3rdParty: rebase rt64 on latest master

Rosalie241 · Jul 8, 2024 · 9899df4 · 9899df4
1 parent 2ac0594
commit 9899df4
Show file tree

Hide file tree

Showing 36 changed files with 178 additions and 63 deletions.
diff --git a/Source/3rdParty/rt64/include/rt64_extended_gbi.h b/Source/3rdParty/rt64/include/rt64_extended_gbi.h
@@ -72,7 +72,8 @@
 #define G_EX_PUSHGEOMETRYMODE_V1        0x000029
 #define G_EX_POPGEOMETRYMODE_V1         0x00002A
 #define G_EX_SETDITHERNOISESTRENGTH_V1  0x00002B
-#define G_EX_MAX                        0x00002C
+#define G_EX_SETRDRAMEXTENDED_V1        0x00002C
+#define G_EX_MAX                        0x00002D
 
 #define G_EX_ORIGIN_NONE            0x800
 #define G_EX_ORIGIN_LEFT            0x0
@@ -508,4 +509,10 @@ typedef union {
         PARAM((value) * 1024, 16, 0) \
     )
 
+#define gEXSetRDRAMExtended(cmd, isExtended) \
+    G_EX_COMMAND1(cmd, \
+        PARAM(RT64_EXTENDED_OPCODE, 8, 24) | PARAM(G_EX_SETRDRAMEXTENDED_V1, 24, 0), \
+        PARAM(isExtended, 1, 0) \
+    )
+
 #endif // RT64_EXTENDED_GBI
diff --git a/Source/3rdParty/rt64/src/common/rt64_thread.cpp b/Source/3rdParty/rt64/src/common/rt64_thread.cpp
@@ -60,6 +60,7 @@ namespace RT64 {
         // that are available without root privileges are lower priority. Instead you can set the thread's "nice" value, which ranges
         // from -20 to 19 (lower being higher priority). However, by strict POSIX spec "nice" is meant to be per-process instead of
         // per-thread. Therefore to avoid issues in case Linux is modified to match the spec in the future, this function does nothing.
+        (void)priority;
 #   else
         static_assert(false, "Unimplemented");
 #   endif

diff --git a/Source/3rdParty/rt64/src/d3d12/rt64_d3d12.cpp b/Source/3rdParty/rt64/src/d3d12/rt64_d3d12.cpp
@@ -2966,12 +2966,9 @@ namespace RT64 {
 
             // Pick this adapter and device if it has better feature support than the current one.
             bool preferOverNothing = (adapter == nullptr) || (d3d == nullptr);
-            bool preferShaderModel = (dataShaderModel.HighestShaderModel > shaderModel);
-            bool preferRtSupport = rtSupportOption && !capabilities.raytracing;
-            bool preferRtStateUpdateSupport = rtStateUpdateSupportOption && !capabilities.raytracingStateUpdate;
             bool preferVideoMemory = adapterDesc.DedicatedVideoMemory > dedicatedVideoMemory;
             bool preferUserChoice = false;//wcsstr(adapterDesc.Description, L"AMD") != nullptr;
-            bool preferOption = preferOverNothing || preferShaderModel || preferRtSupport || preferRtStateUpdateSupport || preferVideoMemory || preferUserChoice;
+            bool preferOption = preferOverNothing || preferVideoMemory || preferUserChoice;
             if (preferOption) {
                 if (d3d != nullptr) {
                     d3d->Release();
@@ -2987,6 +2984,7 @@ namespace RT64 {
                 capabilities.raytracing = rtSupportOption;
                 capabilities.raytracingStateUpdate = rtStateUpdateSupportOption;
                 capabilities.sampleLocations = samplePositionsOption;
+                description.name = win32::Utf16ToUtf8(adapterDesc.Description);
                 dedicatedVideoMemory = adapterDesc.DedicatedVideoMemory;
 
                 if (preferUserChoice) {
@@ -3290,6 +3288,10 @@ namespace RT64 {
         return capabilities;
     }
 
