Threaded GRC blocks

I’m new to gnu radio, but I am trying to port a thread SDR application
in
to a GRC block. My application buffers a bunch of data and then
performs
some signal processing on it that can take up to 500ms. Once processing
is
complete, the processing thread waits a certain amount of time before
reading the buffer and then processing again, meanwhile the main trhread
is
consuming samples and advancing a sample counter. I was wondering what
the
best way to implement this as a GRC block. Currently I am creating the
thread in the the block constructor and killing it in the destructor.
I’ve
done a search to try and find a solution to this but not finding much,
or
constantly be directed at the same results that don’t help

Hi Jon,

I’ll try to structure this and reply in-text, so we can get to a mutual
understanding faster :slight_smile:

My application buffers a bunch of data and then performs some signal
processing on it that can take up to 500ms.

Does that mean it takes up to 0.5s worth of sampled signal, or does just
the computation take that long? Is there something like a minimum block
size of samples that your algorithm needs?
Here, a bit of info on what you’re actually doing would be nice.

Once processing is complete, the processing thread waits a certain
amount of time before reading the buffer and then processing again,
meanwhile the main trhread is consuming samples and advancing a sample
counter.

GNU Radio will do exactly that for you: you just write a block that
transforms a set of input items to a set of output items, and GNU Radio
cares about how to fill your input buffer, when to call you, how to
inform you how much items there are to process, and how to notify your
downstream flowgraph neighbors about new data.

I was wondering what the best way to implement this as a GRC block.

Depends on what you do in that block. I have my doubts about your 500ms
computation step not being split into smaller processing steps; but the
feasibility of that completely depends on the actual thing you want to
do…

Currently I am creating the thread in the the block constructor and
killing it in the destructor.

That sounds a bit like you’re doing GNU Radio wrong. Your block is
already running in a thread of its own – that’s what the
thread-per-block scheduler does for you :wink:

Greetings,
Marcus

Thanks for the reply.
I’m doing a ranging application, and the wake up times are related to
the
signal period of my ranging signal. The application can take up to 500ms
to
search for my reference signal in certain conditions and that can be
improved with some prediction, but the first go round can take a while.
I
have no doubt I’m doing GNU radio wrong, but I’m a bit confused as to
how
to implement this otherwise, as I don’t want my processing completed on
every block coming in and I don’t want samples to stop while I’m
processing

On Fri, Jan 16, 2015 at 12:24 PM, Marcus Müller
[email protected]

Thanks for the reply again, I’m still a little confused. I have a
process
interval of 1 second, meaning every 1 seconds I grab n samples and run
my
ranging algorithm which can take a while a little while, in the mean
time I
need to keep track of the absolute sample count so I know the sample
number
of the first sample in those n samples so that I ca keep track of time ,
everything I see says every chunk of data coming in is processed. So I’d
still need my block to keep track of every sample coming in, but only
use a
buffer when I need to based on time so something needs to keep counting
samples while something else waits, and I don’t know how to do that in
two
blocks

Hi Jon,

On 01/16/2015 08:19 PM, Jon W. wrote:

Thanks for the reply again, I’m still a little confused. I have a
process interval of 1 second, meaning every 1 seconds I grab n samples
and run my ranging algorithm which can take a while a little while, in
the mean time I need to keep track of the absolute sample count so I
know the sample number of the first sample in those n samples so that
I ca keep track of time , everything I see says every chunk of data
coming in is processed.
No, you don’t have to keep track of anything; GNU Radio does that for
you. If you’re in a block, you can call the nitems_read (or
nitems_written) methods, which will give you the number of consumed (or
produced) so far.

So I’d still need my block to keep track of every sample coming in,
but only use a buffer when I need to based on time so something needs
to keep counting samples while something else waits, and I don’t know
how to do that in two blocks
I think I don’t really understand your point, I’m afraid.
If I get you correctly, though, then what you want to do is not to
continously stream samples, and process every one of them, but look at
short “bursts” of n samples in regular intervals of let’s say 500ms.
You could do something

RF source block (@f_sample) -> stream_to_vector (vectors of n) ->
keep_n_in_m (keep 1 out of every 500ms/f_sample/n vectors) -> your block
-> visualization_or_file_sink_or_something_of_the_like

Would that fit what you would need, somehow?

