Update xgl from commit 454d7b37

Update Khronos Vulkan Headers to 1.3.277
Expose the extension VK_EXT_graphics_pipeline_library
Expose the extension VK_EXT_primitives_generated_query
Expose the extension VK_KHR_calibrated_timestamps
Expose the extension VK_KHR_index_type_uint8
Expose the extension VK_KHR_line_rasterization
Expose the extension VK_KHR_load_store_op_none
Expose the extension VK_KHR_maintenance6
Expose the extension VK_KHR_shader_expect_assume
Expose the extension VK_KHR_shader_float_controls2
Expose the extension VK_KHR_shader_quad_control
Expose the extension VK_KHR_shader_subgroup_rotate
Expose the extension VK_KHR_vertex_attribute_divisor
VK_KHR_dynamic_rendering_local_read  implementation
Add ClientGetTemporaryGpuMemory
Add code to reset dirty graphics states.
Add GPURT gpudebug Flags
Add llpc version component before gpurt library cmake configuration.
Add settings EnableEarlyPairCompression and TrianglePairingSearchRadius
Add supported op bits for KHR_subgroup_rotate extension
Bump Gpurt Version to 43
Update PAL Version in XGL 846
Check and create directory automatically for rt, rgp and renderpass
Clean up warning in vk_ray_tracing_pipeline.cpp
Cleanup PalCmdResetEvent and PalCmdSetEvent
Cleanup UpdateRayTracingFunctionNames
Collect info on more calls for RGP SQTT
Exclude collecting info on certain events for RGP SQTT
Force flatten thread group for continuations
Memory object map/unmap optimization for Detroit become human
Offset TLAS instance address.
Re-enable the capability flag transformFeedbackRasterizationStreamSelect
Refine Graphics Pipeline Library object memory footprint
Remove PipelineOptions::enableScratchAccessBoundsChecks
Specific internal hash for color export shader
Fix Ray Tracing Trace Playback Issues for Ray History Data Dump

cmake/XglOverrides.cmake

@@ -106,39 +106,9 @@ macro(xgl_overrides_pal)
 
     set(PAL_BUILD_GPUOPEN ${ICD_GPUOPEN_DEVMODE_BUILD} CACHE BOOL "${PROJECT_NAME} override." FORCE)
 
-    set(PAL_BUILD_NAVI12 ${XGL_BUILD_NAVI12} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_NAVI14 ON CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_GFX103 ${XGL_BUILD_GFX103} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_NAVI21 ON CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_NAVI22 ON CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_NAVI23 ON CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_NAVI24 ON CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_REMBRANDT ${XGL_BUILD_REMBRANDT} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_RAPHAEL ${XGL_BUILD_RAPHAEL} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-    set(PAL_BUILD_MENDOCINO ${XGL_BUILD_MENDOCINO} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-
-#if VKI_BUILD_NAVI31
-    set(PAL_BUILD_NAVI31 ${XGL_BUILD_NAVI31} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-#endif
-
-#if VKI_BUILD_NAVI32
-    set(PAL_BUILD_NAVI32 ${XGL_BUILD_NAVI32} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-#endif
-
-#if VKI_BUILD_NAVI33
-    set(PAL_BUILD_NAVI33 ${XGL_BUILD_NAVI33} CACHE BOOL "${PROJECT_NAME} override." FORCE)
-#endif
-
-    set(PAL_BUILD_PHOENIX1 ${XGL_BUILD_PHOENIX1} CACHE BOOL "${PROJECT_NAME} override." FORCE)
+    if(XGL_BUILD_NAVI31 OR XGL_BUILD_NAVI32 OR XGL_BUILD_NAVI33 OR XGL_BUILD_PHOENIX1)
+        set(PAL_BUILD_GFX11 1 CACHE BOOL "${PROJECT_NAME} override." FORCE)
+    endif()
 
     # Wayland
     set(PAL_BUILD_WAYLAND ${BUILD_WAYLAND_SUPPORT} CACHE BOOL "Build PAL with Wayland support" FORCE)

cmake/XglVersions.cmake

@@ -28,7 +28,7 @@ include_guard()
 # This will become the value of PAL_CLIENT_INTERFACE_MAJOR_VERSION.  It describes the version of the PAL interface
 # that the ICD supports.  PAL uses this value to enable backwards-compatibility for older interface versions.
 # It must be updated on each PAL promotion after handling all of the interface changes described in palLib.h.
-set(ICD_PAL_CLIENT_MAJOR_VERSION "844")
+set(ICD_PAL_CLIENT_MAJOR_VERSION "846")
 
 # This will become the value of GPUOPEN_CLIENT_INTERFACE_MAJOR_VERSION if ICD_GPUOPEN_DEVMODE_BUILD=1.
 # It describes the interface version of the gpuopen shared module (part of PAL) that the ICD supports.
@@ -37,7 +37,7 @@ set(ICD_GPUOPEN_CLIENT_MAJOR_VERSION "42")
 #if VKI_RAY_TRACING
 # This will become the value of GPURT_CLIENT_INTERFACE_MAJOR_VERSION if VKI_RAY_TRACING=1.
 # It describes the interface version of the GpuRT shared module that the ICD supports.
-set(ICD_GPURT_CLIENT_MAJOR_VERSION "41")
+set(ICD_GPURT_CLIENT_MAJOR_VERSION "43")
 #endif
 
 # This will become the value of LLPC_CLIENT_INTERFACE_MAJOR_VERSION if ICD_BUILD_LLPC=1.

icd/CMakeLists.txt

@@ -68,6 +68,10 @@ endif()
 # PAL
 add_subdirectory(${XGL_PAL_PATH} ${PROJECT_BINARY_DIR}/pal)
 
+# Add LLPC version before gpurt
+include("${XGL_VKGC_PATH}/cmake/llpc_version.cmake")
+add_llpc_version_projects()
+
 #if VKI_RAY_TRACING
 # GPURT
 if (VKI_RAY_TRACING)

icd/Loader/LunarG/Lnx/amd-icd.json

@@ -2,13 +2,13 @@
   "file_format_version": "1.0.0",
   "ICD": {
     "library_path": "@AMDVLK_INSTALL_PATH@/amdvlk@[email protected]",
-    "api_version": "1.3.273"
+    "api_version": "1.3.277"
   },
   "layer": {
     "name": "VK_LAYER_AMD_switchable_graphics_@ISABITS@",
     "type": "GLOBAL",
     "library_path": "@AMDVLK_INSTALL_PATH@/amdvlk@[email protected]",
-    "api_version": "1.3.273",
+    "api_version": "1.3.277",
     "implementation_version": "1",
     "description": "AMD switchable graphics layer",
     "functions": {

icd/api/app_profile.cpp

@@ -725,12 +725,24 @@ constexpr AppProfilePatternEntry AppNameDetroit =
     "detroitbecomehuman"
 };
 
-constexpr AppProfilePatternEntry AppEngineQuanticDream =
+constexpr AppProfilePatternEntry AppEngineQuanticDream
 {
     PatternEngineNameLower,
     "quantic dream engine"
 };
 
+constexpr AppProfilePatternEntry AppNameEnshrouded =
+{
+    PatternAppNameLower,
+    "enshrouded"
+};
+
+constexpr AppProfilePatternEntry AppEngineHolistic =
+{
+    PatternEngineNameLower,
+    "holistic"
+};
+
 constexpr AppProfilePatternEntry PatternEnd = {};
 
 // This is a table of patterns.  The first matching pattern in this table will be returned.
@@ -1465,6 +1477,23 @@ AppProfilePattern AppPatternTable[] =
         }
     },
 
+    {
+        AppProfile::Enshrouded,
+        {
+            AppNameEnshrouded,
+            AppEngineHolistic,
+            PatternEnd
+        }
+    },
+
+    {
+        AppProfile::HolisticEngine,
+        {
+            AppEngineHolistic,
+            PatternEnd
+        }
+    },
+
     {
         AppProfile::Zink,
         {

icd/api/barrier_policy.cpp

@@ -112,6 +112,7 @@ class LayoutUsageHelper
 
         InitEntry(VK_IMAGE_LAYOUT_ATTACHMENT_FEEDBACK_LOOP_OPTIMAL_EXT,
                   Pal::LayoutShaderRead | Pal::LayoutShaderWrite);
+
     }
 
