5.3 The write function

The write_particles functions begins to gather all npart. Then depending if we will write in parallel (self.pio = 'yes') or not, we have to reduce or not the values of ngas and nstar. This is important to write a consistent header.

  def write_particles(self,f):

    npart_all = array(mpi.mpi_Allgather(self.npart))
    
    if self.pio == 'yes':
      ngas  = self.ngas
      nstar = self.nstar
    else:
      ngas  = mpi.mpi_reduce(self.ngas)
      nstar = mpi.mpi_reduce(self.nstar)

The header is written with the function io.WriteBlock() using the variables tpl telling what to write with wich format.

    tpl = ((self.tnow,float32),(ngas,int32),(nstar,int32))
    io.WriteBlock(f,tpl,byteorder=self.byteorder)

And finally, write the vectors, using the function io.WriteArray()

 	
    io.WriteArray(f,self.pos.astype(float32),byteorder=self.byteorder,pio=self.pio,nlocal=npart_all)																												   
    io.WriteArray(f,self.vel.astype(float32),byteorder=self.byteorder,pio=self.pio,nlocal=npart_all)
    io.WriteArray(f,self.mass.astype(float32),byteorder=self.byteorder,pio=self.pio,nlocal=npart_all)																												   
    io.WriteArray(f,self.meta[:self.npart[0]].astype(float32),byteorder=self.byteorder,pio=self.pio,nlocal=npart_all)

See About this document... for information on suggesting changes.