Activating/Deactivating Blocks in real time

I have the following problem that I would like your opinion on how to
solve
elegantly:

I have a block A (say a standard sync block with some memory–eg an fir
filter) which
has input x(t) and output y(t) and is pretty computationally intensive.

I would like to add the following functionality to it:

Add a new input s(t) to A which can be 0 or 1.
When s(t)=1 the block operates as before ie, it processes x(t) to
generate
y(t).
If s(t)=0 I would like it to output y(t)=0 and consume the appropriate
x(t)'s from the input.
This way when s(t)=0 block A essentially “does not work”.

This is pretty straightforward to code if I modify the work function of
block A.

However, block A for me is a pretty complicated hierarchical block, so I
don’t have access to its “work” function.

One way to do this is to rewrite the whole hierarchical block A as a
flat
block and then do as suggested above.

Is there a better way?

Thanks
Achilleas

I guess I forgot to make this one thing more clear:

I would like the input stream of block A to be consumed even
when s(t)=1.

With Tim’s suggestion, when s(t)=1 we will have the right output, but
the
input will be waiting in block’s A queue to be processed the
moment s(t) turns 0.
Is this correct, or am I missing something?

But this gives me the following idea:

I guess what i need in addition to that is a block that based
on s(t) either sends the input to the output when s(t)==0 (which is then
connected to A),
or just consumes the input (when s(t)==1).
Is there such a block available? (even if not, this is pretty easy to
write!)

Any comments on this?

thanks for the hints and ideas!
Achilleas

On Wed, Oct 2, 2013 at 5:51 PM, Monahan-Mitchell, Tim <

Can you do this with a 2:1 mux block? Input 2 = constant 0, control
input is s(t)?

From: discuss-gnuradio-bounces+tmonahan=removed_email_address@domain.invalid
[mailto:discuss-gnuradio-bounces+tmonahan=removed_email_address@domain.invalid] On
Behalf Of Achilleas A.
Sent: Wednesday, October 02, 2013 3:48 PM
To: [email protected]
Subject: [Discuss-gnuradio] Activating/Deactivating Blocks in real time

I have the following problem that I would like your opinion on how to
solve elegantly:
I have a block A (say a standard sync block with some memory–eg an fir
filter) which
has input x(t) and output y(t) and is pretty computationally intensive.
I would like to add the following functionality to it:

Add a new input s(t) to A which can be 0 or 1.
When s(t)=1 the block operates as before ie, it processes x(t) to
generate y(t).
If s(t)=0 I would like it to output y(t)=0 and consume the appropriate
x(t)'s from the input.
This way when s(t)=0 block A essentially “does not work”.

This is pretty straightforward to code if I modify the work function of
block A.

However, block A for me is a pretty complicated hierarchical block, so I
don’t have access to its “work” function.
One way to do this is to rewrite the whole hierarchical block A as a
flat block and then do as suggested above.

Is there a better way?

Thanks
Achilleas

Perhaps the gr.copy() block will accomplish what you need.

Help on function copy in module gnuradio.gr.gnuradio_core_general:

copy(*args, **kwargs)
copy(size_t itemsize) -> gr_copy_sptr

output[i] = input[i]

When enabled (default), this block copies its input to its output. 

When
disabled, this block drops its input on the floor.

enabled(self) method of gnuradio.gr.gnuradio_core_general.gr_copy_sptr
instance
enabled(self) -> bool

set_enabled(self, *args, **kwargs) method of
gnuradio.gr.gnuradio_core_general.gr_copy_sptr instance
set_enabled(self, bool enabled)

Thanks for all the suggestions; they helped a lot.

Here is my progress and the BIG problem that I have not solved…

Recall, my original big block is A (it is a complex hier block with 1:1
input:output rate)
as shown below

in --> [ A ] --> out (overall hier “sync” block)

I want to prepend it with a block PRE and append it with a block POST,
so
that
the overall block behaves as A when a control input s(t)=1 and outputs 0
(while consuming
input) when s(t)=1

in --> [ PRE–> A --> POST ] --> out (overall
“sync” block)
^ ^
| |
ctr-------------------------

My motivation for all this is that with s(t) I can control when A works
because it is
really complex and its output is not needed always, while not disturbing
the remaining
graph…

I created one block (call it PRE) that essentially does what the “copy”
block does, with the only difference it is controlled by another control
input s(t).
When s(t)=1 it copies its input to the output and consumes the input
when s(t)=0 it just consumes the input.

I also created another block (call it POST) which is based on the MUX
idea that was suggested (only simpler) that (is supposed to) behave as
follows:
It has two inputs: 0) the output of A and 1) the SAME control signal
s(t).
When s(t)=1 it copies the input to the output.
When s(t)=0 is outputs 0’s

The problem is that the POST block (which is a general block) is
problematic:
I set the ninput_items_required[1] (the control port) to noutput_items
in
the forecast method
BUT
I cannot set the ninput_items_required[0] to any meaningful value,
since the number of input items required depends on the control signal
s(t).
I cannot just set it to noutput_items, because when s(t)=0 block A is
not
producing anything
(that’s the whole point!!!) and thus the POST block will be just waiting
for its input0 queue to fill but
nothing comes to that!!!

Any ideas?

thanks again for all the help,
Achilleas

After some more thought I decided that what i want to accomplish can
only
be done by rearranging the graph
in real time as for example in the “selector” and “valve” blocks.

I have tried a simple example and it works fine this way.

I will post a complete solution once I iron out the details.

thanks
Achilleas

On Thu, Oct 3, 2013 at 8:11 PM, Achilleas A.

On Wed, Oct 2, 2013 at 11:06 PM, Achilleas A.
<[email protected]

wrote:

But this gives me the following idea:
thanks for the hints and ideas!
Achilleas

Achilleas,

I solved a similar problem by creating a general block that consumes an
input stream on some condition, and passes it otherwise. If you pair
this
with another block that tags a sample where passing or rejecting should
begin (correlate_access_code_tag?) it works quite nicely. With this
approach you could use the control pin idea, but I would recommend tags
to
keep it cleaner.

Looking at this might help you get going a little faster if this sounds
like it would work for you:
https://github.com/n-west/gr-west_3_6/blob/master/lib/filter_payload_impl.cc
It’s a little messy because I was trying (and more or less failied) to
make
it dynamic and allow multiple streams coming in to be synchronized and
let
any input stream control output of all streams. It also still uses 3.6
API
(I know, the shame!)

-Nathan