Re: [Discuss-gnuradio] Re: Flowgraph slows down and hangs when using benchmark_tx with a custom block

[John Andrews]

An update on this block. I am now using a gr_sync_interpolator to build the block but the performance is still the same. The flowgraph slows down and hangs. I have to force stop it using the 'kill' command on the linux terminal.

What should I do so that the flowgraph works smoothly like benchmark_tx.py normally does with the other modulation schemes?

Thanks

On Mon, Nov 15, 2010 at 12:17 PM, John Andrews <address@hidden> wrote:

This is what I am doing in general_work

1. I read an item from the input stream.
2. Check if its 0x01 or 0x00.
3. If its 0x01 I output the contents of d_pn_array[0], d_n_pn times. (Basically I am spreading the input data by the ( pn_length * d_n_pn times ))
4. But if its 0x00 I output the contents of d_pn_array[1], d_n_pn times. (Basically I am spreading the input data by the ( pn_length * d_n_pn times ))

The arrays d_pn_array[0] and d_pn_array[1] were initialised in the constructor.

I only read the contents of the arrays and set the values to out[i]. This shouldn't take such a long time although I must say that the lengths of the arrays d_pn_array are 1023 and d_n_pn is 5 i.e. I output 1023*5 = 5115 items for each input item.


int
dsss_spreading_b::general_work(int noutput_items,gr_vector_int &ninput_items,gr_vector_const_void_star &input_items,gr_vector_void_star &output_items)
{ 
  const unsigned char *in = (const unsigned char *)input_items[0];
  unsigned char *out = (unsigned char *)output_items[0];
  int data_items=noutput_items/(d_length_PN*d_n_pn);
  int nout=0;
  for(int i=0;i<data_items;i++){
      if(in[i]&0x01){
      for(int j=0;j<d_length_PN*d_n_pn;j++){
          out[nout]=d_pn_array1[j%d_length_PN];
          nout++;
      }
      }
      else{
      for(int j=0;j<d_length_PN*d_n_pn;j++){
          out[nout]=d_pn_array0[j%d_length_PN];
          nout++;
      }
      }
  }
 
   consume(0,data_items);
   return noutput_items;

}

On Mon, Nov 15, 2010 at 12:00 PM, Tom Rondeau <address@hidden> wrote:

On Mon, Nov 15, 2010 at 11:43 AM, John Andrews <address@hidden> wrote:
> On another  note I use 'gr_block' to build this custom block
>
> On Mon, Nov 15, 2010 at 11:37 AM, John Andrews <address@hidden> wrote:
>>
>> Hi,
>> I have a modified dbpsk.py in which I use a custom block after the
>> self.diffenc (differential encoder block). This custom C++ block outputs
>> 1000 output_items of size 'gr_sizeof_char' for each input_item of size
>> 'gr_sizeof_char'. I then use benchmark_tx.py to test the functioning of the
>> modified dbpsk.py and upon doing so the flowgraph slows down incredibly.
>> What can I do to speed up the process?
>>
>> DiffEnc --> Custom Block --> Chunks2Symbols
>> ( n ouputs) --> (n * 1000 outputs) --> (n *1000 outputs)
>>
>> Thanks

You're really going to have to provide a lot more information about
the block you've created. Posting the general_work function would be
useful.

Tom