Merge branch 'cad-text-rendering'

Author: Erfan-Ahmadi

.gitmodules

@@ -70,6 +70,9 @@
 [submodule "examples_tests"]
 	path = examples_tests
 	url = [email protected]:Devsh-Graphics-Programming/Nabla-Examples-and-Tests.git
+[submodule "3rdparty/msdfgen"]
+	path = 3rdparty/msdfgen
+	url = https://github.com/Chlumsky/msdfgen
 	branch = master
 [submodule "3rdparty/dxc/dxc"]
 	path = 3rdparty/dxc/dxc

3rdparty/CMakeLists.txt

@@ -22,6 +22,25 @@ add_subdirectory(simdjson simdjson EXCLUDE_FROM_ALL)
 set(BUILD_SHARED_LIBS ${_OLD_BUILD_SHARED_LIBS})
 set(BUILD_STATIC_LIBS ${_OLD_BUILD_STATIC_LIBS})
 set(BUILD_TESTING ${_OLD_BUILD_TESTING})
+
+# freetype2
+set(FT_WITH_HARFBUZZ OFF)
+add_subdirectory(freetype2 freetype2 EXCLUDE_FROM_ALL)
+add_library(Freetype::Freetype ALIAS freetype) # alias target for MSDFGEN not trying to find freetype system-wide but rather to use the logical target available in our build tree
+
+# msdfgen
+set(MSDFGEN_CORE_ONLY OFF) # if turned on then MSDFGEN will be built without dependencies like freetype or skia
+set(MSDFGEN_INSTALL OFF)
+set(MSDFGEN_BUILD_STANDALONE OFF)
+set(MSDFGEN_USE_CPP11 ON)
+add_subdirectory(msdfgen msdfgen EXCLUDE_FROM_ALL)
+
+if(MSDFGEN_CORE_ONLY)
+	set(NBL_MSDFGEN_TARGETS msdfgen-core)
+else()
+	set(NBL_MSDFGEN_TARGETS msdfgen-ext)
+endif()
+
 set_property(TARGET simdjson PROPERTY CXX_STANDARD 17)
 
 # DXC
@@ -359,6 +378,7 @@ set(NBL_3RDPARTY_TARGETS
 				jpeg-static
 				bzip2
 				simdjson
+				msdfgen-core
 				nlohmann_json
 				glslang
 				OGLCompiler 
@@ -373,6 +393,8 @@ set(NBL_3RDPARTY_TARGETS
 				Iex
 				IlmThread
 				Imath
+				freetype
+				${NBL_MSDFGEN_TARGETS}
 				blake3
 				${NBL_BOOST_TARGETS}
 )
@@ -425,6 +447,13 @@ nbl_adjust_definitions()
 add_custom_target(3rdparty)
 add_dependencies(3rdparty ${NBL_3RDPARTY_TARGETS})
 
-set(_NBL_3RDPARTY_TARGETS_ ${NBL_3RDPARTY_TARGETS} PARENT_SCOPE)
+nbl_install_dir("${CMAKE_CURRENT_SOURCE_DIR}/parallel-hashmap/parallel_hashmap")
+
+# parent scope exports, must be at the end of the file
+set(_NBL_3RDPARTY_TARGETS_ 
+	${NBL_3RDPARTY_TARGETS}
+PARENT_SCOPE)
 
-nbl_install_dir("${CMAKE_CURRENT_SOURCE_DIR}/parallel-hashmap/parallel_hashmap")
+set(NBL_MSDFGEN_TARGETS
+	${NBL_MSDFGEN_TARGETS}
+PARENT_SCOPE)

3rdparty/msdfgen

@@ -148,10 +148,7 @@ function(nbl_adjust_flags)
 					set(NBL_TARGET_MSVC_DEBUG_INFORMATION_FORMAT "$<$<OR:$<STREQUAL:${MAPPED_CONFIG},DEBUG>,$<STREQUAL:${MAPPED_CONFIG},RELWITHDEBINFO>>:ProgramDatabase>")
 				else()
 					set(NBL_TARGET_MSVC_DEBUG_INFORMATION_FORMAT "$<$<STREQUAL:${MAPPED_CONFIG},DEBUG>:EditAndContinue>$<$<STREQUAL:${MAPPED_CONFIG},RELWITHDEBINFO>:ProgramDatabase>")
-				endif()
-				
-				# test
-				file(GENERATE OUTPUT "${CMAKE_CURRENT_BINARY_DIR}/TEST/GEN/${NBL_TARGET_ITEM}/test.cmake" CONTENT "TEST\nNBL_TARGET_MSVC_DEBUG_INFORMATION_FORMAT: \"${NBL_TARGET_MSVC_DEBUG_INFORMATION_FORMAT}\"\nmapped config: \"${MAPPED_CONFIG}\"" CONDITION $<CONFIG:DEBUG>)
+				endif()	
 			endif()
 