+    const RenderDeviceDescription &D3D12Device::getDescription() const {
+        return description;
+    }
+
     RenderSampleCounts D3D12Device::getSampleCountsSupported(RenderFormat format) const {
         HRESULT res;
         RenderSampleCounts countsSupported = RenderSampleCount::COUNT_0;

diff --git a/Source/3rdParty/rt64/src/d3d12/rt64_d3d12.h b/Source/3rdParty/rt64/src/d3d12/rt64_d3d12.h
@@ -27,7 +27,7 @@ namespace RT64 {
     struct D3D12Texture;
 
     struct D3D12DescriptorHeapAllocator {
-        enum {
+        enum : uint32_t {
             INVALID_OFFSET = 0xFFFFFFFFU
         };
 
@@ -382,6 +382,7 @@ namespace RT64 {
         std::unique_ptr<D3D12DescriptorHeapAllocator> colorTargetHeapAllocator;
         std::unique_ptr<D3D12DescriptorHeapAllocator> depthTargetHeapAllocator;
         RenderDeviceCapabilities capabilities;
+        RenderDeviceDescription description;
 
         D3D12Device(D3D12Interface *renderInterface);
         ~D3D12Device() override;
@@ -404,6 +405,7 @@ namespace RT64 {
         void setTopLevelASBuildInfo(RenderTopLevelASBuildInfo &buildInfo, const RenderTopLevelASInstance *instances, uint32_t instanceCount, bool preferFastBuild, bool preferFastTrace) override;
         void setShaderBindingTableInfo(RenderShaderBindingTableInfo &tableInfo, const RenderShaderBindingGroups &groups, const RenderPipeline *pipeline, RenderDescriptorSet **descriptorSets, uint32_t descriptorSetCount) override;
         const RenderDeviceCapabilities &getCapabilities() const override;
+        const RenderDeviceDescription &getDescription() const override;
         RenderSampleCounts getSampleCountsSupported(RenderFormat format) const override;
         void release();
         bool isValid() const;

diff --git a/Source/3rdParty/rt64/src/gbi/rt64_gbi.cpp b/Source/3rdParty/rt64/src/gbi/rt64_gbi.cpp
@@ -5,6 +5,7 @@
 #include "rt64_gbi.h"
 
 #include <cassert>
+#include <cinttypes>
 
 #include "rt64_gbi_extended.h"
 #include "rt64_gbi_f3d.h"
@@ -171,7 +172,7 @@ namespace RT64 {
 
             // If we used an unknown value in the database, print the hash.
             if (gbiSegment.hashValue == GBI_UNKNOWN_HASH_NUMB) {
-                fprintf(stdout, "HASH: 0x%016llXULL SIZE: 0x%X\n", rdramHash, rdramHashed);
+                fprintf(stdout, "HASH: 0x%016" PRIX64 "ULL SIZE: 0x%X\n", rdramHash, rdramHashed);
             }
         };
 
@@ -290,8 +291,8 @@ namespace RT64 {
                 const uint8_t *twoPointResult = std::search(searchResult, searchResultEnd, twoPointPattern, twoPointPattern + std::size(twoPointPattern) - 1);
                 if ((f3dexResult != searchResultEnd) && (twoPointResult != searchResultEnd)) {
                     const uint32_t lastOverlay = *reinterpret_cast<const uint32_t *>(&RDRAM[dataAddress + 0x410]);
-                    const uint16_t lastOverlaySize = *reinterpret_cast<const uint16_t *>(&RDRAM[dataAddress + 0x414 ^ 2]);
-                    const uint16_t estimatedDataSize = *reinterpret_cast<const uint16_t *>(&RDRAM[dataAddress + 0x380 ^ 2]);
+                    const uint16_t lastOverlaySize = *reinterpret_cast<const uint16_t *>(&RDRAM[(dataAddress + 0x414) ^ 2]);
+                    const uint16_t estimatedDataSize = *reinterpret_cast<const uint16_t *>(&RDRAM[(dataAddress + 0x380) ^ 2]);
                     const uint32_t textSize = lastOverlay + lastOverlaySize + 1;
                     fprintf(stderr, "Detected text size is 0x%X\n", textSize);
                     fprintf(stderr, "Estimated data size is 0x%X\n", estimatedDataSize);
@@ -300,7 +301,7 @@ namespace RT64 {
                 const uint8_t *s2dexResult = std::search(searchResult, searchResultEnd, s2dexPattern, s2dexPattern + std::size(s2dexPattern) - 1);
                 if ((s2dexResult != searchResultEnd) && (twoPointResult != searchResultEnd)) {
                     const uint32_t lastOverlay = *reinterpret_cast<const uint32_t *>(&RDRAM[dataAddress + 0x31C]);
-                    const uint16_t lastOverlaySize = *reinterpret_cast<const uint16_t *>(&RDRAM[dataAddress + 0x320 ^ 2]);
+                    const uint16_t lastOverlaySize = *reinterpret_cast<const uint16_t *>(&RDRAM[(dataAddress + 0x320) ^ 2]);
                     const uint32_t textSize = lastOverlay + lastOverlaySize + 1;
                     fprintf(stderr, "Detected text size is 0x%X\n", textSize);
                 }

