Re: [Discuss-gnuradio] Possible bug in keep_m_in_n and patch

As was pointed out to me in IRC, we have a webpage describing how to submit patches. See this page:

http://gnuradio.org/redmine/projects/gnuradio/wiki/Development#How-do-I-submit-patches

But I forgot that you actually submitted a patch in the email, so don't worry about opening an issue on this right now. I'll take the patch and check it in.

(btw, in a simple test I ran, this gave a 10x speedup on that block.)

Thanks!

Tom

On Thu, Sep 25, 2014 at 10:35 AM, Tom Rondeau <tom@trondeau.com> wrote:

Zhe,

To follow up on Martin's point. First, yes, please either use the Issue tracker or submit a patch/pull request via git and github.

Also, just pointing out a bug is one thing, but it's also much easier for us to test and verify if you provide us with an example file.

That having been said, yes, I think your patch is correct. The current behavior does not produce the wrong answer, but it's wasting a ton of computation time.

Please send a patch.

Tom

On Wed, Sep 24, 2014 at 3:49 PM, Martin Braun <martin.braun@ettus.com> wrote:

No comment on the patch, but in general, it helps us a lot if you do this via github & pull request.

M

On 24.09.2014 12:21, Zhe Feng wrote:

Dear all,

I checked the keep_m_in_n block and found a possible bug in it. In the
work function, the code wrote:

consume_each(d_n);
return d_m;

which I think it should be

consume_each(blks*d_n);
return blks*d_m.

while blks=std::min(noutput_items/d_m, ninput_items[0]/d_n).

Since both m and n of the block might change, which means the output to
input rate might change. I changed the set_output_multiple(m) to
set_relative_rate(d_n/d_m). And we made the set_relative_rate(d_n/d_m)
be called in the set_m and set_n function to make sure the rate is
updated on time.

The patch is attached. Could you check it?

Thanks!
Best,
Zhe