 			set_target_properties(${NBL_TARGET_ITEM} PROPERTIES

cmake/adjust/flags.cmake

@@ -314,7 +314,10 @@ macro(nbl_create_ext_library_project EXT_NAME LIB_HEADERS LIB_SOURCES LIB_INCLUD
 		)
 	endif()
 
-	nbl_install_file_spec(${LIB_HEADERS} "nbl/ext/${EXT_NAME}")	
+	if(LIB_HEADERS)
+		nbl_install_file_spec(${LIB_HEADERS} "nbl/ext/${EXT_NAME}")
+	endif()	
+	
 	nbl_install_lib_spec(${LIB_NAME} "nbl/ext/${EXT_NAME}")
 
 	set("NBL_EXT_${EXT_NAME}_INCLUDE_DIRS"

cmake/common.cmake

@@ -65,23 +65,16 @@ class CSwizzleAndConvertImageFilterBase : public CSwizzleableAndDitherableFilter
 						constexpr auto MaxPlanes = 4;
 						const void* srcPix[MaxPlanes] = { commonExecuteData.inData+readBlockArrayOffset,nullptr,nullptr,nullptr };
 
-						const uint32_t inChannelCount = [&]() {
-							if constexpr (kInFormat != EF_UNKNOWN)
-								return asset::getFormatChannelCount<kInFormat>();
-							else
-								return asset::getFormatChannelCount(rInFormat);
-						}();
-
 						for (auto blockY=0u; blockY<blockDims.y; blockY++)
 						for (auto blockX=0u; blockX<blockDims.x; blockX++)
 						{						
 							constexpr auto maxChannels = 4;
 							decodeBufferType decodeBuffer[maxChannels] = {};
 
 							if constexpr (kInFormat!=EF_UNKNOWN)
-								base_t::template onDecode<kInFormat>(state, srcPix, decodeBuffer, blockX, blockY, inChannelCount);
+								base_t::template onDecode<kInFormat>(state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 							else
-								base_t::template onDecode<encodeBufferType>(rInFormat, state, srcPix, decodeBuffer, blockX, blockY, inChannelCount);
+								base_t::template onDecode<encodeBufferType>(rInFormat, state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 
 							state->normalization.prepass(decodeBuffer,readBlockPos*blockDims+commonExecuteData.offsetDifferenceInTexels,blockX,blockY,4u/*TODO: figure this out*/);
 						}
@@ -143,7 +136,6 @@ class CSwizzleAndConvertImageFilter : public CImageFilter<CSwizzleAndConvertImag
 			base_t::template normalizationPrepass<inFormat,ExecutionPolicy,decodeBufferType,encodeBufferType>(EF_UNKNOWN,policy,state,blockDims);
 			auto perOutputRegion = [policy,&blockDims,&state](const CMatchedSizeInOutImageFilterCommon::CommonExecuteData& commonExecuteData, CBasicImageFilterCommon::clip_region_functor_t& clip) -> bool
 			{
-				constexpr uint32_t inChannelsAmount = asset::getFormatChannelCount<inFormat>();
 				constexpr uint32_t outChannelsAmount = asset::getFormatChannelCount<outFormat>();
 
 				auto swizzle = [&commonExecuteData,&blockDims,&state,&outChannelsAmount](uint32_t readBlockArrayOffset, core::vectorSIMDu32 readBlockPos)
@@ -160,7 +152,7 @@ class CSwizzleAndConvertImageFilter : public CImageFilter<CSwizzleAndConvertImag
 						constexpr auto maxChannels = 4;
 						decodeBufferType decodeBuffer[maxChannels] = {};
 
