diff --git a/WickedEngine/WickedEngine_SHADERS.vcxproj b/WickedEngine/WickedEngine_SHADERS.vcxproj index f0a1fc8ca..a7ab47df4 100644 --- a/WickedEngine/WickedEngine_SHADERS.vcxproj +++ b/WickedEngine/WickedEngine_SHADERS.vcxproj @@ -133,6 +133,10 @@ Compute 5.0 + + Compute + 5.0 + Compute 5.0 diff --git a/WickedEngine/WickedEngine_SHADERS.vcxproj.filters b/WickedEngine/WickedEngine_SHADERS.vcxproj.filters index f50f21ef8..878c27cdd 100644 --- a/WickedEngine/WickedEngine_SHADERS.vcxproj.filters +++ b/WickedEngine/WickedEngine_SHADERS.vcxproj.filters @@ -702,6 +702,9 @@ PS + + CS + diff --git a/WickedEngine/emittedparticle_nbodyCS.hlsl b/WickedEngine/emittedparticle_nbodyCS.hlsl new file mode 100644 index 000000000..24fcd34e2 --- /dev/null +++ b/WickedEngine/emittedparticle_nbodyCS.hlsl @@ -0,0 +1,201 @@ +#include "globals.hlsli" +#include "ShaderInterop_EmittedParticle.h" + +RWSTRUCTUREDBUFFER(particleBuffer, Particle, 0); +RWSTRUCTUREDBUFFER(aliveBuffer_CURRENT, uint, 1); +RWSTRUCTUREDBUFFER(aliveBuffer_NEW, uint, 2); +RWSTRUCTUREDBUFFER(deadBuffer, uint, 3); +RWSTRUCTUREDBUFFER(counterBuffer, ParticleCounters, 4); + +struct LDSParticle +{ + float3 position; + float size; + float3 v; // velocity + float m; // mass + float p; // density + float P; // Pressure +}; +groupshared LDSParticle LDSParticles[THREADCOUNT_SIMULATION]; + +[numthreads(THREADCOUNT_SIMULATION, 1, 1)] +void main( uint3 DTid : SV_DispatchThreadID, uint groupIndex : SV_GroupIndex, uint3 Gid : SV_GroupID ) +{ + uint aliveCount = counterBuffer[0].aliveCount; + + if (DTid.x < aliveCount) + { + uint particleIndexA = aliveBuffer_CURRENT[DTid.x]; + Particle particleA = particleBuffer[particleIndexA]; + + float lifeLerpA = 1 - particleA.life / particleA.maxLife; + float particleSizeA = lerp(particleA.sizeBeginEnd.x, particleA.sizeBeginEnd.y, lifeLerpA); + + LDSParticles[groupIndex].position = particleA.position; + LDSParticles[groupIndex].size = particleSizeA; + LDSParticles[groupIndex].v = particleA.velocity; + LDSParticles[groupIndex].m = 1; + LDSParticles[groupIndex].p = 0; + LDSParticles[groupIndex].P = 0; + + } + + GroupMemoryBarrierWithGroupSync(); + + const uint LDSParticleCount = /*clamp(aliveCount - Gid.x * THREADCOUNT_SIMULATION, 0, THREADCOUNT_SIMULATION)*/ 256; + + + uint particleIndexA = groupIndex; + LDSParticle particleA = LDSParticles[particleIndexA]; + + + // Compute density field: + + const float h = 1.0f; // smoothing radius + const float h2 = h*h; + const float h3 = h2 * h; + const float h9 = h3 * h3; + + uint i; + + for (i = 0; i < LDSParticleCount; ++i) + { + if (i != particleIndexA) + { + uint particleIndexB = i; + LDSParticle particleB = LDSParticles[particleIndexB]; + + float3 diff = particleA.position - particleB.position; + float r2 = dot(diff, diff); // distance squared + + float range = particleA.size + particleB.size; // range of affection + float range2 = range * range; // range squared + + if (r2 < range2) + { + float W = (315.0f / (64.0f * PI * h9)) * pow(h2 - r2, 3); // poly6 smoothing kernel + + particleA.p += particleB.m * W; + } + + } + } + + // Compute particle pressure: + const float K = 20; // pressure constant + const float p0 = 20; // reference density + particleA.P = max(p0, K * (particleA.p - p0)); + + // Store the results: + LDSParticles[particleIndexA].p = particleA.p; + LDSParticles[particleIndexA].P = particleA.P; + + + // Wait for all particles to compute pressure + GroupMemoryBarrierWithGroupSync(); + + + if (particleA.p > 0) + { + + // Compute acceleration: + float3 a = 0; // pressure force + float3 av = 0; // viscosity force + const float e = 0.018f; // viscosity constant + + for (i = 0; i < LDSParticleCount; ++i) + { + if (i != particleIndexA) + { + uint particleIndexB = i; + LDSParticle particleB = LDSParticles[particleIndexB]; + + float3 diff = particleA.position - particleB.position; + float r2 = dot(diff, diff); // distance squared + float r = sqrt(r2); + + float range = particleA.size + particleB.size; // range of affection + + if (r < range) + { + float3 rNorm = normalize(diff); + float W = (-45 / (PI * pow(h, 6))) * pow(h - r, 2); // spiky kernel smoothing function + + a += -(particleB.m / particleA.m) * ((particleA.P + particleB.P) / (2 * particleA.p * particleB.p)) * W * rNorm; + + float r3 = r2 * r; + float h2 = h * h; + float h3 = h2 * h; + W = -(r3 / (2 * h3)) + (r2 / h2) + (h / (2 * r)) - 1; + av += e * (particleB.m / particleA.m) * (1.0f / particleB.p) * (particleB.v - particleA.v) * W * rNorm; + } + + } + } + + //a *= -1; + + //av *= e; + + float3 force = a + av; + + const float dt = g_xFrame_DeltaTime; + particleA.v += dt * force / particleA.p; + + } + + float elastic = 0.9; + + if (particleA.position.y - particleA.size < 0) + { + particleA.position.y = particleA.size; + particleA.v.y *= -elastic; + } + + //// box collision: + //float extent = 4; + //if (particleA.position.x + particleA.size > extent) + //{ + // particleA.position.x = extent - particleA.size; + // particleA.v.x *= -elastic; + //} + //if (particleA.position.x - particleA.size < -extent) + //{ + // particleA.position.x = -extent + particleA.size; + // particleA.v.x *= -elastic; + //} + //if (particleA.position.z + particleA.size > extent) + //{ + // particleA.position.z = extent - particleA.size; + // particleA.v.z *= -elastic; + //} + //if (particleA.position.z - particleA.size < -extent) + //{ + // particleA.position.z = -extent + particleA.size; + // particleA.v.z *= -elastic; + //} + + particleA.v *= 0.99f; + + particleA.v.y -= 0.8f; + + + if (DTid.x < aliveCount) + { + uint writeIndex = aliveBuffer_CURRENT[DTid.x]; + particleBuffer[writeIndex].position = particleA.position; + particleBuffer[writeIndex].velocity = particleA.v; + + particleBuffer[writeIndex].color_mirror = 0x00FFFFFF; + //particleBuffer[writeIndex].color_mirror |= ((uint)particleA.p) & 0xFF; + + if (particleA.p > 0) + { + particleBuffer[writeIndex].color_mirror = 0xFF; + } + + } + + +} + diff --git a/WickedEngine/emittedparticle_sortCS.hlsl b/WickedEngine/emittedparticle_sortCS.hlsl index f26b6917b..3cbf06be9 100644 --- a/WickedEngine/emittedparticle_sortCS.hlsl +++ b/WickedEngine/emittedparticle_sortCS.hlsl @@ -91,7 +91,7 @@ void main(uint3 Gid : SV_GroupID, float2 a = g_LDS[index]; float2 b = g_LDS[nSwapElem]; - if (a.x > b.x) + if (a.x < b.x) { g_LDS[index] = b; g_LDS[nSwapElem] = a; diff --git a/WickedEngine/emittedparticle_sortInnerCS.hlsl b/WickedEngine/emittedparticle_sortInnerCS.hlsl index bc9617087..9c6ea65a8 100644 --- a/WickedEngine/emittedparticle_sortInnerCS.hlsl +++ b/WickedEngine/emittedparticle_sortInnerCS.hlsl @@ -88,7 +88,7 @@ void main(uint3 Gid : SV_GroupID, float2 a = g_LDS[index]; float2 b = g_LDS[nSwapElem]; - if (a.x > b.x) + if (a.x < b.x) { g_LDS[index] = b; g_LDS[nSwapElem] = a; diff --git a/WickedEngine/emittedparticle_sortStepCS.hlsl b/WickedEngine/emittedparticle_sortStepCS.hlsl index 214088ef1..6994f90f7 100644 --- a/WickedEngine/emittedparticle_sortStepCS.hlsl +++ b/WickedEngine/emittedparticle_sortStepCS.hlsl @@ -55,7 +55,7 @@ void main(uint3 Gid : SV_GroupID, float a = distanceBuffer[index]; float b = distanceBuffer[nSwapElem]; - if (a > b) + if (a < b) { distanceBuffer[index] = b; distanceBuffer[nSwapElem] = a; diff --git a/WickedEngine/wiEmittedParticle.cpp b/WickedEngine/wiEmittedParticle.cpp index 0276190cc..0500afd6a 100644 --- a/WickedEngine/wiEmittedParticle.cpp +++ b/WickedEngine/wiEmittedParticle.cpp @@ -14,7 +14,7 @@ using namespace wiGraphicsTypes; VertexShader *wiEmittedParticle::vertexShader = nullptr; PixelShader *wiEmittedParticle::pixelShader[PARTICLESHADERTYPE_COUNT] = {}; -ComputeShader *wiEmittedParticle::kickoffUpdateCS, *wiEmittedParticle::emitCS = nullptr, *wiEmittedParticle::simulateCS = nullptr, +ComputeShader *wiEmittedParticle::kickoffUpdateCS, *wiEmittedParticle::emitCS = nullptr, *wiEmittedParticle::nbodyCS = nullptr, *wiEmittedParticle::simulateCS = nullptr, *wiEmittedParticle::simulateCS_SORTING = nullptr, *wiEmittedParticle::simulateCS_DEPTHCOLLISIONS = nullptr, *wiEmittedParticle::simulateCS_SORTING_DEPTHCOLLISIONS = nullptr; ComputeShader *wiEmittedParticle::kickoffSortCS = nullptr, *wiEmittedParticle::sortCS = nullptr, *wiEmittedParticle::sortInnerCS = nullptr, *wiEmittedParticle::sortStepCS = nullptr; GPUBuffer *wiEmittedParticle::sortCB = nullptr; @@ -23,7 +23,7 @@ RasterizerState wiEmittedParticle::rasterizerState, wiEmittedParticle::wireFram DepthStencilState wiEmittedParticle::depthStencilState; GraphicsPSO wiEmittedParticle::PSO[BLENDMODE_COUNT][PARTICLESHADERTYPE_COUNT]; GraphicsPSO wiEmittedParticle::PSO_wire; -ComputePSO wiEmittedParticle::CPSO_kickoffUpdate, wiEmittedParticle::CPSO_emit, wiEmittedParticle::CPSO_simulate, +ComputePSO wiEmittedParticle::CPSO_kickoffUpdate, wiEmittedParticle::CPSO_emit, wiEmittedParticle::CPSO_nbody, wiEmittedParticle::CPSO_simulate, wiEmittedParticle::CPSO_simulate_SORTING, wiEmittedParticle::CPSO_simulate_DEPTHCOLLISIONS, wiEmittedParticle::CPSO_simulate_SORTING_DEPTHCOLLISIONS; ComputePSO wiEmittedParticle::CPSO_kickoffSort, wiEmittedParticle::CPSO_sort, wiEmittedParticle::CPSO_sortInner, wiEmittedParticle::CPSO_sortStep; @@ -342,6 +342,11 @@ void