diff --git a/Source/3rdParty/rt64/src/gbi/rt64_gbi_extended.cpp b/Source/3rdParty/rt64/src/gbi/rt64_gbi_extended.cpp
@@ -286,14 +286,19 @@ namespace RT64 {
             state->setDitherNoiseStrength(noiseStrength / 1024.0f);
         }
 
+        void setRDRAMExtendedV1(State *state, DisplayList **dl) {
+            const uint8_t extended = (*dl)->p1(0, 1);
+            state->setExtendedRDRAM(extended);
+        }
+
         void noOpHook(State *state, DisplayList **dl) {
             uint32_t magicNumber = (*dl)->p0(0, 24);
             if (magicNumber == RT64_HOOK_MAGIC_NUMBER) {
                 uint32_t hookValue = (*dl)->p1(0, 28);
                 uint32_t hookOp = (*dl)->p1(28, 4);
                 switch (hookOp) {
                 case RT64_HOOK_OP_GETVERSION: {
-                    const uint32_t rdramAddress = state->rsp->fromSegmented(hookValue);
+                    const uint32_t rdramAddress = state->rsp->fromSegmentedMasked(hookValue);
                     uint32_t *returnRDRAM = reinterpret_cast<uint32_t *>(state->fromRDRAM(rdramAddress));
                     *returnRDRAM = G_EX_VERSION;
                     break;
@@ -317,7 +322,7 @@ namespace RT64 {
                         state->pushReturnAddress(*dl);
                     }
 
-                    const uint32_t rdramAddress = state->rsp->fromSegmented(hookValue);
+                    const uint32_t rdramAddress = state->rsp->fromSegmentedMasked(hookValue);
                     *dl = reinterpret_cast<DisplayList *>(state->fromRDRAM(rdramAddress)) - 1;
                     break;
                 }
@@ -386,6 +391,7 @@ namespace RT64 {
             Map[G_EX_PUSHGEOMETRYMODE_V1] = &pushGeometryModeV1;
             Map[G_EX_POPGEOMETRYMODE_V1] = &popGeometryModeV1;
             Map[G_EX_SETDITHERNOISESTRENGTH_V1] = &setDitherNoiseStrengthV1;
+            Map[G_EX_SETRDRAMEXTENDED_V1] = &setRDRAMExtendedV1;
             MapInitialized = true;
         }
     }

diff --git a/Source/3rdParty/rt64/src/gbi/rt64_gbi_f3d.cpp b/Source/3rdParty/rt64/src/gbi/rt64_gbi_f3d.cpp
@@ -78,7 +78,7 @@ namespace RT64 {
                 state->pushReturnAddress(*dl);
             }
 
-            const uint32_t rdramAddress = state->rsp->fromSegmented((*dl)->w1);
+            const uint32_t rdramAddress = state->rsp->fromSegmentedMasked((*dl)->w1);
             *dl = reinterpret_cast<DisplayList *>(state->fromRDRAM(rdramAddress)) - 1;
         }
 
