fix for nvdec disabled, cleanup host1x

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp

@@ -73,14 +73,15 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     offset = SpliceVectors(input, wait_checks, params.syncpoint_count, offset);
     offset = SpliceVectors(input, fences, params.fence_count, offset);
 
-    for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
-        SyncptIncr syncpt_incr = syncpt_increments[i];
-
-        fences[i].id = syncpt_incr.id;
-        fences[i].value =
-            syncpoint_manager.IncreaseSyncpoint(syncpt_incr.id, syncpt_incr.increments);
-    }
     auto& gpu = system.GPU();
+    if (gpu.UseNvdec()) {
+        for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
+            const SyncptIncr& syncpt_incr = syncpt_increments[i];
+            fences[i].id = syncpt_incr.id;
+            fences[i].value =
+                syncpoint_manager.IncreaseSyncpoint(syncpt_incr.id, syncpt_incr.increments);
+        }
+    }
     for (const auto& cmd_buffer : command_buffers) {
         auto object = nvmap_dev->GetObject(cmd_buffer.memory_id);
         ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
@@ -95,11 +96,13 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
                                       cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(cmdlist);
     }
-    fences[0].value = syncpoint_manager.IncreaseSyncpoint(fences[0].id, 1);
+    if (gpu.UseNvdec()) {
 
-    Tegra::ChCommandHeaderList cmdlist{{(4 << 28) | fences[0].id}};
-    gpu.PushCommandBuffer(cmdlist);
+        fences[0].value = syncpoint_manager.IncreaseSyncpoint(fences[0].id, 1);
 
+        Tegra::ChCommandHeaderList cmdlist{{(4 << 28) | fences[0].id}};
+        gpu.PushCommandBuffer(cmdlist);
+    }
     std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
     // Some games expect command_buffers to be written back
     offset = sizeof(IoctlSubmit);
@@ -118,7 +121,7 @@ NvResult nvhost_nvdec_common::GetSyncpoint(const std::vector<u8>& input, std::ve
     std::memcpy(&params, input.data(), sizeof(IoctlGetSyncpoint));
     LOG_DEBUG(Service_NVDRV, "called GetSyncpoint, id={}", params.param);
 
-    if (device_syncpoints[params.param] == 0) {
+    if (device_syncpoints[params.param] == 0 && system.GPU().UseNvdec()) {
         device_syncpoints[params.param] = syncpoint_manager.AllocateSyncpoint();
     }
     params.value = device_syncpoints[params.param];

src/video_core/command_classes/host1x.cpp

@@ -10,22 +10,14 @@ Tegra::Host1x::Host1x(GPU& gpu_) : gpu(gpu_) {}
 
 Tegra::Host1x::~Host1x() = default;
 
-void Tegra::Host1x::StateWrite(u32 offset, u32 arguments) {
-    u8* const state_offset = reinterpret_cast<u8*>(&state) + offset * sizeof(u32);
-    std::memcpy(state_offset, &arguments, sizeof(u32));
-}
-
-void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& arguments) {
-    StateWrite(static_cast<u32>(method), arguments[0]);
+void Tegra::Host1x::ProcessMethod(Method method, u32 argument) {
     switch (method) {
-    case Method::WaitSyncpt:
-        Execute(arguments[0]);
-        break;
     case Method::LoadSyncptPayload32:
-        syncpoint_value = arguments[0];
+        syncpoint_value = argument;
         break;
+    case Method::WaitSyncpt:
     case Method::WaitSyncpt32:
-        Execute(arguments[0]);
+        Execute(argument);
         break;
     default:
         UNIMPLEMENTED_MSG("Host1x method 0x{:X}", static_cast<u32>(method));
@@ -34,8 +26,5 @@ void Tegra::Host1x::ProcessMethod(Method method, const std::vector<u32>& argumen
 }
 
 void Tegra::Host1x::Execute(u32 data) {
-    u32 syncpointId = (data & 0xFF);
-    u32 threshold = state.load_syncpoint_payload32;
-
-    gpu.WaitFence(syncpointId, threshold);
+    gpu.WaitFence(data, syncpoint_value);
 }

src/video_core/command_classes/host1x.h

@@ -14,64 +14,23 @@ class Nvdec;
 
 class Host1x {
 public:
-    struct Host1xClassRegisters {
-        u32 incr_syncpt{};
-        u32 incr_syncpt_ctrl{};
-        u32 incr_syncpt_error{};
-        INSERT_PADDING_WORDS(5);
-        u32 wait_syncpt{};
-        u32 wait_syncpt_base{};
-        u32 wait_syncpt_incr{};
-        u32 load_syncpt_base{};
-        u32 incr_syncpt_base{};
-        u32 clear{};
-        u32 wait{};
-        u32 wait_with_interrupt{};
-        u32 delay_use{};
-        u32 tick_count_high{};
-        u32 tick_count_low{};
-        u32 tick_ctrl{};
-        INSERT_PADDING_WORDS(23);
-        u32 ind_ctrl{};
-        u32 ind_off2{};
-        u32 ind_off{};
-        std::array<u32, 31> ind_data{};
-        INSERT_PADDING_WORDS(1);
-        u32 load_syncpoint_payload32{};
-        u32 stall_ctrl{};
-        u32 wait_syncpt32{};
-        u32 wait_syncpt_base32{};
-        u32 load_syncpt_base32{};
-        u32 incr_syncpt_base32{};
-        u32 stall_count_high{};
-        u32 stall_count_low{};
-        u32 xref_ctrl{};
-        u32 channel_xref_high{};
-        u32 channel_xref_low{};
-    };
-    static_assert(sizeof(Host1xClassRegisters) == 0x164, "Host1xClassRegisters is an invalid size");
-
     enum class Method : u32 {
-        WaitSyncpt = offsetof(Host1xClassRegisters, wait_syncpt) / 4,
-        LoadSyncptPayload32 = offsetof(Host1xClassRegisters, load_syncpoint_payload32) / 4,
-        WaitSyncpt32 = offsetof(Host1xClassRegisters, wait_syncpt32) / 4,
+        WaitSyncpt = 0x8,
+        LoadSyncptPayload32 = 0x4e,
+        WaitSyncpt32 = 0x50,
     };
 
     explicit Host1x(GPU& gpu);
     ~Host1x();
 
     /// Writes the method into the state, Invoke Execute() if encountered
-    void ProcessMethod(Method method, const std::vector<u32>& arguments);
+    void ProcessMethod(Method method, u32 argument);
 
 private:
     /// For Host1x, execute is waiting on a syncpoint previously written into the state
     void Execute(u32 data);
 
-    /// Write argument into the provided offset
-    void StateWrite(u32 offset, u32 arguments);
-
     u32 syncpoint_value{};
-    Host1xClassRegisters state{};
     GPU& gpu;
 };