void TriangleMesh::interpolate(unsigned primID, float u, float v, RTCBufferType buffer, float* P, float* dPdu, float* dPdv, size_t numFloats)
{
/* test if interpolation is enabled */
#if defined(DEBUG)
if ((parent->aflags & RTC_INTERPOLATE) == 0)
throw_RTCError(RTC_INVALID_OPERATION,"rtcInterpolate can only get called when RTC_INTERPOLATE is enabled for the scene");
#endif
/* calculate base pointer and stride */
assert((buffer >= RTC_VERTEX_BUFFER0 && buffer <= RTC_VERTEX_BUFFER1) ||
(buffer >= RTC_USER_VERTEX_BUFFER0 && buffer <= RTC_USER_VERTEX_BUFFER1));
const char* src = nullptr;
size_t stride = 0;
if (buffer >= RTC_USER_VERTEX_BUFFER0) {
src = userbuffers[buffer&0xFFFF]->getPtr();
stride = userbuffers[buffer&0xFFFF]->getStride();
} else {
src = vertices[buffer&0xFFFF].getPtr();
stride = vertices[buffer&0xFFFF].getStride();
}
#if !defined(__MIC__)
for (size_t i=0; i<numFloats; i+=4)
{
size_t ofs = i*sizeof(float);
const float w = 1.0f-u-v;
const Triangle& tri = triangle(primID);
const vfloat4 p0 = vfloat4::loadu((float*)&src[tri.v[0]*stride+ofs]);
const vfloat4 p1 = vfloat4::loadu((float*)&src[tri.v[1]*stride+ofs]);
const vfloat4 p2 = vfloat4::loadu((float*)&src[tri.v[2]*stride+ofs]);
const vbool4 valid = vint4(i)+vint4(step) < vint4(numFloats);
if (P ) vfloat4::storeu(valid,P+i,w*p0 + u*p1 + v*p2);
if (dPdu) vfloat4::storeu(valid,dPdu+i,p1-p0);
if (dPdv) vfloat4::storeu(valid,dPdv+i,p2-p0);
}
#else
for (size_t i=0; i<numFloats; i+=16)
{
size_t ofs = i*sizeof(float);
vbool16 mask = (i+16 > numFloats) ? (vbool16)(((unsigned int)1 << (numFloats-i))-1) : vbool16( true );
const float w = 1.0f-u-v;
const Triangle& tri = triangle(primID);
const vfloat16 p0 = vfloat16::loadu(mask,(float*)&src[tri.v[0]*stride+ofs]);
const vfloat16 p1 = vfloat16::loadu(mask,(float*)&src[tri.v[1]*stride+ofs]);
const vfloat16 p2 = vfloat16::loadu(mask,(float*)&src[tri.v[2]*stride+ofs]);
if (P ) vfloat16::storeu_compact(mask,P+i,w*p0 + u*p1 + v*p2);
if (dPdu) vfloat16::storeu_compact(mask,dPdu+i,p1-p0);
if (dPdv) vfloat16::storeu_compact(mask,dPdv+i,p2-p0);
}
#endif
}
void BezierCurves::interpolate(unsigned primID, float u, float v, RTCBufferType buffer, float* P, float* dPdu, float* dPdv, size_t numFloats)
{
/* test if interpolation is enabled */
#if defined(DEBUG)
if ((parent->aflags & RTC_INTERPOLATE) == 0)
throw_RTCError(RTC_INVALID_OPERATION,"rtcInterpolate can only get called when RTC_INTERPOLATE is enabled for the scene");
#endif
/* calculate base pointer and stride */
assert((buffer >= RTC_VERTEX_BUFFER0 && buffer <= RTC_VERTEX_BUFFER1) ||
(buffer >= RTC_USER_VERTEX_BUFFER0 && buffer <= RTC_USER_VERTEX_BUFFER1));
const char* src = nullptr;
size_t stride = 0;
if (buffer >= RTC_USER_VERTEX_BUFFER0) {
src = userbuffers[buffer&0xFFFF]->getPtr();
stride = userbuffers[buffer&0xFFFF]->getStride();
} else {
src = vertices[buffer&0xFFFF].getPtr();
stride = vertices[buffer&0xFFFF].getStride();
}
#if !defined(__MIC__)
for (size_t i=0; i<numFloats; i+=4)
{
size_t ofs = i*sizeof(float);
const size_t curve = curves[primID];
const vfloat4 p0 = vfloat4::loadu((float*)&src[(curve+0)*stride+ofs]);
const vfloat4 p1 = vfloat4::loadu((float*)&src[(curve+1)*stride+ofs]);
const vfloat4 p2 = vfloat4::loadu((float*)&src[(curve+2)*stride+ofs]);
const vfloat4 p3 = vfloat4::loadu((float*)&src[(curve+3)*stride+ofs]);
const vbool4 valid = vint4(i)+vint4(step) < vint4(numFloats);
const BezierCurveT<vfloat4> bezier(p0,p1,p2,p3,0.0f,1.0f,0);
vfloat4 Q, dQdu; bezier.eval(u,Q,dQdu);
if (P ) vfloat4::storeu(valid,P+i,Q);
if (dPdu) vfloat4::storeu(valid,dPdu+i,dQdu);
}
#else
for (size_t i=0; i<numFloats; i+=16)
{
size_t ofs = i*sizeof(float);
vbool16 mask = (i+16 > numFloats) ? (vbool16)(((unsigned int)1 << (numFloats-i))-1) : vbool16( true );
const size_t curve = curves[primID];
const vfloat16 p0 = vfloat16::loadu(mask,(float*)&src[(curve+0)*stride+ofs]);
const vfloat16 p1 = vfloat16::loadu(mask,(float*)&src[(curve+1)*stride+ofs]);
const vfloat16 p2 = vfloat16::loadu(mask,(float*)&src[(curve+2)*stride+ofs]);
const vfloat16 p3 = vfloat16::loadu(mask,(float*)&src[(curve+3)*stride+ofs]);
const BezierCurveT<vfloat16> bezier(p0,p1,p2,p3,0.0f,1.0f,0);
vfloat16 Q, dQdu; bezier.eval(u,Q,dQdu);
if (P ) vfloat16::storeu_compact(mask,P+i,Q);
if (dPdu) vfloat16::storeu_compact(mask,dPdu+i,dQdu);
}
#endif
}
