Merge pull request #137 from shadps4-emu/video_core/multi_submits

amdgpu: multiple GFX submits

src/core/libraries/gnmdriver/gnmdriver.cpp

@@ -1359,13 +1359,13 @@ static inline s32 PatchFlipRequest(u32* cmdbuf, u32 size, u32 vo_handle, u32 buf
     return ORBIS_OK;
 }
 
-s32 PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffers(u32 count, void* dcb_gpu_addrs[],
+s32 PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffers(u32 count, u32* dcb_gpu_addrs[],
-                                                   u32* dcb_sizes_in_bytes, void* ccb_gpu_addrs[],
+                                                   u32* dcb_sizes_in_bytes, u32* ccb_gpu_addrs[],
                                                    u32* ccb_sizes_in_bytes, u32 vo_handle,
                                                    u32 buf_idx, u32 flip_mode, u32 flip_arg) {
     LOG_INFO(Lib_GnmDriver, "called [buf = {}]", buf_idx);
 
-    auto* cmdbuf = reinterpret_cast<u32*>(dcb_gpu_addrs[count - 1]);
+    auto* cmdbuf = dcb_gpu_addrs[count - 1];
     const auto size_dw = dcb_sizes_in_bytes[count - 1] / 4;
 
     const s32 patch_result =
@@ -1374,7 +1374,8 @@ s32 PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffers(u32 count, void* dcb_gpu_addr
         return patch_result;
     }
 
-    return sceGnmSubmitCommandBuffers(count, dcb_gpu_addrs, dcb_sizes_in_bytes, ccb_gpu_addrs,
+    return sceGnmSubmitCommandBuffers(count, const_cast<const u32**>(dcb_gpu_addrs),
+                                      dcb_sizes_in_bytes, const_cast<const u32**>(ccb_gpu_addrs),
                                       ccb_sizes_in_bytes);
 }
 
@@ -1383,11 +1384,10 @@ int PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffersForWorkload() {
     return ORBIS_OK;
 }
 
-s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, void* dcb_gpu_addrs[],
+s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, const u32* dcb_gpu_addrs[],
-                                            u32* dcb_sizes_in_bytes, void* ccb_gpu_addrs[],
+                                            u32* dcb_sizes_in_bytes, const u32* ccb_gpu_addrs[],
                                             u32* ccb_sizes_in_bytes) {
     LOG_INFO(Lib_GnmDriver, "called");
-    ASSERT_MSG(count == 1, "Multiple command buffer submission is unsupported!");
 
     if (!dcb_gpu_addrs || !dcb_sizes_in_bytes) {
         LOG_ERROR(Lib_GnmDriver, "dcbGpuAddrs and dcbSizesInBytes must not be NULL");
@@ -1411,7 +1411,12 @@ s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, void* dcb_gpu_addrs[],
         }
     }
 
-    liverpool->Submit(reinterpret_cast<u32*>(dcb_gpu_addrs[0]), dcb_sizes_in_bytes[0]);
+    for (auto cbpair = 0u; cbpair < count; ++cbpair) {
+        const auto* ccb = ccb_gpu_addrs ? ccb_gpu_addrs[cbpair] : nullptr;
+        const auto ccb_size = ccb_sizes_in_bytes ? ccb_sizes_in_bytes[cbpair] : 0;
+
+        liverpool->SubmitGfx({dcb_gpu_addrs[cbpair], dcb_sizes_in_bytes[cbpair]}, {ccb, ccb_size});
+    }
 
     return ORBIS_OK;
 }

src/core/libraries/gnmdriver/gnmdriver.h

@@ -194,13 +194,13 @@ int PS4_SYSV_ABI sceGnmSqttStopTrace();
 int PS4_SYSV_ABI sceGnmSqttSwitchTraceBuffer();
 int PS4_SYSV_ABI sceGnmSqttSwitchTraceBuffer2();
 int PS4_SYSV_ABI sceGnmSqttWaitForEvent();
-s32 PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffers(u32 count, void* dcb_gpu_addrs[],
+s32 PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffers(u32 count, u32* dcb_gpu_addrs[],
-                                                   u32* dcb_sizes_in_bytes, void* ccb_gpu_addrs[],
+                                                   u32* dcb_sizes_in_bytes, u32* ccb_gpu_addrs[],
                                                    u32* ccb_sizes_in_bytes, u32 vo_handle,
                                                    u32 buf_idx, u32 flip_mode, u32 flip_arg);
 int PS4_SYSV_ABI sceGnmSubmitAndFlipCommandBuffersForWorkload();
-s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, void* dcb_gpu_addrs[],
+s32 PS4_SYSV_ABI sceGnmSubmitCommandBuffers(u32 count, const u32* dcb_gpu_addrs[],
-                                            u32* dcb_sizes_in_bytes, void* ccb_gpu_addrs[],
+                                            u32* dcb_sizes_in_bytes, const u32* ccb_gpu_addrs[],
                                             u32* ccb_sizes_in_bytes);
 int PS4_SYSV_ABI sceGnmSubmitCommandBuffersForWorkload();
 int PS4_SYSV_ABI sceGnmSubmitDone();

