Assets/Shaders/ParticleUpdate_CS.glsl

@@ -118,6 +118,9 @@ void main()
 
   if (particle.life > 0.0f)
   {
+    // update position from velocity
+    // particle.position += particle.velocity * genericDataBuffer.data.dt;
+    // particle.life -= genericDataBuffer.data.dt;
     
     if (particle.life < 0.0f || particle.scaleAndDecay.x < 0.0f || particle.scaleAndDecay.y < 0.0f)
     {

Assets/Shaders/Particle_VS.glsl

@@ -78,8 +78,11 @@ void main()
   vec2 particleScaleData = particle.scaleAndDecay.xz; // x and y
 
 	mat4 localModel = mat4 (1.0f);
-  localModel[0][0] = particleScaleData.x;
+  localModel[0][0] = 1.0f;
-  localModel[1][1] = particleScaleData.y;
+  localModel[1][1] = 1.0f;
+
+  // localModel[0][0] = particleScaleData.x;
+  // localModel[1][1] = particleScaleData.y;
 
   mat4 rotate = mat4(1.0f);
 	rotate[0][0] = cos(angle);
@@ -89,6 +92,6 @@ void main()
 	// localModel = rotate * localModel;
 
 	localModel[3] = vec4 (particle.position.xyz, 1.0f);
-  gl_Position = cameraData.vpMat * localModel * vec4(vertexPos, 1.0f);
+  // gl_Position = cameraData.vpMat * localModel * vec4(vertexPos, 1.0f);
-  // gl_Position =  vec4(vertexPos, 1.0f);
+  gl_Position =  vec4(vertexPos, 1.0f);
 }

SHADE_Engine/src/Graphics/MiddleEnd/GlobalData/SHGraphicsPredefinedData.cpp

@@ -251,7 +251,7 @@ namespace SHADE
     // particle draw call binding
     SHVkDescriptorSetLayout::Binding particleDrawDataBinding
     {
-      .Type = vk::DescriptorType::eStorageBufferDynamic, // UBO (Because lesser data), dynamic (1 set for each frame)
+      .Type = vk::DescriptorType::eStorageBufferDynamic, // SSBO, dynamic (1 set for each frame)
       .Stage = vk::ShaderStageFlagBits::eCompute,
       .BindPoint = SHGraphicsConstants::DescriptorSetBindings::PARTICLE_DRAW_DATA,
       .DescriptorCount = 1,

SHADE_Engine/src/Graphics/MiddleEnd/Interface/SHGraphicsSystem.cpp

@@ -707,9 +707,11 @@ namespace SHADE
 #endif
     }
 
+
     renderGraph->Begin(frameIndex);
     auto cmdBuffer = renderGraph->GetCommandBuffer(frameIndex);
 
+    particleSubSystem->ResetInstanceCounts(cmdBuffer, frameIndex);
     particleSubSystem->Run(cmdBuffer, frameIndex);
 
     renderGraph->Execute(frameIndex, descPool, MESH_DATA);

SHADE_Engine/src/Graphics/MiddleEnd/Particles/SHParticleSubSystem.cpp

@@ -14,6 +14,8 @@
 #include "Graphics/Pipeline/SHVkPipeline.h"
 #include "Graphics/RenderGraph/SHSubpass.h"
 #include "Graphics/SHVkUtil.h"
+#include "Graphics/Synchronization/SHVkFence.h"
+
 
 namespace SHADE
 {
@@ -68,7 +70,7 @@ namespace SHADE
       }
 
       // buffer to store draw call data. Non-indexed, host-visible mapped, triple buffered. 
-      comp.drawCallData = logicalDevice->CreateBuffer(SHGraphicsConstants::NUM_FRAME_BUFFERS * sizeofIndirectCmdAligned, indirectCommands.data(), SHGraphicsConstants::NUM_FRAME_BUFFERS * sizeofIndirectCmdAligned, vk::BufferUsageFlagBits::eStorageBuffer | vk::BufferUsageFlagBits::eIndirectBuffer, VMA_MEMORY_USAGE_AUTO, VmaAllocationCreateFlagBits::VMA_ALLOCATION_CREATE_HOST_ACCESS_SEQUENTIAL_WRITE_BIT | VmaAllocationCreateFlagBits::VMA_ALLOCATION_CREATE_MAPPED_BIT);
+      comp.drawCallData = logicalDevice->CreateBuffer(SHGraphicsConstants::NUM_FRAME_BUFFERS * sizeofIndirectCmdAligned, indirectCommands.data(), SHGraphicsConstants::NUM_FRAME_BUFFERS * sizeofIndirectCmdAligned, vk::BufferUsageFlagBits::eStorageBuffer | vk::BufferUsageFlagBits::eIndirectBuffer, VMA_MEMORY_USAGE_AUTO, VmaAllocationCreateFlagBits::VMA_ALLOCATION_CREATE_HOST_ACCESS_RANDOM_BIT | VmaAllocationCreateFlagBits::VMA_ALLOCATION_CREATE_MAPPED_BIT);
 
