Dynamitos/Seele
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Many stupid changes

Browse Source
This commit is contained in:
Dynamitos
2024-10-18 19:01:04 +02:00
parent 7623d647ee
commit 1193406dd8
13 changed files with 419 additions and 117 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.vscode/settings.json
Vendored
+7 -1
View File
@@ -24,5 +24,11 @@
"lldb.dereferencePointers": true,
"lldb.consoleMode": "commands",
"cmake.generator": "Xcode",
"editor.tabSize": 4
"editor.tabSize": 4,
"slang.additionalSearchPaths": [
"res/shaders/",
"res/shaders/lib"
],
"slang.slangdLocation": "C:\\Users\\Dynamitos\\slang\\build\\Release\\bin\\slangd.exe",
"slang.searchInAllWorkspaceDirectories": false
}
res/shaders/terrain/CBTCompute.slang
+1 -1
View File
@@ -41,7 +41,7 @@ void Split(uint dispatchID : SV_DispatchThreadID)
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + dispatchID];
// Split the element
SplitElement(currentID, pParams.geometry.baseDepth);
SplitElement(currentID, pParams.geometry.baseDepth, dispatchID);
}
[numthreads(1, 1, 1)]
res/shaders/terrain/CompileTest.slang
+203 -26
View File
@@ -1,4 +1,65 @@
// Pointer to an invalid neighbor or index
const static int INVALID_POINTER = 4294967295;
// Possible culling state
const static int BACK_FACE_CULLED = -3;
const static int FRUSTUM_CULLED = -2;
const static int TOO_SMALL = -1;
const static int UNCHANGED_ELEMENT = 0;
const static int BISECT_ELEMENT = 1;
const static int SIMPLIFY_ELEMENT = 2;
const static int MERGED_ELEMENT = 3;
// Bisector flags
const static int VISIBLE_BISECTOR = 0x1;
const static int MODIFIED_BISECTOR = 0x2;
// Possible splits
static const uint64_t NO_SPLIT = 0x00;
static const uint64_t CENTER_SPLIT = 0x01;
static const uint64_t RIGHT_SPLIT = 0x02;
static const uint64_t LEFT_SPLIT = 0x04;
static const uint64_t RIGHT_DOUBLE_SPLIT = (CENTER_SPLIT | RIGHT_SPLIT);
static const uint64_t LEFT_DOUBLE_SPLIT = (CENTER_SPLIT | LEFT_SPLIT);
static const uint64_t TRIPLE_SPLIT = (CENTER_SPLIT | RIGHT_SPLIT | LEFT_SPLIT);
// Split buffer slots
static const uint64_t SPLIT_COUNTER = 0;
static const uint64_t SIMPLIFY_COUNTER = 1;
static const uint64_t CLASSIFY_COUNTER_OFFSET = 2;
struct BisectorData
{
// Allocated indices for this bisector
uint32_t indices[3];
// Subvision that should be applied to this bisector
uint32_t subdivisionPattern;
// Neighbor that should be processed
uint32_t problematicNeighbor;
// State of this bisector (split, merge, etc)
uint32_t bisectorState;
// Visibility and modification flags of a bisector
uint32_t flags;
// ID used for the propagation
uint32_t propagationID;
};
uint HeapIDDepth(uint64_t x)
{
uint depth = 0;
while (x > 0u) {
++depth;
x >>= 1u;
}
return depth;
}
struct ComputeParams
{
RWStructuredBuffer<uint> indirectDrawBuffer;
@@ -6,45 +67,161 @@ struct ComputeParams
RWStructuredBuffer<uint> classificationBuffer;
RWStructuredBuffer<int> allocateBuffer;
RWStructuredBuffer<int> memoryBuffer;
RWStructuredBuffer<uint4> neighboursBuffer;
RWStructuredBuffer<BisectorData> bisectorDataBuffer;
RWStructuredBuffer<int> propagateBuffer;
RWStructuredBuffer<uint> simplifyBuffer;
RWStructuredBuffer<uint32_t> bitFieldBuffer;
};
ParameterBlock<ComputeParams> pParams;
// Split buffer slots
static const uint64_t SPLIT_COUNTER = 0;
static const uint64_t SIMPLIFY_COUNTER = 1;
static const uint64_t CLASSIFY_COUNTER_OFFSET = 2;
void ResetBuffers()
void SplitElement(uint currentID, uint baseDepth, uint dispatchID)
{
pParams.memoryBuffer[0] = 0;
pParams.memoryBuffer[1] = 0;
// Get the neighbors information
uint4 cNeighbors = pParams.neighboursBuffer[currentID];
pParams.classificationBuffer[SPLIT_COUNTER] = 0;
pParams.classificationBuffer[SIMPLIFY_COUNTER] = 0;
// If there is a neighbor X
if (cNeighbors.x != INVALID_POINTER)
{
// This is on the path of it's neighbor X
uint4 xNeighbors = pParams.neighboursBuffer[cNeighbors.x];
if (xNeighbors.z == currentID && pParams.bisectorDataBuffer[cNeighbors.x].bisectorState != UNCHANGED_ELEMENT)
return;
}
pParams.allocateBuffer[0] = 0;
// If there is a neighbor Y
if (cNeighbors.y != INVALID_POINTER)
{
// This is on the path of it's neighbor Y
uint4 yNeighbors = pParams.neighboursBuffer[cNeighbors.y];
if (yNeighbors.z == currentID && pParams.bisectorDataBuffer[cNeighbors.y].bisectorState != UNCHANGED_ELEMENT)
return;
}
pParams.propagateBuffer[0] = 0;
pParams.propagateBuffer[1] = 0;
// Depth of the current triangle
uint64_t heapID = pParams.heapIDBuffer[currentID];
uint currentDepth = HeapIDDepth(heapID);
pParams.simplifyBuffer[0] = 0;
// Compute the maximal required memory for this subdivision
int maxRequiredMemory = 2 * (currentDepth - baseDepth) - 1;
pParams.indirectDrawBuffer[0] = 0;
pParams.indirectDrawBuffer[1] = 1;
pParams.indirectDrawBuffer[2] = 0;
pParams.indirectDrawBuffer[3] = 0;
// Get the twin information
uint twinID = cNeighbors.z;
pParams.indirectDrawBuffer[4] = 0;
pParams.indirectDrawBuffer[5] = 1;
pParams.indirectDrawBuffer[6] = 0;
pParams.indirectDrawBuffer[7] = 0;
// This avoid the massive over-reservation and saves a bunch of artifacts
if (twinID == INVALID_POINTER)
maxRequiredMemory = 1;
else if (pParams.neighboursBuffer[twinID].z == currentID)
maxRequiredMemory = 2;
pParams.indirectDrawBuffer[8] = 0;
// Try to reserve
int remainingMemory;
InterlockedAdd(pParams.memoryBuffer[1], -maxRequiredMemory, remainingMemory);
// Did someone manage to sneak-in while we were trying to pick the memory, add it back and try again
if (remainingMemory < maxRequiredMemory)
{
// Then add back the required memory and stop
InterlockedAdd(pParams.memoryBuffer[1], maxRequiredMemory, remainingMemory);
return;
}
// Let's actually count the memory that we will be using
uint usedMemory = 1;
uint prevPattern;
InterlockedOr(pParams.bisectorDataBuffer[currentID].subdivisionPattern, CENTER_SPLIT, prevPattern);
// If this is not zero, it means an other neighbor went faster than us, we restore the memory and leave.
