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

Removing tons of debugging and sync code

Browse Source
This commit is contained in:
Dynamitos
2024-10-21 21:40:48 +02:00
parent 6506287474
commit f1316eb213
8 changed files with 285 additions and 298 deletions
Download Patch File Download Diff File Expand all files Collapse all files
res/shaders/terrain/CBT.slang
+44 -21
View File
@@ -228,35 +228,58 @@ uint bit_count(uint depth, uint element)
// decodes the position of the i-th one in the bitfield
uint decode_bit(uint handle)
{
uint bitID = 1;
for (uint currentDepth = 0; currentDepth < WAVE_TREE_DEPTH; ++currentDepth)
//uint bitID = 1;
//for (uint currentDepth = 0; currentDepth < WAVE_TREE_DEPTH; ++currentDepth)
//{
// uint heapValue = get_heap_element(2 * bitID);
// uint b = handle < heapValue ? 0 : 1;
//
// bitID = 2 * bitID + b;
// handle -= heapValue * b;
//}
//
//return (bitID ^ OCBT_NUM_ELEMENTS);
for(uint i = 0; i < OCBT_NUM_ELEMENTS; ++i)
{
uint heapValue = get_heap_element(2 * bitID);
uint b = handle < heapValue ? 0 : 1;
bitID = 2 * bitID + b;
handle -= heapValue * b;
if(get_bit(i) == 1)
{
if(handle == 0)
{
return i;
}
handle--;
}
}
return (bitID ^ OCBT_NUM_ELEMENTS);
}
// decodes the position of the i-th zero in the bitfield
uint decode_bit_complement(uint handle)
{
uint bitID = 1u;
uint c = OCBT_NUM_ELEMENTS / 2u;
while (bitID < OCBT_NUM_ELEMENTS) {
uint heapValue = c - get_heap_element(2u * bitID);
uint b = handle < heapValue ? 0u : 1u;
bitID = 2u * bitID + b;
handle -= heapValue * b;
c /= 2u;
//uint bitID = 1u;
//uint c = OCBT_NUM_ELEMENTS / 2u;
//
//while (bitID < OCBT_NUM_ELEMENTS) {
// uint heapValue = c - get_heap_element(2u * bitID);
// uint b = handle < heapValue ? 0u : 1u;
//
// bitID = 2u * bitID + b;
// handle -= heapValue * b;
// c /= 2u;
//}
//
//return (bitID ^ OCBT_NUM_ELEMENTS);
for(uint i = 0; i < OCBT_NUM_ELEMENTS; ++i)
{
if(get_bit(i) == 0)
{
if(handle == 0)
{
return i;
}
handle--;
}
}
return (bitID ^ OCBT_NUM_ELEMENTS);
}
void reduce(uint groupIndex)
res/shaders/terrain/CBTCompute.slang
+100 -102
View File
@@ -10,48 +10,46 @@ void Reset()
ResetBuffers();
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void Classify(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
for(uint i = 0; i < pParams.indirectDrawBuffer[9]; ++i)
{
//if(dispatchID > 0)
// return;
//for(uint i = 0; i < pParams.indirectDrawBuffer[9]; ++i)
//{
// This thread doesn't have any work to do, we're done
//if (currentID >= pParams.indirectDrawBuffer[9])
// return;
if (dispatchID >= pParams.indirectDrawBuffer[9])
return;
// Operate the indirection
uint currentID = pParams.indexedBisectorBuffer[i];
// Operate the indirection
uint currentID = pParams.indexedBisectorBuffer[dispatchID];
// Read the current geometry data
BisectorGeometry bis;
bis.p[0] = pParams.currentVertexBuffer[3 * currentID].xyz;
bis.p[1] = pParams.currentVertexBuffer[3 * currentID + 1].xyz;
bis.p[2] = pParams.currentVertexBuffer[3 * currentID + 2].xyz;
bis.p[3] = pParams.currentVertexBuffer[3 * pParams.geometry.totalNumElements + currentID].xyz;
// Read the current geometry data
BisectorGeometry bis;
bis.p[0] = pParams.currentVertexBuffer[3 * currentID].xyz;
bis.p[1] = pParams.currentVertexBuffer[3 * currentID + 1].xyz;
bis.p[2] = pParams.currentVertexBuffer[3 * currentID + 2].xyz;
bis.p[3] = pParams.currentVertexBuffer[3 * pParams.geometry.totalNumElements + currentID].xyz;
// Classify the element
ClassifyElement(currentID, bis, pParams.geometry.totalNumElements, pParams.geometry.baseDepth);
