Indeed, if branch-prediction misses in the CPU caused actual visible effects to data integrity, thousands and thousands of bits of software would be negatively impacted, not just Gnu Radio and not just multi-threaded applications.
All of the CPU-implementation "tricks" that are buried in the CPU, like branch-prediction, out-of-order execution, pipelining, are completely and utterly transparent to executing code. If that weren't the case, then there'd be chaos. > >I've been over the thread-per-block scheduler code a lot in the past >few years. What you are suggesting here is, frankly impossible. And I >don't think that branch prediction has anything to do with it. I >haven't looked at the link you sent, but branch prediction misses >should only slow down the processing but have no effect on the data. > >I think you'll have to look closer at your own blocks for the problem. >Have you written QA for each of the blocks? Best to verify that each >one works as expected on its own before trying to put them all in line >together. > >Tom > >_______________________________________________ >Discuss-gnuradio mailing list >[email protected] >https://lists.gnu.org/mailman/listinfo/discuss-gnuradio > > _______________________________________________ Discuss-gnuradio mailing list [email protected] https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