     // Return layout usage index corresponding to the specified layout.

icd/api/compiler_solution.cpp

@@ -287,8 +287,9 @@ void CompilerSolution::SetRayTracingFunctionName(
 // =====================================================================================================================
 // Parse and update RayTracingFunctionName of all funcTypes
 void CompilerSolution::UpdateRayTracingFunctionNames(
-    const Device*   pDevice,
-    Vkgc::RtState*  pRtState)
+    const Device*          pDevice,
+    Pal::RayTracingIpLevel rayTracingIp,
+    Vkgc::RtState*         pRtState)
 {
     VkResult result              = VK_ERROR_UNKNOWN;
     GpuRt::IDevice* pGpurtDevice = pDevice->RayTrace()->GpuRt(DefaultDeviceIndex);
@@ -298,28 +299,6 @@ void CompilerSolution::UpdateRayTracingFunctionNames(
         const RuntimeSettings& settings = pDevice->VkPhysicalDevice(DefaultDeviceIndex)->GetRuntimeSettings();
         auto pTable                     = &pRtState->gpurtFuncTable;
 
-        Pal::RayTracingIpLevel rayTracingIp =
-            pDevice->VkPhysicalDevice(DefaultDeviceIndex)->PalProperties().gfxipProperties.rayTracingIp;
-
-        // Optionally, override RTIP level based on software emulation setting
-        switch (settings.emulatedRtIpLevel)
-        {
-        case EmulatedRtIpLevelNone:
-            break;
-        case HardwareRtIpLevel1_1:
-        case EmulatedRtIpLevel1_1:
-            rayTracingIp = Pal::RayTracingIpLevel::RtIp1_1;
-            break;
-#if VKI_BUILD_GFX11
-        case EmulatedRtIpLevel2_0:
-            rayTracingIp = Pal::RayTracingIpLevel::RtIp2_0;
-            break;
-#endif
-        default:
-            VK_ASSERT(false);
-            break;
-        }
-
         GpuRt::EntryFunctionTable entryFuncTable = {};
         result = PalToVkResult(pGpurtDevice->QueryRayTracingEntryFunctionTable(rayTracingIp, &entryFuncTable));
 

icd/api/compiler_solution_llpc.cpp

@@ -132,8 +132,6 @@ VkResult CompilerSolutionLlpc::BuildShaderModule(
     VkShaderModuleCreateFlags    flags,
     VkShaderModuleCreateFlags    internalShaderFlags,
     const Vkgc::BinaryData&      shaderBinary,
-    const bool                   adaptForFastLink,
-    bool                         isInternal,
     ShaderModuleHandle*          pShaderModule,
     const PipelineOptimizerKey&  profileKey)
 {
@@ -409,6 +407,7 @@ VkResult CompilerSolutionLlpc::CreateColorExportBinary(
     auto pPipelineBuildInfo                 = &pCreateInfo->pipelineInfo;
     pPipelineBuildInfo->pInstance           = pInstance;
     pPipelineBuildInfo->pfnOutputAlloc      = AllocateShaderOutput;
+    pPipelineBuildInfo->pipelineApiHash     = pCreateInfo->libraryHash[GraphicsLibraryColorExport];
 
     VK_ASSERT(pCreateInfo->pBinaryMetadata->pFsOutputMetaData != nullptr);
     Vkgc::Result llpcResult = m_pLlpc->BuildColorExportShader(pPipelineBuildInfo,
@@ -439,15 +438,14 @@ VkResult CompilerSolutionLlpc::CreateColorExportBinary(
 VkResult CompilerSolutionLlpc::CreateGraphicsShaderBinary(
     const Device*                     pDevice,
     PipelineCache*                    pPipelineCache,
-    const ShaderStage                 stage,
+    GraphicsLibraryType               gplType,
     GraphicsPipelineBinaryCreateInfo* pCreateInfo,
     void*                             pPipelineDumpHandle,
-    ShaderModuleHandle*               pShaderModule)
+    GplModuleState*                   pModuleState)
 {
     VkResult result = VK_SUCCESS;
     Util::MetroHash::Hash cacheId = {};
     Vkgc::BinaryData shaderLibraryBinary = {};
-    GraphicsLibraryType gplType = GetGraphicsLibraryType(stage);
 
