lightmap improvements

WickedEngine/shaders/lightmap_expandCS.hlsl

@@ -14,7 +14,7 @@ static const int2 offsets[] = {
 	int2(-1, -1),
 	int2(1, -1),
 	int2(1, 1),
-	int2(-1, -1),
+	int2(-1, 1),
 };
 
 [numthreads(POSTPROCESS_BLOCKSIZE, POSTPROCESS_BLOCKSIZE, 1)]

WickedEngine/shaders/renderlightmapPS.hlsl

@@ -11,7 +11,6 @@ static const uint ANYTHIT_CUTOFF_AFTER_BOUNCE_COUNT = 4;
 struct Input
 {
 	float4 pos : SV_POSITION;
-	centroid float2 uv : TEXCOORD;
 	centroid float3 pos3D : WORLDPOSITION;
 	centroid float3 normal : NORMAL;
 };
@@ -106,6 +105,8 @@ void BakeryPixelPush(inout float3 P, in float3 N, in float2 UV, inout RNG rng, i
 
 float4 main(Input input) : SV_TARGET
 {
+	float2 uv = input.pos.xy * xTraceResolution_rcp;
+
 	Surface surface;
 	surface.init();
 	surface.N = normalize(input.normal);
@@ -116,9 +117,8 @@ float4 main(Input input) : SV_TARGET
 	float3 P = input.pos3D;
 
 	float bakerydebug = 0;
-	BakeryPixelPush(P, surface.N, input.uv, rng, bakerydebug);
-
-	float2 uv = input.uv;
+	BakeryPixelPush(P, surface.N, uv, rng, bakerydebug);
+	
 	RayDesc ray;
 	ray.Origin = P;
 	ray.Direction = sample_hemisphere_cos(surface.N, rng);
@@ -247,7 +247,7 @@ float4 main(Input input) : SV_TARGET
 					newRay.Origin = surface.P;
 					newRay.TMin = 0.0001;
 					newRay.TMax = dist;
-					newRay.Direction = L + max3(surface.sss);
+					newRay.Direction = normalize(L + max3(surface.sss));
 
 #ifdef RTAPI
 					uint flags = RAY_FLAG_SKIP_PROCEDURAL_PRIMITIVES | RAY_FLAG_CULL_FRONT_FACING_TRIANGLES;

WickedEngine/shaders/renderlightmapVS.hlsl

@@ -6,7 +6,6 @@ PUSHCONSTANT(push, LightmapPushConstants);
 struct Output
 {
 	float4 pos : SV_POSITION;
-	centroid float2 uv : TEXCOORD;
 	centroid float3 pos3D : WORLDPOSITION;
 	centroid float3 normal : NORMAL;
 };
@@ -25,8 +24,6 @@ Output main(uint vertexID : SV_VertexID)
 	output.pos.y *= -1;
 	output.pos.xy += xTracePixelOffset;
 
-	output.uv = atl;
-
 	output.pos3D = mul(inst.transform.GetMatrix(), float4(pos, 1)).xyz;
 
 	output.normal = mul(inst.transformRaw.GetMatrixAdjoint(), nor);

WickedEngine/wiRenderer.cpp

@@ -1445,11 +1445,7 @@ void LoadShaders()
 		desc.bs = &blendStates[BSTYPE_TRANSPARENT];
 		desc.dss = &depthStencils[DSSTYPE_DEPTHDISABLED];
 
-		RenderPassInfo renderpass_info;
-		renderpass_info.rt_count = 1;
-		renderpass_info.rt_formats[0] = Format::R32G32B32A32_FLOAT;
-
-		device->CreatePipelineState(&desc, &PSO_renderlightmap, &renderpass_info);
+		device->CreatePipelineState(&desc, &PSO_renderlightmap);
 		});
 	wi::jobsystem::Execute(ctx, [](wi::jobsystem::JobArgs args) {
 		PipelineStateDesc desc;
@@ -2304,7 +2300,7 @@ void SetUpStates()
 
 
 	rs = rasterizers[RSTYPE_DOUBLESIDED];
-	// Note: conservative raster can produce bright lightmap pixels, so now it's disabled!
+	// Note: conservative rasterization can cause GPU hang sometimes
 	//if (device->CheckCapability(GraphicsDeviceCapability::CONSERVATIVE_RASTERIZATION))
 	//{
 	//	rs.conservative_rasterization_enable = true;
@@ -4609,6 +4605,8 @@ void UpdateRenderData(
 	auto prof_updatebuffer_cpu = wi::profiler::BeginRangeCPU("Update Buffers (CPU)");
 	auto prof_updatebuffer_gpu = wi::profiler::BeginRangeGPU("Update Buffers (GPU)", cmd);
 
