Merge branch 'vulkan_1_3' into suballocdescriptorset

Author: deprilula28

3rdparty/dxc/dxc

@@ -145,14 +145,15 @@ NBL_ENUM_ADD_BITWISE_OPERATORS(ACCESS_FLAGS)
 
 struct SMemoryBarrier
 {
+    // TODO: pack these up into just `src` and `dst` with another struct
     core::bitflag<PIPELINE_STAGE_FLAGS> srcStageMask = PIPELINE_STAGE_FLAGS::NONE;
     core::bitflag<ACCESS_FLAGS> srcAccessMask = ACCESS_FLAGS::NONE;
     core::bitflag<PIPELINE_STAGE_FLAGS> dstStageMask = PIPELINE_STAGE_FLAGS::NONE;
     core::bitflag<ACCESS_FLAGS> dstAccessMask = ACCESS_FLAGS::NONE;
 
     auto operator<=>(const SMemoryBarrier&) const = default;
 
-    // utilities
+    // Make the immediately previous barrier after this one
     inline SMemoryBarrier prevBarrier(const core::bitflag<PIPELINE_STAGE_FLAGS> prevStageMask, const core::bitflag<ACCESS_FLAGS> prevAccessMask) const
     {
         return {
@@ -162,6 +163,12 @@ struct SMemoryBarrier
             .dstAccessMask = srcAccessMask
         };
     }
+    // Make the immediately previous barrier if you know the penultimate, as in one before the last (previous)
+    inline SMemoryBarrier prevBarrier(const SMemoryBarrier& penultimate) const
+    {
+        return prevBarrier(penultimate.dstStageMask,penultimate.dstAccessMask);
+    }
+    // Make the immediately next barrier after this one
     inline SMemoryBarrier nextBarrier(const core::bitflag<PIPELINE_STAGE_FLAGS> nextStageMask, const core::bitflag<ACCESS_FLAGS> nextAccessMask) const
     {
         return {
@@ -171,6 +178,11 @@ struct SMemoryBarrier
             .dstAccessMask = nextAccessMask
         };
     }
+    // Make the immediately next barrier, if you know the barrier thats after the immediately next one
+    inline SMemoryBarrier nextBarrier(const SMemoryBarrier& twoAhead) const
+    {
+        return prevBarrier(twoAhead.srcStageMask,twoAhead.srcAccessMask);
+    }
 };
 
 inline core::bitflag<ACCESS_FLAGS> allAccessesFromStages(core::bitflag<PIPELINE_STAGE_FLAGS> stages)

examples_tests

@@ -1,18 +1,17 @@
-#ifndef __NBL_I_CPU_FRAMEBUFFER_H_INCLUDED__
-#define __NBL_I_CPU_FRAMEBUFFER_H_INCLUDED__
+#ifndef _NBL_I_CPU_FRAMEBUFFER_H_INCLUDED_
+#define _NBL_I_CPU_FRAMEBUFFER_H_INCLUDED_
 
 #include "nbl/asset/IFramebuffer.h"
 #include "nbl/asset/IAsset.h"
 #include "nbl/asset/ICPUImageView.h"
 #include "nbl/asset/ICPURenderpass.h"
 
-namespace nbl {
-namespace asset
+namespace nbl::asset
 {
 
-class ICPUFramebuffer final : public IAsset, public IFramebuffer<ICPURenderpass, ICPUImageView>
+class ICPUFramebuffer final : public IAsset, public IFramebuffer<ICPURenderpass,ICPUImageView>
 {
-    using base_t = IFramebuffer<ICPURenderpass, ICPUImageView>;
+    using base_t = IFramebuffer<ICPURenderpass,ICPUImageView>;
 
 public:
     using base_t::base_t;
@@ -49,6 +48,4 @@ class ICPUFramebuffer final : public IAsset, public IFramebuffer<ICPURenderpass,
 };
 
 }
-}
-
 #endif

include/nbl/asset/ECommonEnums.h

@@ -169,8 +169,11 @@ class IRenderpass
                     using SColorAttachmentsRef = SRenderAttachmentsRef<SColorAttachmentRef>;
 
 
-                    auto operator<=>(const SSubpassDescription&) const = default;
+                    //
+                    bool operator!=(const SSubpassDescription&) const;
+                    inline bool operator==(const SSubpassDescription& rhs) const {return !((*this)!=rhs);}
 
+                    //
                     bool valid(const SCreationParams& params, const uint32_t depthStencilAttachmentCount, const uint32_t colorAttachmentCount) const;
 
