Hair Simulation in TressFX - GPU Pro: Advanced Rendering Techniques - page 416

// Local shape constraints
// One thread computes one strand
[ numthreads ( GROUP_SIZE ,1,1)]
void SimulateHair_B ( uint GIndex : SV_GroupIndex ,
uint3 GId : SV_GroupID ,
uint3 DTid : SV_DispatchThreadID )

{

uint globalStrandIndex = GROUP_SIZE ￿ GId . x ;
globalStrandIndex += GIndex ;

int offset =0;
int strandType =0;

if ( globalStrandIndex > 0)
{

offset =

g_GuideHairVerticesOffsetsSRV
. Load ( globalStrandIndex − 1) ;

strandType = g_GuideHairStrandType
. Load ( globalStrandIndex − 1) ;

}

uint numVerticesInTheStrand = g_GuideHairVerticesOffsetsSRV
. Load ( globalStrandIndex ) − offset ;

// Local shape constraint for bending/twisting
{ float4 pos_minus_one = g_GuideHairVertexPositions
[ offset ];
float4 pos = g_GuideHairVertexPositions [ offset +1];
float4 pos_plus_one ;
uint globalVertexIndex =0;
float4 rotGlobal = g_GlobalRotations [ offset ];

for ( uint localVertexIndex =1; localVertexIndex <
numVerticesInTheStrand − 1; localVertexIndex ++ )

{

globalVertexIndex = offset + localVertexIndex ;
pos_plus_one = g_GuideHairVertexPositions
[ globalVertexIndex +1];

// Update position i and i plus 1

// Update local/global frames

}

}

return ;

}

Listing 1.2. SimulateHair_B compute shader kernel.

1.9 Conclusion

The hair simulation of TressFX is optimized for real-time games in many as-
pects, such as using shape constraints and limits on the number of vertices per
strand. Therefore, it is required to understand its basic approaches to maximize
its performance and quality.