+	barrier_stack.push_back(GPUBarrier::Buffer(&vis.scene->meshletBuffer, ResourceState::SHADER_RESOURCE, ResourceState::UNORDERED_ACCESS));
+
 	barrier_stack.push_back(GPUBarrier::Buffer(&buffers[BUFFERTYPE_FRAMECB], ResourceState::CONSTANT_BUFFER, ResourceState::COPY_DST));
 	if (vis.scene->instanceBuffer.IsValid())
 	{
@@ -15963,7 +15961,7 @@ void CreateVolumetricCloudResources(VolumetricCloudResources& res, XMUINT2 resol
 	desc.width = resolution.x / 4;
 	desc.height = resolution.y / 4;
 	desc.format = Format::R16G16B16A16_FLOAT;
-	desc.layout = ResourceState::SHADER_RESOURCE;
+	desc.layout = ResourceState::SHADER_RESOURCE_COMPUTE;
 	device->CreateTexture(&desc, nullptr, &res.texture_cloudRender);
 	device->SetName(&res.texture_cloudRender, "texture_cloudRender");
 	desc.format = Format::R32G32_FLOAT;
@@ -16191,7 +16189,23 @@ void Postprocess_VolumetricClouds_Upsample(
 		device->BindResource(&res.texture_reproject[temporal_output], 0, cmd);
 		device->BindResource(&res.texture_reproject_depth[temporal_output], 1, cmd);
 
+		{
+			GPUBarrier barriers[] = {
+				GPUBarrier::Image(&res.texture_reproject[temporal_output], res.texture_reproject[temporal_output].desc.layout, ResourceState::SHADER_RESOURCE),
+				GPUBarrier::Image(&res.texture_reproject_depth[temporal_output], res.texture_reproject_depth[temporal_output].desc.layout, ResourceState::SHADER_RESOURCE),
+			};
+			device->Barrier(barriers, arraysize(barriers), cmd);
+		}
+
 		device->Draw(3, 0, cmd);
+
+		{
+			GPUBarrier barriers[] = {
+				GPUBarrier::Image(&res.texture_reproject[temporal_output], ResourceState::SHADER_RESOURCE, res.texture_reproject[temporal_output].desc.layout),
+				GPUBarrier::Image(&res.texture_reproject_depth[temporal_output], ResourceState::SHADER_RESOURCE, res.texture_reproject_depth[temporal_output].desc.layout),
+			};
+			device->Barrier(barriers, arraysize(barriers), cmd);
+		}
 	}
 
 	wi::profiler::EndRange(range);

WickedEngine/wiScene.cpp

@@ -4577,6 +4577,9 @@ namespace wi::scene
 						object.lightmapWidth = wi::graphics::AlignTo(object.lightmapWidth, 4u);
 						object.lightmapHeight = wi::graphics::AlignTo(object.lightmapHeight, 4u);
 
+						object.lightmapWidth = clamp(object.lightmapWidth, 16u, 16384u);
+						object.lightmapHeight = clamp(object.lightmapHeight, 16u, 16384u);
+
 						TextureDesc desc;
 						desc.width = object.lightmapWidth;
 						desc.height = object.lightmapHeight;

WickedEngine/wiVersion.cpp

@@ -9,7 +9,7 @@ namespace wi::version
 	// minor features, major updates, breaking compatibility changes
 	const int minor = 71;
 	// minor bug fixes, alterations, refactors, updates
-	const int revision = 638;
+	const int revision = 639;
 
 	const std::string version_string = std::to_string(major) + "." + std::to_string(minor) + "." + std::to_string(revision);