So the failures are a separate section? So a quad core should get something approaching 3X barring cache contention. Question would be how often in real life do failures occur? And would you make the # of trials dependent on # of cores?
On Tue, Sep 29, 2015 at 3:50 PM, Joe Taylor <j...@princeton.edu> wrote: > On 9/29/2015 4:28 PM, Michael Black wrote: > > What kind of time is it taking now? The overhead of splitting the work > can > > kill any advantage as I'm sure you know from the previous multi-threading > > of JT9. > > Increasing ntrials from 10^4 to 10^5 increased running time by x 6.4: > about 22 minutes for 1000 decoding attempts, or an average of 1.3 s per > attempt. Most successful decodes very quick, typically less than 0.1 s. > But the failures take a long time, up to 9 s for ntrials=100000 (and > linear in ntrials). > > The gain in speed from multi-threading was very significant for > jt9[.exe]. It should be even better for sfrsd, since the task in > question is simple, easy to set up, and almost fully parallelizable. > > -- Joe, K1JT > > > ------------------------------------------------------------------------------ > _______________________________________________ > wsjt-devel mailing list > wsjt-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/wsjt-devel >
------------------------------------------------------------------------------
_______________________________________________ wsjt-devel mailing list wsjt-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/wsjt-devel