if (prevPattern != 0)
{
InterlockedAdd(pParams.memoryBuffer[1], maxRequiredMemory, remainingMemory);
return;
}
// Mark this for allocation
uint targetLocation = 0;
InterlockedAdd(pParams.allocateBuffer[0], 1, targetLocation);
pParams.allocateBuffer[1 + targetLocation] = currentID;
// While we're not done (up the tree or everything is subdivided properly)
bool done = false;
while (!done)
{
// If this neighbor is not allocated, we're done.
if (twinID == INVALID_POINTER)
break;
// Grab the bisector of the neighbor
uint64_t nHeapID = pParams.heapIDBuffer[twinID];
BisectorData nBisectorData = pParams.bisectorDataBuffer[twinID];
uint nDepth = HeapIDDepth(nHeapID);
uint4 nNeighbors = pParams.neighboursBuffer[twinID];
// If both triangles have the same depth
if (nDepth == currentDepth)
{
// Raised the center split
InterlockedOr(pParams.bisectorDataBuffer[twinID].subdivisionPattern, CENTER_SPLIT, prevPattern);
// Only account for it if it was not raised before.
if (prevPattern == 0)
{
// Mark this for allocation
uint targetLocation = 0;
InterlockedAdd(pParams.allocateBuffer[0], 1, targetLocation);
pParams.allocateBuffer[1 + targetLocation] = twinID;
usedMemory++;
}
// And we're done
done = true;
}
// If this node has already been subdivided, it means that we need to add the third subdivision and we're done
else
{
if (nNeighbors[0] == currentID)
InterlockedOr(pParams.bisectorDataBuffer[twinID].subdivisionPattern, RIGHT_DOUBLE_SPLIT, prevPattern);
else // if (nNeighbors[1] == currentID)
InterlockedOr(pParams.bisectorDataBuffer[twinID].subdivisionPattern, LEFT_DOUBLE_SPLIT, prevPattern);
if (prevPattern != 0)
{
usedMemory++;
done = true;
}
else
{
// Mark this for allocation
uint targetLocation = 0;
InterlockedAdd(pParams.allocateBuffer[0], 1, targetLocation);
pParams.allocateBuffer[1 + targetLocation] = twinID;
// Account for two splits
usedMemory += 2;
// the new bisector that needs to be propagated
currentID = twinID;
currentDepth = nDepth;
twinID = pParams.neighboursBuffer[currentID].z;
}
}
}
int change = maxRequiredMemory - usedMemory;
if(change > 0)
{
// Add back the unused memory (in case)
InterlockedAdd(pParams.memoryBuffer[1], change, remainingMemory);
}
}
[numthreads(1, 1, 1)]
void Reset()
[numthreads(64, 1, 1)]
void Split(uint dispatchID : SV_DispatchThreadID)
{
ResetBuffers();
}
if (dispatchID >= pParams.classificationBuffer[SPLIT_COUNTER])
return;
// Grab the real elementID
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + dispatchID];
// Split the element
SplitElement(currentID, 7, dispatchID);
}
res/shaders/terrain/Parameters.slang
+7 -4
View File
@@ -29,9 +29,12 @@ struct UpdateCB
struct DebugStruct
{
uint3 neighbours;
int status;
}
int maxRequiredMemory;
int usedMemory;
uint memoryChange;
uint twinID;
};
struct ComputeParams
{
@@ -59,6 +62,6 @@ struct ComputeParams
RWStructuredBuffer<uint> modifiedBisectorIndices;
RWStructuredBuffer<float4> lebPositionBuffer;
StructuredBuffer<float3x3> lebMatrixCache;
RWStructuredBuffer<DebugStruct> debugBuffer;
globallycoherent RWStructuredBuffer<DebugStruct> debugBuffer;
};
ParameterBlock<ComputeParams> pParams;
res/shaders/terrain/update_utils.slang
+64 -34
View File
@@ -23,13 +23,30 @@ static const uint64_t SPLIT_COUNTER = 0;
static const uint64_t SIMPLIFY_COUNTER = 1;
static const uint64_t CLASSIFY_COUNTER_OFFSET = 2;
bool FrustumAABBIntersect(in Frustum frustum, float3 aabbMin, float3 aabbMax)
{
float3 center = (aabbMax + aabbMin) * 0.5;
float3 extents = (aabbMax - aabbMin) * 0.5;
for (int i = 0; i < 4; i++)
{
Plane plane = frustum.sides[i];
float3 normal_sign = sign(plane.n);
float3 test_point = center + extents * normal_sign;
float dotProd = dot(test_point, plane.n);
if (dotProd + plane.d < 0)
return false;
}
return true;
}
int ClassifyBisector(in BisectorGeometry tri, uint depth)
{
// Check the triangle's visibility
float3 triNormal = normalize(cross(tri.p[2] - tri.p[1], tri.p[0] - tri.p[1]));
float3 triCenter = (tri.p[0] + tri.p[1] + tri.p[2]) / 3.0;
float3 viewDir = normalize(-triCenter);
float FdotV = dot(viewDir, pViewParams.cameraForward_WS.xyz);
float FdotV = dot(viewDir, -pViewParams.cameraForward_WS.xyz);
float VdotN = dot(viewDir, triNormal);
// Here we don't use 0 as it introduces stability issues at grazing angles
@@ -41,11 +58,11 @@ int ClassifyBisector(in BisectorGeometry tri, uint depth)
float3 aabbMax = float3(max(max(tri.p[0].x, tri.p[1].x), tri.p[2].x), max(max(tri.p[0].y, tri.p[1].y), tri.p[2].y), max(max(tri.p[0].z, tri.p[1].z), tri.p[2].z));
// First we do a frustum culling pass
//if (!FrustumAABBIntersect(_FrustumPlanes, aabbMin, aabbMax))
// return FRUSTUM_CULLED;
if (!FrustumAABBIntersect(pViewParams.viewFrustum, aabbMin, aabbMax))
return FRUSTUM_CULLED;
// Project the points on screen
float4x4 viewProjectionMatrix = mul(pViewParams.projectionMatrix, pViewParams.viewMatrix);
float4x4 viewProjectionMatrix = pParams.update.viewProjectionMatrix;//mul(pViewParams.projectionMatrix, pViewParams.viewMatrix);
