Thank you, Avadh! This bug has been the bane of my existence and I can only imagine how much of a pain in the butt it was to track down.
I haven't had a chance to sit down and test it, but Ishwar from our group also says that this fixes some problems with the scheduler. On Wed, May 4, 2011 at 4:33 PM, avadh patel <[email protected]> wrote: > > On Mon, Apr 18, 2011 at 3:56 PM, Zhe Wang <[email protected]> wrote: > >> Hi, >> >> I am not sure if my simple java test problem is really a bug with QEMU's >> handling of IDE interrupt requests. Since keep restarting simulations does >> not work for me. It always get "stuck" , never get a chance to pass the >> test. If someone used to run java application on marss successfully, please >> give me some suggestion. >> >> Finally I have debugged this issue and tested Java applications on > 'qemu-0.14' branch and it works. The issue was related to VM-Clock offset, > which was not set properly in simulation mode. > > I have also updated 'master' branch with the fixes but haven't tested Java > applications yet. > > - Avadh > > >> Thanks >> zhe >> >> >> On Mon, Apr 18, 2011 at 11:33 AM, avadh patel <[email protected]>wrote: >> >>> I guess finally we found the application that will trigger this bug, and >>> we can try some debugging with generating execution traces to find the >>> issues. Give me few days to work on 0.14 upgrade and see if this bug >>> disappears or not. >>> >>> - Avadh >>> >>> >>> On Sat, Apr 16, 2011 at 10:33 PM, DRAM Ninjas <[email protected]>wrote: >>> >>>> Unfortunately the only way I've been able to get around this bug is just >>>> to keep restarting the simulations until by some chance they don't get >>>> stuck. It's a non-deterministic event so it's very hard to track down and I >>>> don't think there's a real fix/workaround yet. >>>> >>>> I'm hoping that the move to qemu 0.14 might alleviate the issue. >>>> >>>> >>>> >>>> On Sun, Apr 17, 2011 at 1:06 AM, Zhe Wang <[email protected]> wrote: >>>> >>>>> Hi, >>>>> >>>>> You are right. I check the logfile. The RIPs are 0xffffff for a long >>>>> time. It all appear to be in kernel mode. Would you please let me know is >>>>> there any way to solve this problem? Since it seems some people could run >>>>> java benchmark on Marss. >>>>> >>>>> Thanks >>>>> zhe >>>>> >>>>> >>>>> On Sat, Apr 16, 2011 at 11:53 PM, DRAM Ninjas <[email protected]>wrote: >>>>> >>>>>> Actually one thing I've encountered is that sometimes my simulations >>>>>> also get "stuck" ... if you look at your log file, the last field is an >>>>>> RIP >>>>>> value. If you see all your RIPs for a long along time being 0xffffff..... >>>>>> (instead of 0x00000004....), then that means that the simulation got >>>>>> stuck >>>>>> in kernel mode. According to Avadh, this is potentially a bug with QEMU's >>>>>> handling of IDE interrupt requests. Can you look at your log file and >>>>>> see if >>>>>> the RIPs all appear to be in kernel mode? >>>>>> >>>>>> On Sun, Apr 17, 2011 at 12:16 AM, Zhe Wang <[email protected]>wrote: >>>>>> >>>>>>> Hi, >>>>>>> I already tried /start_sim; java HelloWorld; ./stop_sim >>>>>>> >>>>>>> I run it more than 3 hours. It still did not stop. Have no idea what >>>>>>> happened. >>>>>>> >>>>>>> Thanks >>>>>>> zhe >>>>>>> >>>>>>> >>>>>>> On Sat, Apr 16, 2011 at 11:11 PM, DRAM Ninjas >>>>>>> <[email protected]>wrote: >>>>>>> >>>>>>>> how long have you let it run? perhaps the java startup time is >>>>>>>> significant when running in simulation mode so it will take some time >>>>>>>> to get >>>>>>>> the ball rolling. >>>>>>>> >>>>>>>> Try running: >>>>>>>> >>>>>>>> ./start_sim; java HelloWorld; ./stop_sim >>>>>>>> >>>>>>>> And just let it go ... if you get to the stop, then you know that >>>>>>>> the HelloWorld program ran successfully >>>>>>>> >>>>>>>> >>>>>>>> On Sat, Apr 16, 2011 at 11:15 PM, Zhe Wang <[email protected]>wrote: >>>>>>>> >>>>>>>>> Hi, >>>>>>>>> >>>>>>>>> The ptlsim keep printing the status update into the logfile like: >>>>>>>>> Completed 