wiEmittedParticle::UpdateRenderData(GRAPHICSTHREAD threadID) device->DispatchIndirect(indirectBuffers, ARGUMENTBUFFER_OFFSET_DISPATCHEMIT, threadID); device->UAVBarrier(uavs, ARRAYSIZE(uavs), threadID); + // perform N-body collision response simulation: + device->BindComputePSO(&CPSO_nbody, threadID); + device->DispatchIndirect(indirectBuffers, ARGUMENTBUFFER_OFFSET_DISPATCHSIMULATION, threadID); + device->UAVBarrier(uavs, ARRAYSIZE(uavs), threadID); + // update CURRENT alive list, write NEW alive list if (SORTING) { @@ -528,6 +533,7 @@ void wiEmittedParticle::LoadShaders() kickoffUpdateCS = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_kickoffUpdateCS.cso", wiResourceManager::COMPUTESHADER)); emitCS = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_emitCS.cso", wiResourceManager::COMPUTESHADER)); + nbodyCS = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_nbodyCS.cso", wiResourceManager::COMPUTESHADER)); simulateCS = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_simulateCS.cso", wiResourceManager::COMPUTESHADER)); simulateCS_SORTING = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_simulateCS_SORTING.cso", wiResourceManager::COMPUTESHADER)); simulateCS_DEPTHCOLLISIONS = static_cast(wiResourceManager::GetShaderManager()->add(wiRenderer::SHADERPATH + "emittedparticle_simulateCS_DEPTHCOLLISIONS.cso", wiResourceManager::COMPUTESHADER)); @@ -581,6 +587,9 @@ void wiEmittedParticle::LoadShaders() desc.cs = emitCS; device->CreateComputePSO(&desc, &CPSO_emit); + desc.cs = nbodyCS; + device->CreateComputePSO(&desc, &CPSO_nbody); + desc.cs = simulateCS; device->CreateComputePSO(&desc, &CPSO_simulate); diff --git a/WickedEngine/wiEmittedParticle.h b/WickedEngine/wiEmittedParticle.h index fb45e3c1d..89eae494d 100644 --- a/WickedEngine/wiEmittedParticle.h +++ b/WickedEngine/wiEmittedParticle.h @@ -34,7 +34,7 @@ private: wiGraphicsTypes::GPUBuffer* constantBuffer; void CreateSelfBuffers(); - static wiGraphicsTypes::ComputeShader *kickoffUpdateCS, *emitCS, *simulateCS, *simulateCS_SORTING, *simulateCS_DEPTHCOLLISIONS, *simulateCS_SORTING_DEPTHCOLLISIONS; + static wiGraphicsTypes::ComputeShader *kickoffUpdateCS, *emitCS, *nbodyCS, *simulateCS, *simulateCS_SORTING, *simulateCS_DEPTHCOLLISIONS, *simulateCS_SORTING_DEPTHCOLLISIONS; static wiGraphicsTypes::ComputeShader *kickoffSortCS, *sortCS, *sortInnerCS, *sortStepCS; static wiGraphicsTypes::GPUBuffer *sortCB; static wiGraphicsTypes::VertexShader *vertexShader; @@ -45,7 +45,7 @@ private: static wiGraphicsTypes::GraphicsPSO PSO[BLENDMODE_COUNT][PARTICLESHADERTYPE_COUNT]; static wiGraphicsTypes::GraphicsPSO PSO_wire; - static wiGraphicsTypes::ComputePSO CPSO_kickoffUpdate, CPSO_emit, CPSO_simulate, CPSO_simulate_SORTING, CPSO_simulate_DEPTHCOLLISIONS, CPSO_simulate_SORTING_DEPTHCOLLISIONS; + static wiGraphicsTypes::ComputePSO CPSO_kickoffUpdate, CPSO_emit, CPSO_nbody, CPSO_simulate, CPSO_simulate_SORTING, CPSO_simulate_DEPTHCOLLISIONS, CPSO_simulate_SORTING_DEPTHCOLLISIONS; static wiGraphicsTypes::ComputePSO CPSO_kickoffSort, CPSO_sort, CPSO_sortInner, CPSO_sortStep; public: