Speed up HLSL preprocessing and prepared SPIR-V hot paths

.gitmodules

@@ -72,7 +72,7 @@
 [submodule "3rdparty/dxc/dxc"]
 	path = 3rdparty/dxc/dxc
 	url = git@github.com:Devsh-Graphics-Programming/DirectXShaderCompiler.git
-	branch = devshFixes
+	branch = unroll-devshFixes
 [submodule "3rdparty/imgui"]
 	path = 3rdparty/imgui
 	url = git@github.com:Devsh-Graphics-Programming/imgui.git

cmake/common.cmake

@@ -1191,6 +1191,7 @@ option(NSC_DEBUG_EDIF_LINE_BIT "Add \"-fspv-debug=line\" to NSC Debug CLI" OFF)
 option(NSC_DEBUG_EDIF_TOOL_BIT "Add \"-fspv-debug=tool\" to NSC Debug CLI" ON)
 option(NSC_DEBUG_EDIF_NON_SEMANTIC_BIT "Add \"-fspv-debug=vulkan-with-source\" to NSC Debug CLI" OFF)
 option(NSC_USE_DEPFILE "Generate depfiles for NSC custom commands" ON)
+option(NBL_DISABLE_EXPERIMENTAL_OPTS "Disable -O1experimental for all NSC compile rules globally" OFF)
 
 function(NBL_CREATE_NSC_COMPILE_RULES)
     set(COMMENT "this code has been autogenerated with Nabla CMake NBL_CREATE_HLSL_COMPILE_RULES utility")
@@ -1230,10 +1231,7 @@ struct DeviceConfigCaps
 		-fspv-target-env=vulkan1.3 
 		-Wshadow 
 		-Wconversion 
-		-Wno-local-type-template-args 
-		$<$<CONFIG:Debug>:-O0> 
-		$<$<CONFIG:Release>:-O3> 
-		$<$<CONFIG:RelWithDebInfo>:-O3>
+		-Wno-local-type-template-args
 	)
 
 	if(NSC_DEBUG_EDIF_FILE_BIT)
@@ -1259,18 +1257,32 @@ struct DeviceConfigCaps
 	if(NOT NBL_EMBED_BUILTIN_RESOURCES)
 		list(APPEND REQUIRED_OPTIONS
 			-no-nbl-builtins
-			-I "${NBL_ROOT_PATH}/include"
-			-I "${NBL_ROOT_PATH}/3rdparty/dxc/dxc/external/SPIRV-Headers/include"
-			-I "${NBL_ROOT_PATH}/3rdparty/boost/superproject/libs/preprocessor/include"
-			-I "${NBL_ROOT_PATH_BINARY}/src/nbl/device/include"
+			-isystem "${NBL_ROOT_PATH}/include"
+			-isystem "${NBL_ROOT_PATH}/3rdparty/dxc/dxc/external/SPIRV-Headers/include"
+			-isystem "${NBL_ROOT_PATH}/3rdparty/boost/superproject/libs/preprocessor/include"
+			-isystem "${NBL_ROOT_PATH_BINARY}/src/nbl/device/include"
 		)
 	endif()
 
     set(REQUIRED_SINGLE_ARGS TARGET BINARY_DIR OUTPUT_VAR INPUTS INCLUDE NAMESPACE MOUNT_POINT_DEFINE)
     set(OPTIONAL_SINGLE_ARGS GLOB_DIR)
-    cmake_parse_arguments(IMPL "DISCARD_DEFAULT_GLOB" "${REQUIRED_SINGLE_ARGS};${OPTIONAL_SINGLE_ARGS};LINK_TO" "COMMON_OPTIONS;DEPENDS" ${ARGV})
+    cmake_parse_arguments(IMPL "DISCARD_DEFAULT_GLOB;DISABLE_EXPERIMENTAL_OPTS" "${REQUIRED_SINGLE_ARGS};${OPTIONAL_SINGLE_ARGS};LINK_TO" "COMMON_OPTIONS;DEPENDS" ${ARGV})
     NBL_PARSE_REQUIRED(IMPL ${REQUIRED_SINGLE_ARGS})
 
