I agree that it is possibly due to the environment rather than gem5 itself. I could not reproduce the segmentation fault when I run it under valgrind. The default gcc installed in the cluster is gcc 4.1.2, which doesn't meet the requirement. I pointed CC and CXX to a newer version of gcc and g++ (4.4.3), it can build successfully but I'm not sure the library meets the requirement.
________________________________ From: gem5-users-boun...@gem5.org [gem5-users-boun...@gem5.org] on behalf of Anthony Gutierrez [atgut...@umich.edu] Sent: Tuesday, October 02, 2012 11:27 AM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 I would recommend that you test your runs in a more controlled environment. This will make it easier to determine if there is truly a problem with gem5, or if it's your setup - running on a cluster. Ensure that the checkpoint is being taken correctly and that all images/binaries/checkpoints are in the proper location and that your path is setup correctly. -Tony On Tue, Oct 2, 2012 at 10:36 AM, Fangfei Liu <fangf...@princeton.edu<mailto:fangf...@princeton.edu>> wrote: It turned out gem5.debug works find so far (after running about 20 minutes). I was wondering how much slower is gem5.debug than gem5.opt? ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Ali Saidi [sa...@umich.edu<mailto:sa...@umich.edu>] Sent: Monday, October 01, 2012 9:54 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 That is very odd... Could you run the debug binary and see if you get a better trace. Also could you run it under valgrind? Thanks, Ali On 01.10.2012 19:33, Fangfei Liu wrote: Hi, This is the back trace of the stack for the segmentation fault: command line: /tigress-hsm/fangfeil/gem5/build/ARM/gem5.opt --outdir=bbench2 configs/example/fs.py -b bbench-gb --kernel=vmlinux.smp.mouse.arm --frame-capture --caches --cpu-type=detailed Global frequency set at 1000000000000 ticks per second info: kernel located at: /tigress-hsm/fangfeil/gem5/system/binaries/vmlinux.smp.mouse.arm Listening for system connection on port 5900 Listening for system connection on port 3456 0: system.remote_gdb.listener: listening for remote gdb #0 on port 7000 info: Using bootloader at address 0x80000000 **** REAL SIMULATION **** info: Entering event queue @ 0. Starting simulation... warn: The clidr register always reports 0 caches. warn: clidr LoUIS field of 0b001 to match current ARM implementations. warn: The csselr register isn't implemented. Program received signal SIGSEGV, Segmentation fault. 0x00002aaaabafcbab in __dynamic_cast () from /usr/lib64/libstdc++.so.6 (gdb) bt #0 0x00002aaaabafcbab in __dynamic_cast () from /usr/lib64/libstdc++.so.6 #1 0x00000000016a8b15 in MSHR::TargetList::clearDownstreamPending (this=0x3c16f30) at build/ARM/mem/cache/mshr.cc:133 #2 0x00000000016a8d90 in MSHR::markInService (this=, pkt=) at build/ARM/mem/cache/mshr.cc:238 #3 0x00000000016ab88f in MSHRQueue::markInService (this=0x3bfab90, mshr=0x3bfd2e8, pkt=) at build/ARM/mem/cache/mshr_queue.cc:202 #4 0x0000000001681871 in Cache::markInService(MSHR*, Packet*) () #5 0x00000000016964a2 in Cache::MemSidePacketQueue::sendDeferredPacket() () #6 0x0000000001863d44 in EventQueue::serviceOne (this=) at build/ARM/sim/eventq.cc:204 #7 0x00000000018abf82 in simulate (num_cycles=) at build/ARM/sim/simulate.cc:71 #8 0x000000000182738b in _wrap_simulate__SWIG_0 (self=, args=) at build/ARM/python/swig/event_wrap.cc:4371 #9 _wrap_simulate (self=, args=) at build/ARM/python/swig/event_wrap.cc:4421 #10 0x00002aaaab3226c0 in PyObject_Call () from /usr/lib64/libpython2.4.so.1.0 #11 0x00002aaaab380b1c in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #12 0x00002aaaab3835d5 in PyEval_EvalCodeEx () from /usr/lib64/libpython2.4.so.1.0 #13 0x00002aaaab381e08 in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #14 0x00002aaaab381fbb in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #15 0x00002aaaab381fbb in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #16 0x00002aaaab3835d5 in PyEval_EvalCodeEx () from /usr/lib64/libpython2.4.so.1.0 #17 0x00002aaaab383622 in PyEval_EvalCode () from /usr/lib64/libpython2.4.so.1.0 #18 0x00002aaaab381f2d in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #19 0x00002aaaab3835d5 in PyEval_EvalCodeEx () from /usr/lib64/libpython2.4.so.1.0 #20 0x00002aaaab381e08 in PyEval_EvalFrame () from /usr/lib64/libpython2.4.so.1.0 #21 0x00002aaaab3835d5 in PyEval_EvalCodeEx () from /usr/lib64/libpython2.4.so.1.0 #22 0x00002aaaab383622 in PyEval_EvalCode () from /usr/lib64/libpython2.4.so.1.0 #23 0x00002aaaab39fd19 in ?? () from /usr/lib64/libpython2.4.so.1.0 #24 0x000000000186be5c in m5Main (argc=, argv=) at build/ARM/sim/init.cc:256 #25 0x0000000000409c11 in main (argc=9, argv=0x7fffffffe4f8) at build/ARM/sim/main.cc:57 Do you have any idea how to solve this problem? Thanks! Best regards Fangfei ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Ali Saidi [sa...@umich.edu<mailto:sa...