Hi Xiaojun, sorry for the slow reply. You're correct that the segfault
is because activeThreads isn't being checked to be empty, and is getting
a garbage value back when accessing the front of the list. The
underlying problem is that the activeThreads object should not be made
empty on the same cycle that the CPU is still running, which
unfortunately happens in a few cases in our 2.0 beta 1 release. What
command line are you running to get this error? I've fixed problems
related to activeThreads being deallocated improperly in the code that
will be released shortly, but I want to make sure your exact case has
been fixed.
Kevin
xiaojun.chen wrote:
Hi,m5-users,
When I use DetailedO3CPU on FS mode, there is always a segmentation fault.
And gdb infomation as bellow:
Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread -1208240448 (LWP 900)]
0x082b85bb in DefaultFetch<AlphaSimpleImpl>::getFetchingThread (this=0x9da9564,
[EMAIL PROTECTED]) at build/ALPHA_FS/cpu/o3/fetch_impl.hh:1334
1334 if (fetchStatus[tid] == Running ||
(gdb) p tid
$1 = 165293664
(gdb) p (*activeThreads).size()
$6 = 0
(gdb) bt
#0 0x082b85bb in DefaultFetch<AlphaSimpleImpl>::getFetchingThread
(this=0x9da9564,
[EMAIL PROTECTED]) at build/ALPHA_FS/cpu/o3/fetch_impl.hh:1334
#1 0x082b8614 in DefaultFetch<AlphaSimpleImpl>::fetch (this=0x9da9564, [EMAIL
PROTECTED])
at build/ALPHA_FS/cpu/o3/fetch_impl.hh:1007
#2 0x082ba28d in DefaultFetch<AlphaSimpleImpl>::tick (this=0x9da9564)
at build/ALPHA_FS/cpu/o3/fetch_impl.hh:834
#3 0x082c95c2 in FullO3CPU<AlphaSimpleImpl>::tick (this=0x9da9160)
at build/ALPHA_FS/cpu/o3/cpu.cc:414
build/ALPHA_FS/cpu/o3/fetch_impl.hh:
int tid = *((*activeThreads).begin());
if (fetchStatus[tid] == Running ||
fetchStatus[tid] == IcacheAccessComplete ||
fetchStatus[tid] == Idle) {
return tid;
} else {
return -1;
}
It seems that there didn't check whether the activeThreads is empty or not. Is
that the exact reason for this error?
Thanks and BestRegards!
xiaojun.chen
[EMAIL PROTECTED]
2006-11-22
------------------------------------------------------------------------
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