+	if(NBL_DISABLE_EXPERIMENTAL_OPTS OR IMPL_DISABLE_EXPERIMENTAL_OPTS)
+		list(APPEND REQUIRED_OPTIONS
+			$<$<CONFIG:Debug>:-O0>
+			$<$<CONFIG:Release>:-O3>
+			$<$<CONFIG:RelWithDebInfo>:-O3>
+		)
+	else()
+		list(APPEND REQUIRED_OPTIONS
+			$<$<CONFIG:Debug>:-O1experimental>
+			$<$<CONFIG:Release>:-O3>
+			$<$<CONFIG:RelWithDebInfo>:-O1experimental>
+		)
+	endif()
+
 	set(IMPL_HLSL_GLOB "")
 	if(NOT IMPL_DISCARD_DEFAULT_GLOB)
 		set(GLOB_ROOT "${CMAKE_CURRENT_SOURCE_DIR}")
@@ -1306,12 +1318,6 @@ struct DeviceConfigCaps
 			TARGET ${IMPL_TARGET}
 		)
 
-		target_sources(${IMPL_TARGET} PUBLIC ${INCLUDE_FILE})
-		set_source_files_properties(${INCLUDE_FILE} PROPERTIES 
-			HEADER_FILE_ONLY ON
-			VS_TOOL_OVERRIDE None
-		)
-
 		target_compile_definitions(${IMPL_TARGET} INTERFACE $<TARGET_PROPERTY:${IMPL_TARGET},NBL_MOUNT_POINT_DEFINES>)
 		target_include_directories(${IMPL_TARGET} INTERFACE ${INCLUDE_DIR})
 		set_target_properties(${IMPL_TARGET} PROPERTIES NBL_HEADER_GENERATED_RULE ON)

include/nbl/asset/utils/ISPIRVEntryPointTrimmer.h

@@ -7,10 +7,12 @@
 
 #include "nbl/system/ILogger.h"
 
+#include <mutex>
+
 namespace nbl::asset
 {
 
-class ISPIRVEntryPointTrimmer final : public core::IReferenceCounted
+class NBL_API2 ISPIRVEntryPointTrimmer final : public core::IReferenceCounted
 {
     public:
         ISPIRVEntryPointTrimmer();
@@ -46,6 +48,8 @@ class ISPIRVEntryPointTrimmer final : public core::IReferenceCounted
         };
 
         Result trim(const ICPUBuffer* spirvBuffer, const core::set<EntryPoint>& entryPoints, system::logger_opt_ptr logger = nullptr) const;
+        bool ensureValidated(const ICPUBuffer* spirvBuffer, system::logger_opt_ptr logger = nullptr) const;
+        void markValidated(const ICPUBuffer* spirvBuffer) const;
 
         inline core::smart_refctd_ptr<const IShader> trim(const IShader* shader, const core::set<EntryPoint>& entryPoints, system::logger_opt_ptr logger = nullptr) const
         {
@@ -72,6 +76,8 @@ class ISPIRVEntryPointTrimmer final : public core::IReferenceCounted
 
     private:
         core::smart_refctd_ptr<ISPIRVOptimizer> m_optimizer;
+        mutable std::mutex m_validationCacheMutex;
+        mutable core::unordered_set<core::blake3_hash_t> m_validatedSpirvHashes;
 };
 
 }

include/nbl/asset/utils/IShaderCompiler.h

@@ -17,6 +17,8 @@
 #include "nbl/builtin/hlsl/enums.hlsl"
 
 #include <functional>
+#include <mutex>
+#include <unordered_set>
 
 namespace nbl::asset
 {
@@ -26,6 +28,25 @@ class NBL_API2 IShaderCompiler : public core::IReferenceCounted
 	public:
 		IShaderCompiler(core::smart_refctd_ptr<system::ISystem>&& system);
 
