quick/Quake-III-Arena-VS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

DX12: frames synchronization.

Browse Source 
We have a sync with GPU at the beginning of dx_begin_frame but in
practise we never wait since we usually CPU limited for such simple graphics.
This commit is contained in:
Artem Kharytoniuk 2017-12-18 20:50:00 +01:00
parent 79545d341f
commit 6e4a2f13aa
1 changed files with 19 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

54
src/engine/renderer/dx.cpp
View File

@ -60,37 +60,13 @@ static void get_hardware_adapter(IDXGIFactory4* p_factory, IDXGIAdapter1** pp_ad
*pp_adapter = adapter.Detach();
}
void wait_for_previous_frame()
{
// WAITING FOR THE FRAME TO COMPLETE BEFORE CONTINUING IS NOT BEST PRACTICE.
// This is code implemented as such for simplicity. The D3D12HelloFrameBuffering
// sample illustrates how to use fences for efficient resource usage and to
// maximize GPU utilization.
// Signal and increment the fence value.
const UINT64 fence = dx.fence_value;
DX_CHECK(dx.command_queue->Signal(dx.fence, fence));
dx.fence_value++;
// Wait until the previous frame is finished.
if (dx.fence->GetCompletedValue() < fence)
{
DX_CHECK(dx.fence->SetEventOnCompletion(fence, dx.fence_event));
WaitForSingleObject(dx.fence_event, INFINITE);
}
dx.frame_index = dx.swapchain->GetCurrentBackBufferIndex();
}
static void wait_for_queue_idle(ID3D12CommandQueue* command_queue) {
const UINT64 fence = dx.fence_value;
DX_CHECK(command_queue->Signal(dx.fence, fence));
if (dx.fence->GetCompletedValue() == dx.fence_value) {
dx.fence_value++;
if (dx.fence->GetCompletedValue() < fence) {
DX_CHECK(dx.fence->SetEventOnCompletion(fence, dx.fence_event));
WaitForSingleObject(dx.fence_event, INFINITE);
DX_CHECK(command_queue->Signal(dx.fence, dx.fence_value));
}
DX_CHECK(dx.fence->SetEventOnCompletion(dx.fence_value, dx.fence_event));
WaitForSingleObject(dx.fence_event, INFINITE);
}
static void record_and_run_commands(ID3D12CommandQueue* command_queue, std::function<void(ID3D12GraphicsCommandList*)> recorder) {
@ -329,7 +305,7 @@ void dx_initialize() {
// Create synchronization objects.
{
DX_CHECK(dx.device->CreateFence(0, D3D12_FENCE_FLAG_NONE, IID_PPV_ARGS(&dx.fence)));
dx.fence_value = 1;
dx.fence_value = 0;
// Create an event handle to use for frame synchronization.
dx.fence_event = CreateEvent(nullptr, FALSE, FALSE, nullptr);
@ -341,7 +317,7 @@ void dx_initialize() {
// Wait for the command list to execute; we are reusing the same command
// list in our main loop but for now, we just want to wait for setup to
// complete before continuing.
wait_for_previous_frame();
wait_for_queue_idle(dx.command_queue);
}
//
@ -1350,6 +1326,13 @@ void dx_shade_geometry(ID3D12PipelineState* pipeline_state, bool multitexture, V
}
void dx_begin_frame() {
if (dx.fence->GetCompletedValue() < dx.fence_value) {
DX_CHECK(dx.fence->SetEventOnCompletion(dx.fence_value, dx.fence_event));
WaitForSingleObject(dx.fence_event, INFINITE);
}
dx.frame_index = dx.swapchain->GetCurrentBackBufferIndex();
// Command list allocators can only be reset when the associated
// command lists have finished execution on the GPU; apps should use
// fences to determine GPU execution progress.
@ -1398,12 +1381,11 @@ void dx_end_frame() {
DX_CHECK(dx.command_list->Close());
// Execute the command list.
ID3D12CommandList* ppCommandLists[] = { dx.command_list };
dx.command_queue->ExecuteCommandLists(_countof(ppCommandLists), ppCommandLists);
ID3D12CommandList* command_list = dx.command_list;
dx.command_queue->ExecuteCommandLists(1, &command_list);
// Present the frame.
DX_CHECK(dx.swapchain->Present(1, 0));
dx.fence_value++;
DX_CHECK(dx.command_queue->Signal(dx.fence, dx.fence_value));
wait_for_previous_frame();
DX_CHECK(dx.swapchain->Present(0, 0));
}