float4 p0P = mul(viewProjectionMatrix, float4(tri.p[0], 1.0));
p0P.xy = p0P.xy / p0P.w;
p0P.xy = (p0P.xy * 0.5 + 0.5);
@@ -166,16 +183,20 @@ void ClassifyElement(uint currentID, BisectorGeometry bis, uint totalNumElements
pParams.bisectorDataBuffer[currentID] = cbisectorData;
}
void SplitElement(uint currentID, uint baseDepth)
void SplitElement(uint currentID, uint baseDepth, uint dispatchID)
{
DebugStruct debug;
debug.maxRequiredMemory = 0;
debug.usedMemory = 0;
debug.memoryChange = 0;
// Get the neighbors information
uint3 cNeighbors = pParams.neighboursBuffer[currentID].xyz;
uint4 cNeighbors = pParams.neighboursBuffer[currentID];
// If there is a neighbor X
if (cNeighbors.x != INVALID_POINTER)
{
// This is on the path of it's neighbor X
uint3 xNeighbors = pParams.neighboursBuffer[cNeighbors.x].xyz;
uint4 xNeighbors = pParams.neighboursBuffer[cNeighbors.x];
if (xNeighbors.z == currentID && pParams.bisectorDataBuffer[cNeighbors.x].bisectorState != UNCHANGED_ELEMENT)
return;
}
@@ -184,7 +205,7 @@ void SplitElement(uint currentID, uint baseDepth)
if (cNeighbors.y != INVALID_POINTER)
{
// This is on the path of it's neighbor Y
uint3 yNeighbors = pParams.neighboursBuffer[cNeighbors.y].xyz;
uint4 yNeighbors = pParams.neighboursBuffer[cNeighbors.y];
if (yNeighbors.z == currentID && pParams.bisectorDataBuffer[cNeighbors.y].bisectorState != UNCHANGED_ELEMENT)
return;
}
@@ -205,10 +226,13 @@ void SplitElement(uint currentID, uint baseDepth)
else if (pParams.neighboursBuffer[twinID].z == currentID)
maxRequiredMemory = 2;
debug.maxRequiredMemory = maxRequiredMemory;
debug.twinID = twinID;
// Try to reserve
int remainingMemory;
InterlockedAdd(pParams.memoryBuffer[1], -maxRequiredMemory, remainingMemory);
debug.memoryChange = -maxRequiredMemory;
// Did someone manage to sneak-in while we were trying to pick the memory, add it back and try again
if (remainingMemory < maxRequiredMemory)
{
@@ -218,7 +242,7 @@ void SplitElement(uint currentID, uint baseDepth)
}
// Let's actually count the memory that we will be using
uint usedMemory = 1;
int usedMemory = 1;
uint prevPattern;
InterlockedOr(pParams.bisectorDataBuffer[currentID].subdivisionPattern, CENTER_SPLIT, prevPattern);
@@ -246,7 +270,7 @@ void SplitElement(uint currentID, uint baseDepth)
uint64_t nHeapID = pParams.heapIDBuffer[twinID];
BisectorData nBisectorData = pParams.bisectorDataBuffer[twinID];
uint nDepth = HeapIDDepth(nHeapID);
uint3 nNeighbors = pParams.neighboursBuffer[twinID].xyz;
uint4 nNeighbors = pParams.neighboursBuffer[twinID];
// If both triangles have the same depth
if (nDepth == currentDepth)
@@ -297,9 +321,13 @@ void SplitElement(uint currentID, uint baseDepth)
}
}
}
int change = maxRequiredMemory - usedMemory;
// Add back the unused memory (in case)
InterlockedAdd(pParams.memoryBuffer[1], max(maxRequiredMemory - usedMemory, 0), remainingMemory);
InterlockedAdd(pParams.memoryBuffer[1], change, remainingMemory);
debug.memoryChange += change;
debug.usedMemory = usedMemory;
pParams.debugBuffer[dispatchID] = debug;
}
void AllocateElement(uint currentID)
@@ -335,7 +363,7 @@ void AllocateElement(uint currentID)
void evaluate_neighbors(uint currentID, uint bisectorID, out uint resX, out uint resY)
{
BisectorData nBisectorData = pParams.bisectorDataBuffer[bisectorID];
uint3 nNeighbors = pParams.neighboursBuffer[bisectorID].xyz;
uint4 nNeighbors = pParams.neighboursBuffer[bisectorID];
if (nBisectorData.subdivisionPattern == 0x01)
{
resX = nBisectorData.indices[SUBLING0_ID];
@@ -399,7 +427,7 @@ void BisectElement(uint currentID)
uint currentSubdiv = cBisectorData.subdivisionPattern;
// neighbors of the parent
uint3 cNeighbors = pParams.neighboursBuffer[currentID].xyz;
uint4 cNeighbors = pParams.neighboursBuffer[currentID];
uint p_n0 = cNeighbors[0];
uint p_n1 = cNeighbors[1];
uint p_n2 = cNeighbors[2];
@@ -421,15 +449,15 @@ void BisectElement(uint currentID)
pParams.heapIDBuffer[siblingID0] = 2 * baseHeapID + 1;
// Update the neighbors
uint3 modifiedNeighbors;
uint4 modifiedNeighbors;
modifiedNeighbors[0] = siblingID0;
modifiedNeighbors[1] = resX;
modifiedNeighbors[2] = p_n0;
pParams.neighboursOutputBuffer[currentID] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[currentID] = modifiedNeighbors;
modifiedNeighbors[0] = resY;
modifiedNeighbors[1] = currentID;
modifiedNeighbors[2] = p_n1;
pParams.neighboursOutputBuffer[siblingID0] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID0] = modifiedNeighbors;
// Keep track of the parent
BisectorData modifiedBisector = cBisectorData;
@@ -444,7 +472,7 @@ void BisectElement(uint currentID)
pParams.bisectorDataBuffer[siblingID0] = modifiedBisector;
// Mark this for propagation
uint targetLocation = 0;
uint targetLocation;
InterlockedAdd(pParams.propagateBuffer[0], 1, targetLocation);
pParams.propagateBuffer[2 + targetLocation] = siblingID0;
}
@@ -463,19 +491,19 @@ void BisectElement(uint currentID)
pParams.heapIDBuffer[siblingID0] = 2 * baseHeapID + 1;
pParams.heapIDBuffer[siblingID1] = 4 * baseHeapID + 1;