@umich.edu>] Sent: Monday, October 01, 2012 5:46 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Without a back trace of where the segmentation fault is happening and perhaps the output of valgrind we're not going to be much help. Ali On 01.10.2012 15:51, Fangfei Liu wrote: I find that it may not be related to checkpoint but segmentation occurs for all the configuration using detailed/arm_detailed cpu type. Is it because of the version of runtime envorinment? ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Fangfei Liu [fangf...@princeton.edu] Sent: Monday, October 01, 2012 2:56 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Actually no. But I think the directory where checkpoint is located is shared by all nodes via GPFS. There are no compiler but only runtime environment installed on the nodes. Will this be a potential problem? Thanks! ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Anthony Gutierrez [atgut...@umich.edu<mailto:atgut...@umich.edu>] Sent: Monday, October 01, 2012 12:44 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Are you copying the checkpoint to the node on which it's running? On Mon, Oct 1, 2012 at 12:11 AM, Fangfei Liu <fangf...@princeton.edu<mailto:fangf...@princeton.edu>> wrote: Sorry I may use the wrong command when I run the executable with gdbto get previous information. The problem I met is a little bit strange. I'm running gem5on a cluster. It turns out that it works fine at the head node of the cluster with gdb(at least no segmentation fault after running for more than 10 minutes.) But when I submitted the job into a job queue and run it on other nodes, segmentation fault occurredafter it run for only several seconds. ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Ali Saidi [sa...@umich.edu<mailto:sa...@umich.edu>] Sent:Sunday, September 30, 2012 11:32 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Could you get a back trace when it happens. It's not a segmentation fault it's a SIGABRT. You're running into a assertion and we need to know who is creating this event that is causing you trouble. Ali On Sep 30, 2012, at 9:44 PM, Fangfei Liu wrote: I tried several different configurations. It works in atomic mode even with caches. But segmentation fault occurs for both arm_detailed and detailed cpu type. The checkpoint is taken in atomic mode. Should the checkpoint be taken with the same cpu type as it runs by restoring the checkpoint? ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Fangfei Liu [fangf...@princeton.edu<mailto:fangf...@princeton.edu>] Sent: Sunday, September 30, 2012 4:16 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Hi, This is the information I got from running it in gdb: Switch at curTick count:10000 info: Entering event queue @ 19318968082000. Starting simulation... gem5.opt: build/ARM/sim/simulate.cc:66: SimLoopExitEvent* simulate(Tick): Assertion `curTick() Program received signal SIGABRT, Aborted. 0x0000003127430285 in raise () from /lib64/libc.so.6 It seems that it failed at the assertion. Do you have any idea what's wrong with it? Thanks! Best regards Fangfei ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Anthony Gutierrez [atgut...@umich.edu<mailto:atgut...@umich.edu>] Sent: Sunday, September 30, 2012 3:32 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 You will need to use gdb or a similar tool to find out where that seg fault is coming from. -Tony On Sun, Sep 30, 2012 at 3:29 PM, Fangfei Liu <fangf...@princeton.edu<mailto:fangf...@princeton.edu>> wrote: I didn't use caches when taking a checkpoint. This is the command line I used to take the checkpoint. build/ARM/gem5.opt --outdir=bbench configs/example/fs.py -b bbench-gb --kernel=vmlinux.smp.mouse.arm --frame-capture --checkpoint-dir=bbench ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Anthony Gutierrez [atgut...@umich.edu<mailto:atgut...@umich.edu>] Sent: Sunday, September 30, 2012 3:04 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 Did you take a checkpoint with caches? If so, you must take a checkpoint without caches. -Tony On Sun, Sep 30, 2012 at 2:58 PM, Fangfei Liu <fangf...@princeton.edu<mailto:fangf...@princeton.edu>> wrote: I see, probably I had some interacting with vncviewer with the mouse when the keyboard is open. I tried to run the benchmark again and now it can run to completion. Thank you very much for your kind help! I have another problem. I took a checkpoint after booting the OS in atomic mode with the default configuration. I did this by inserting a m5 instruction in the bbench-gb.rcS file: /sbin/m5 checkpoint (right after sleep 10). The checkpoint can be taken correctly. Then I want to resume the checkpoint and simulate BBench in arm_detailed mode, but I got segmentation fault. The command line and output is as follows: command line: build/ARM/gem5.opt --outdir=bbench2 configs/example/fs.py -b bbench-gb --kernel=vmlinux.smp.mouse.arm --frame-capture --checkpoint-dir=bbench -r 1 --caches --cpu-type=arm_detailed --l2cache --cacheline_size=64 --l1d_size=32kB --l1d_assoc=4 --l2_size=512kB --l2_assoc=8 Global frequency set at 1000000000000 ticks per second info: kernel located at: /tigress-hsm/fangfeil/gem5/system/binaries/vmlinux.smp.mouse.arm Switch at curTick count:10000 info: Entering event queue @ 19318968082000. Starting simulation... Switched CPUS @ tick 19318968092000 Changing memory mode to timing switching cpus **** REAL SIMULATION **** info: Entering event queue @ 19318968092000. Starting simulation... Segmentation fault I still use the same rcS file except comment the checkpoint command. I was wondering whether I did this correctly. Thanks in advance! Best regards Fangfei ________________________________ From: gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org> [gem5-users-boun...