+		enum class IncludeRootOrigin : uint8_t
+		{
+			User,
+			Builtin,
+			Generated
+		};
+
+		enum class HeaderClass : uint8_t
+		{
+			User,
+			System
+		};
+
+		struct IncludeClassification
+		{
+			IncludeRootOrigin origin = IncludeRootOrigin::User;
+			HeaderClass headerClass = HeaderClass::User;
+		};
+
 		class NBL_API2 IIncludeLoader : public core::IReferenceCounted
 		{
 			public:
@@ -34,12 +55,13 @@ class NBL_API2 IShaderCompiler : public core::IReferenceCounted
 					system::path absolutePath = {};
 					std::string contents = {};
 					core::blake3_hash_t hash = {}; // TODO: we're not yet using IFile::getPrecomputedHash(), so for builtins we can maybe use that in the future
+					IncludeClassification classification = {};
 					// Could be used in the future for early rejection of cache hit
 					//nbl::system::IFileBase::time_point_t lastWriteTime = {};
 
 					explicit inline operator bool() const {return !absolutePath.empty();}
 				};
-				virtual found_t getInclude(const system::path& searchPath, const std::string& includeName) const = 0;
+				virtual found_t getInclude(const system::path& searchPath, const std::string& includeName, bool needHash = true) const = 0;
 		};
 
 		class NBL_API2 IIncludeGenerator : public core::IReferenceCounted
@@ -65,7 +87,7 @@ class NBL_API2 IShaderCompiler : public core::IReferenceCounted
 			public:
 				CFileSystemIncludeLoader(core::smart_refctd_ptr<system::ISystem>&& system);
 
-				IIncludeLoader::found_t getInclude(const system::path& searchPath, const std::string& includeName) const override;
+				IIncludeLoader::found_t getInclude(const system::path& searchPath, const std::string& includeName, bool needHash = true) const override;
 
 			protected:
 				core::smart_refctd_ptr<system::ISystem> m_system;
@@ -74,37 +96,88 @@ class NBL_API2 IShaderCompiler : public core::IReferenceCounted
 		class NBL_API2 CIncludeFinder : public core::IReferenceCounted
 		{
 			public:
+				struct SSessionCache
+				{
+					struct Stats
+					{
+						uint64_t lookupFound = 0ull;
+						uint64_t lookupMissing = 0ull;
+						uint64_t lookupMiss = 0ull;
+						uint64_t storeFound = 0ull;
+						uint64_t storeMissing = 0ull;
+					};
+
+					enum class LookupResult : uint8_t
+					{
+						Miss,
+						Missing,
+						Found
+					};
+
+					explicit SSessionCache(const bool threadSafe = false) : threadSafe(threadSafe) {}
+
+					void clear();
+					LookupResult lookup(const std::string& key, IIncludeLoader::found_t& result) const;
+					void store(const std::string& key, IIncludeLoader::found_t result);
+					Stats snapshotStats() const;
+
+					bool threadSafe = false;
+
+					mutable std::mutex mutex;
+					mutable Stats stats;
+					core::unordered_map<std::string, IIncludeLoader::found_t> found;
+					core::unordered_set<std::string> missing;
+				};
+
 				CIncludeFinder(core::smart_refctd_ptr<system::ISystem>&& system);
 
 				// ! includes within <>
 				// @param requestingSourceDir: the directory where the incude was requested
 				// @param includeName: the string within <> of the include preprocessing directive
-				IIncludeLoader::found_t getIncludeStandard(const system::path& requestingSourceDir, const std::string& includeName) const;
+				IIncludeLoader::found_t getIncludeStandard(const system::path& requestingSourceDir, const std::string& includeName, bool needHash = true, SSessionCache* readSessionCache = nullptr, SSessionCache* writeSessionCache = nullptr) const;
 