     }
 
@@ -151,12 +153,6 @@ namespace SHADE
   {
     auto const& mappings = SHGraphicsPredefinedData::GetMappings(SHGraphicsPredefinedData::SystemType::PARTICLE_RENEDERING);
 
-    uint32_t instanceCountOffset = sizeof (vk::DrawIndirectCommand) * frameIndex + offsetof(vk::DrawIndirectCommand, instanceCount);
-    uint32_t ZERO = 0;
-
-    // reset instance count to 0
-    comp.drawCallData->WriteToMemory (&ZERO, sizeof(uint32_t), 0, instanceCountOffset);
-
     // bind the descriptor sets required for emitting particles
     cmdBuffer->BindDescriptorSet(comp.particleDescriptorSet, SH_PIPELINE_TYPE::COMPUTE, mappings.at(SHPredefinedDescriptorTypes::PARTICLES), comp.dynamicOffsets[frameIndex]);
 
@@ -170,7 +166,7 @@ namespace SHADE
     //cmdBuffer->DrawArrays(4, 1, 0, 0);
   }
 
-  void SHParticleSubSystem::PreparePrePostUpdateBarriers(std::vector<vk::BufferMemoryBarrier>& preUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& postUpdateBarriers, SHParticleEmitterComponent const& emitter, uint32_t const EMITTER_INDEX, uint32_t const FRAME_INDEX) noexcept
+  void SHParticleSubSystem::PreparePrePostUpdateBarriers(std::vector<vk::BufferMemoryBarrier>& preUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& postUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& preDrawBarriers, SHParticleEmitterComponent const& emitter, uint32_t const EMITTER_INDEX, uint32_t const FRAME_INDEX) noexcept
   {
     // pre-update particles data barrier. Note that this is for input because we want the input to be available before we use it.
     vk::BufferMemoryBarrier particleDataBarrier
@@ -219,18 +215,17 @@ namespace SHADE
     // make new barrier on stack...
     vk::BufferMemoryBarrier drawDataBarrier
     {
-      .srcAccessMask = vk::AccessFlagBits::eShaderWrite,
+      .srcAccessMask = vk::AccessFlagBits::eShaderWrite | vk::AccessFlagBits::eMemoryWrite,
-      .dstAccessMask = vk::AccessFlagBits::eShaderRead,
+      .dstAccessMask = vk::AccessFlagBits::eIndirectCommandRead,
       .buffer = emitter.drawCallData->GetVkBuffer(),
       .offset = emitter.dynamicOffsets[FRAME_INDEX][DYOFF_INDEX_DRAW_DATA],
       .size = static_cast<uint32_t>(sizeof(vk::DrawIndirectCommand))
     };
 
     // ...copy assign barriers on heap 
-    postUpdateBarriers[EMITTER_INDEX * 3] = particleDataBarrierPost;
+    postUpdateBarriers[EMITTER_INDEX * 2] = particleDataBarrierPost;
-    postUpdateBarriers[(EMITTER_INDEX * 3) + 1] = indicesDataBarrier;
+    postUpdateBarriers[(EMITTER_INDEX * 2) + 1] = indicesDataBarrier;
-    postUpdateBarriers[(EMITTER_INDEX * 3) + 2] = drawDataBarrier;
+    preDrawBarriers[EMITTER_INDEX] = drawDataBarrier;
-
   }
 
   void SHParticleSubSystem::Init(Handle<SHVkLogicalDevice> device, Handle<SHVkDescriptorPool> inDescPool, Handle<SHVkRenderpass> compatibleRenderpass, Handle<SHSubpass> subpass, Handle<SHVkShaderModule> VS, Handle<SHVkShaderModule> FS, Handle<SHVkShaderModule> emitCS, Handle<SHVkShaderModule> defaultUpdateCS) noexcept
@@ -310,7 +305,7 @@ namespace SHADE
     SHComponentManager::CreateComponentSparseSet<SHParticleEmitterComponent>();
   }
 