-						base_t::template onDecode<inFormat>(state, srcPix, decodeBuffer, blockX, blockY, inChannelsAmount);
+						base_t::template onDecode<inFormat>(state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 						base_t::template onEncode<outFormat>(state, dstPix, decodeBuffer, localOutPos, blockX, blockY, outChannelsAmount);
 					}
 				};
@@ -211,9 +203,7 @@ class CSwizzleAndConvertImageFilter<EF_UNKNOWN,EF_UNKNOWN,Swizzle,Dither,Normali
 			base_t::template normalizationPrepass<EF_UNKNOWN,ExecutionPolicy,double,double>(inFormat,policy,state,blockDims);
 			auto perOutputRegion = [policy,&blockDims,inFormat,outFormat,outChannelsAmount,&state](const CMatchedSizeInOutImageFilterCommon::CommonExecuteData& commonExecuteData, CBasicImageFilterCommon::clip_region_functor_t& clip) -> bool
 			{
-				const uint32_t inChannelsAmount = asset::getFormatChannelCount(inFormat);
-
-				auto swizzle = [&commonExecuteData,&blockDims,inFormat,outFormat,outChannelsAmount,&state,inChannelsAmount](uint32_t readBlockArrayOffset, core::vectorSIMDu32 readBlockPos)
+				auto swizzle = [&commonExecuteData,&blockDims,inFormat,outFormat,outChannelsAmount,&state](uint32_t readBlockArrayOffset, core::vectorSIMDu32 readBlockPos)
 				{
 					constexpr auto MaxPlanes = 4;
 					const void* srcPix[MaxPlanes] = { commonExecuteData.inData+readBlockArrayOffset,nullptr,nullptr,nullptr };
@@ -227,7 +217,7 @@ class CSwizzleAndConvertImageFilter<EF_UNKNOWN,EF_UNKNOWN,Swizzle,Dither,Normali
 						constexpr auto maxChannels = 4;
 						double decodeBuffer[maxChannels] = {};
 
-						base_t::template onDecode(inFormat, state, srcPix, decodeBuffer, blockX, blockY, inChannelsAmount);
+						base_t::template onDecode(inFormat, state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 						base_t::template onEncode(outFormat, state, dstPix, decodeBuffer, localOutPos, blockX, blockY, outChannelsAmount);
 					}
 				};
@@ -285,7 +275,6 @@ class CSwizzleAndConvertImageFilter<EF_UNKNOWN,outFormat,Swizzle,Dither,Normaliz
 			normalizationPrepass<EF_UNKNOWN,ExecutionPolicy,double,encodeBufferType>(inFormat,policy,state,blockDims);
 			auto perOutputRegion = [policy,&blockDims,inFormat,&state](const CMatchedSizeInOutImageFilterCommon::CommonExecuteData& commonExecuteData, CBasicImageFilterCommon::clip_region_functor_t& clip) -> bool
 			{
-				const uint32_t inChannelsAmount = asset::getFormatChannelCount(inFormat);
 				constexpr uint32_t outChannelsAmount = asset::getFormatChannelCount<outFormat>();
 
 				auto swizzle = [&commonExecuteData,&blockDims,inFormat,&outChannelsAmount,&state](uint32_t readBlockArrayOffset, core::vectorSIMDu32 readBlockPos)
@@ -302,7 +291,7 @@ class CSwizzleAndConvertImageFilter<EF_UNKNOWN,outFormat,Swizzle,Dither,Normaliz
 						constexpr auto maxChannels = 4;
 						double decodeBuffer[maxChannels] = {};
 
-						base_t::template onDecode(inFormat, state, srcPix, decodeBuffer, blockX, blockY, inChannelsAmount);
+						base_t::template onDecode(inFormat, state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 						base_t::template onEncode<outFormat>(state, dstPix, decodeBuffer, localOutPos, blockX, blockY, outChannelsAmount);
 					}
 				};
@@ -361,7 +350,6 @@ class CSwizzleAndConvertImageFilter<inFormat,EF_UNKNOWN,Swizzle,Dither,Normaliza
 			normalizationPrepass<inFormat,ExecutionPolicy,decodeBufferType,double>(EF_UNKNOWN,policy,state,blockDims);
 			auto perOutputRegion = [policy,&blockDims,&outFormat,outChannelsAmount,&state](const CMatchedSizeInOutImageFilterCommon::CommonExecuteData& commonExecuteData, CBasicImageFilterCommon::clip_region_functor_t& clip) -> bool
 			{
-				constexpr uint32_t inChannelsAmount = asset::getFormatChannelCount<inFormat>();
 				const uint32_t outChannelsAmount = asset::getFormatChannelCount(outFormat);
 
 				auto swizzle = [&commonExecuteData,&blockDims,&outFormat,&outChannelsAmount,&state](uint32_t readBlockArrayOffset, core::vectorSIMDu32 readBlockPos)
@@ -378,7 +366,7 @@ class CSwizzleAndConvertImageFilter<inFormat,EF_UNKNOWN,Swizzle,Dither,Normaliza
 							constexpr auto maxChannels = 4;
 							decodeBufferType decodeBuffer[maxChannels] = {};
 