 				// ! includes within ""
 				// @param requestingSourceDir: the directory where the incude was requested
 				// @param includeName: the string within "" of the include preprocessing directive
-				IIncludeLoader::found_t getIncludeRelative(const system::path& requestingSourceDir, const std::string& includeName) const;
+				IIncludeLoader::found_t getIncludeRelative(const system::path& requestingSourceDir, const std::string& includeName, bool needHash = true, SSessionCache* readSessionCache = nullptr, SSessionCache* writeSessionCache = nullptr) const;
 
 				inline core::smart_refctd_ptr<CFileSystemIncludeLoader> getDefaultFileSystemLoader() const { return m_defaultFileSystemLoader; }
 
-				void addSearchPath(const std::string& searchPath, const core::smart_refctd_ptr<IIncludeLoader>& loader);
+				void addSearchPath(const std::string& searchPath, const core::smart_refctd_ptr<IIncludeLoader>& loader, IncludeClassification classification = {});
 
-				void addGenerator(const core::smart_refctd_ptr<IIncludeGenerator>& generator);
+				void addGenerator(const core::smart_refctd_ptr<IIncludeGenerator>& generator, IncludeClassification classification = {IncludeRootOrigin::Generated,HeaderClass::System});
+
+				bool isKnownGlobalInclude(std::string_view includeName) const;
+				IIncludeLoader::found_t classifyFound(IIncludeLoader::found_t found) const;
 
 			protected:
-				IIncludeLoader::found_t trySearchPaths(const std::string& includeName) const;
+				IIncludeLoader::found_t trySearchPaths(const std::string& includeName, bool needHash) const;
 
 				IIncludeLoader::found_t tryIncludeGenerators(const std::string& includeName) const;
+				void registerHeaderRoot(std::string rootPath, IncludeClassification classification);
 
 				struct LoaderSearchPath
 				{
 					core::smart_refctd_ptr<IIncludeLoader> loader = nullptr;
 					std::string searchPath = {};
+					IncludeClassification classification = {};
+				};
+
+				struct GeneratorEntry
+				{
+					core::smart_refctd_ptr<IIncludeGenerator> generator = nullptr;
+					IncludeClassification classification = {IncludeRootOrigin::Generated,HeaderClass::System};
+				};
+
+				struct HeaderRoot
+				{
+					std::string path = {};
+					IncludeClassification classification = {};
 				};
 
 				std::vector<LoaderSearchPath> m_loaders;
-				std::vector<core::smart_refctd_ptr<IIncludeGenerator>> m_generators;
+				std::vector<GeneratorEntry> m_generators;
+				std::vector<HeaderRoot> m_headerRoots;
 				core::smart_refctd_ptr<CFileSystemIncludeLoader> m_defaultFileSystemLoader;
 		};
 
@@ -134,9 +207,12 @@ class NBL_API2 IShaderCompiler : public core::IReferenceCounted
 			std::string_view sourceIdentifier = "";
 			system::logger_opt_ptr logger = nullptr;
 			const CIncludeFinder* includeFinder = nullptr;
+			CIncludeFinder::SSessionCache* readIncludeSessionCache = nullptr;
+			CIncludeFinder::SSessionCache* writeIncludeSessionCache = nullptr;
 			std::span<const SMacroDefinition> extraDefines = {};
 			E_SPIRV_VERSION targetSpirvVersion = E_SPIRV_VERSION::ESV_1_6;
 			bool depfile = false;
+			bool preserveComments = false;
 			system::path depfilePath = {};
 			std::function<void(std::string_view)> onPartialOutputOnFailure = {};
 		};

include/nbl/builtin/hlsl/bxdf/reflection/beckmann.hlsl

@@ -4,9 +4,7 @@
 #ifndef _NBL_BUILTIN_HLSL_BXDF_REFLECTION_BECKMANN_INCLUDED_
 #define _NBL_BUILTIN_HLSL_BXDF_REFLECTION_BECKMANN_INCLUDED_
 