@@ -121,7 +121,7 @@ namespace RT64 {
                 // TODO
                 break;
             case G_MW_SEGMENT:
-                state->rsp->setSegment((*dl)->p0(10, 4), (*dl)->w1 & 0x00FFFFFF);
+                state->rsp->setSegment((*dl)->p0(10, 4), (*dl)->w1);
                 break;
             case G_MW_FOG:
                 state->rsp->setFog((int16_t)((*dl)->p1(16, 16)), (int16_t)((*dl)->p1(0, 16)));

diff --git a/Source/3rdParty/rt64/src/gbi/rt64_gbi_f3dex2.cpp b/Source/3rdParty/rt64/src/gbi/rt64_gbi_f3dex2.cpp
@@ -77,7 +77,7 @@ namespace RT64 {
                 state->rsp->setClipRatio((*dl)->w1);
                 break;
             case G_MW_SEGMENT:
-                state->rsp->setSegment((*dl)->p0(2, 4), (*dl)->w1 & 0x00FFFFFF);
+                state->rsp->setSegment((*dl)->p0(2, 4), (*dl)->w1);
                 break;
             case G_MW_FOG:
                 state->rsp->setFog((int16_t)((*dl)->p1(16, 16)), (int16_t)((*dl)->p1(0, 16)));

diff --git a/Source/3rdParty/rt64/src/gbi/rt64_gbi_s2dex.cpp b/Source/3rdParty/rt64/src/gbi/rt64_gbi_s2dex.cpp
@@ -133,7 +133,7 @@ namespace RT64 {
 
             RDP *rdp = state->rdp.get();
             RSP *rsp = state->rsp.get();
-            const uint32_t rdramAddress = state->rsp->fromSegmented((*dl)->w1);
+            const uint32_t rdramAddress = state->rsp->fromSegmentedMasked((*dl)->w1);
             // TODO load this into the S2D struct buffer for a more accurate implementation in case there's ever command state bleed.
             const uObjBg *bgObject = reinterpret_cast<const uObjBg *>(state->fromRDRAM(rdramAddress));
             const uObjBg_t &bg = bgObject->bg;
@@ -398,7 +398,7 @@ namespace RT64 {
 
             RDP *rdp = state->rdp.get();
             RSP* rsp = state->rsp.get();
-            const uint32_t rdramAddress = state->rsp->fromSegmented((*dl)->w1);
+            const uint32_t rdramAddress = state->rsp->fromSegmentedMasked((*dl)->w1);
             // TODO load this into the S2D struct buffer for a more accurate implementation in case there's ever command state bleed.
             const uObjBg *bgObject = reinterpret_cast<const uObjBg *>(state->fromRDRAM(rdramAddress));
             const uObjBg_t &bg = bgObject->bg;
@@ -451,7 +451,7 @@ namespace RT64 {
 
         void readS2DStruct(State *state, uint32_t ptr, uint32_t loadSize) {
             // Convert the segmented obj pointer
-            uint32_t rdramAddress = state->rsp->fromSegmented(ptr);
+            uint32_t rdramAddress = state->rsp->fromSegmentedMasked(ptr);
             // Mask the address as the RSP DMA hardware would
             rdramAddress &= RSP_DMA_MASK;
             // Truncate the load size as the ucode does

diff --git a/Source/3rdParty/rt64/src/gui/rt64_debugger_inspector.cpp b/Source/3rdParty/rt64/src/gui/rt64_debugger_inspector.cpp
@@ -261,7 +261,7 @@ namespace RT64 {
         uint32_t totalSyncCount = 0;
         for (uint32_t f = 0; f < workload.fbPairCount; f++) {
             auto &fbPair = workload.fbPairs[f];
-            if (renderer.framebufferIndex == f) {
+            if ((uint32_t)renderer.framebufferIndex == f) {
                 renderer.framebufferAddress = renderer.framebufferDepth ? fbPair.depthImage.address : fbPair.colorImage.address;
             }
 
@@ -896,7 +896,7 @@ namespace RT64 {
                                     ImGui::PushID(l);
 
