Re: [gem5-users] Ruby functional failed for MESI

Hello Malek, I logged the RubyPort trace file, and these are the tail lines of it: 1249879500: system.ruby.l1_cntrl9.sequencer-slave1: Hit callback needs response 1 1249879500: system.ruby.l1_cntrl9.sequencer-slave1: Sending packet back over port 1249879500: system.ruby.l1_cntrl9.sequencer-sl

Re: [gem5-users] library error after compilation libhooks.so.0

Hi Fulya, This error occurs when the hooks library is linked as a shared library. Though I don't recall the details to build the hooks library statically, I suggest that you try doing that before building the benchmarks (building this statically is suggested by the ".la" on the end of "libhooks.

[gem5-users] library error after compilation libhooks.so.0

Hi, I am running Parsec on X86 and I modified something in hooks, so I compiled benchmarks myself and mounted the new versions on the x86root-parsec.img that I got from Parsec website. I did not change anything else in the x86root-parsec image except replacing the newly compiled binaries. However

Re: [gem5-users] Memory controller count has little impact on tested bandwidth.

Does that mean I cannot simulate impact of memory controller count with current Ruby system? Thanks! Runjie On Wed, Sep 18, 2013 at 12:18 PM, Andreas Hansson wrote: > Hi Runjie, > > The memory controllers are still instantiated, and are used by the > devices I suspect. It would be great if Ru

[gem5-users] Running multi-threaded program on GEM5 in SE mode

Hi All, I want to run a multi-threaded program on GEM5 in SE mode. The program can be as simple as this: https://computing.llnl.gov/tutorials/pthreads/samples/hello.c My goal: I want to manually map each thread on different core. What I know and do not know: 1. Pthread is not supported: "Alpha_

Re: [gem5-users] Memory controller count has little impact on tested bandwidth.

Hi Runjie, The memory controllers are still instantiated, and are used by the devices I suspect. It would be great if Ruby could split the directory and the actual memory controller so we would attached e.g. The SimpleDRAM model to Ruby, but at the moment that is not possible. Andreas From: R

Re: [gem5-users] Memory controller count has little impact on tested bandwidth.

Hi, Andreas Thanks for your respond. I noticed that if I change --mem-type in the cmd line, the item [system.mem_ctrls] in config.ini would change accordingly. So when you say --mem-type does not work with Ruby, what do you mean exactly? Does system.mem_ctrls get bypassed and system.ruby.dir_cnt