 
@@ -255,16 +258,16 @@ class IRenderpass
 
         inline const SCreationParams& getCreationParameters() const { return m_params; }
 
-        inline uint32_t getDepthStencilAttachmentCount() const { return m_depthStencilAttachments->size(); }
-        inline uint32_t getColorAttachmentCount() const { return m_colorAttachments->size(); }
-        inline uint32_t getDepthStencilLoadOpAttachmentEnd() const { return m_loadOpDepthStencilAttachmentEnd; }
-        inline uint32_t getColorLoadOpAttachmentEnd() const { return m_loadOpColorAttachmentEnd; }
+        inline uint32_t getDepthStencilAttachmentCount() const {return m_depthStencilAttachments ? (m_depthStencilAttachments->size()-1):0;}
+        inline uint32_t getColorAttachmentCount() const {return m_colorAttachments ? (m_colorAttachments->size()-1):0;}
+        inline uint32_t getDepthStencilLoadOpAttachmentEnd() const {return m_loadOpDepthStencilAttachmentEnd;}
+        inline uint32_t getColorLoadOpAttachmentEnd() const {return m_loadOpColorAttachmentEnd;}
 
-        inline uint32_t getSubpassCount() const { return m_subpasses->size(); }
-        inline uint32_t getDependencyCount() const { return m_subpassDependencies->size(); }
+        inline uint32_t getSubpassCount() const {return m_subpasses->size()-1;}
+        inline uint32_t getDependencyCount() const {return m_subpassDependencies ? (m_subpassDependencies->size()-1):0;}
 
-        inline bool hasViewMasks() const { return m_viewMaskMSB>0; }
-        inline int8_t getViewMaskMSB() const { return m_viewMaskMSB; }
+        inline bool hasViewMasks() const {return m_viewMaskMSB>0;}
+        inline int8_t getViewMaskMSB() const {return m_viewMaskMSB;}
 
 
         struct SCreationParamValidationResult final
@@ -367,7 +370,11 @@ inline IRenderpass::SCreationParamValidationResult IRenderpass::validateCreation
         return retval;
 
     retval.subpassCount = 0xdeadbeefu;
-    auto setRetvalFalse = [&retval]()->bool{retval.subpassCount = 0; return false;};     
+    auto setRetvalFalse = [&retval]()->bool
+    {
+        retval.subpassCount = 0; return false;
+    };
+
     core::visit_token_terminated_array(params.depthStencilAttachments,SCreationParams::DepthStencilAttachmentsEnd,[&params,setRetvalFalse,&retval](const SCreationParams::SDepthStencilAttachmentDescription& attachment)->bool
     {
         if (!attachment.valid())
@@ -500,6 +507,27 @@ inline bool IRenderpass::SCreationParams::SColorAttachmentDescription::valid() c
 }
 
 
+inline bool IRenderpass::SCreationParams::SSubpassDescription::operator!=(const SSubpassDescription& other) const
+{
+    if (depthStencilAttachment!=other.depthStencilAttachment)
+        return true;
+    for (auto i=0u; i<MaxColorAttachments; i++)
+    if (colorAttachments[i]!=other.colorAttachments[i])
+        return true;
+    auto tokenTerminatedSequenceUnequal = []<typename T>(const T* lhs, const T* rhs, const T& endToken) -> bool
+    {
+        while ((*lhs)!=endToken && (*rhs)!=endToken)
+        if ((*lhs)!=(*rhs))
+            return true;
+        return (*lhs)!=(*rhs);
+    };
+    if (tokenTerminatedSequenceUnequal(inputAttachments,other.inputAttachments,InputAttachmentsEnd))
+        return true;
+    if (tokenTerminatedSequenceUnequal(preserveAttachments,other.preserveAttachments,PreserveAttachmentsEnd))
+        return true;
+    return viewMask!=other.viewMask || flags!=other.flags;
+}
+
 inline bool IRenderpass::SCreationParams::SSubpassDescription::valid(const SCreationParams& params, const uint32_t depthStencilAttachmentCount, const uint32_t colorAttachmentCount) const
 {
     // TODO: https://registry.khronos.org/vulkan/specs/1.3-extensions/html/vkspec.html#VUID-VkSubpassDescription2-pNext-06870

include/nbl/asset/ICPUFramebuffer.h

@@ -1,79 +1,76 @@
 // Copyright (C) 2018-2020 - 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_ASSET_E_COLOR_SPACE_H_INCLUDED_
+#define _NBL_ASSET_E_COLOR_SPACE_H_INCLUDED_
 
