/d3d11/d3d11_shader.cc b/src/xenia/gpu/d3d11/d3d11_shader.cc
index a60a7bdf3..97e0cb295 100644
--- a/src/xenia/gpu/d3d11/d3d11_shader.cc
+++ b/src/xenia/gpu/d3d11/d3d11_shader.cc
@@ -145,6 +145,8 @@ void D3D11Shader::set_translated_src(char* value) {
 }
 
 ID3D10Blob* D3D11Shader::Compile(const char* shader_source) {
+  SCOPE_profile_cpu_f("gpu");
+
   // TODO(benvanik): pick shared runtime mode defines.
   D3D10_SHADER_MACRO defines[] = {
     "TEST_DEFINE", "1",
@@ -256,6 +258,7 @@ D3D11VertexShader::~D3D11VertexShader() {
 }
 
 int D3D11VertexShader::Prepare(xe_gpu_program_cntl_t* program_cntl) {
+  SCOPE_profile_cpu_f("gpu");
   if (handle_) {
     return 0;
   }
@@ -411,6 +414,8 @@ int D3D11VertexShader::Prepare(xe_gpu_program_cntl_t* program_cntl) {
 }
 
 const char* D3D11VertexShader::Translate(xe_gpu_program_cntl_t* program_cntl) {
+  SCOPE_profile_cpu_f("gpu");
+
   Output* output = new Output();
   xe_gpu_translate_ctx_t ctx;
   ctx.output  = output;
@@ -599,6 +604,7 @@ D3D11PixelShader::~D3D11PixelShader() {
 
 int D3D11PixelShader::Prepare(xe_gpu_program_cntl_t* program_cntl,
                               D3D11VertexShader* input_shader) {
+  SCOPE_profile_cpu_f("gpu");
   if (handle_) {
     return 0;
   }
@@ -641,6 +647,7 @@ int D3D11PixelShader::Prepare(xe_gpu_program_cntl_t* program_cntl,
 
 const char* D3D11PixelShader::Translate(
     xe_gpu_program_cntl_t* program_cntl, D3D11VertexShader* input_shader) {
+  SCOPE_profile_cpu_f("gpu");
   Output* output = new Output();
   xe_gpu_translate_ctx_t ctx;
   ctx.output  = output;
diff --git a/src/xenia/gpu/d3d11/d3d11_shader_cache.cc b/src/xenia/gpu/d3d11/d3d11_shader_cache.cc
index 7f6a5a722..be9352b50 100644
--- a/src/xenia/gpu/d3d11/d3d11_shader_cache.cc
+++ b/src/xenia/gpu/d3d11/d3d11_shader_cache.cc
@@ -31,6 +31,7 @@ Shader* D3D11ShaderCache::CreateCore(
     xenos::XE_GPU_SHADER_TYPE type,
     const uint8_t* src_ptr, size_t length,
     uint64_t hash) {
+  SCOPE_profile_cpu_f("gpu");
   switch (type) {
   case XE_GPU_SHADER_TYPE_VERTEX:
     return new D3D11VertexShader(
diff --git a/src/xenia/gpu/d3d11/d3d11_window.cc b/src/xenia/gpu/d3d11/d3d11_window.cc
index 64a9b0df6..da33ab6bb 100644
--- a/src/xenia/gpu/d3d11/d3d11_window.cc
+++ b/src/xenia/gpu/d3d11/d3d11_window.cc
@@ -114,6 +114,8 @@ int D3D11Window::Initialize(const char* title, uint32_t width, uint32_t height)
 }
 
 void D3D11Window::Swap() {
+  SCOPE_profile_cpu_f("gpu");
+
   // Present profiler.
   context_->OMSetRenderTargets(1, &render_target_view_, NULL);
   Profiler::Present();
diff --git a/src/xenia/gpu/ring_buffer_worker.cc b/src/xenia/gpu/ring_buffer_worker.cc
index 3792c0b61..9999601bb 100644
--- a/src/xenia/gpu/ring_buffer_worker.cc
+++ b/src/xenia/gpu/ring_buffer_worker.cc
@@ -125,6 +125,8 @@ void RingBufferWorker::Pump() {
 
 void RingBufferWorker::ExecutePrimaryBuffer(
     uint32_t start_index, uint32_t end_index) {
+  SCOPE_profile_cpu_f("gpu");
+
   // Adjust pointer base.
   uint32_t ptr = primary_buffer_ptr_ + start_index * 4;
   ptr = (primary_buffer_ptr_ & ~0x1FFFFFFF) | (ptr & 0x1FFFFFFF);
diff --git a/src/xenia/gpu/shader_cache.cc b/src/xenia/gpu/shader_cache.cc
index 9aee3e2b7..33033bc36 100644
--- a/src/xenia/gpu/shader_cache.cc
+++ b/src/xenia/gpu/shader_cache.cc
@@ -55,6 +55,8 @@ Shader* ShaderCache::Find(
 Shader* ShaderCache::FindOrCreate(
     XE_GPU_SHADER_TYPE type,
     const uint8_t* src_ptr, size_t length) {
+  SCOPE_profile_cpu_f("gpu");
+
   uint64_t hash = Hash(src_ptr, length);
   unordered_map<uint64_t, Shader*>::iterator it = map_.find(hash);
   if (it != map_.end()) {
diff --git a/src/xenia/hid/input_system.cc b/src/xenia/hid/input_system.cc
index b82ca11af..6ad1ab177 100644
--- a/src/xenia/hid/input_system.cc
+++ b/src/xenia/hid/input_system.cc
@@ -42,6 +42,8 @@ void InputSystem::AddDriver(InputDriver* driver) {
 
 X_RESULT InputSystem::GetCapabilities(
     uint32_t user_index, uint32_t flags, X_INPUT_CAPABILITIES& out_caps) {
+  SCOPE_profile_cpu_f("hid");
+
   for (auto it = drivers_.begin(); it != drivers_.end(); ++it) {
     InputDriver* driver = *it;
     if (XSUCCEEDED(driver->GetCapabilities(user_index, flags, out_caps))) {
@@ -52,6 +54,8 @@ X_RESULT InputSystem::GetCapabilities(
 }
 
 X_RESULT InputSystem::GetState(uint32_t user_index, X_INPUT_STATE& out_state) {
+  SCOPE_profile_cpu_f("hid");
+
   for (auto it = drivers_.begin(); it != drivers_.end(); ++it) {
     InputDriver* driver = *it;
     if (driver->GetState(user_index, out_state) == X_ERROR_SUCCESS) {
@@ -63,6 +67,8 @@ X_RESULT InputSystem::GetState(uint32_t user_index, X_INPUT_STATE& out_state) {
 
 X_RESULT InputSystem::SetState(
     uint32_t user_index, X_INPUT_VIBRATION& vibration) {
+  SCOPE_profile_cpu_f("hid");
+
   for (auto it = drivers_.begin(); it != drivers_.end(); ++it) {
     InputDriver* driver = *it;
     if (XSUCCEEDED(driver->SetState(user_index, vibration))) {
@@ -74,6 +80,8 @@ X_RESULT InputSystem::SetState(
 
 X_RESULT InputSystem::GetKeystroke(
     uint32_t user_index, uint32_t flags, X_INPUT_KEYSTROKE& out_keystroke) {
+  SCOPE_profile_cpu_f("hid");
+
   for (auto it = drivers_.begin(); it != drivers_.end(); ++it) {
     InputDriver* driver = *it;
     if (XSUCCEEDED(driver->GetKeystroke(user_index, flags, out_keystroke))) {