uint3 modifiedNeighbors;
uint4 modifiedNeighbors;
modifiedNeighbors[0] = siblingID1;
modifiedNeighbors[1] = res0X;
modifiedNeighbors[2] = siblingID0;
pParams.neighboursOutputBuffer[currentID] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[currentID] = modifiedNeighbors;
modifiedNeighbors[0] = res1Y;
modifiedNeighbors[1] = currentID;
modifiedNeighbors[2] = p_n1;
pParams.neighboursOutputBuffer[siblingID0] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID0] = modifiedNeighbors;
modifiedNeighbors[0] = res0Y;
modifiedNeighbors[1] = currentID;
modifiedNeighbors[2] = res1X;
pParams.neighboursOutputBuffer[siblingID1] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID1] = modifiedNeighbors;
// Keep track of the parent
BisectorData modifiedBisector = cBisectorData;
@@ -497,7 +525,7 @@ void BisectElement(uint currentID)
pParams.bisectorDataBuffer[siblingID1] = modifiedBisector;
// Mark this for propagation
uint targetLocation = 0;
uint targetLocation;
InterlockedAdd(pParams.propagateBuffer[0], 1, targetLocation);
pParams.propagateBuffer[2 + targetLocation] = siblingID0;
}
@@ -516,19 +544,19 @@ void BisectElement(uint currentID)
pParams.heapIDBuffer[siblingID0] = 4 * baseHeapID + 2;
pParams.heapIDBuffer[siblingID1] = 4 * baseHeapID + 3;
uint3 modifiedNeighbors;
uint4 modifiedNeighbors;
modifiedNeighbors[0] = siblingID1;
modifiedNeighbors[1] = res1X;
modifiedNeighbors[2] = p_n0;
pParams.neighboursOutputBuffer[currentID] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[currentID] = modifiedNeighbors;
modifiedNeighbors[0] = siblingID1;
modifiedNeighbors[1] = res0X;
modifiedNeighbors[2] = res1Y;
pParams.neighboursOutputBuffer[siblingID0] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID0] = modifiedNeighbors;
modifiedNeighbors[0] = res0Y;
modifiedNeighbors[1] = siblingID0;
modifiedNeighbors[2] = currentID;
pParams.neighboursOutputBuffer[siblingID1] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID1] = modifiedNeighbors;
// Keep track of the parent
BisectorData modifiedBisector = cBisectorData;
@@ -568,23 +596,23 @@ void BisectElement(uint currentID)
pParams.heapIDBuffer[siblingID1] = 4 * baseHeapID + 1;
pParams.heapIDBuffer[siblingID2] = 4 * baseHeapID + 3;
uint3 modifiedNeighbors;
uint4 modifiedNeighbors;
modifiedNeighbors[0] = siblingID1;
modifiedNeighbors[1] = res0X;
modifiedNeighbors[2] = siblingID2;
pParams.neighboursOutputBuffer[currentID] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[currentID] = modifiedNeighbors;
modifiedNeighbors[0] = siblingID2;
modifiedNeighbors[1] = res1X;
modifiedNeighbors[2] = res2Y;
pParams.neighboursOutputBuffer[siblingID0] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID0] = modifiedNeighbors;
modifiedNeighbors[0] = res0Y;
modifiedNeighbors[1] = currentID;
modifiedNeighbors[2] = res2X;
pParams.neighboursOutputBuffer[siblingID1] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID1] = modifiedNeighbors;
modifiedNeighbors[0] = res1Y;
modifiedNeighbors[1] = siblingID0;
modifiedNeighbors[2] = currentID;
pParams.neighboursOutputBuffer[siblingID2] = uint4(modifiedNeighbors, 0);
pParams.neighboursOutputBuffer[siblingID2] = modifiedNeighbors;
// Keep track of the parent
BisectorData modifiedBisector = cBisectorData;
@@ -930,7 +958,8 @@ void ValidateBisector(uint currentID)
return;
// Load the bisector data of the target element
uint3 cNeighbors = pParams.neighboursBuffer[currentID].xyz;
uint4 cNeighbors = pParams.neighboursBuffer[currentID];
uint4 tempnNeighbours[3];
bool failed = false;
uint targetNeighbor = INVALID_POINTER;
@@ -941,7 +970,8 @@ void ValidateBisector(uint currentID)
if (neighborID != INVALID_POINTER)
{
bool found = false;
uint3 nNeighbors = pParams.neighboursBuffer[neighborID].xyz;
uint4 nNeighbors = pParams.neighboursBuffer[neighborID];
tempnNeighbours[i] = nNeighbors;
for (uint j = 0; j < 3; ++j)
{
if (nNeighbors[j] == currentID)
src/Engine/Graphics/CBT/CBT.cpp
+46 -19
View File
@@ -124,7 +124,7 @@ CPUMesh Seele::generateCPUMesh(uint32 cbtNumElements) {
neighbours.x = halfedge.prevID != -1 ? cbtNumElements + halfedge.prevID : UINT32_MAX;
neighbours.y = halfedge.nextID != -1 ? cbtNumElements + halfedge.nextID : UINT32_MAX;
neighbours.z = halfedge.twinID != -1 ? cbtNumElements + halfedge.twinID : UINT32_MAX;
result.neighborsArray[elementID] = UVector4(neighbours, 1);
result.neighborsArray[elementID] = UVector4(neighbours, 0);
result.basePoints[3 * halfedgeIdx + 2] = basePoints[halfedge.vertexID];
@@ -404,6 +404,7 @@ void MeshUpdater::evaluateLeb(const BaseMesh& baseMesh, CBTMesh& mesh, Gfx::PDes
Gfx::PUniformBuffer geometryBuffer, Gfx::PUniformBuffer updateBuffer, Gfx::PShaderBuffer lebMatrixCache,
bool clear, bool complete) {
if (clear) {
graphics->beginDebugRegion("ClearLEB");
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryBuffer);
set->updateBuffer(LEB_POSITION_BUFFER, 0, mesh.currentVertexBuffer);
@@ -415,8 +416,10 @@ void MeshUpdater::evaluateLeb(const BaseMesh& baseMesh, CBTMesh& mesh, Gfx::PDes