-#ifndef __NBL_ASSET_E_COLOR_SPACE_H_INCLUDED__
-#define __NBL_ASSET_E_COLOR_SPACE_H_INCLUDED__
-
-namespace nbl
+// TODO: move to HLSL and turn into enum classes (when 8bit int comes)
+namespace nbl::asset
 {
-	namespace asset
-	{
-        //! Specifies a color space of an image
-        enum E_COLOR_PRIMARIES
-        {
-            //! Specifies support for the sRGB color space. The primaries are the same for scRGB and BT709, only EOTFs differ.
-            ECP_SRGB,
-
-            //! Specifies support for the Display-P3 color space to be displayed using an sRGB-like EOTF.
-            ECP_DISPLAY_P3,
-
-            //! Specifies support for the DCI-P3 color space to be displayed using the DCI-P3 EOTF. Note that values in such an image are interpreted as XYZ encoded color data by the presentation engine.
-            ECP_DCI_P3,
-
-            //! Specifies support for the BT2020 color space to be displayed using a linear EOTF. Same primaries are used for HDR10 and DolbyVision
-            ECP_BT2020,
-
-            //! Specifies support for the AdobeRGB color space to be displayed.
-            ECP_ADOBERGB,
+//! Specifies a color space of an image
+enum E_COLOR_PRIMARIES : uint8_t
+{
+    //! Specifies support for the sRGB color space. The primaries are the same for scRGB and BT709, only EOTFs differ.
+    ECP_SRGB,
 
-            //! The reference ACES color space, not really supported by any graphics API for display/swapchain
-            ECP_ACES,
+    //! Specifies support for the Display-P3 color space to be displayed using an sRGB-like EOTF.
+    ECP_DISPLAY_P3,
 
-            //! The slightly different primary space for ACES with quantization (ACEScc and ACEScct use same primaries)
-            ECP_ACES_CC_T,
+    //! Specifies support for the DCI-P3 color space to be displayed using the DCI-P3 EOTF. Note that values in such an image are interpreted as XYZ encoded color data by the presentation engine.
+    ECP_DCI_P3,
 
-            //! Specifies that color components are used �as is�. This is intended to allow applications to supply data for color spaces not described here.
-            ECP_PASS_THROUGH,
+    //! Specifies support for the BT2020 color space to be displayed using a linear EOTF. Same primaries are used for HDR10 and DolbyVision
+    ECP_BT2020,
 
-            //! For internal 
-            ECP_COUNT
-        };
+    //! Specifies support for the AdobeRGB color space to be displayed.
+    ECP_ADOBERGB,
 
-        //! Data to linear value for images
-        enum ELECTRO_OPTICAL_TRANSFER_FUNCTION
-        {
-            EOTF_IDENTITY,
-            EOTF_sRGB,
-            EOTF_DCI_P3_XYZ,
-            EOTF_SMPTE_170M,
-            EOTF_SMPTE_ST2084,
-            EOTF_HDR10_HLG,
-            EOTF_GAMMA_2_2,
-            EOTF_ACEScc,
-            EOTF_ACEScct,
+    //! The reference ACES color space, not really supported by any graphics API for display/swapchain
+    ECP_ACES,
 
-            EOTF_UNKNOWN
-        };
+    //! The slightly different primary space for ACES with quantization (ACEScc and ACEScct use same primaries)
+    ECP_ACES_CC_T,
 
-        //! Linear value to data for displays and swapchains
-        enum OPTICO_ELECTRICAL_TRANSFER_FUNCTION
-        {
-            OETF_IDENTITY,
-            OETF_sRGB,
-            OETF_DCI_P3_XYZ,
-            OETF_SMPTE_170M,
-            OETF_SMPTE_ST2084,
-            OETF_HDR10_HLG,
-            OETF_GAMMA_2_2,
-            OETF_ACEScc,
-            OETF_ACEScct,
+    //! Specifies that color components are used �as is�. This is intended to allow applications to supply data for color spaces not described here.
+    ECP_PASS_THROUGH,
 
-            OETF_UNKNOWN
-        };
+    //! For internal 
+    ECP_COUNT
+};
 