-  void SHParticleSubSystem::Run(Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex) noexcept
+  void SHParticleSubSystem::Run(Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex, Handle<SHVkFence> waitFence) noexcept
   {
     float dt = static_cast<float>(SHFrameRateController::GetRawDeltaTime());
 
@@ -332,7 +327,11 @@ namespace SHADE
     // After we invoke the updates for the emitters, we need to make sure all particles and indices data are done updating
     // before we issue them for rendering. 
     std::vector<vk::BufferMemoryBarrier> postUpdateBarriers{};
-    postUpdateBarriers.resize(emitters.size() * 3);
+    postUpdateBarriers.resize(emitters.size() * 2);
+
+    std::vector<vk::BufferMemoryBarrier> preDrawBarriers{};
+    preDrawBarriers.resize(emitters.size());
+
 
     // If we wanted to be VERY safe, a barrier would be good here to make sure output particles have finish reading input particles in 
     // the update compute. HOWEVER since a NUM_FRAME_BUFFERS frames would have passed by then, we will not insert 1 here.
@@ -350,6 +349,10 @@ namespace SHADE
     SHGlobalDescriptorSets::BindGenericAndTextureData(logicalDevice, cmdBuffer, SH_PIPELINE_TYPE::COMPUTE, mappings.at(SHPredefinedDescriptorTypes::STATIC_DATA), frameIndex);
 
     uint32_t i = 0;
+
+    if (waitFence)
+      waitFence->Wait(true);
+
     for (auto& emitter : emitters)
     {
       if (!emitter.initialized)
@@ -382,7 +385,7 @@ namespace SHADE
       }
 
       // prepare barriers
-      PreparePrePostUpdateBarriers(preUpdateBarriers, postUpdateBarriers, emitter, i, frameIndex);
+      PreparePrePostUpdateBarriers(preUpdateBarriers, postUpdateBarriers, preDrawBarriers, emitter, i, frameIndex);
 
       // Emitter will emit once and stop emitting next frame (unless specified before reaching here to do so).
       emitter.toEmit = false;
@@ -398,6 +401,7 @@ namespace SHADE
     /* EMITTING PARTICLES DONE, BEGIN UPDATES....                            */
     /*-----------------------------------------------------------------------*/
 
+
     // bind the pipeline for updating 
     cmdBuffer->BindPipeline(defaultUpdatePipelineData.pipeline);
 
@@ -411,9 +415,53 @@ namespace SHADE
     /*-----------------------------------------------------------------------*/
     // issue the barrier to wait for output particles to be done updating and indices data to finish being modified.
     cmdBuffer->PipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, vk::PipelineStageFlagBits::eComputeShader, {}, {}, postUpdateBarriers, {});
+    cmdBuffer->PipelineBarrier(vk::PipelineStageFlagBits::eComputeShader, vk::PipelineStageFlagBits::eDrawIndirect, {}, {}, preDrawBarriers, {});
 