graphics->executeCommands(std::move(clearCmd));
mesh.currentVertexBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
graphics->endDebugRegion();
}
graphics->waitDeviceIdle();
graphics->beginDebugRegion("EvaluateLEB");
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryBuffer);
set->updateBuffer(UPDATE_CB, 0, updateBuffer);
@@ -436,6 +439,7 @@ void MeshUpdater::evaluateLeb(const BaseMesh& baseMesh, CBTMesh& mesh, Gfx::PDes
graphics->executeCommands(std::move(evalCmd));
mesh.lebVertexBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
graphics->endDebugRegion();
}
void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::PUniformBuffer geometryCB, Gfx::PUniformBuffer updateCB) {
@@ -444,8 +448,8 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
Gfx::PShaderBuffer nextNeighborsBuffer = mesh.neighborsBuffers[nextNeighborsBufferIdx];
resetBuffers(mesh);
// classify
graphics->waitDeviceIdle();
graphics->beginDebugRegion("Classify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -467,8 +471,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.updateBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// prepare indirect pass
graphics->beginDebugRegion("PrepareIndirect");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(ALLOCATE_BUFFER, 0, mesh.classificationBuffer);
@@ -482,9 +488,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
indirectBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// split pass
graphics->beginDebugRegion("Split");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -508,9 +515,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.allocateBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// prepare indirect pass
graphics->beginDebugRegion("PrepareIndirect");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(ALLOCATE_BUFFER, 0, mesh.allocateBuffer);
@@ -524,9 +532,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
indirectBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Allocate Pass
graphics->beginDebugRegion("Allocate");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -546,17 +555,19 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.updateBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// copy
graphics->beginDebugRegion("Copy");
currentNeighborsBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_TRANSFER_READ_BIT, Gfx::SE_PIPELINE_STAGE_TRANSFER_BIT);
graphics->copyBuffer(currentNeighborsBuffer, nextNeighborsBuffer);
nextNeighborsBuffer->pipelineBarrier(Gfx::SE_ACCESS_TRANSFER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_TRANSFER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// bisect
graphics->beginDebugRegion("Bisect");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -579,9 +590,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
nextNeighborsBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Prepare indirect pass
graphics->beginDebugRegion("PrepareIndirect");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(ALLOCATE_BUFFER, 0, mesh.propagateBuffer);
@@ -595,9 +607,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
indirectBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// propagate split pass
graphics->beginDebugRegion("PropagateBisect");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -615,9 +628,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.updateBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// prepare simplify
graphics->beginDebugRegion("PrepareSimplify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -637,9 +651,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.simplificationBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Prepare Indirect Simplify
graphics->beginDebugRegion("PrepareIndirectSimplify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(ALLOCATE_BUFFER, 0, mesh.simplificationBuffer);
@@ -653,9 +668,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
indirectBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Simplify Pass
graphics->beginDebugRegion("Simplify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -677,9 +693,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
nextNeighborsBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Prepare Indirect Propagate Simplify
graphics->beginDebugRegion("PrepareIndirectPropagateSimplify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(ALLOCATE_BUFFER, 0, mesh.propagateBuffer);
@@ -693,9 +710,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
indirectBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Propagate Simplify Pass
graphics->beginDebugRegion("PropagateSimplify");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -714,9 +732,10 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
nextNeighborsBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
// Update Tree
graphics->beginDebugRegion("Update Tree");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(CBT_BUFFER0, 0, mesh.gpuCBT.bufferArray[0]);
@@ -747,6 +766,7 @@ void MeshUpdater::update(CBTMesh& mesh, Gfx::PDescriptorSet viewParamsSet, Gfx::
