Drastically reduce cpu time wasted by XMADecoderThread spinning, went from 13% of all cpu time to about 0.6% in my tests
Commented out lock in WatchMemoryRange, lock is always held by caller properly set the value/check the irql for spinlocks in xboxkrnl_threading
This commit is contained in:
@@ -680,9 +680,6 @@ class D3D12CommandProcessor final : public CommandProcessor {
|
||||
ID3D12Resource* readback_buffer_ = nullptr;
|
||||
uint32_t readback_buffer_size_ = 0;
|
||||
|
||||
std::atomic<bool> pix_capture_requested_ = false;
|
||||
bool pix_capturing_;
|
||||
|
||||
// The current fixed-function drawing state.
|
||||
D3D12_VIEWPORT ff_viewport_;
|
||||
D3D12_RECT ff_scissor_;
|
||||
@@ -776,6 +773,9 @@ class D3D12CommandProcessor final : public CommandProcessor {
|
||||
// scratch memexport data
|
||||
MemExportRange memexport_ranges_[512];
|
||||
uint32_t memexport_range_count_ = 0;
|
||||
|
||||
std::atomic<bool> pix_capture_requested_ = false;
|
||||
bool pix_capturing_;
|
||||
};
|
||||
|
||||
} // namespace d3d12
|
||||
|
||||
Reference in New Issue
Block a user