-							base_t::template onDecode<inFormat>(state, srcPix, decodeBuffer, blockX, blockY, inChannelsAmount);
+							base_t::template onDecode<inFormat>(state, srcPix, decodeBuffer, blockX, blockY, maxChannels);
 							base_t::template onEncode(outFormat, state, dstPix, decodeBuffer, localOutPos, blockX, blockY, outChannelsAmount);
 						}
 				};

include/nbl/asset/filters/CSwizzleAndConvertImageFilter.h

@@ -74,12 +74,6 @@ inline matrix<T,M,N> transpose(const matrix<T,N,M>& m)
     return reinterpret_cast<matrix<T,M,N>&>(glm::transpose(reinterpret_cast<typename matrix<T,N,M>::Base const&>(m)));
 }
 
-template<typename T>
-inline T bitfieldExtract( T val, uint32_t offsetBits, uint32_t numBits )
-{
-    return glm::bitfieldExtract( val, int32_t( offsetBits ), int32_t( numBits ) );
-}
-
 #undef NBL_BIT_OP_GLM_PASSTHROUGH
 #undef NBL_SIMPLE_GLM_PASSTHROUGH
 

include/nbl/builtin/hlsl/cpp_compat/intrinsics.h

@@ -15,7 +15,22 @@ namespace hlsl
 namespace glsl
 {
 
-#ifdef __HLSL_VERSION
+#ifndef __HLSL_VERSION
+
+// GLM Aliases
+template<typename genIUType>
+genIUType bitfieldExtract(genIUType Value, int Offset, int Bits)
+{
+	return glm::bitfieldExtract<genIUType>(Value, Offset, Bits);
+}
+
+template<typename genIUType>
+genIUType bitfieldInsert(genIUType const& Base, genIUType const& Insert, int Offset, int Bits)
+{
+	return glm::bitfieldInsert<genIUType>(Base, Insert, Offset, Bits);
+}
+
+#else
 /**
 * Generic SPIR-V
 */
@@ -189,7 +204,7 @@ struct bitfieldExtract<T, false, true>
 template<typename T>
 T bitfieldExtract( T val, uint32_t offsetBits, uint32_t numBits )
 {
-    return impl::bitfieldExtract<T, is_signed<T>::value, is_integral<T>::value>::template  __call(val,offsetBits,numBits);
+    return impl::bitfieldExtract<T, is_signed<T>::value, is_integral<T>::value>::__call(val,offsetBits,numBits);
 }
 
 
@@ -199,7 +214,7 @@ namespace impl
 template<typename T>
 struct bitfieldInsert
 {
-    enable_if_t<is_integral_v<T>, T> __call( T base, T insert, uint32_t offset, uint32_t count )
+    static enable_if_t<is_integral_v<T>, T> __call( T base, T insert, uint32_t offset, uint32_t count )
     {
         return spirv::bitFieldInsert<T>( base, insert, offset, count );
     }
@@ -210,7 +225,7 @@ struct bitfieldInsert
 template<typename T>
 T bitfieldInsert( T base, T insert, uint32_t offset, uint32_t count )
 {
-    return impl::bitfieldInsert<T>::template  __call(base, insert, offset, count);
+    return impl::bitfieldInsert<T>::__call(base, insert, offset, count);
 }
 
 #endif
@@ -219,4 +234,4 @@ T bitfieldInsert( T base, T insert, uint32_t offset, uint32_t count )
 }
 }
 