Greetings,
Marcus

:slight_smile:

I have no doubt I’m doing GNU radio wrong,
I might have put that a little harsh; sorry. You’re not doing it wrong,
you just weren’t aware of a few core concepts of the GNU Radio
scheduler:
but I’m a bit confused as to how to implement this otherwise, as I
don’t want my processing completed on every block coming in and I
don’t want samples to stop while I’m processing
And exactly that’s the kind of worries GNU Radio strives to take away
from you.
All blocks in your flow graph can run in parallel. So while your block
is still working, the upstream block is already processing what is going
to be the input of your block’s next iteration, while your downstream
block is busy processing what your block produced the last time.
Also, GNU Radio uses input and output buffers, which are identical to
your upstream block’s output and your downstream blocks’ input buffers,
respectively, and makes sure you know how many space there is in these
buffers (or how much samples there are for you to process).

This all is done transparently in separate threads, so you don’t have to
worry about it.

I hope that illustrates why I think that spawning your own thread is not
necessary; it’s already being done for you, and in a manner that allows
you not to care about the correct transportation of data, notifying
threads, ensuring data flow and proper multiprocessor scaling – this
all happens behind the curtains. To the user, each block only has to
care about processing its input as fast as possible to produce output;
the scheduler will coordinate everything else.

In fact, GNU Radio even encourages you to think about how you can
further break down your algorithm, to as well avoid re-inventing the
wheel, and to use optimized algorithms.
Maybe you do an FFT inside? Well, then use the existing GNU Radio FFT.
Are you multiplying to sample streams? Do that with a GNU radio multiply
block, and you’ll profit from SIMD-optimized routines.
Also, when breaking down an algorithm into existing and to-be-written
blocks, you increase the level of parallelity, which generally is a good
thing, because buffering necessary to avoid congestion when a step takes
especially long will then be automatically distributed between a lot of
blocks.

Greetings,
Marcus

:slight_smile: Looking forward to your application (if you plan to release it)!

Since you were talking about ranging:
Do you know of gr-radar [1]? It already comes with quite a set of
algorithms and visualizations, and is generally considered to be awesome
[2]. Installation is quite straightforward[3], and the source is
available under [4].

Greetings,
Marcus

[1] https://grradar.wordpress.com/
[2] by me, at least
[3] either use pybombs, and do a “pybombs install gr-radar”, or just
“git clone [4]”
[4] GitHub - kit-cel/gr-radar: GNU Radio Radar Toolbox

Hi Jon,
On 01/16/2015 09:10 PM, Jon W. wrote:

So I think I should probably say I have a working function that I want
to port in to a Grc block, and given my time constraints, and other
existing frame work, it does not lend itself to be completely broken
down and reconstituted in Grc properly.
That’s pretty understandable :slight_smile:
The way our frameworkworks lends itself well to the structure of
blocks except in this case where our “work” function is processed in a
new thread that reads time and uses it to wake up and read samples
from a buffer while the function that starts the thread keeps time and
costumes samples to copy them in to the buffer the work thread uses
That is really what GNU Radio does: Wake the thread of a block when new
data is available, and makes that thread execute the block’s work
function.
So all you’d had to do is replace the logic that gets you the right
samples by just using the pointer to the input items that you get when
your work() is called.

At this point, though, this all feels a bit like idle chatter, because
it’s hard to discuss the concepts of two frameworks with the other not
really having had a chance to get a feeling for how things work in the
respective environment.

I think the quickest way for you to actually get going is to write your
first block (which should ideally be a bit simpler than what you want to
do in the end); I really suggest doing the guided tutorials [1]; if
installation is already done, after a couple of minutes you should be at
the point where you can start your own block in python, and a couple
minutes more and you’re in the part where C++ blocks are introduced. I
recommend going through the tutorials in order, though, since the C++
stuff refers to the python stuff, which refers to the GRC stuff.

Greetings,
Marcus

[1] https://gnuradio.org/redmine/projects/gnuradio/wiki/Guided_Tutorials


Discuss-gnuradio mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio