Update xgl from commit 0e90e78b

* Add missing Pal::CoherSampleRate flag on queue family policy mask and images used as VRS attachment * Use new cache ID for internal pipelines * Bump LLPC version to 68 * VK_EXT_primitives_generated_query-Driver Implementation * VK_EXT_frame_boundary-Expose the extension * Remove *CreateFlags typedef replace with Vk*CreateFlags2KHR after maintenance5 is released * Update PAL Version in XGL 834 * Fix performance drop observed in X-Plane with Resize Bar Enabled * Don't attempt to dump a missing binary * Expose PAL's cmdBufBatchedSubmitChainLimit setting in XGL * Fix dEQP-VK.pipeline.shader_object*.samples_1.* fail * Fix memory size for uber fetch internal data * Remove all bltmsaastate related function and parameter * Remove prebltmsaa and postbltrestoremsaa functions for cmdbuffer * Support the debug printf output in the hang state * Add Get64BitInstanceNodePtr from gpurt for compiler * Send Pal with msaastate as non-nullptr * Expose cooperative matrix * Fragment Shading Rate bug fix * Update Khronos Vulkan Headers to 1.3.269 * Fix TriangleCompressionMode setting * Fix failure in dEQP-VK.ray_tracing_pipeline.misc.* tests * Refine graphic pipeline library fast link related code * Enable RT triangle pair compression * Skip trivial task-mesh dispatch * Bump Gpurt Version to 40 * Fix dynamic rendering partial binding depth/stencil * VK_KHR_maintenance5 implementation * Add support for VK_EXT_fragment_shader_interlock * Update default navi3 RT settings, Move VGPR limit from Validate to Override and Remove Quake2 Settings * WWZ Chopped Present after Alt+Tab * Fix an invalid assertion for Pal::ShaderLibraryFunctionInfo::symbolName * Fix ANGLE test failure in dEQP-GLES31.functional.image_load_store.3d*
GPUOpen-Drivers · Nov 21, 2023 · 194a181 · 194a181
1 parent c9f6c1c
commit 194a181
Show file tree

Hide file tree

Showing 59 changed files with 1,722 additions and 966 deletions.
diff --git a/cmake/XglOverrides.cmake b/cmake/XglOverrides.cmake
@@ -91,6 +91,7 @@ endmacro()
 
 macro(xgl_overrides_pal)
 ### For PAL ###########################################################################################################
+
     set(PAL_BUILD_JEMALLOC OFF CACHE BOOL "Force jemalloc off" FORCE)
 
     set(PAL_CLIENT_INTERFACE_MAJOR_VERSION ${ICD_PAL_CLIENT_MAJOR_VERSION} CACHE STRING "${PROJECT_NAME} override." FORCE)

diff --git a/cmake/XglVersions.cmake b/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 "827")
+set(ICD_PAL_CLIENT_MAJOR_VERSION "834")
 
 # 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.
@@ -37,9 +37,9 @@ 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 "39")
+set(ICD_GPURT_CLIENT_MAJOR_VERSION "40")
 #endif
 
 # This will become the value of LLPC_CLIENT_INTERFACE_MAJOR_VERSION if ICD_BUILD_LLPC=1.
 # It describes the version of the interface version of LLPC that the ICD supports.
-set(ICD_LLPC_CLIENT_MAJOR_VERSION "66")
+set(ICD_LLPC_CLIENT_MAJOR_VERSION "68")
diff --git a/icd/Loader/LunarG/Lnx/amd-icd.json b/icd/Loader/LunarG/Lnx/amd-icd.json
@@ -2,13 +2,13 @@
   "file_format_version": "1.0.0",
   "ICD": {
     "library_path": "@AMDVLK_INSTALL_PATH@/amdvlk@ISABITS@.so",
-    "api_version": "1.3.267"
+    "api_version": "1.3.269"
   },
   "layer": {
     "name": "VK_LAYER_AMD_switchable_graphics_@ISABITS@",
     "type": "GLOBAL",
     "library_path": "@AMDVLK_INSTALL_PATH@/amdvlk@ISABITS@.so",
-    "api_version": "1.3.267",
+    "api_version": "1.3.269",
     "implementation_version": "1",
     "description": "AMD switchable graphics layer",
     "functions": {

diff --git a/icd/api/app_shader_optimizer.cpp b/icd/api/app_shader_optimizer.cpp
@@ -86,6 +86,8 @@ void ShaderOptimizer::CreateShaderOptimizerKey(
 
     if (settings.pipelineUseShaderHashAsProfileHash)
     {
+        VK_ASSERT(pModuleData != nullptr);
+
         pShaderKey->codeHash.lower = Vkgc::IPipelineDumper::GetShaderHash(pModuleData);
         pShaderKey->codeHash.upper = 0;
     }
@@ -249,7 +251,6 @@ void ShaderOptimizer::ApplyProfileToShaderCreateInfo(
                 {
                     options.pOptions->disableFMA = shaderCreate.tuningOptions.disableFMA;
                 }
-
                 if (shaderCreate.apply.workaroundStorageImageFormats)
                 {
                     options.pOptions->workaroundStorageImageFormats = true;

diff --git a/icd/api/barrier_policy.cpp b/icd/api/barrier_policy.cpp
@@ -705,6 +705,7 @@ void DeviceBarrierPolicy::InitQueueFamilyPolicy(
                                              | Pal::CoherClear
                                              | Pal::CoherIndirectArgs
                                              | Pal::CoherIndexData
+                                             | Pal::CoherSampleRate
                                              | Pal::CoherPresent;
         pPolicy->supportedLayoutUsageMask   |= Pal::LayoutColorTarget
                                              | Pal::LayoutDepthStencilTarget
@@ -1041,6 +1042,12 @@ void ImageBarrierPolicy::InitImageCachePolicy(
         supportedInputCacheMask  |= Pal::CoherDepthStencilTarget | Pal::CoherClear;
     }
 
+    if (usage & VK_IMAGE_USAGE_FRAGMENT_SHADING_RATE_ATTACHMENT_BIT_KHR)
+    {
+        supportedOutputCacheMask |= Pal::CoherSampleRate;
+        supportedInputCacheMask  |= Pal::CoherSampleRate;
+    }
+
     // Apply device specific supported cache masks to limit the scope.
     supportedOutputCacheMask &= pDevice->GetBarrierPolicy().GetSupportedOutputCacheMask();
     supportedInputCacheMask  &= pDevice->GetBarrierPolicy().GetSupportedInputCacheMask();
@@ -1255,7 +1262,7 @@ uint32_t ImageBarrierPolicy::GetQueueFamilyLayoutEngineMask(
 // Constructor for buffer barrier policies.
 BufferBarrierPolicy::BufferBarrierPolicy(
     Device*                             pDevice,
-    BufferUsageFlagBits                 usage,
+    VkBufferUsageFlagBits2KHR           usage,
     VkSharingMode                       sharingMode,
     uint32_t                            queueFamilyIndexCount,
     const uint32_t*                     pQueueFamilyIndices)
@@ -1268,7 +1275,7 @@ BufferBarrierPolicy::BufferBarrierPolicy(
 // Initialize the cache policy of the buffer according to the input parameters.
 void BufferBarrierPolicy::InitBufferCachePolicy(
     Device*                             pDevice,
-    BufferUsageFlagBits                 usage,
+    VkBufferUsageFlagBits2KHR           usage,
     VkSharingMode                       sharingMode,
     uint32_t                            queueFamilyIndexCount,
     const uint32_t*                     pQueueFamilyIndices)

diff --git a/icd/api/compiler_solution.cpp b/icd/api/compiler_solution.cpp
@@ -347,6 +347,8 @@ void CompilerSolution::UpdateRayTracingFunctionNames(
                                   pTable->pFunc[Vkgc::RT_ENTRY_FETCH_HIT_TRIANGLE_FROM_NODE_POINTER]);
         SetRayTracingFunctionName(entryFuncTable.intrinsic.pFetchTrianglePositionFromRayQuery,
                                   pTable->pFunc[Vkgc::RT_ENTRY_FETCH_HIT_TRIANGLE_FROM_RAY_QUERY]);
+        SetRayTracingFunctionName(entryFuncTable.rayQuery.pGet64BitInstanceNodePtr,
+                                  pTable->pFunc[Vkgc::RT_ENTRY_GET_INSTANCE_NODE]);
     }
 }
 

diff --git a/icd/api/compiler_solution_llpc.cpp b/icd/api/compiler_solution_llpc.cpp
@@ -609,11 +609,14 @@ VkResult CompilerSolutionLlpc::CreateGraphicsShaderBinary(
                 m_pPhysicalDevice->Manager()->VkInstance()->FreeMem(const_cast<void*>(finalBinary.pCode));
             }
 
-            if (pPipelineDumpHandle != nullptr)
+            if ((pPipelineDumpHandle != nullptr) && (pCreateInfo->earlyElfPackage[gplType].pCode != nullptr))
             {
-                Vkgc::BinaryData elfBinary = ExtractPalElfBinary(pCreateInfo->earlyElfPackage[gplType]);
-                Vkgc::IPipelineDumper::DumpPipelineBinary(
-                    pPipelineDumpHandle, m_gfxIp, &elfBinary);
+                if (pCreateInfo->earlyElfPackage[gplType].pCode != nullptr)
+                {
+                    Vkgc::BinaryData elfBinary = ExtractPalElfBinary(pCreateInfo->earlyElfPackage[gplType]);
+                    Vkgc::IPipelineDumper::DumpPipelineBinary(
+                        pPipelineDumpHandle, m_gfxIp, &elfBinary);
+                }
             }
         }
         m_gplCacheMatrix.totalBinaries++;