-#endif
+#endif

include/nbl/builtin/hlsl/glsl_compat/core.hlsl

@@ -0,0 +1,45 @@
+// Copyright (C) 2023 - DevSH Graphics Programming Sp. z O.O.
+// This file is part of the "Nabla Engine".
+// For conditions of distribution and use, see copyright notice in nabla.h
+#ifndef _NBL_BUILTIN_HLSL_TEXT_RENDERING_MSDF_INCLUDED_
+#define _NBL_BUILTIN_HLSL_TEXT_RENDERING_MSDF_INCLUDED_
+
+namespace nbl 
+{
+namespace hlsl
+{
+namespace text
+{
+
+float median(float r, float g, float b) {
+    return max(min(r, g), min(max(r, g), b));
+}
+
+/*
+
+Returns the distance to the shape in the given MSDF.
+This can then be processed using smoothstep to provide anti-alising when rendering the shape.
+
+@params:
+- msdfSample: sampled SNORM value from the MSDF texture generated by msdfgen library.
+
+- msdfPixelRange: specifies the width of the range around the shape between the minimum and maximum representable signed distance in shape units or distance field pixels, respectivelly.
+  for example if msdfPixelRange is 4, then the range of distance values are [-2, +2], and it can be computed by snormValue * MSDFPixelRange/2.0
+    so an snorm value of 1.0 means a distance of 2 pixels outside the shape (in msdf texture space)
+    This value is set when rendering the MSDF with MSDFgen.
+
+- screenPxRangeValue: the value used to convert the distance values in the msdf texture/atlas to distance in screen space. 
+  In other words it's DistanceInScreenSpace/DistanceInMSDFTextureSpace, the larger the glyph (or more zoomed in) the larger this value is.
+  In 2D Text Rendering it is computed by `GlyphScreenSpaceSize/GlyphTextureSpaceSize`
+  where GlyphTextureSpaceSize is the size of the glyph inside the msdf texture/atlas
+*/
+float msdfDistance(float3 msdfSample, float msdfPixelRange, float screenPxRangeValue) {
+    float texelDist = median(msdfSample.r, msdfSample.g, msdfSample.b) * msdfPixelRange / 2.0f;
+    return texelDist * screenPxRangeValue;
+}
+
+}
+}
+}
+
+#endif

include/nbl/builtin/hlsl/text_rendering/msdf.hlsl

@@ -32,8 +32,6 @@ namespace impl
 			MapEquals m_equals;
 			const mutable Key* searchedKey;
 
-			LRUCacheBase() = default;
-
 			LRUCacheBase(const uint32_t capacity, MapHash&& _hash, MapEquals&& _equals, disposal_func_t&& df) : m_list(capacity, std::move(df)), m_hash(std::move(_hash)), m_equals(std::move(_equals)), searchedKey(nullptr)
 			{ }
 
@@ -56,7 +54,7 @@ namespace impl
 // Stores fixed size amount of elements. 
 // When the cache is full inserting will remove the least used entry
 template<typename Key, typename Value, typename MapHash=std::hash<Key>, typename MapEquals=std::equal_to<Key> >
-class LRUCache : protected impl::LRUCacheBase<Key,Value,MapHash,MapEquals>
+class LRUCache : protected impl::LRUCacheBase<Key,Value,MapHash,MapEquals>, public core::Unmovable, public core::Uncopyable
 {
 		// typedefs
 		typedef impl::LRUCacheBase<Key,Value,MapHash,MapEquals> base_t;
@@ -118,21 +116,7 @@ class LRUCache : protected impl::LRUCacheBase<Key,Value,MapHash,MapEquals>
 			assert(capacity > 1);
 			m_shortcut_map.reserve(capacity);
 		}
-
-		LRUCache() = default;
-		
-		LRUCache& operator=(LRUCache&& other)
-		{
-			m_shortcut_map = std::move(other.m_shortcut_map);
-			base_t::m_list = std::move(other.m_list);
-			base_t::m_hash = std::move(other.m_hash);
-			base_t::m_equals = std::move(other.m_equals);
-			base_t::searchedKey = other.searchedKey;
-
-			// Nullify other
-			other.searchedKey = nullptr;
-			return *this;
-		}
+		LRUCache() = delete;
 
 		inline void print(core::smart_refctd_ptr<system::ILogger> logger)
 		{
@@ -165,7 +149,7 @@ class LRUCache : protected impl::LRUCacheBase<Key,Value,MapHash,MapEquals>
 				const bool overflow = m_shortcut_map.size()>=base_t::m_list.getCapacity();
 				if (overflow)
 				{
-					// evictCallback(base_t::m_list.getBack()->data.second);
+					evictCallback(base_t::m_list.getBack()->data.second);
 					m_shortcut_map.erase(base_t::m_list.getLastAddress());
 					base_t::m_list.popBack();
 				}