                                     if (openCall) {
-                                        bool open = (openLoadIndex == (callDesc.loadIndex + l));
+                                        bool open = ((uint32_t)openLoadIndex == (callDesc.loadIndex + l));
                                         ImGui::SetNextItemOpen(open);
                                         if (open) {
                                             ImGui::SetScrollHereY(0.0f);
@@ -984,7 +984,7 @@ namespace RT64 {
                                     const DrawCallTile &callTile = drawData.callTiles[callDesc.tileIndex + t];
                                     const LoadTile &loadTile = callTile.loadTile;
                                     if (openCall) {
-                                        bool open = (openTileIndex == (callDesc.tileIndex + t));
+                                        bool open = ((uint32_t)openTileIndex == (callDesc.tileIndex + t));
                                         ImGui::SetNextItemOpen(open);
                                         if (open) {
                                             ImGui::SetScrollHereY(0.0f);

diff --git a/Source/3rdParty/rt64/src/gui/rt64_inspector.cpp b/Source/3rdParty/rt64/src/gui/rt64_inspector.cpp
@@ -181,11 +181,12 @@ namespace RT64 {
 
 #   if defined(_WIN32) && !defined(RT64_BUILD_PLUGIN)
         ImGui_ImplWin32_NewFrame();
-#   else
+#   elif defined(RT64_BUILD_PLUGIN)
         // ????????
         // if mupen64plus == true, dont assert....
         ImGuiIO &io = ImGui::GetIO();
         io.DisplaySize = ImVec2((float)window_width, (float)window_height);
+#   else
         //assert(false && "Unimplemented.");
 #   endif
 

diff --git a/Source/3rdParty/rt64/src/hle/rt64_application.cpp b/Source/3rdParty/rt64/src/hle/rt64_application.cpp
@@ -175,6 +175,29 @@ namespace RT64 {
             return SetupResult::GraphicsDeviceNotFound;
         }
 
+        // Driver workarounds.
+        //
+        // Wireframe artifacts have been reported when using a high-precision color format on RDNA3 GPUs in D3D12. The workaround is to switch to Vulkan if this is the case.
+        bool isRDNA3 = device->getDescription().name.find("AMD Radeon RX 7") != std::string::npos;
+        bool useHDRinD3D12 = (userConfig.graphicsAPI == UserConfiguration::GraphicsAPI::D3D12) && (userConfig.internalColorFormat == UserConfiguration::InternalColorFormat::Automatic) && device->getCapabilities().preferHDR;
+        if (isRDNA3 && useHDRinD3D12) {
+            device.reset();
+            renderInterface.reset();
+            renderInterface = CreateVulkanInterface();
+            if (renderInterface == nullptr) {
+                fprintf(stderr, "Unable to initialize graphics API.\n");
+                return SetupResult::GraphicsAPINotFound;
+            }
+
+            createdGraphicsAPI = UserConfiguration::GraphicsAPI::Vulkan;
+
+            device = renderInterface->createDevice();
+            if (device == nullptr) {
+                fprintf(stderr, "Unable to find compatible graphics device.\n");
+                return SetupResult::GraphicsDeviceNotFound;
+            }
+        }
+
         // Call the init hook if one was attached.
         RenderHookInit *initHook = GetRenderHookInit();
         if (initHook != nullptr) {

diff --git a/Source/3rdParty/rt64/src/hle/rt64_application_window.cpp b/Source/3rdParty/rt64/src/hle/rt64_application_window.cpp
@@ -285,13 +285,18 @@ namespace RT64 {
         RRMode activeModeID = 0;
         for (int i = 0; i < screenResources->ncrtc; ++i) {
             XRRCrtcInfo *crtcInfo = XRRGetCrtcInfo(windowHandle.display, screenResources, screenResources->crtcs[i]);
-            if ((crtcInfo != nullptr) && (crtcInfo->mode != 0L)) {
-                activeModeID = crtcInfo->mode; 
+            if (crtcInfo != nullptr) {
+                if (crtcInfo->mode != 0L) {
+                    activeModeID = crtcInfo->mode; 
+                }
+
+                XRRFreeCrtcInfo(crtcInfo);
             }
         }
 
         if (activeModeID == 0L) {
             fprintf(stderr, "Unable to find active mode through XRRGetScreenResources and XRRGetCrtcInfo.\n");
+            XRRFreeScreenResources(screenResources);
             return;
         }
 
