> It appears that we are back in business on SourceForge. Yep! I just committed an update to the experimental wspr decoder (wsprd_exp) that can use the tried-and-true Fano decoder or a new experimental “stack” decoder.
My initial experiments indicate that the stack decoder as currently implemented has about the same performance as the Fano decoder when it is set up to use a maximum of 5000 cycles of the algorithm before timing out. This is half as many cycles (10000) as we give the Fano decoder in our default wsprd configuration. That’s the good news! The bad news is that each cycle of the stack algorithm is more expensive than a cycle of the Fano algorithm. Based on experiments on my fairly speedy laptop and also on a Raspberry Pi, I find that the average number of decodes and execution time of the Fano and stack algorithms are equal when the stack algorithm is run with a 5000 cycle limit. The following command will run the stack decoder with 5000 cycles: ./wsprd_exp -JC 5000 <filename> So why bother with the stack decoder if the performance is the same as what we already have? Good question! Mainly just because I wanted to learn about the stack decoder. I also think that it may pull ahead of the Fano algorithm in a deeper search, and it could serve as a starting point for other variants that might have some real advantages. Steve k9an ------------------------------------------------------------------------------ _______________________________________________ wsjt-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/wsjt-devel
