diff options
| -rw-r--r-- | src/CMakeLists.txt | 1 | ||||
| -rw-r--r-- | src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp | 30 | ||||
| -rw-r--r-- | src/core/hle/service/nvdrv/devices/nvhost_ctrl.h | 5 | ||||
| -rw-r--r-- | src/video_core/gpu.cpp | 5 |
4 files changed, 35 insertions, 6 deletions
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index 54de1dc94..3575a3cb3 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -121,6 +121,7 @@ else() if (ARCHITECTURE_x86_64) add_compile_options("-mcx16") + add_compile_options("-fwrapv") endif() if (APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL Clang) diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp index abde2a6d3..0e22d9273 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp @@ -52,6 +52,8 @@ NvResult nvhost_ctrl::Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>& return IocCtrlEventRegister(input, output); case 0x20: return IocCtrlEventUnregister(input, output); + case 0x21: + return IocCtrlEventUnregisterBatch(input, output); } break; default: @@ -249,6 +251,25 @@ NvResult nvhost_ctrl::IocCtrlEventUnregister(const std::vector<u8>& input, return FreeEvent(event_id); } +NvResult nvhost_ctrl::IocCtrlEventUnregisterBatch(const std::vector<u8>& input, + std::vector<u8>& output) { + IocCtrlEventUnregisterBatchParams params{}; + std::memcpy(¶ms, input.data(), sizeof(params)); + u64 event_mask = params.user_events; + LOG_DEBUG(Service_NVDRV, " called, event_mask: {:X}", event_mask); + + auto lock = NvEventsLock(); + while (event_mask != 0) { + const u64 event_id = std::countr_zero(event_mask); + event_mask &= ~(1ULL << event_id); + const auto result = FreeEvent(static_cast<u32>(event_id)); + if (result != NvResult::Success) { + return result; + } + } + return NvResult::Success; +} + NvResult nvhost_ctrl::IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output) { IocCtrlEventClearParams params{}; std::memcpy(¶ms, input.data(), sizeof(params)); @@ -362,10 +383,11 @@ u32 nvhost_ctrl::FindFreeNvEvent(u32 syncpoint_id) { } void nvhost_ctrl::SignalNvEvent(u32 syncpoint_id, u32 value) { - const u32 max = MaxNvEvents - std::countl_zero(events_mask); - const u32 min = std::countr_zero(events_mask); - for (u32 i = min; i < max; i++) { - auto& event = events[i]; + u64 signal_mask = events_mask; + while (signal_mask != 0) { + const u64 event_id = std::countr_zero(signal_mask); + signal_mask &= ~(1ULL << event_id); + auto& event = events[event_id]; if (event.assigned_syncpt != syncpoint_id || event.assigned_value != value) { continue; } diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h index f2fc5d047..9bd10257b 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h +++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h @@ -177,16 +177,17 @@ private: static_assert(sizeof(IocCtrlEventUnregisterParams) == 4, "IocCtrlEventUnregisterParams is incorrect size"); - struct IocCtrlEventKill { + struct IocCtrlEventUnregisterBatchParams { u64_le user_events{}; }; - static_assert(sizeof(IocCtrlEventKill) == 8, "IocCtrlEventKill is incorrect size"); + static_assert(sizeof(IocCtrlEventUnregisterBatchParams) == 8, "IocCtrlEventKill is incorrect size"); NvResult NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output); NvResult IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_allocation); NvResult IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output); NvResult IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output); + NvResult IocCtrlEventUnregisterBatch(const std::vector<u8>& input, std::vector<u8>& output); NvResult IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output); NvResult FreeEvent(u32 slot); diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 80a1c69e0..8c0ff0094 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -249,6 +249,11 @@ struct GPU::Impl { void RegisterSyncptInterrupt(u32 syncpoint_id, u32 value) { std::scoped_lock lock{sync_mutex}; + u32 current_value = syncpoints.at(syncpoint_id).load(); + if ((static_cast<s32>(current_value) - static_cast<s32>(value)) >= 0) { + TriggerCpuInterrupt(syncpoint_id, value); + return; + } auto& interrupt = syncpt_interrupts.at(syncpoint_id); bool contains = std::any_of(interrupt.begin(), interrupt.end(), [value](u32 in_value) { return in_value == value; }); |