-        static_assert(EOTF_UNKNOWN == static_cast<ELECTRO_OPTICAL_TRANSFER_FUNCTION>(OETF_UNKNOWN), "Definitions of transfer functions don't match");
-	}
+//! Data to linear value for images
+enum ELECTRO_OPTICAL_TRANSFER_FUNCTION : uint8_t
+{
+    EOTF_IDENTITY,
+    EOTF_sRGB,
+    EOTF_DCI_P3_XYZ,
+    EOTF_SMPTE_170M,
+    EOTF_SMPTE_ST2084,
+    EOTF_HDR10_HLG,
+    EOTF_GAMMA_2_2,
+    EOTF_ACEScc,
+    EOTF_ACEScct,
+
+    EOTF_UNKNOWN
+};
+
+//! Linear value to data for displays and swapchains
+enum OPTICO_ELECTRICAL_TRANSFER_FUNCTION : uint8_t
+{
+    OETF_IDENTITY,
+    OETF_sRGB,
+    OETF_DCI_P3_XYZ,
+    OETF_SMPTE_170M,
+    OETF_SMPTE_ST2084,
+    OETF_HDR10_HLG,
+    OETF_GAMMA_2_2,
+    OETF_ACEScc,
+    OETF_ACEScct,
+
+    OETF_UNKNOWN
+};
+
+static_assert(EOTF_UNKNOWN == static_cast<ELECTRO_OPTICAL_TRANSFER_FUNCTION>(OETF_UNKNOWN), "Definitions of transfer functions don't match");
 }
 
 #endif

include/nbl/asset/IFramebuffer.h

@@ -71,6 +71,11 @@ inline matrix<T,M,N> transpose(const matrix<T,N,M>& m)
     return 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/asset/IRenderpass.h

@@ -6,6 +6,7 @@
 
 #include "nbl/builtin/hlsl/cpp_compat.hlsl"
 #include "nbl/builtin/hlsl/spirv_intrinsics/core.hlsl"
+#include "nbl/builtin/hlsl/type_traits.hlsl"
 
 namespace nbl 
 {
@@ -18,7 +19,7 @@ namespace glsl
 template<typename T>
 T atomicAdd(NBL_REF_ARG(T) ptr, T value)
 {
-    return spirv::atomicAnd<T>(ptr, spv::ScopeDevice, spv::DecorationRelaxedPrecision, value);
+    return spirv::atomicAdd<T>(ptr, spv::ScopeDevice, spv::DecorationRelaxedPrecision, value);
 }
 template<typename T>
 T atomicAnd(NBL_REF_ARG(T) ptr, T value)
@@ -83,6 +84,50 @@ void tess_ctrl_barrier() {
 void memoryBarrierShared() {
     spirv::memoryBarrier(spv::ScopeDevice, spv::MemorySemanticsAcquireReleaseMask | spv::MemorySemanticsWorkgroupMemoryMask);
 }
+
+namespace impl 
+{
+
+template<typename T, bool isSigned, bool isIntegral>
+struct bitfieldExtract {};
+
+template<typename T, bool isSigned>
+struct bitfieldExtract<T, isSigned, false>
+{
+    T operator()( T val, uint32_t offsetBits, uint32_t numBits )
+    {
+        static_assert( is_integral<T>::value, "T is not an integral type!" );
+        return val;
+    }
+};
+
+template<typename T>
+struct bitfieldExtract<T, true, true>
+{
+    T operator()( T val, uint32_t offsetBits, uint32_t numBits )
+    {
+        return spirv::bitFieldSExtract<T>( val, offsetBits, numBits );
+    }
+};
+
+template<typename T>
+struct bitfieldExtract<T, false, true>
+{
+    T operator()( T val, uint32_t offsetBits, uint32_t numBits )
+    {
+        return spirv::bitFieldUExtract<T>( val, offsetBits, numBits );
+    } 
+};
+
+}
+
+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 
+        ( val, offsetBits, numBits );
+}
+
 #endif
 
 }

include/nbl/asset/format/EColorSpace.h

@@ -129,6 +129,15 @@ void memoryBarrier(uint32_t memoryScope, uint32_t memorySemantics);
 template<class T, class U>
 [[vk::ext_instruction(spv::OpBitcast)]]
 T bitcast(U);
+
+template<typename Unsigned>
+[[vk::ext_instruction( spv::OpBitFieldUExtract )]]
+Unsigned bitFieldUExtract( Unsigned val, uint32_t offsetBits, uint32_t numBits );
+
+template<typename Signed>
+[[vk::ext_instruction( spv::OpBitFieldSExtract )]]
+Signed bitFieldSExtract( Signed val, uint32_t offsetBits, uint32_t numBits );
+
 }
 #endif
 }