     bool hitCache = false;
     bool hitAppCache = false;
@@ -484,7 +482,7 @@ VkResult CompilerSolutionLlpc::CreateGraphicsShaderBinary(
             if (pPipelineCache != nullptr)
             {
                 // Update the shader feedback
-                PipelineCreationFeedback* pStageFeedBack = &pCreateInfo->stageFeedback[stage];
+                PipelineCreationFeedback* pStageFeedBack = &pCreateInfo->stageFeedback[pModuleState->stage];
                 pStageFeedBack->feedbackValid = true;
                 pStageFeedBack->hitApplicationCache = hitAppCache;
             }
@@ -497,11 +495,11 @@ VkResult CompilerSolutionLlpc::CreateGraphicsShaderBinary(
             Vkgc::UnlinkedShaderStage unlinkedStage = UnlinkedStageCount;
 
             // Belong to vertexProcess stage before fragment
-            if (stage < ShaderStage::ShaderStageFragment)
+            if (gplType == GraphicsLibraryPreRaster)
             {
                 unlinkedStage = UnlinkedShaderStage::UnlinkedStageVertexProcess;
             }
-            else if (stage == ShaderStage::ShaderStageFragment)
+            else if (gplType == GraphicsLibraryFragment)
             {
                 unlinkedStage = UnlinkedShaderStage::UnlinkedStageFragment;
             }
@@ -548,21 +546,41 @@ VkResult CompilerSolutionLlpc::CreateGraphicsShaderBinary(
                     &shaderLibraryBinary);
             }
 
-            pShaderModule->elfPackage = shaderLibraryBinary;
-            pCreateInfo->earlyElfPackage[gplType] = pShaderModule->elfPackage;
+            pModuleState->elfPackage                  = shaderLibraryBinary;
+            pModuleState->pFsOutputMetaData           = nullptr;
+            pCreateInfo->earlyElfPackage[gplType]     = pModuleState->elfPackage;
             pCreateInfo->earlyElfPackageHash[gplType] = cacheId;
 
-            if (stage == ShaderStage::ShaderStageFragment)
+            if (gplType == GraphicsLibraryFragment)
             {
                 if (shaderLibraryBinary.pCode != nullptr)
                 {
                     const auto* pShaderLibraryHeader =
                         reinterpret_cast<const ShaderLibraryBlobHeader*>(shaderLibraryBinary.pCode);
 
-                    pCreateInfo->pBinaryMetadata->fsOutputMetaDataSize = pShaderLibraryHeader->fragMetaLength;
-                    pCreateInfo->pBinaryMetadata->pFsOutputMetaData    = (pShaderLibraryHeader->fragMetaLength > 0) ?
-                        const_cast<void*>(VoidPtrInc(pShaderLibraryHeader + 1, pShaderLibraryHeader->binaryLength)) :
-                        nullptr;
+                    pCreateInfo->pBinaryMetadata->fsOutputMetaDataSize = 0;
+                    pCreateInfo->pBinaryMetadata->pFsOutputMetaData    = nullptr;
+                    if (pShaderLibraryHeader->fragMetaLength > 0)
+                    {
+                        void* pFsOutputMetaData = m_pPhysicalDevice->Manager()->VkInstance()->AllocMem(
+                            pShaderLibraryHeader->fragMetaLength,
+                            VK_DEFAULT_MEM_ALIGN,
+                            VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+
+                        if (pFsOutputMetaData != nullptr)
+                        {
+                            memcpy(pFsOutputMetaData,
+                                VoidPtrInc(pShaderLibraryHeader + 1, pShaderLibraryHeader->binaryLength),
+                                pShaderLibraryHeader->fragMetaLength);
+                            pCreateInfo->pBinaryMetadata->fsOutputMetaDataSize = pShaderLibraryHeader->fragMetaLength;
+                            pCreateInfo->pBinaryMetadata->pFsOutputMetaData    = pFsOutputMetaData;
+                            pModuleState->pFsOutputMetaData                    = pFsOutputMetaData;
+                        }
+                        else
+                        {
+                            result = VK_ERROR_OUT_OF_HOST_MEMORY;
+                        }
+                    }
                 }
             }
 
