Merge branch 'master' into nahim_complex_fft_workgroup

Author: Fletterio

CONTRIBUTING.md

@@ -60,7 +60,7 @@ Below are some tasks that are open for contribution. If you start working on a t
 
  **[Issues](https://github.com/Devsh-Graphics-Programming/Nabla/issues) also count as free tasks.** We also have old issues from previous generation **[here](https://github.com/buildaworldnet/IrrlichtBAW/issues?q=is%3Aissue+is%3Aopen+sort%3Aupdated-desc)**.
 
-If you have questions about the list, you can ask directly on our [Discord server](https://discord.gg/SsJzqS23) or open an issue [here](https://github.com/Devsh-Graphics-Programming/Nabla/issues).
+If you have questions about the list, you can ask directly on our [Discord server](https://discord.gg/Dx83Cu7suD) or open an issue [here](https://github.com/Devsh-Graphics-Programming/Nabla/issues).
 
 ## Pull Request Process
 
@@ -78,7 +78,7 @@ If you have questions about the list, you can ask directly on our [Discord serve
 
 ## Connect with Other Project Contributors
 
-Join our [Discord server](https://discord.gg/SsJzqS23) to connect with other contributors. It's the first place to go for questions and discussions about the project.
+Join our [Discord server](https://discord.gg/Dx83Cu7suD) to connect with other contributors. It's the first place to go for questions and discussions about the project.
 
 ## License
 

examples_tests

@@ -46,10 +46,10 @@ class IDescriptorSetLayout : public virtual core::IReferenceCounted  // TODO: tr
 		enum class E_CREATE_FLAGS : uint8_t
 		{
 			ECF_NONE = 0,
-			ECF_UPDATE_AFTER_BIND_BIT = 1u << 1,
-			ECF_UPDATE_UNUSED_WHILE_PENDING_BIT = 1u << 2,
-			ECF_PARTIALLY_BOUND_BIT = 1u << 3,
-			ECF_VARIABLE_DESCRIPTOR_COUNT_BIT = 1u << 4
+			ECF_UPDATE_AFTER_BIND_BIT = 1u << 0,
+			ECF_UPDATE_UNUSED_WHILE_PENDING_BIT = 1u << 1,
+			ECF_PARTIALLY_BOUND_BIT = 1u << 2,
+			ECF_VARIABLE_DESCRIPTOR_COUNT_BIT = 1u << 3
 		};
 
 		uint32_t binding;

include/nbl/asset/IDescriptorSetLayout.h

@@ -6,6 +6,7 @@
 
 #include <nbl/builtin/hlsl/type_traits.hlsl>
 #include <nbl/builtin/hlsl/bit.hlsl>
+#include <nbl/builtin/hlsl/macros.h>
 
 // C++ headers
 #ifndef __HLSL_VERSION
@@ -127,7 +128,7 @@ struct num_base : type_identity<T>
     NBL_CONSTEXPR_STATIC_INLINE int32_t float_max_decimal_exponent = 4*S16 + 30*S32 + 232*S64;
 
     NBL_CONSTEXPR_STATIC_INLINE int32_t float_exponent_bits = 8 * size - float_digits - 1;
-    NBL_CONSTEXPR_STATIC_INLINE int32_t float_max_exponent = 1l << float_exponent_bits;
+    NBL_CONSTEXPR_STATIC_INLINE int32_t float_max_exponent = 1 << float_exponent_bits;
     NBL_CONSTEXPR_STATIC_INLINE int32_t float_min_exponent = 3 - float_max_exponent;
     NBL_CONSTEXPR_STATIC_INLINE bool is_bool = is_same<T, bool>::value;
 
@@ -235,9 +236,9 @@ struct num_traits<float32_t> : num_base<float32_t>
 template<> 
 struct num_traits<float64_t> : num_base<float64_t>
 {
-    NBL_CONSTEXPR_STATIC_INLINE float64_t max           = 1.7976931348623158e+308;
-    NBL_CONSTEXPR_STATIC_INLINE float64_t min           = 2.2250738585072014e-308;
-    NBL_CONSTEXPR_STATIC_INLINE float64_t denorm_min    = 4.9406564584124654e-324;
+    NBL_CONSTEXPR_STATIC_INLINE float64_t max           = NBL_FP64_LITERAL(1.7976931348623158e+308);
+    NBL_CONSTEXPR_STATIC_INLINE float64_t min           = NBL_FP64_LITERAL(2.2250738585072014e-308);
+    NBL_CONSTEXPR_STATIC_INLINE float64_t denorm_min    = NBL_FP64_LITERAL(4.9406564584124654e-324);
     NBL_CONSTEXPR_STATIC_INLINE uint64_t  quiet_NaN     = 0x7FF8000000000000ull;
     NBL_CONSTEXPR_STATIC_INLINE uint64_t  signaling_NaN = 0x7FF0000000000001ull;
     NBL_CONSTEXPR_STATIC_INLINE uint64_t  infinity      = 0x7FF0000000000000ull;

include/nbl/builtin/hlsl/limits.hlsl

@@ -17,6 +17,12 @@
 } 
 #endif
 
+#ifdef __HLSL_VERSION // cause DXC is insane
+#define NBL_FP64_LITERAL(LIT) LIT##l
+#else // and nobody except GCC supports C++23 `f64` suffix on float literals
+#define NBL_FP64_LITERAL(LIT) LIT
+#endif
+
 // basics
 #define NBL_EVAL(...) __VA_ARGS__
 #define NBL_CONCAT_IMPL2(X,Y) X ## Y
@@ -159,4 +165,4 @@
 
 #define NBL_FOREACH(WHAT, ... ) NBL_EVAL(NBL_CONCATENATE(NBL_FOREACH_,NBL_VA_ARGS_COUNT(__VA_ARGS__))(WHAT, __VA_ARGS__))
 
-#endif
+#endif

include/nbl/builtin/hlsl/macros.h

@@ -1027,6 +1027,16 @@ inline constexpr VkDescriptorType getVkDescriptorTypeFromDescriptorType(const as
     }
 }
 
+inline VkDescriptorBindingFlagBits getVkDescriptorBindingFlagsFrom(const core::bitflag<IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS> flags)
+{
+    // Wait for C++23
+    //static_assert(std::to_underlying(IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS::ECF_UPDATE_AFTER_BIND_BIT)==VK_DESCRIPTOR_BINDING_UPDATE_AFTER_BIND_BIT);
+    //static_assert(IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS::ECF_UPDATE_AFTER_BIND_BIT==VK_DESCRIPTOR_BINDING_UPDATE_UNUSED_WHILE_PENDING_BIT);
+    //static_assert(IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS::ECF_UPDATE_AFTER_BIND_BIT==VK_DESCRIPTOR_BINDING_PARTIALLY_BOUND_BIT);
+    //static_assert(IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS::ECF_UPDATE_AFTER_BIND_BIT==VK_DESCRIPTOR_BINDING_VARIABLE_DESCRIPTOR_COUNT_BIT);
+    return static_cast<VkDescriptorBindingFlagBits>(flags.value);
+}
+
 inline IPhysicalDevice::E_DRIVER_ID getDriverIdFromVkDriverId(const VkDriverId in)
 {
     if(in == VK_DRIVER_ID_AMD_PROPRIETARY) return IPhysicalDevice::E_DRIVER_ID::EDI_AMD_PROPRIETARY;

include/nbl/video/CVulkanCommon.h

@@ -102,12 +102,14 @@ class NBL_API2 IDescriptorPool : public IBackendObject
             switch (type)
             {
             case asset::IDescriptor::E_TYPE::ET_SAMPLER:
-                baseAddress = reinterpret_cast<core::smart_refctd_ptr<asset::IDescriptor>*>(m_mutableStandaloneSamplerStorage.get());
+                baseAddress = reinterpret_cast<core::smart_refctd_ptr<asset::IDescriptor>*>(m_samplerStorage.get());
                 break;
             case asset::IDescriptor::E_TYPE::ET_COMBINED_IMAGE_SAMPLER:
-            case asset::IDescriptor::E_TYPE::ET_SAMPLED_IMAGE:
                 baseAddress = reinterpret_cast<core::smart_refctd_ptr<asset::IDescriptor>*>(m_textureStorage.get());
                 break;
+            case asset::IDescriptor::E_TYPE::ET_SAMPLED_IMAGE:
+                baseAddress = reinterpret_cast<core::smart_refctd_ptr<asset::IDescriptor>*>(m_textureStorage.get()) + m_creationParameters.maxDescriptorCount[static_cast<uint32_t>(asset::IDescriptor::E_TYPE::ET_COMBINED_IMAGE_SAMPLER)];
+                break;
             case asset::IDescriptor::E_TYPE::ET_STORAGE_IMAGE:
                 baseAddress = reinterpret_cast<core::smart_refctd_ptr<asset::IDescriptor>*>(m_storageImageStorage.get());
                 break;
@@ -145,7 +147,7 @@ class NBL_API2 IDescriptorPool : public IBackendObject
 
         inline core::smart_refctd_ptr<IGPUSampler>* getMutableCombinedSamplerStorage() const
         {
-            return reinterpret_cast<core::smart_refctd_ptr<IGPUSampler>*>(m_mutableCombinedSamplerStorage.get());
+            return reinterpret_cast<core::smart_refctd_ptr<IGPUSampler>*>(m_samplerStorage.get()) + m_creationParameters.maxDescriptorCount[static_cast<uint32_t>(asset::IDescriptor::E_TYPE::ET_SAMPLER)];
         }
 
         friend class IGPUDescriptorSet;
@@ -222,8 +224,7 @@ class NBL_API2 IDescriptorPool : public IBackendObject
         std::unique_ptr<IGPUDescriptorSet* []> m_allocatedDescriptorSets = nullptr; // This array might be sparse.
 