-#include "nbl/builtin/hlsl/bxdf/common.hlsl"
 #include "nbl/builtin/hlsl/bxdf/bxdf_traits.hlsl"
-#include "nbl/builtin/hlsl/sampling/cos_weighted_spheres.hlsl"
 #include "nbl/builtin/hlsl/bxdf/ndf/beckmann.hlsl"
 #include "nbl/builtin/hlsl/bxdf/base/cook_torrance_base.hlsl"
 

include/nbl/builtin/hlsl/bxdf/reflection/ggx.hlsl

@@ -4,9 +4,7 @@
 #ifndef _NBL_BUILTIN_HLSL_BXDF_REFLECTION_GGX_INCLUDED_
 #define _NBL_BUILTIN_HLSL_BXDF_REFLECTION_GGX_INCLUDED_
 
-#include "nbl/builtin/hlsl/bxdf/common.hlsl"
 #include "nbl/builtin/hlsl/bxdf/bxdf_traits.hlsl"
-#include "nbl/builtin/hlsl/sampling/cos_weighted_spheres.hlsl"
 #include "nbl/builtin/hlsl/bxdf/ndf/ggx.hlsl"
 #include "nbl/builtin/hlsl/bxdf/base/cook_torrance_base.hlsl"
 

include/nbl/builtin/hlsl/bxdf/transmission/beckmann.hlsl

@@ -4,10 +4,8 @@
 #ifndef _NBL_BUILTIN_HLSL_BXDF_TRANSMISSION_BECKMANN_INCLUDED_
 #define _NBL_BUILTIN_HLSL_BXDF_TRANSMISSION_BECKMANN_INCLUDED_
 
-#include "nbl/builtin/hlsl/bxdf/common.hlsl"
 #include "nbl/builtin/hlsl/bxdf/bxdf_traits.hlsl"
-#include "nbl/builtin/hlsl/sampling/cos_weighted_spheres.hlsl"
-#include "nbl/builtin/hlsl/bxdf/reflection.hlsl"
+#include "nbl/builtin/hlsl/bxdf/ndf/beckmann.hlsl"
 #include "nbl/builtin/hlsl/bxdf/base/cook_torrance_base.hlsl"
 
 namespace nbl

include/nbl/builtin/hlsl/bxdf/transmission/ggx.hlsl

@@ -4,10 +4,7 @@
 #ifndef _NBL_BUILTIN_HLSL_BXDF_TRANSMISSION_GGX_INCLUDED_
 #define _NBL_BUILTIN_HLSL_BXDF_TRANSMISSION_GGX_INCLUDED_
 
-#include "nbl/builtin/hlsl/bxdf/common.hlsl"
 #include "nbl/builtin/hlsl/bxdf/bxdf_traits.hlsl"
-#include "nbl/builtin/hlsl/sampling/cos_weighted_spheres.hlsl"
-#include "nbl/builtin/hlsl/bxdf/reflection.hlsl"
 #include "nbl/builtin/hlsl/bxdf/base/cook_torrance_base.hlsl"
 
 namespace nbl

include/nbl/builtin/hlsl/emulated/float64_t.hlsl

@@ -2,7 +2,6 @@
 #define _NBL_BUILTIN_HLSL_EMULATED_FLOAT64_T_HLSL_INCLUDED_
 
 #include <nbl/builtin/hlsl/emulated/float64_t_impl.hlsl>
-#include <nbl/builtin/hlsl/concepts/core.hlsl>
 
 namespace nbl
 {

include/nbl/builtin/hlsl/member_test_macros.hlsl

@@ -5,7 +5,10 @@
 #define _NBL_BUILTIN_HLSL_MEMBER_TEST_MACROS_INCLUDED_
 
 #include <nbl/builtin/hlsl/utility.hlsl>
-#include <boost/preprocessor.hpp>
+#include <boost/preprocessor/arithmetic/inc.hpp>
+#include <boost/preprocessor/comparison/not_equal.hpp>
+#include <boost/preprocessor/punctuation/comma_if.hpp>
+#include <boost/preprocessor/repetition/repeat.hpp>
 
 #ifdef __HLSL_VERSION
 
@@ -123,4 +126,4 @@ GENERATE_METHOD_TESTER(set)
 
 
 #endif
-#endif
+#endif