Performance of 3.7

Hello,

I have noticed that if I run the ofdm benchmark_tx.py and
benchmark_rx.py
programs on both 3.6 and 3.7 versions of gnuradio, I get drastically
different performance. I have been using the 3.7 live dvd to evaluate
the
newer version.

When running on 3.6 I can use a bandwidth of 5MHz and the receiver is
able
to keep up. If I use the same settings on a clean install of 3.7, after
a
few seconds, the buffers for multiple blocks start to overflow.

Also, on 3.6 if I use the top command to observe the CPU usage, I notice
that it sits around 500-600% on an i7 machine. On the same machine with
3.7, the program only uses 200-300%.

It would appear that for some reason, in 3.7, gnuradio is not making
full
use of the available processors and thus is unable to keep up with the
large sampling rate of 5MHz. If I drop the sampling rate to 500KHz,
everything works fine on 3.7. I’ve used the control port programs to
observe the buffers and the runtime graphs and it appears that a lot of
time is being taken up by FFT filters. Did something change in 3.7 that
would account for the observed differences?

Thank you in advance.

Regards,

Dirk

Thanks Steve.

I ran volk_profile and the performance seemed to improve a bit. That
is,
CPU usage increased to 700% using many more cores of the processor.
However, I still see buffers filling up and taking longer than realtime
to
process. In 3.6 the benchmark_rx program could keep up at 5MHz and
process
data in real time.

Regards,

Dirk

This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.

| Privacy Policy | Terms of Service | Remote Ruby Jobs