On Saturday, March 31, 2012 at 9:32 AM, Maurizio De Tommaso wrote: > Hi Mark, > > I see you are able to reproduce the behavior of slowness of new SIMH code > that I observed on my rx1620 Integrity server. > > As by your suggestion, I modified my configuration file to include the > directive "Set Noasync" and I removed the directive "set xq type=delqa". > Now the new SIMH code runs with acceptable performance, similar to V3.8-1 > release. I turned on the debugging and I captured its output (see attached > file), as per your instructions. During these tests I used EIB0 network > adapter, > Lat and Tcp/Ip protocols . As reported before, my impression is that the > simulated network adapter is able to transmit but it is not able to receive > packets. > > I look forward for your updates
Hi Maurizio, This turned out to be a simple missing /THREADS_ENABLE on the link line for the simulator. Without that linker switch, all threads actually shared the single process and only one could run at a time. The thread executing instructions eventually consumed enough CPU that the pthreads scheduler decided to finally run the disk I/O thread and do the disk I/O. This is what caused the excessively slow behavior. If you pull the latest code from the github master branch, things will work MUCH better. https://github.com/markpizz/simh/zipball/master - Mark _______________________________________________ Simh mailing list [email protected] http://mailman.trailing-edge.com/mailman/listinfo/simh