@gem5.org<mailto:gem5-users-boun...@gem5.org>] on behalf of Anthony Gutierrez [atgut...@umich.edu<mailto:atgut...@umich.edu>] Sent: Sunday, September 30, 2012 12:59 PM To: gem5 users mailing list Subject: Re: [gem5-users] questions on running benchmark on gem5 BBench is a web-page rendering benchmark, however, it shouldn't interact with the web-pages in any way. It's doing a google search for something called ewbay and the keyboard is open, that shouldn't happen. Were you interacting with the vncviewer with the mouse? Is it possible that this query could have been entered in error? There is nothing in the BBench source called "ewbay", so it's hard to tell how that url is being produced otherwise. -Tony On Sat, Sep 29, 2012 at 2:15 PM, Fangfei Liu <fangf...@princeton.edu<mailto:fangf...@princeton.edu>> wrote: Thanks for your reply. I didn’t change anything and the command line I used was exactly the same as that in the BBench instruction page. The exact URL is http://www.google.com/m?hl=md&gl=us&source=android-browser-type&q=ewbay<https://owa.princeton.edu/owa/redir.aspx?C=7HmSRhGXSk2J12xsesLMwH3TuxeMcc8IjThnd3hkIWuO2qoM84GhCA0ZlhUlBjlhcqAIq2Ta4iw.&URL=http%3a%2f%2fwww.google.com%2fm%3fhl%3dmd%26gl%3dus%26source%3dandroid-browser-type%26q%3dewbay>. It is actually doing google search for ebay. Please find attached a snapshot for this page. Best regards Fangfei Date: Fri, 28 Sep 2012 14:49:55 -0400 From: Anthony Gutierrez <atgut...@umich.edu<mailto:atgut...@umich.edu>> To: gem5 users mailing list <gem5-users@gem5.org<mailto:gem5-users@gem5.org>> Subject: Re: [gem5-users] questions on running benchmark on gem5 Message-ID: <caeqebpsxtdzuouczrvt6q1-3ckj5mkwzgicnwdo9qbpwume...@mail.gmail.com<mailto:caeqebpsxtdzuouczrvt6q1-3ckj5mkwzgicnwdo9qbpwume...@mail.gmail.com>> Content-Type: text/plain; charset="iso-8859-1" I started a BBench run this morning with a fresh check out of the latest dev repo, the Android kernel and an unmodified Gingerbread disk image with BBench, both from gem5.org/bbench-gem5<http://gem5.org/bbench-gem5>. I do not see this problem, eBay renders fine and the simulation continues. I will need more info on your setup/how you're running the simulation if you'd like help with this problem. -Tony On Fri, Sep 28, 2012 at 11:31 AM, Anthony Gutierrez <atgut...@umich.edu<mailto:atgut...@umich.edu>>wrote: > I've never seen this problem before, and the BBench source for eBay does > not show anything called ewbay in the source, so I don't know how it's > getting redirected to that URL. Did you modify the BBench source in any > way? Can you send me your exact command line? And, can you tell me the > exact URL that the browser shows when it hangs? > > -Tony > > On Thu, Sep 27, 2012 at 10:25 AM, Fangfei Liu > <fangf...@princeton.edu<mailto:fangf...@princeton.edu>>wrote: > >> Hi,**** >> >> ** ** >> >> I tried to run Bbench on gem5 with Gingerbread disk image (atomic mode >> and default configuration). The simulator took about 140 minutes to boot >> Android and start running the benchmark. But it seems that the benchmark >> stalled at certain web page after running for a while. I find that it is >> supposed to render a web page in ebay but used the wrong url with ewbay. I >> terminated the simulation manually after 97 hours but find that the >> statistics shows only 87 billion instructions were simulated. I was >> wondering what caused the problem and what is supposed to be the end of >> running of Bbench? Will the simulation exit automatically? How many >> instructions are there if I want to run Bbench to completion? I also >> downloaded an EEMBC benchmark: AndEbench, which is an Android app (.apk >> file). Can I run this app directly on gem5? If so, does anyone know how to >> install and run it on gem5? Thanks!**** >> >> ** ** >> >> ** ** >> >> Best regards**** >> >> Fangfei**** >> >> ** ** >> >> _______________________________________________ >> gem5-users mailing list >> gem5-users@gem5.org<mailto:gem5-users@gem5.org> >> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users >> > > _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users _______________________________________________ gem5-users mailing list gem5-users@gem5.org<mailto:gem5-users@gem5.org> http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
_______________________________________________ gem5-users mailing list gem5-users@gem5.org http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