src/video_core/amdgpu/liverpool.cpp

@@ -9,16 +9,50 @@
 
 namespace AmdGpu {
 
-Liverpool::Liverpool() = default;
+Liverpool::Liverpool() {
+    process_thread = std::jthread{std::bind_front(&Liverpool::Process, this)};
+}
 
-void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
+Liverpool::~Liverpool() {
+    process_thread.request_stop();
+    cv_submit.notify_one();
+}
+
+void Liverpool::Process(std::stop_token stoken) {
+    while (!stoken.stop_requested()) {
+        std::span<const u32> dcb{};
+        {
+            std::unique_lock lock{m_ring_access};
+            cv_submit.wait(lock, stoken, [&]() { return !gfx_ring.empty(); });
+
+            if (stoken.stop_requested()) {
+                break;
+            }
+
+            dcb = gfx_ring.front();
+            gfx_ring.pop();
+        }
+
+        ASSERT_MSG(dcb.size() != 0, "Empty command list received");
+        ProcessCmdList(dcb.data(), dcb.size());
+
+        cv_complete.notify_all();
+    }
+}
+
+void Liverpool::WaitGpuIdle() {
+    std::unique_lock lock{m_ring_access};
+    cv_complete.wait(lock, [this]() { return gfx_ring.empty(); });
+}
+
+void Liverpool::ProcessCmdList(const u32* cmdbuf, u32 size_in_bytes) {
     Common::SetCurrentThreadName("CommandProcessor_Gfx");
 
-    auto* header = reinterpret_cast<PM4Header*>(cmdbuf);
+    auto* header = reinterpret_cast<const PM4Header*>(cmdbuf);
     u32 processed_cmd_size = 0;
 
