diff options
| author | Markus Wick <markus@selfnet.de> | 2021-04-07 13:57:49 +0200 |
|---|---|---|
| committer | Markus Wick <markus@selfnet.de> | 2021-04-07 22:38:52 +0200 |
| commit | e6fb49fa4bb2864702abcefc14f6bb62eaba7a7e (patch) | |
| tree | 6cad1f7c35b6f9c539fbb9ad0fa2a1359a543a97 /src/video_core/gpu_thread.cpp | |
| parent | 5145133a604f626c05f832465ac22019b003c32a (diff) | |
video_core/gpu_thread: Keep the write lock for allocating the fence.
Else the fence might get submited out-of-order into the queue, which makes testing them pointless.
Overhead should be tiny as the mutex is just moved from the queue to the writing code.
Diffstat (limited to 'src/video_core/gpu_thread.cpp')
| -rw-r--r-- | src/video_core/gpu_thread.cpp | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 6b8f06f780..9488bf5444 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -151,11 +151,13 @@ void ThreadManager::OnCommandListEnd() { } u64 ThreadManager::PushCommand(CommandData&& command_data) { + std::unique_lock lk(state.write_lock); const u64 fence{++state.last_fence}; state.queue.Push(CommandDataContainer(std::move(command_data), fence)); if (!is_async) { // In synchronous GPU mode, block the caller until the command has executed + lk.unlock(); WaitIdle(); } |