}
// Classify the element
ClassifyElement(currentID, bis, pParams.geometry.totalNumElements, pParams.geometry.baseDepth);
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void Split(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
//if(dispatchID > 0)
// return;
//for(dispatchID = 0; dispatchID < pParams.classificationBuffer[SPLIT_COUNTER]; ++dispatchID)
//{
if (dispatchID >= pParams.classificationBuffer[SPLIT_COUNTER])
return;
for(dispatchID = 0; dispatchID < pParams.classificationBuffer[SPLIT_COUNTER]; ++dispatchID)
{
//if (dispatchID >= pParams.classificationBuffer[SPLIT_COUNTER])
// return;
// Grab the real elementID
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + dispatchID];
// Grab the real elementID
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + dispatchID];
// Split the element
SplitElement(currentID, pParams.geometry.baseDepth);
}
// Split the element
SplitElement(currentID, pParams.geometry.baseDepth);
}
[numthreads(1, 1, 1)]
@@ -62,102 +60,102 @@ void PrepareIndirect(uint currentID : SV_DispatchThreadID)
pParams.indirectDispatchBuffer[currentID * 3 + 2] = 1;
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void Allocate(uint groupIndex : SV_GroupIndex, uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
for(uint i = 0; i < 64; ++i)
{
//if(dispatchID > 0)
// return;
//
//for(uint i = 0; i < 64; ++i)
//{
// Load the CBT to the LDS
load_buffer_to_shared_memory(i);
}
load_buffer_to_shared_memory(groupIndex);
// If this element doesn't need to be processed, we're done
//if (dispatchID >= pParams.allocateBuffer[0])
// return;
for(uint i = 0; i < pParams.allocateBuffer[0]; ++i)
{
if (dispatchID >= pParams.allocateBuffer[0])
return;
//for(uint i = 0; i < pParams.allocateBuffer[0]; ++i)
//{
// Allocate the required bits
AllocateElement(pParams.allocateBuffer[1 + i]);
}
AllocateElement(pParams.allocateBuffer[1 + dispatchID]);
//}
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void Bisect(uint groupIndex : SV_GroupIndex, uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
// If this element doesn't need to be processed, we're done
//if (dispatchID >= pParams.allocateBuffer[0])
//if(dispatchID > 0)
// return;
for(uint i = 0; i < pParams.allocateBuffer[0]; ++i)
{
// If this element doesn't need to be processed, we're done
if (dispatchID >= pParams.allocateBuffer[0])
return;
//for(uint i = 0; i < pParams.allocateBuffer[0]; ++i)
//{
// Operation the bisection of this element
BisectElement(pParams.allocateBuffer[1 + i], i);
}
BisectElement(pParams.allocateBuffer[1 + dispatchID]);
//}
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void PropagateBisect(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
// If this element doesn't need to be processed, we're done
//if (dispatchID >= pParams.propagateBuffer[0])
//if(dispatchID > 0)
// return;
for(uint i = 0; i < pParams.propagateBuffer[0]; ++i)
{
PropagateBisectElement(pParams.propagateBuffer[2 + i]);
}
// If this element doesn't need to be processed, we're done
if (dispatchID >= pParams.propagateBuffer[0])
return;
//for(uint i = 0; i < pParams.propagateBuffer[0]; ++i)
//{
PropagateBisectElement(pParams.propagateBuffer[2 + dispatchID]);
//}
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void PrepareSimplify(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
// If this element doesn't need to be processed, we're done
//if (dispatchID >= pParams.classificationBuffer[SIMPLIFY_COUNTER])
//if(dispatchID > 0)
// return;
