squash reallocs in command buffers by using large prealloced buffer, directly use virtual memory with it so os allocs on demand

mark raw clock functions as noinline, the way msvc was inlining them and ordering the branches meant that rdtsc would often be speculatively executed
add alternative clock impl for win, instead of using queryperformancecounter we grab systemtime from kusershared. it does not have the same precision as queryperformancecounter, we only have 100 nanosecond precision, but we round to milliseconds so it never made sense to use the performance counter in the first place
stubbed out the "guest clock mutex"... (the entirety of clock.cc needs a rewrite)
added some helpers for minf/maxf without the nan handling behavior

src/xenia/gpu/d3d12/deferred_command_list.cc

@@ -31,8 +31,8 @@ void DeferredCommandList::Execute(ID3D12GraphicsCommandList* command_list,
 #if XE_UI_D3D12_FINE_GRAINED_DRAW_SCOPES
   SCOPE_profile_cpu_f("gpu");
 #endif  // XE_UI_D3D12_FINE_GRAINED_DRAW_SCOPES
-  const uintmax_t* stream = command_stream_.data();
-  size_t stream_remaining = command_stream_.size();
+  const uintmax_t* stream = (const uintmax_t*)command_stream_.data();
+  size_t stream_remaining = command_stream_.size() / sizeof(uintmax_t);
   ID3D12PipelineState* current_pipeline_state = nullptr;
   while (stream_remaining != 0) {
     const CommandHeader& header =
@@ -266,8 +266,12 @@ void DeferredCommandList::Execute(ID3D12GraphicsCommandList* command_list,
 
 void* DeferredCommandList::WriteCommand(Command command,
                                         size_t arguments_size_bytes) {
+
   size_t arguments_size_elements =
-      (arguments_size_bytes + sizeof(uintmax_t) - 1) / sizeof(uintmax_t);
+      round_up(arguments_size_bytes, sizeof(uintmax_t), false);
+
+      //(arguments_size_bytes + sizeof(uintmax_t) - 1) / sizeof(uintmax_t);
+  #if 0
   size_t offset = command_stream_.size();
   command_stream_.resize(offset + kCommandHeaderSizeElements +
                          arguments_size_elements);
@@ -276,6 +280,19 @@ void* DeferredCommandList::WriteCommand(Command command,
   header.command = command;
   header.arguments_size_elements = uint32_t(arguments_size_elements);
   return command_stream_.data() + (offset + kCommandHeaderSizeElements);
+  #else
+
+  size_t offset = command_stream_.size();
+  constexpr size_t kCommandHeaderSizeBytes =
+      kCommandHeaderSizeElements * sizeof(uintmax_t);
+  command_stream_.resize(offset + kCommandHeaderSizeBytes +
+                         arguments_size_elements);
+  CommandHeader& header =
+      *reinterpret_cast<CommandHeader*>(command_stream_.data() + offset);
+  header.command = command;
+  header.arguments_size_elements = uint32_t(arguments_size_elements) / sizeof(uintmax_t);
+  return command_stream_.data() + (offset + kCommandHeaderSizeBytes);
+  #endif
 }
 
 }  // namespace d3d12

src/xenia/gpu/d3d12/deferred_command_list.h

@@ -19,7 +19,7 @@
 #include "xenia/base/literals.h"
 #include "xenia/base/math.h"
 #include "xenia/ui/d3d12/d3d12_api.h"
-
+#include "xenia/base/memory.h"
 namespace xe {
 namespace gpu {
 namespace d3d12 {
@@ -30,11 +30,12 @@ class D3D12CommandProcessor;
 
 class DeferredCommandList {
  public:
+  static constexpr size_t MAX_SIZEOF_COMMANDLIST = 65536 * 128; //around 8 mb
   /*
 	chrispy: upped from 1_MiB to 4_MiB, m:durandal hits frequent resizes in large open maps
   */
   DeferredCommandList(const D3D12CommandProcessor& command_processor,
-                      size_t initial_size_bytes = 4_MiB);
+                      size_t initial_size_bytes = MAX_SIZEOF_COMMANDLIST);
 
   void Reset();
   void Execute(ID3D12GraphicsCommandList* command_list,
@@ -565,7 +566,8 @@ class DeferredCommandList {
   const D3D12CommandProcessor& command_processor_;
 
   // uintmax_t to ensure uint64_t and pointer alignment of all structures.
-  std::vector<uintmax_t> command_stream_;
+  //std::vector<uintmax_t> command_stream_;
+  fixed_vmem_vector<MAX_SIZEOF_COMMANDLIST> command_stream_;
 };
 
 }  // namespace d3d12