Re: [Discuss-gnuradio] Stream tags and None 1 to 1 processing blocks. Failure to propagate

2011-07-19 Thread Tom Rondeau
On Mon, Jul 18, 2011 at 10:28 PM, Colby Boyer colby.bo...@gmail.com wrote:

 Hi All,

 I have seem to run into an issue with the stream tags and interp/decimation
 processing blocks. When computing to add the tag to the proper output
 offset, it always rounds down. In certain cases, computed offset is X.YZ,
 where the first input item to the next block is at X+1. So the stream tag is
 *never* copied over.

 One fix might be the ceil function?

 --Colby



Eric and I went over this pretty carefully, and I thought we handled it
correctly. I wasn't able to follow your XYZ logic in your email. Could you
be more specific about what the problem is that you are seeing?

Tom
___
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio


Re: [Discuss-gnuradio] Stream tags and None 1 to 1 processing blocks. Failure to propagate

2011-07-19 Thread Colby Boyer
Yea, sorry for vague email. Was tired and in a hurry. I encountered the
error when dealing with a bank of polyphase resamplers. I traced the tag
copying to
gnuradio-core/src/lib/runtime/gr_block_executor.cc line 121 for none 1 to 1
(fractional rate) blocks.

The logic written originally is pmt::pmt_from_uint64(newcount * rrate) where
rrate is the block's rate and newcount is the absolute position of the tag
on the input.

I run into this following situation. Say input sample to the resampler is
tagged at index 800. With a resample rate of 0.9984 we get the product of
798.72. This will round down to 798. However the next block has already
processed samples 0 to 798. It is now looking at 799 and forward. So the tag
never propagates.

-Colby

On Tue, Jul 19, 2011 at 5:03 PM, Tom Rondeau trondeau1...@gmail.com wrote:

 On Mon, Jul 18, 2011 at 10:28 PM, Colby Boyer colby.bo...@gmail.comwrote:

 Hi All,

 I have seem to run into an issue with the stream tags and
 interp/decimation processing blocks. When computing to add the tag to the
 proper output offset, it always rounds down. In certain cases, computed
 offset is X.YZ, where the first input item to the next block is at X+1. So
 the stream tag is *never* copied over.

 One fix might be the ceil function?

 --Colby



 Eric and I went over this pretty carefully, and I thought we handled it
 correctly. I wasn't able to follow your XYZ logic in your email. Could you
 be more specific about what the problem is that you are seeing?

 Tom


___
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio


[Discuss-gnuradio] Stream tags and None 1 to 1 processing blocks. Failure to propagate

2011-07-18 Thread Colby Boyer
Hi All,

I have seem to run into an issue with the stream tags and interp/decimation
processing blocks. When computing to add the tag to the proper output
offset, it always rounds down. In certain cases, computed offset is X.YZ,
where the first input item to the next block is at X+1. So the stream tag is
*never* copied over.

One fix might be the ceil function?

--Colby
___
Discuss-gnuradio mailing list
Discuss-gnuradio@gnu.org
https://lists.gnu.org/mailman/listinfo/discuss-gnuradio