@@ -302,6 +307,8 @@ namespace RT64 {
                 break;
             }
         }
+
+        XRRFreeScreenResources(screenResources);
 #   endif
     }
 

diff --git a/Source/3rdParty/rt64/src/hle/rt64_light_manager.cpp b/Source/3rdParty/rt64/src/hle/rt64_light_manager.cpp
@@ -109,7 +109,7 @@ namespace RT64 {
         hlslpp::float3 sunCol = { 0.8f, 0.7f, 0.6f };
         float biggestIntensity = 0;
         int biggestDirLight = -1;
-        for (int i = 0; i < directionalLights.size(); i++) {
+        for (size_t i = 0; i < directionalLights.size(); i++) {
             if (directionalLights[i].intensityTotal > biggestIntensity) {
                 biggestDirLight = i;
                 biggestIntensity = directionalLights[i].intensityTotal;

diff --git a/Source/3rdParty/rt64/src/hle/rt64_present_queue.cpp b/Source/3rdParty/rt64/src/hle/rt64_present_queue.cpp
@@ -92,7 +92,7 @@ namespace RT64 {
         presentThread = new std::thread(&PresentQueue::threadLoop, this);
     }
 
-    void PresentQueue::threadPresent(const Present &present) {
+    void PresentQueue::threadPresent(const Present &present, bool &swapChainValid) {
         FramebufferManager &fbManager = ext.sharedResources->framebufferManager;
         RenderTargetManager &targetManager = ext.sharedResources->renderTargetManager;
         const bool usingMSAA = (targetManager.multisampling.sampleCount > 1);
@@ -264,7 +264,6 @@ namespace RT64 {
             }
         }
 
-        bool swapChainValid = true;
         for (int32_t i = 0; i < framesToPresent; i++) {
             uint32_t frameCountersNextPresented = 0;
             if ((framesToPresent > 1) && (usingMSAA || (i > 0))) {
@@ -432,6 +431,7 @@ namespace RT64 {
         bool skipPresent = false;
         uint32_t displayTimingRate = UINT32_MAX;
         const bool displayTiming = ext.device->getCapabilities().displayTiming;
+        bool swapChainValid = false;
         while (presentThreadRunning) {
             {
                 std::unique_lock<std::mutex> cursorLock(cursorMutex);
@@ -448,13 +448,13 @@ namespace RT64 {
 
             if (processCursor >= 0) {
                 std::unique_lock<std::mutex> threadLock(threadMutex);
-                const bool needsResize = ext.swapChain->needsResize();
+                const bool needsResize = ext.swapChain->needsResize() || !swapChainValid;
                 if (needsResize) {
                     ext.presentGraphicsWorker->commandList->begin();
                     ext.presentGraphicsWorker->commandList->end();
                     ext.presentGraphicsWorker->execute();
                     ext.presentGraphicsWorker->wait();
-                    ext.swapChain->resize();
+                    swapChainValid = ext.swapChain->resize();
                     swapChainFramebuffers.clear();
                     ext.sharedResources->setSwapChainSize(ext.swapChain->getWidth(), ext.swapChain->getHeight());
                 }
@@ -490,7 +490,7 @@ namespace RT64 {
                     notifyPresentId(present);
                 }
                 else {
-                    threadPresent(present);
+                    threadPresent(present, swapChainValid);
                 }
 
                 if (!present.fbOperations.empty()) {

diff --git a/Source/3rdParty/rt64/src/hle/rt64_present_queue.h b/Source/3rdParty/rt64/src/hle/rt64_present_queue.h
@@ -63,7 +63,7 @@ namespace RT64 {
         void waitForIdle();
         void waitForPresentId(uint64_t waitId);
         void setup(const External &ext);
-        void threadPresent(const Present &present);
+        void threadPresent(const Present &present, bool &swapChainValid);
         void skipInterpolation();
         void notifyPresentId(const Present &present);
         void threadAdvanceBarrier();