mesh.gpuCBT.bufferArray[0]->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
mesh.currentNeighborsBufferIdx = nextNeighborsBufferIdx;
@@ -778,6 +798,7 @@ void MeshUpdater::validation(const CBTMesh& mesh, Gfx::PUniformBuffer geometryCB
}
void MeshUpdater::resetBuffers(const CBTMesh& mesh) {
graphics->beginDebugRegion("ResetBuffers");
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(CBT_BUFFER0, 0, mesh.gpuCBT.bufferArray[0]);
set->updateBuffer(CBT_BUFFER1, 0, mesh.gpuCBT.bufferArray[1]);
@@ -796,12 +817,14 @@ void MeshUpdater::resetBuffers(const CBTMesh& mesh) {
graphics->executeCommands(std::move(resetCmd));
memoryBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
graphics->endDebugRegion();
}
void MeshUpdater::prepareIndirection(CBTMesh& mesh, Gfx::PUniformBuffer geometryCB) {
uint32 numGroups = (mesh.totalNumElements + WORKGROUP_SIZE - 1) / WORKGROUP_SIZE;
// Bitsector indexation
graphics->waitDeviceIdle();
graphics->beginDebugRegion("BisectorIndexation");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryCB);
@@ -821,7 +844,9 @@ void MeshUpdater::prepareIndirection(CBTMesh& mesh, Gfx::PUniformBuffer geometry
}
mesh.indirectDrawBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
// Prepare bisector indirect dispatch
graphics->endDebugRegion();
graphics->waitDeviceIdle();
graphics->beginDebugRegion("PrepareBisectorIndirectDispatch");
{
Gfx::PDescriptorSet set = layout->allocateDescriptorSet();
set->updateBuffer(INDIRECT_DRAW_BUFFER, 0, mesh.indirectDrawBuffer);
@@ -836,6 +861,8 @@ void MeshUpdater::prepareIndirection(CBTMesh& mesh, Gfx::PUniformBuffer geometry
mesh.indirectDispatchBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
}
graphics->endDebugRegion();
graphics->waitDeviceIdle();
}
bool MeshUpdater::checkIfValid() { return true; }
src/Engine/Graphics/Graphics.h
+3
View File
@@ -97,6 +97,9 @@ class Graphics {
virtual Gfx::OPipelineStatisticsQuery createPipelineStatisticsQuery(const std::string& name = "") = 0;
virtual Gfx::OTimestampQuery createTimestampQuery(uint64 numTimestamps, const std::string& name = "") = 0;
virtual void beginDebugRegion(const std::string& name) = 0;
virtual void endDebugRegion() = 0;
virtual void resolveTexture(PTexture source, PTexture destination) = 0;
virtual void copyTexture(PTexture src, PTexture dst) = 0;
src/Engine/Graphics/Initializer.h
+1 -1
View File
@@ -34,7 +34,7 @@ struct WindowCreateInfo {
};
struct ViewportCreateInfo {
URect dimensions;
float fieldOfView = 1.222f; // 70 deg
float fieldOfView = glm::radians(70.0f);
Gfx::SeSampleCountFlags numSamples;
};
// doesnt own the data, only proxy it
src/Engine/Graphics/RenderPass/RenderPass.cpp
+22
View File
@@ -25,6 +25,28 @@ RenderPass::~RenderPass() {}
void RenderPass::beginFrame(const Component::Camera& cam) {
auto screenDim = Vector2(static_cast<float>(viewport->getWidth()), static_cast<float>(viewport->getHeight()));
viewParams = {
.viewFrustum =
{
.planes =
{
Plane{
.n = Vector(-0.62628, 0, 0.7796),
.d = 3.898,
},
Plane{
.n = Vector(0.62628, 0, 0.7796),
.d = 3.898,
},
Plane{
.n = Vector(0, 0.81915, 0.57358),
.d = 2.86788,
},
Plane{
.n = Vector(0, -0.81915, 0.57358),
.d = 2.86788,
},
},
},
.viewMatrix = cam.getViewMatrix(),
.inverseViewMatrix = glm::inverse(cam.getViewMatrix()),
.projectionMatrix = viewport->getProjectionMatrix(),
src/Engine/Graphics/RenderPass/TerrainRenderer.cpp
+17 -11
View File
@@ -10,13 +10,13 @@ using namespace Seele;
TerrainRenderer::TerrainRenderer(Gfx::PGraphics graphics, PScene scene, Gfx::PDescriptorLayout viewParamsLayout,
Gfx::PDescriptorSet viewParamsSet)
: graphics(graphics), scene(scene) {
//Gfx::OPipelineLayout test = graphics->createPipelineLayout();
//graphics->beginShaderCompilation(ShaderCompilationInfo{
// .modules = {"CompileTest"},
// .entryPoints = {{"Reset", "CompileTest"}},
// .rootSignature = test,
//});
//graphics->createComputeShader({0});
Gfx::OPipelineLayout test = graphics->createPipelineLayout();
graphics->beginShaderCompilation(ShaderCompilationInfo{
.modules = {"CompileTest"},
.entryPoints = {{"Split", "CompileTest"}},
.rootSignature = test,
});
graphics->createComputeShader({0});
meshUpdater.init(graphics, viewParamsLayout);
lebCache.init(graphics, 5);
CBT<18> cbt;
@@ -223,9 +223,9 @@ TerrainRenderer::TerrainRenderer(Gfx::PGraphics graphics, PScene scene, Gfx::PDe
Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT);
UpdateCB updateCB = {
.viewProjectionMatrix = Matrix4(0),
.triangleSize = 10.0f,
.triangleSize = 60.0f,
.maxSubdivisionDepth = 63,
.fov = 70.0f,
.fov = 1.22173f,
.farPlaneDistance = 1000.0f,
};
updateBuffer = graphics->createUniformBuffer(UniformBufferCreateInfo{
@@ -265,6 +265,8 @@ TerrainRenderer::TerrainRenderer(Gfx::PGraphics graphics, PScene scene, Gfx::PDe
graphics->waitDeviceIdle();
applyDeformation(viewParamsSet);
graphics->waitDeviceIdle();
meshUpdater.validation(plainMesh, geometryBuffer);
graphics->waitDeviceIdle();
}
TerrainRenderer::~TerrainRenderer() {}