         std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<video::IGPUImageView>>[]> m_textureStorage;
-        std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<video::IGPUSampler>>[]> m_mutableCombinedSamplerStorage;
-        std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<video::IGPUSampler>>[]> m_mutableStandaloneSamplerStorage;
+        std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<video::IGPUSampler>>[]> m_samplerStorage;
         std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<IGPUImageView>>[]> m_storageImageStorage; // storage image | input attachment
         std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<IGPUBuffer>>[]> m_UBO_SSBOStorage; // ubo | ssbo | ubo dynamic | ssbo dynamic
         std::unique_ptr<core::StorageTrivializer<core::smart_refctd_ptr<IGPUBufferView>>[]> m_UTB_STBStorage; // utb | stb

include/nbl/video/IDescriptorPool.h

@@ -631,6 +631,7 @@ class NBL_API2 ILogicalDevice : public core::IReferenceCounted, public IDeviceMe
             {
                 if (layout)
                 {
+                    // TODO: when creating the layouts, cache if they have any update after bindingings, and patch `createInfo.flags` with that here
                     const auto setCount = setCounts ? *(setCountsIt):1u;
                     createInfo.maxSets += setCount;
                     for (uint32_t t=0; t<static_cast<uint32_t>(asset::IDescriptor::E_TYPE::ET_COUNT); ++t)

include/nbl/video/ILogicalDevice.h

@@ -71,7 +71,7 @@ struct SPhysicalDeviceFeatures
     bool largePoints = false;
 
     // Some AMD don't support
-    bool alphaToOne = true;
+    bool alphaToOne = false;
 
     // [REQUIRE] good device support
     //bool multiViewport = true;

include/nbl/video/SPhysicalDeviceFeatures.h

@@ -539,9 +539,12 @@ core::smart_refctd_ptr<IGPUDescriptorSetLayout> CVulkanLogicalDevice::createDesc
 {
     std::vector<VkSampler> vk_samplers;
     std::vector<VkDescriptorSetLayoutBinding> vk_dsLayoutBindings;
+    std::vector<VkDescriptorBindingFlags> vk_bindingFlags;
     vk_samplers.reserve(maxSamplersCount); // Reserve to avoid resizing and pointer change while iterating 
     vk_dsLayoutBindings.reserve(bindings.size());
+    vk_bindingFlags.reserve(bindings.size());
 
+    bool updateAfterBindFound = false;
     for (const auto& binding : bindings)
     {
         auto& vkDescSetLayoutBinding = vk_dsLayoutBindings.emplace_back();
@@ -560,13 +563,22 @@ core::smart_refctd_ptr<IGPUDescriptorSetLayout> CVulkanLogicalDevice::createDesc
                 vk_samplers.push_back(static_cast<const CVulkanSampler*>(binding.immutableSamplers[i].get())->getInternalObject());
             vkDescSetLayoutBinding.pImmutableSamplers = vk_samplers.data()+samplerOffset;
         }
+
+        if (binding.createFlags.hasFlags(IGPUDescriptorSetLayout::SBinding::E_CREATE_FLAGS::ECF_UPDATE_AFTER_BIND_BIT))
+            updateAfterBindFound = true;
+        vk_bindingFlags.emplace_back() = getVkDescriptorBindingFlagsFrom(binding.createFlags);
     }
 
-    VkDescriptorSetLayoutCreateInfo vk_createInfo = { VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO };
-    vk_createInfo.pNext = nullptr; // pNext of interest:  VkDescriptorSetLayoutBindingFlagsCreateInfo
-    vk_createInfo.flags = 0; // Todo(achal): I would need to create a IDescriptorSetLayout::SCreationParams for this
-    vk_createInfo.bindingCount = vk_dsLayoutBindings.size();
+    VkDescriptorSetLayoutBindingFlagsCreateInfo vk_bindingFlagsInfo = { VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_BINDING_FLAGS_CREATE_INFO, nullptr };
+    VkDescriptorSetLayoutCreateInfo vk_createInfo = { VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO, &vk_bindingFlagsInfo };
+    // Todo(achal): I would need to create a IDescriptorSetLayout::SCreationParams for this
+    // Answer: We don't actually support any extensions/features that would necessitate exposing any other flag than update_after_bind
+    vk_createInfo.flags = 0;
+    if (updateAfterBindFound)
+        vk_createInfo.flags |= VK_DESCRIPTOR_SET_LAYOUT_CREATE_UPDATE_AFTER_BIND_POOL_BIT;
+    vk_createInfo.bindingCount = vk_bindingFlagsInfo.bindingCount = vk_dsLayoutBindings.size();
     vk_createInfo.pBindings = vk_dsLayoutBindings.data();
+    vk_bindingFlagsInfo.pBindingFlags = vk_bindingFlags.data();
 
     VkDescriptorSetLayout vk_dsLayout;
     if (m_devf.vk.vkCreateDescriptorSetLayout(m_vkdev,&vk_createInfo,nullptr,&vk_dsLayout)==VK_SUCCESS)