@@ -1164,7 +1182,8 @@ void CompilerSolutionLlpc::UpdateStageCreationFeedback(
 bool CompilerSolutionLlpc::IsGplFastLinkCompatible(
     const Device*                           pDevice,
     uint32_t                                deviceIdx,
-    const GraphicsPipelineBinaryCreateInfo* pCreateInfo)
+    const GraphicsPipelineBinaryCreateInfo* pCreateInfo,
+    const GraphicsPipelineLibraryInfo&      libInfo)
 {
     return (pCreateInfo->pipelineInfo.iaState.enableMultiView == false) &&
            ((pCreateInfo->flags & VK_PIPELINE_CREATE_CAPTURE_INTERNAL_REPRESENTATIONS_BIT_KHR) == 0);

icd/api/devmode/devmode_mgr.cpp

@@ -638,6 +638,11 @@ Pal::Result DevModeMgr::TraceEndingToIdleStep(TraceState* pState)
 
                 if (settings.devModeEnableRgpTraceDump)
                 {
+                    // The min length of both win and linux is 260, so we use this number.
+                    char pathBuffer[260];
+                    Util::SplitFilePath(settings.devModeRgpTraceDumpFile, pathBuffer, 260, nullptr, 0);
+                    Util::MkDirRecursively(pathBuffer);
+
                     Util::File dumpFile;
                     if (dumpFile.Open(settings.devModeRgpTraceDumpFile, Util::FileAccessMode::FileAccessWrite | Util::FileAccessMode::FileAccessBinary) == Util::Result::Success)
                     {

icd/api/entry.cpp

@@ -121,6 +121,61 @@ VKAPI_ATTR void VKAPI_CALL vkCmdBindIndexBuffer2KHR(
         indexType);
 }
 
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdBindDescriptorSets2KHR(
+    VkCommandBuffer                             commandBuffer,
+    const VkBindDescriptorSetsInfoKHR*          pBindDescriptorSetsInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->VkDevice()->GetEntryPoints().vkCmdBindDescriptorSets2KHR(
+        commandBuffer,
+        pBindDescriptorSetsInfo);
+}
+
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdPushConstants2KHR(
+    VkCommandBuffer                             commandBuffer,
+    const VkPushConstantsInfoKHR*               pPushConstantsInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->PushConstants2KHR(pPushConstantsInfo);
+}
+
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdPushDescriptorSet2KHR(
+    VkCommandBuffer                             commandBuffer,
+    const VkPushDescriptorSetInfoKHR*           pPushDescriptorSetInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->VkDevice()->GetEntryPoints().vkCmdPushDescriptorSet2KHR(
+        commandBuffer,
+        pPushDescriptorSetInfo);
+}
+
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdPushDescriptorSetWithTemplate2KHR(
+    VkCommandBuffer                             commandBuffer,
+    const VkPushDescriptorSetWithTemplateInfoKHR* pPushDescriptorSetWithTemplateInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->VkDevice()->GetEntryPoints().vkCmdPushDescriptorSetWithTemplate2KHR(
+        commandBuffer,
+        pPushDescriptorSetWithTemplateInfo);
+}
+
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdSetDescriptorBufferOffsets2EXT(
+    VkCommandBuffer                             commandBuffer,
+    const VkSetDescriptorBufferOffsetsInfoEXT*  pSetDescriptorBufferOffsetsInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->SetDescriptorBufferOffsets2EXT(pSetDescriptorBufferOffsetsInfo);
+}
+
+// =====================================================================================================================
+VKAPI_ATTR void VKAPI_CALL vkCmdBindDescriptorBufferEmbeddedSamplers2EXT(
+    VkCommandBuffer                             commandBuffer,
+    const VkBindDescriptorBufferEmbeddedSamplersInfoEXT* pBindDescriptorBufferEmbeddedSamplersInfo)
+{
+    ApiCmdBuffer::ObjectFromHandle(commandBuffer)->BindDescriptorBufferEmbeddedSamplers2EXT(
+        pBindDescriptorBufferEmbeddedSamplersInfo);
+}
+
 // =====================================================================================================================
 VKAPI_ATTR void VKAPI_CALL vkCmdBindVertexBuffers(
     VkCommandBuffer                             cmdBuffer,