@@ -274,9 +276,9 @@ static bool first = true;
void TerrainRenderer::beginFrame(Gfx::PDescriptorSet viewParamsSet, const Component::Camera& cam) {
UpdateCB updateCB = {
.viewProjectionMatrix = viewport->getProjectionMatrix() * cam.getViewMatrix(),
.triangleSize = 10.0f,
.triangleSize = 60.0f,
.maxSubdivisionDepth = 63,
.fov = 70.0f,
.fov = 1.22173f,
.farPlaneDistance = 1000.0f,
};
updateBuffer = graphics->createUniformBuffer(UniformBufferCreateInfo{
@@ -296,6 +298,8 @@ void TerrainRenderer::beginFrame(Gfx::PDescriptorSet viewParamsSet, const Compon
graphics->waitDeviceIdle();
applyDeformation(viewParamsSet);
graphics->waitDeviceIdle();
meshUpdater.validation(plainMesh, geometryBuffer);
graphics->waitDeviceIdle();
}
Gfx::ORenderCommand TerrainRenderer::render(Gfx::PDescriptorSet viewParamsSet) {
@@ -344,6 +348,7 @@ void TerrainRenderer::setViewport(Gfx::PViewport _viewport, Gfx::PRenderPass ren
}
void TerrainRenderer::applyDeformation(Gfx::PDescriptorSet viewParamsSet) {
graphics->beginDebugRegion("ApplyDeformation");
Gfx::PDescriptorSet set = meshUpdater.layout->allocateDescriptorSet();
set->updateBuffer(GEOMETRY_CB, 0, geometryBuffer);
set->updateBuffer(INDIRECT_DRAW_BUFFER, 0, plainMesh.indirectDrawBuffer);
@@ -358,4 +363,5 @@ void TerrainRenderer::applyDeformation(Gfx::PDescriptorSet viewParamsSet) {
graphics->executeCommands(std::move(command));
plainMesh.currentVertexBuffer->pipelineBarrier(Gfx::SE_ACCESS_SHADER_WRITE_BIT, Gfx::SE_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
Gfx::SE_ACCESS_SHADER_READ_BIT, Gfx::SE_PIPELINE_STAGE_VERTEX_SHADER_BIT);
graphics->endDebugRegion();
}
src/Engine/Graphics/Vulkan/Graphics.cpp
+36 -11
View File
@@ -33,6 +33,8 @@ thread_local PCommandPool Seele::Vulkan::Graphics::transferCommands = nullptr;
PFN_vkCmdDrawMeshTasksEXT cmdDrawMeshTasks;
PFN_vkCmdDrawMeshTasksIndirectEXT cmdDrawMeshTasksIndirect;
PFN_vkSetDebugUtilsObjectNameEXT setDebugUtilsObjectName;
PFN_vkQueueBeginDebugUtilsLabelEXT queueBeginDebugUtilsLabelEXT;
PFN_vkQueueEndDebugUtilsLabelEXT queueEndDebugUtilsLabelEXT;
PFN_vkCreateAccelerationStructureKHR createAccelerationStructure;
PFN_vkCmdBuildAccelerationStructuresKHR cmdBuildAccelerationStructures;
PFN_vkGetAccelerationStructureBuildSizesKHR getAccelerationStructureBuildSize;
@@ -57,6 +59,18 @@ VkResult vkSetDebugUtilsObjectNameEXT(VkDevice device, const VkDebugUtilsObjectN
return VK_SUCCESS;
}
void vkQueueBeginDebugUtilsLabelEXT(VkQueue queue, const VkDebugUtilsLabelEXT* pLabelInfo) {
if (queueBeginDebugUtilsLabelEXT != nullptr) {
queueBeginDebugUtilsLabelEXT(queue, pLabelInfo);
}
}
void vkQueueEndDebugUtilsLabelEXT(VkQueue queue) {
if (queueEndDebugUtilsLabelEXT != nullptr) {
queueEndDebugUtilsLabelEXT(queue);
}
}
VkResult vkCreateAccelerationStructureKHR(VkDevice device, const VkAccelerationStructureCreateInfoKHR* pCreateInfo,
const VkAllocationCallbacks* pAllocator, VkAccelerationStructureKHR* pAccelerationStructure) {
return createAccelerationStructure(device, pCreateInfo, pAllocator, pAccelerationStructure);
@@ -164,7 +178,7 @@ void Graphics::endRenderPass() { getGraphicsCommands()->getCommands()->endRender
void Graphics::waitDeviceIdle() {
getGraphicsCommands()->submitCommands();
vkDeviceWaitIdle(handle);
vkDeviceWaitIdle(handle);
getGraphicsCommands()->refreshCommands();
}
@@ -288,6 +302,17 @@ Gfx::OTimestampQuery Graphics::createTimestampQuery(uint64 numTimestamps, const
return new TimestampQuery(this, name);
}
void Graphics::beginDebugRegion(const std::string& name) {
VkDebugUtilsLabelEXT label = {
.sType = VK_STRUCTURE_TYPE_DEBUG_UTILS_LABEL_EXT,
.pNext = nullptr,
.pLabelName = name.c_str(),
};
vkQueueBeginDebugUtilsLabelEXT(queues[graphicsQueue]->getHandle(), &label);
}
void Graphics::endDebugRegion() { vkQueueEndDebugUtilsLabelEXT(queues[graphicsQueue]->getHandle()); }
void Graphics::resolveTexture(Gfx::PTexture source, Gfx::PTexture destination) {
PTextureBase sourceTex = source.cast<TextureBase>();
PTextureBase destinationTex = destination.cast<TextureBase>();
@@ -389,7 +414,6 @@ void Graphics::copyBuffer(Gfx::PShaderBuffer srcBuffer, Gfx::PShaderBuffer dstBu
vkCmdCopyBuffer(getGraphicsCommands()->getCommands()->getHandle(), src->getHandle(), dst->getHandle(), 1, &region);
}
Gfx::OBottomLevelAS Graphics::createBottomLevelAccelerationStructure(const Gfx::BottomLevelASCreateInfo& createInfo) {
return new BottomLevelAS(this, createInfo);
}
@@ -401,9 +425,9 @@ Gfx::OTopLevelAS Graphics::createTopLevelAccelerationStructure(const Gfx::TopLev
void Graphics::buildBottomLevelAccelerationStructures(Array<Gfx::PBottomLevelAS> data) {
Gfx::PShaderBuffer verticesBuffer = StaticMeshVertexData::getInstance()->getPositionBuffer();
Gfx::PIndexBuffer indexBuffer = StaticMeshVertexData::getInstance()->getIndexBuffer();
// Setup vertices and indices for a single triangle
// Create buffers for the bottom level geometry
// For the sake of simplicity we won't stage the vertex data to the GPU memory