for(uint i = 0; i < pParams.classificationBuffer[SIMPLIFY_COUNTER]; ++i)
{
// Grab the real elementID
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + pParams.geometry.totalNumElements + i];
// If this element doesn't need to be processed, we're done
if (dispatchID >= pParams.classificationBuffer[SIMPLIFY_COUNTER])
return;
//for(uint i = 0; i < pParams.classificationBuffer[SIMPLIFY_COUNTER]; ++i)
//{
// Grab the real elementID
uint currentID = pParams.classificationBuffer[CLASSIFY_COUNTER_OFFSET + pParams.geometry.totalNumElements + dispatchID];
// Simplify an element
PrepareSimplifyElement(currentID);
}
// Simplify an element
PrepareSimplifyElement(currentID);
//}
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void Simplify(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
// This thread doesn't have any work to do, we're done
//if (currentID >= pParams.simplifyBuffer[0])
//if(dispatchID > 0)
// return;
for(uint i = 0; i < pParams.simplifyBuffer[0]; ++i)
{
// Simplify an element
SimplifyElement(pParams.simplifyBuffer[1 + i]);
}
// This thread doesn't have any work to do, we're done
if (dispatchID >= pParams.simplifyBuffer[0])
return;
//for(uint i = 0; i < pParams.simplifyBuffer[0]; ++i)
//{
// Simplify an element
SimplifyElement(pParams.simplifyBuffer[1 + dispatchID]);
//}
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void PropagateSimplify(uint dispatchID : SV_DispatchThreadID)
{
if(dispatchID > 0)
return;
// If this element doesn't need to be processed, we're done
//if (dispatchID >= pParams.propagateBuffer[1])
//if(dispatchID > 0)
// return;
for(uint i = 0; i < pParams.propagateBuffer[1]; ++i)
{
PropagateElementSimplify(pParams.propagateBuffer[2 + i]);
}
// If this element doesn't need to be processed, we're done
if (dispatchID >= pParams.propagateBuffer[1])
return;
//for(uint i = 0; i < pParams.propagateBuffer[1]; ++i)
//{
PropagateElementSimplify(pParams.propagateBuffer[2 + dispatchID]);
//}
}
[numthreads(WORKGROUP_SIZE, 1, 1)]
@@ -180,20 +178,20 @@ void ReduceSecondPass(uint groupIndex : SV_GroupIndex)
reduce_second_pass(groupIndex);
}
[numthreads(1, 1, 1)]
[numthreads(WORKGROUP_SIZE, 1, 1)]
void BisectorIndexation(uint currentID : SV_DispatchThreadID)
{
if(currentID > 0)
return;
// This thread doesn't have any work to do, we're done
//if (currentID >= pParams.geometry.totalNumElements)
//if(currentID > 0)
// return;
// This thread doesn't have any work to do, we're done
if (currentID >= pParams.geometry.totalNumElements)
return;
for(uint i = 0; i < pParams.geometry.totalNumElements; ++i)
{
// Indexate this bisector
BisectorElementIndexation(i);
}
//for(uint i = 0; i < pParams.geometry.totalNumElements; ++i)
//{
// Indexate this bisector
BisectorElementIndexation(currentID);
//}
}
[numthreads(1, 1, 1)]
res/shaders/terrain/LEB.slang
+4 -10
View File
@@ -237,12 +237,6 @@ void ClearBuffer(uint currentID : SV_DispatchThreadID)
pParams.lebPositionBuffer[currentID] = float4(0.0, 0.0, 0.0, 1.0);
}
float3 TransformToPlanetCoordinate(float3 posPS)
{
// Evaluate the planet position
return normalize(posPS);
}
[[vk::push_constant]]
ConstantBuffer<uint> preRendering;
struct Triangle
@@ -316,9 +310,9 @@ void EvaluateLEB(uint currentID : SV_DispatchThreadID, uint groupIndex: SV_Group
EvaluateElementPosition(cHeapID, 0, pParams.geometry.baseDepth, pParams.currentVertexBuffer, parentTri, childTri);
// Export the child
pParams.lebPositionBuffer[3 * currentID] = float4(TransformToPlanetCoordinate(childTri.p[0]), 1.0f);
pParams.lebPositionBuffer[3 * currentID + 1] = float4(TransformToPlanetCoordinate(childTri.p[1]), 1.0f);
pParams.lebPositionBuffer[3 * currentID + 2] = float4(TransformToPlanetCoordinate(childTri.p[2]), 1.0f);
pParams.lebPositionBuffer[3 * currentID] = float4((childTri.p[0]), 1.0f);
pParams.lebPositionBuffer[3 * currentID + 1] = float4((childTri.p[1]), 1.0f);
pParams.lebPositionBuffer[3 * currentID + 2] = float4((childTri.p[2]), 1.0f);
// Export the fourth element
if (pParams.geometry.baseDepth < depth)
@@ -327,6 +321,6 @@ void EvaluateLEB(uint currentID : SV_DispatchThreadID, uint groupIndex: SV_Group
const uint parentOffset = 3 * pParams.geometry.totalNumElements;
// Transform the coordinate to planet space
pParams.lebPositionBuffer[parentOffset + currentID] = float4(TransformToPlanetCoordinate(cHeapID % 2 == 0 ? parentTri.p[0] : parentTri.p[2]), 1.0f);
pParams.lebPositionBuffer[parentOffset + currentID] = float4((cHeapID % 2 == 0 ? parentTri.p[0] : parentTri.p[2]), 1.0f);
}
}
res/shaders/terrain/Parameters.slang
-1
View File
@@ -20,7 +20,6 @@ struct DeformationCB
struct UpdateCB
{
float4x4 viewProjectionMatrix;
float triangleSize;
uint32_t maxSubdivisionDepth;
float fov;
res/shaders/terrain/update_utils.slang
+2 -15
View File
@@ -62,7 +62,7 @@ int ClassifyBisector(in BisectorGeometry tri, uint depth)
return FRUSTUM_CULLED;
// Project the points on screen
float4x4 viewProjectionMatrix = pParams.update.viewProjectionMatrix;//mul(pViewParams.projectionMatrix, pViewParams.viewMatrix);
float4x4 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);
@@ -402,15 +402,11 @@ void evaluate_neighbors(uint currentID, uint bisectorID, out uint resX, out uint
}
}
void BisectElement(uint currentID, uint dispatchID)
void BisectElement(uint currentID)
{
DebugStruct debug;
// If this bisector is not allocated or not subdivided, stop right away
uint64_t baseHeapID = pParams.heapIDBuffer[currentID];
BisectorData cBisectorData = pParams.bisectorDataBuffer[currentID];
debug.baseHeapID = baseHeapID;
debug.subdivision = cBisectorData.subdivisionPattern;
debug.propagateLocation = 0;
if (baseHeapID == 0 || cBisectorData.subdivisionPattern == NO_SPLIT)
return;
@@ -428,11 +424,6 @@ void BisectElement(uint currentID, uint dispatchID)
uint siblingID1 = cBisectorData.indices[1];
uint siblingID2 = cBisectorData.indices[2];
debug.indices[0] = cBisectorData.indices[0];
debug.indices[1] = cBisectorData.indices[1];
debug.indices[2] = cBisectorData.indices[2];
debug.indices[3] = 0;
// Simple subdivision (along the main axis)
if (currentSubdiv == CENTER_SPLIT)
{
@@ -476,7 +467,6 @@ void BisectElement(uint currentID, uint dispatchID)
uint targetLocation = 0;
InterlockedAdd(pParams.propagateBuffer[0], 1, targetLocation);
pParams.propagateBuffer[2 + targetLocation] = siblingID0;
debug.propagateLocation = targetLocation;
}
else if (currentSubdiv == RIGHT_DOUBLE_SPLIT)
{
@@ -535,7 +525,6 @@ void BisectElement(uint currentID, uint dispatchID)
uint targetLocation = 0;
InterlockedAdd(pParams.propagateBuffer[0], 1, targetLocation);
pParams.propagateBuffer[2 + targetLocation] = siblingID0;
debug.propagateLocation = targetLocation;
}
else if (currentSubdiv == LEFT_DOUBLE_SPLIT)
{
@@ -659,8 +648,6 @@ void BisectElement(uint currentID, uint dispatchID)
// How many bits do we need to raise
uint numSiblings = countbits(currentSubdiv);
debug.numSiblings = numSiblings;
pParams.debugBuffer[dispatchID] = debug;
for (uint siblingIdx = 0; siblingIdx < numSiblings; ++siblingIdx)
{
set_bit_atomic(cBisectorData.indices[siblingIdx], true);