44560000 cycles, 40828598 commits: 134494 >>>>>>>>> Hz, 171620 insns/sec: rip ffffffff81050178 >>>>>>>>> >>>>>>>>> But on the QEMU screen, it freeze like: >>>>>>>>> >>>>>>>>> root@ubuntu:~# ./start_sim; >>>>>>>>> >>>>>>>>> Switching to simulation >>>>>>>>> ptlsim_ptlcall_init: Mapped Ptlcall MMIO page at phys 0x8ffff000, >>>>>>>>> virt 0x7fabc927b000 >>>>>>>>> >>>>>>>>> root@ubuntu:~# java HelloWorld >>>>>>>>> >>>>>>>>> >>>>>>>>> There are no "Hello, World" output as it supposed to be. Any idea >>>>>>>>> about this? >>>>>>>>> >>>>>>>>> Thanks >>>>>>>>> zhe >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Sat, Apr 16, 2011 at 8:56 PM, DRAM Ninjas <[email protected] >>>>>>>>> > wrote: >>>>>>>>> >>>>>>>>>> When you say 'freezing', you mean you aren't getting the ptlsim >>>>>>>>>> prints with the status updates? Is there anything in your log file >>>>>>>>>> after you >>>>>>>>>> run start_sim and the java code? >>>>>>>>>> >>>>>>>>>> >>>>>>>>>> On Sat, Apr 16, 2011 at 8:33 PM, Zhe Wang >>>>>>>>>> <[email protected]>wrote: >>>>>>>>>> >>>>>>>>>>> Hi, >>>>>>>>>>> >>>>>>>>>>> I have encountered problem while runnining simple Java code after >>>>>>>>>>> executing "start_sim" on qemu linux, as follows: >>>>>>>>>>> >>>>>>>>>>> public class HelloWorld { >>>>>>>>>>> public static void main(String[] args) { >>>>>>>>>>> System.out.println("Hello, World"); >>>>>>>>>>> } >>>>>>>>>>> >>>>>>>>>>> } >>>>>>>>>>> >>>>>>>>>>> I tried both open-java6-jdk as well as sun-java6-jdk installed on >>>>>>>>>>> qemu linux, the compiled class file ran very well on Ubuntu on >>>>>>>>>>> Qemu, but >>>>>>>>>>> freeze after starting "start_sim" >>>>>>>>>>> I executed "java -verbose HelloWorld" and got nothing output. >>>>>>>>>>> >>>>>>>>>>> I also tried compile the Java code into executable binary using >>>>>>>>>>> gcj-4.3, and the executable freeze without printing "Hello, World" >>>>>>>>>>> as well. >>>>>>>>>>> Does anybody know what could be the reason cause such issue for >>>>>>>>>>> running Java benchmarks after "start_sim"? >>>>>>>>>>> >>>>>>>>>>> Sincerely >>>>>>>>>>> Zhe >>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>>> On Sat, Apr 16, 2011 at 11:53 AM, DRAM Ninjas < >>>>>>>>>>> [email protected]> wrote: >>>>>>>>>>> >>>>>>>>>>>> To the simulator, it doesn't really matter what type of workload >>>>>>>>>>>> is running inside of the virtual machine. It's all just a machine >>>>>>>>>>>> code >>>>>>>>>>>> instruction stream. Are you asking if the marss disk images have >>>>>>>>>>>> java >>>>>>>>>>>> installed on them? They don't have java installed, but if you have >>>>>>>>>>>> the >>>>>>>>>>>> network working in the virtual machine you can just install java >>>>>>>>>>>> from >>>>>>>>>>>> apt-get. >>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>>> On Sat, Apr 16, 2011 at 11:37 AM, Zhe Wang < >>>>>>>>>>>> [email protected]> wrote: >>>>>>>>>>>> >>>>>>>>>>>>> Hello, all >>>>>>>>>>>>> >>>>>>>>>>>>> I am wondering does Marss support running java application. I >>>>>>>>>>>>> did some simple java test on marss and it seems marss does not >>>>>>>>>>>>> support java >>>>>>>>>>>>> application. Does anybody have idea about this? Thanks. >>>>>>>>>>>>> >>>>>>>>>>>>> zhe >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>>> _______________________________________________ >>>>>>>>>>>>> http://www.marss86.org >>>>>>>>>>>>> Marss86-Devel mailing list >>>>>>>>>>>>> [email protected] >>>>>>>>>>>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >>>>>>>>>>>>> >>>>>>>>>>>>> >>>>>>>>>>>> >>>>>>>>>>> >>>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>> >>>>>> >>>>> >>>> >>>> _______________________________________________ >>>> http://www.marss86.org >>>> Marss86-Devel mailing list >>>> [email protected] >>>> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel >>>> >>>> >>> >> >
_______________________________________________ http://www.marss86.org Marss86-Devel mailing list [email protected] https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