     while (processed_cmd_size < size_in_bytes) {
-        PM4Header* next_header{};
+        const PM4Header* next_header{};
         const u32 type = header->type;
         switch (type) {
         case 3: {
@@ -26,7 +60,7 @@ void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
             const u32 count = header->type3.NumWords();
             switch (opcode) {
             case PM4ItOpcode::Nop: {
-                const auto* nop = reinterpret_cast<PM4CmdNop*>(header);
+                const auto* nop = reinterpret_cast<const PM4CmdNop*>(header);
                 if (nop->header.count.Value() == 0) {
                     break;
                 }
@@ -44,30 +78,30 @@ void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
                 break;
             }
             case PM4ItOpcode::SetContextReg: {
-                const auto* set_data = reinterpret_cast<PM4CmdSetData*>(header);
+                const auto* set_data = reinterpret_cast<const PM4CmdSetData*>(header);
                 std::memcpy(&regs.reg_array[ContextRegWordOffset + set_data->reg_offset],
                             header + 2, (count - 1) * sizeof(u32));
                 break;
             }
             case PM4ItOpcode::SetShReg: {
-                const auto* set_data = reinterpret_cast<PM4CmdSetData*>(header);
+                const auto* set_data = reinterpret_cast<const PM4CmdSetData*>(header);
                 std::memcpy(&regs.reg_array[ShRegWordOffset + set_data->reg_offset], header + 2,
                             (count - 1) * sizeof(u32));
                 break;
             }
             case PM4ItOpcode::SetUconfigReg: {
-                const auto* set_data = reinterpret_cast<PM4CmdSetData*>(header);
+                const auto* set_data = reinterpret_cast<const PM4CmdSetData*>(header);
                 std::memcpy(&regs.reg_array[UconfigRegWordOffset + set_data->reg_offset],
                             header + 2, (count - 1) * sizeof(u32));
                 break;
             }
             case PM4ItOpcode::IndexType: {
-                const auto* index_type = reinterpret_cast<PM4CmdDrawIndexType*>(header);
+                const auto* index_type = reinterpret_cast<const PM4CmdDrawIndexType*>(header);
                 regs.index_buffer_type.raw = index_type->raw;
                 break;
             }
             case PM4ItOpcode::DrawIndex2: {
-                const auto* draw_index = reinterpret_cast<PM4CmdDrawIndex2*>(header);
+                const auto* draw_index = reinterpret_cast<const PM4CmdDrawIndex2*>(header);
                 regs.max_index_size = draw_index->max_size;
                 regs.index_base_address.base_addr_lo = draw_index->index_base_lo;
                 regs.index_base_address.base_addr_hi.Assign(draw_index->index_base_hi);
@@ -77,7 +111,7 @@ void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
                 break;
             }
             case PM4ItOpcode::DrawIndexAuto: {
-                const auto* draw_index = reinterpret_cast<PM4CmdDrawIndexAuto*>(header);
+                const auto* draw_index = reinterpret_cast<const PM4CmdDrawIndexAuto*>(header);
                 regs.num_indices = draw_index->index_count;
                 regs.draw_initiator = draw_index->draw_initiator;
                 // rasterizer->DrawIndex();
@@ -88,21 +122,21 @@ void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
                 break;
             }
             case PM4ItOpcode::EventWriteEos: {
-                const auto* event_eos = reinterpret_cast<PM4CmdEventWriteEos*>(header);
+                const auto* event_eos = reinterpret_cast<const PM4CmdEventWriteEos*>(header);
                 event_eos->SignalFence();
                 break;
             }
             case PM4ItOpcode::EventWriteEop: {
-                const auto* event_eop = reinterpret_cast<PM4CmdEventWriteEop*>(header);
+                const auto* event_eop = reinterpret_cast<const PM4CmdEventWriteEop*>(header);
                 event_eop->SignalFence();
                 break;
             }
             case PM4ItOpcode::DmaData: {
-                const auto* dma_data = reinterpret_cast<PM4DmaData*>(header);
+                const auto* dma_data = reinterpret_cast<const PM4DmaData*>(header);
                 break;
             }
             case PM4ItOpcode::WriteData: {
-                const auto* write_data = reinterpret_cast<PM4CmdWriteData*>(header);
+                const auto* write_data = reinterpret_cast<const PM4CmdWriteData*>(header);
                 ASSERT(write_data->dst_sel.Value() == 2 || write_data->dst_sel.Value() == 5);
                 const u32 data_size = (header->type3.count.Value() - 2) * 4;
                 if (!write_data->wr_one_addr.Value()) {
@@ -117,7 +151,7 @@ void Liverpool::ProcessCmdList(u32* cmdbuf, u32 size_in_bytes) {
                 break;
             }
             case PM4ItOpcode::WaitRegMem: {
-                const auto* wait_reg_mem = reinterpret_cast<PM4CmdWaitRegMem*>(header);
+                const auto* wait_reg_mem = reinterpret_cast<const PM4CmdWaitRegMem*>(header);
                 ASSERT(wait_reg_mem->engine.Value() == PM4CmdWaitRegMem::Engine::Me);
                 while (!wait_reg_mem->Test()) {
                     using namespace std::chrono_literals;

src/video_core/amdgpu/liverpool.h

@@ -11,6 +11,9 @@
 #include <condition_variable>
 #include <functional>
 #include <future>
+#include <span>
+#include <thread>
+#include <queue>
 
 namespace AmdGpu {
 
@@ -614,23 +617,41 @@ struct Liverpool {
 
 public:
     Liverpool();
+    ~Liverpool();
 
-    void Submit(u32* cmdbuf, u32 size_in_bytes) {
+    void SubmitGfx(std::span<const u32> dcb, std::span<const u32> ccb) {
-        ASSERT_MSG(!cp.valid(), "Trying to submit while previous submission is pending");
+        if (submission_lock) {
-        cp = std::async(&Liverpool::ProcessCmdList, this, cmdbuf, size_in_bytes);
+            WaitGpuIdle();
+
+            // Suspend logic goes here
+
+            submission_lock = false;
+        }
+
+        {
+            std::scoped_lock lock{m_ring_access};
+            gfx_ring.emplace(dcb);
+
+            ASSERT_MSG(ccb.size() == 0, "CCBs are not supported yet");
+        }
+        cv_submit.notify_one();
     }
     void SubmitDone() {
-        // This is wrong as `submitDone()` should never be blocking. The behavior will be
+        submission_lock = true;
-        // reworked with mutiple queues introduction
-        if (cp.valid()) {
-            cp.get();
-        }
     }
 
 private:
-    void ProcessCmdList(u32* cmdbuf, u32 size_in_bytes);
+    void ProcessCmdList(const u32* cmdbuf, u32 size_in_bytes);
+    void Process(std::stop_token stoken);
+    void WaitGpuIdle();
 
-    std::future<void> cp{};
+    std::jthread process_thread{};
+    std::queue<std::span<const u32>> gfx_ring{};
+    std::condition_variable_any cv_submit{};
+    std::condition_variable cv_complete{};
+    std::mutex m_ring_access{};
+
+    bool submission_lock{};
 };
 
 static_assert(GFX6_3D_REG_INDEX(ps_program) == 0x2C08);