GPU Pro 4

242 III Image Space

inline __device__

float3 st_integrate_rk2( float2 p0 , float sigma ,

unsigned w, unsigned h,

float step_size , bool adaptive )

{

float radius = 2 * sigma ;

float twoSigma2 = 2 * sigma * sigma ;

float3 c=make_float3( tex2D( texSRC , p0.x , p0.y ));

float sum = 1;

float2 v0 = st_minor_ev(tex2D( texST , p0 .x , p0 .y )) ;

float sign = -1;

float dr = radius / CUDART_PI_F;

do {

float2 v = v0 * sign;

float2 p=p0;

float u=0;

for (;;) {

float2 t = st_minor_ev(tex2D (texST , p .x , p .y ));

if ( dot(v,t)<0)t=-t;

float2 ph = p + 0.5f * step_size * t;

t = st_minor_ev(tex2D (texST , ph .x , ph .y ) );

float vt = dot(v, t);

if (vt < 0) {

t=-t;

vt = - vt ;

}

v=t ...