   }
 
+  void SHParticleSubSystem::ResetInstanceCounts(Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex) noexcept
+  {
+    auto& emitters = SHComponentManager::GetDense<SHParticleEmitterComponent>();
+    std::vector<vk::BufferMemoryBarrier> preResetBarriers{};
+    preResetBarriers.resize(emitters.size());
+
+    uint32_t i = 0;
+    for (auto& emitter : emitters)
+    {
+      if (emitter.initialized)
+      {
+        // pre-update particles data barrier. Note that this is for input because we want the input to be available before we use it.
+        vk::BufferMemoryBarrier drawDataHostWriteBarrier
+        {
+          .srcAccessMask = vk::AccessFlagBits::eIndirectCommandRead | vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
+          .dstAccessMask = vk::AccessFlagBits::eMemoryWrite | vk::AccessFlagBits::eHostWrite,
+          .buffer = emitter.drawCallData->GetVkBuffer(),
+          .offset = emitter.dynamicOffsets[frameIndex][DYOFF_INDEX_DRAW_DATA],
+          .size = static_cast<uint32_t>(sizeof(vk::DrawIndirectCommand))
+        };
+
+        preResetBarriers[i] = drawDataHostWriteBarrier;
+      }
+      ++i;
+    }
+
+    cmdBuffer->PipelineBarrier(vk::PipelineStageFlagBits::eDrawIndirect, vk::PipelineStageFlagBits::eHost, {}, {}, preResetBarriers, {});
+
+
+    for (auto& emitter : emitters)
+    {
+      if (emitter.initialized)
+      {
+        //uint32_t instanceCountOffset = 20;
+        uint32_t instanceCountOffset = sizeof(vk::DrawIndirectCommand) * frameIndex + offsetof(vk::DrawIndirectCommand, instanceCount);
+        uint32_t ZERO = 0;
+
+        // reset instance count to 0
+        emitter.drawCallData->WriteToMemory(&ZERO, sizeof(uint32_t), 0, instanceCountOffset);
+      }
+    }
+  }
+
   void SHParticleSubSystem::Render(Handle<SHVkCommandBuffer> cmdBuffer, Handle<SHRenderer> renderer, uint32_t frameIndex) noexcept
   {
     auto& emitters = SHComponentManager::GetDense<SHParticleEmitterComponent>();
@@ -432,6 +480,7 @@ namespace SHADE
     }
   }
 
+
   void SHParticleSubSystem::Exit(void) noexcept
   {
 

SHADE_Engine/src/Graphics/MiddleEnd/Particles/SHParticleSubSystem.h

@@ -20,6 +20,7 @@ namespace SHADE
   class SHVkShaderModule;
   class SHRenderer;
   class SHParticleEmitterComponent;
+  class SHVkFence;
 
 
 
@@ -81,12 +82,13 @@ namespace SHADE
     void UpdateCompoennt(Handle<SHVkCommandBuffer> cmdBuffer, SHParticleEmitterComponent& comp, uint32_t frameIndex) noexcept;
     void RenderComponent(Handle<SHVkCommandBuffer> cmdBuffer, SHParticleEmitterComponent& comp, uint32_t frameIndex) noexcept;
 
-    void PreparePrePostUpdateBarriers (std::vector<vk::BufferMemoryBarrier>& preUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& postUpdateBarriers, SHParticleEmitterComponent const& emitter, uint32_t const EMITTER_INDEX, uint32_t const FRAME_INDEX) noexcept;
+    void PreparePrePostUpdateBarriers (std::vector<vk::BufferMemoryBarrier>& preUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& postUpdateBarriers, std::vector<vk::BufferMemoryBarrier>& preDrawBarriers, SHParticleEmitterComponent const& emitter, uint32_t const EMITTER_INDEX, uint32_t const FRAME_INDEX) noexcept;
 
   public:
     void Init(Handle<SHVkLogicalDevice> device, Handle<SHVkDescriptorPool> inDescPool, Handle<SHVkRenderpass> compatibleRenderpass, Handle<SHSubpass> subpass, Handle<SHVkShaderModule> VS, Handle<SHVkShaderModule> FS, Handle<SHVkShaderModule> emitCS, Handle<SHVkShaderModule> defaultUpdateCS) noexcept;
 
-    void Run(Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex) noexcept;
+    void Run(Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex, Handle<SHVkFence> waitFence = {}) noexcept;
+    void ResetInstanceCounts (Handle<SHVkCommandBuffer> cmdBuffer, uint32_t frameIndex) noexcept;
 
     void Render(Handle<SHVkCommandBuffer> cmdBuffer, Handle<SHRenderer> renderer, uint32_t frameIndex) noexcept;
     void Exit(void) noexcept;

SHADE_Engine/src/Graphics/Synchronization/SHVkFence.h

@@ -30,7 +30,7 @@ namespace SHADE
     /*-----------------------------------------------------------------------*/
     /* PUBLIC MEMBER FUNCTIONS                                               */
     /*-----------------------------------------------------------------------*/
-    bool Wait  (bool waitAll, uint64_t timer) noexcept;
+    bool Wait  (bool waitAll, uint64_t timer = std::numeric_limits<uint32_t>::max()) noexcept;
     void Reset (void) noexcept;
 
     /*-----------------------------------------------------------------------*/