@@ -411,7 +435,6 @@ void Graphics::buildBottomLevelAccelerationStructures(Array<Gfx::PBottomLevelAS>
const VkBufferUsageFlags buffer_usage_flags = VK_BUFFER_USAGE_ACCELERATION_STRUCTURE_BUILD_INPUT_READ_ONLY_BIT_KHR |
VK_BUFFER_USAGE_SHADER_DEVICE_ADDRESS_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT;
VkBufferCreateInfo transformBufferInfo = {
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
@@ -437,9 +460,9 @@ void Graphics::buildBottomLevelAccelerationStructures(Array<Gfx::PBottomLevelAS>
VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT,
VK_PIPELINE_STAGE_ACCELERATION_STRUCTURE_BUILD_BIT_KHR | VK_PIPELINE_STAGE_TRANSFER_BIT);
verticesBuffer->pipelineBarrier(VK_ACCESS_TRANSFER_WRITE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_SHADER_READ_BIT,
VK_PIPELINE_STAGE_ACCELERATION_STRUCTURE_BUILD_BIT_KHR);
VK_PIPELINE_STAGE_ACCELERATION_STRUCTURE_BUILD_BIT_KHR);
indexBuffer->pipelineBarrier(VK_ACCESS_TRANSFER_WRITE_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT, VK_ACCESS_SHADER_READ_BIT,
VK_PIPELINE_STAGE_ACCELERATION_STRUCTURE_BUILD_BIT_KHR);
VK_PIPELINE_STAGE_ACCELERATION_STRUCTURE_BUILD_BIT_KHR);
Array<VkAccelerationStructureGeometryKHR> geometries(data.size());
Array<VkAccelerationStructureBuildGeometryInfoKHR> buildGeometries(data.size());
@@ -672,7 +695,7 @@ void Graphics::initInstance(GraphicsInitializer initInfo) {
extensions.add("VK_KHR_portability_enumeration");
#endif
Array<const char*> layers = initInfo.layers;
//layers.add("VK_LAYER_KHRONOS_validation");
// layers.add("VK_LAYER_KHRONOS_validation");
VkInstanceCreateInfo info = {
.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO,
.pNext = nullptr,
@@ -885,9 +908,9 @@ void Graphics::createDevice(GraphicsInitializer initializer) {
#ifdef __APPLE__
initializer.deviceExtensions.add("VK_KHR_portability_subset");
#endif
//initializer.deviceExtensions.add(VK_KHR_ACCELERATION_STRUCTURE_EXTENSION_NAME);
//initializer.deviceExtensions.add(VK_KHR_RAY_TRACING_PIPELINE_EXTENSION_NAME);
//initializer.deviceExtensions.add(VK_KHR_DEFERRED_HOST_OPERATIONS_EXTENSION_NAME);
// initializer.deviceExtensions.add(VK_KHR_ACCELERATION_STRUCTURE_EXTENSION_NAME);
// initializer.deviceExtensions.add(VK_KHR_RAY_TRACING_PIPELINE_EXTENSION_NAME);
// initializer.deviceExtensions.add(VK_KHR_DEFERRED_HOST_OPERATIONS_EXTENSION_NAME);
VkDeviceCreateInfo deviceInfo = {
.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO,
.pNext = &features,
@@ -920,6 +943,8 @@ void Graphics::createDevice(GraphicsInitializer initializer) {
cmdDrawMeshTasks = (PFN_vkCmdDrawMeshTasksEXT)vkGetDeviceProcAddr(handle, "vkCmdDrawMeshTasksEXT");
cmdDrawMeshTasksIndirect = (PFN_vkCmdDrawMeshTasksIndirectEXT)vkGetDeviceProcAddr(handle, "vkCmdDrawMeshTasksIndirectEXT");
setDebugUtilsObjectName = (PFN_vkSetDebugUtilsObjectNameEXT)vkGetInstanceProcAddr(instance, "vkSetDebugUtilsObjectNameEXT");
queueBeginDebugUtilsLabelEXT = (PFN_vkQueueBeginDebugUtilsLabelEXT)vkGetInstanceProcAddr(instance, "vkQueueBeginDebugUtilsLabelEXT");
queueEndDebugUtilsLabelEXT = (PFN_vkQueueEndDebugUtilsLabelEXT)vkGetInstanceProcAddr(instance, "vkQueueEndDebugUtilsLabelEXT");
createAccelerationStructure = (PFN_vkCreateAccelerationStructureKHR)vkGetDeviceProcAddr(handle, "vkCreateAccelerationStructureKHR");
cmdBuildAccelerationStructures =
src/Engine/Graphics/Vulkan/Graphics.h
+3
View File
@@ -76,6 +76,9 @@ class Graphics : public Gfx::Graphics {
virtual Gfx::OPipelineStatisticsQuery createPipelineStatisticsQuery(const std::string& name = "") override;
virtual Gfx::OTimestampQuery createTimestampQuery(uint64 numTimestamps, const std::string& name = "") override;
virtual void beginDebugRegion(const std::string& name) override;
virtual void endDebugRegion() override;
virtual void resolveTexture(Gfx::PTexture source, Gfx::PTexture destination) override;
virtual void copyTexture(Gfx::PTexture src, Gfx::PTexture dst) override;
src/Engine/Graphics/slang-compile.cpp
+9 -9
View File
@@ -33,14 +33,14 @@ void Seele::beginCompilation(const ShaderCompilationInfo& info, SlangCompileTarg
slang::SessionDesc sessionDesc;
sessionDesc.flags = 0;
Array<slang::CompilerOptionEntry> option = {
{
.name = slang::CompilerOptionName::Capability,
.value =
{
.kind = slang::CompilerOptionValueKind::Int,
.intValue0 = globalSession->findCapability("GLSL_450"),
},
},
//{
// .name = slang::CompilerOptionName::Capability,
// .value =
// {
// .kind = slang::CompilerOptionValueKind::Int,
// .intValue0 = globalSession->findCapability("GLSL_450"),
// },
//},
{
.name = slang::CompilerOptionName::EmitSpirvDirectly,
.value =
@@ -88,7 +88,7 @@ void Seele::beginCompilation(const ShaderCompilationInfo& info, SlangCompileTarg
sessionDesc.preprocessorMacroCount = macros.size();
sessionDesc.preprocessorMacros = macros.data();
slang::TargetDesc targetDesc;
targetDesc.profile = globalSession->findProfile("glsl_450");
targetDesc.profile = globalSession->findProfile("spirv_1_6");
targetDesc.format = target;
sessionDesc.targetCount = 1;
sessionDesc.targets = &targetDesc;