avtWholeImageCompositerWithZ::avtWholeImageCompositerWithZ()
{
if (avtWholeImageCompositerWithZ::objectCount == 0)
InitializeMPIStuff();
avtWholeImageCompositerWithZ::objectCount++;
}
void
ParallelZComposite(const MPI_Comm &comm,
int npixels,
const float *inz, const unsigned char *inrgba,
float *outz, unsigned char *outrgba,
unsigned char bgr, unsigned char bgg, unsigned char bgb)
{
static bool MPIStuffInitialized = false;
if (!MPIStuffInitialized)
{
InitializeMPIStuff();
MPIStuffInitialized = true;
}
const int chunksize = 1 << 20;
std::vector<Pixel> inpixels(chunksize);
std::vector<Pixel> outpixels(chunksize);
local_bg[0] = bgr;
local_bg[1] = bgg;
local_bg[2] = bgb;
//cerr << "merging "<<npixels<<" pixels, bg="<<int(bgr)<<","<<int(bgg)<<","<<int(bgb)<<"\n";
//cerr << "inpixel[0] = "<<int(inrgba[0])<<","<<int(inrgba[1])<<","<<int(inrgba[2])<<","<<int(inrgba[3])<<"\n";
//cerr << "inzbuff[0] = "<<inz[0]<<endl;
int i_in = 0, i_out = 0;
while (npixels > 0)
{
int chunk = npixels < chunksize ? npixels : chunksize;
for (int i=0; i<chunk; ++i, ++i_in)
{
inpixels[i].z = inz[i_in];
inpixels[i].r = inrgba[i_in*4 + 0];
inpixels[i].g = inrgba[i_in*4 + 1];
inpixels[i].b = inrgba[i_in*4 + 2];
inpixels[i].a = inrgba[i_in*4 + 3];
}
int err = MPI_Allreduce(&inpixels[0], &outpixels[0], chunk,
mpiTypePixel, mpiOpMergePixelBuffers, comm);
if (err != MPI_SUCCESS)
{
int errclass;
MPI_Error_class(err,&errclass);
char err_buffer[4096];
int resultlen;
MPI_Error_string(err,err_buffer,&resultlen);
cerr << err_buffer << endl;
}
for (int i=0; i<chunk; ++i, ++i_out)
{
outz[i_out] = outpixels[i].z;
outrgba[i_out*4 + 0] = outpixels[i].r;
outrgba[i_out*4 + 1] = outpixels[i].g;
outrgba[i_out*4 + 2] = outpixels[i].b;
outrgba[i_out*4 + 3] = outpixels[i].a;
}
npixels -= chunk;
}
}
