On Jun 2, 2008, at 10:47 PM, nathaniel jones wrote:
i've been trying run the spec benchmarks in the default timing
simple cpu. i tried running mgrid and what i get so far after 3
hours is
It appears as though you're using an atomic cpu, not a timing one.
command line: ./build/ALPHA_SE/m5.opt tests/run.py long/92.mgrid/
alpha/tru-64/simple-atomic
Global frequency set at 1000000000000 ticks per second
0: system.remote_gdb.listener: listening for remote gdb #0 on port
7000
warn: Entering event queue @ 0. Starting simulation...
warn: Increasing stack size by one page.
warn: ignoring syscall sigprocmask(1, 1073070158, ...)
warn: ignoring syscall sigprocmask(3, 0, ...)
warn: ignoring syscall sigaction(11, 4831387520, ...)
warn: ignoring syscall sigaction(5, 4831387520, ...)
warn: ignoring syscall sigaction(8, 4831387520, ...)
warn: ignoring syscall sigaction(6, 4831387520, ...)
warn: ignoring syscall sigaction(15, 4831387520, ...)
warn: ignoring syscall sigaction(3, 4831387520, ...)
warn: ignoring syscall sigaction(2, 4831387520, ...)
warn: setsysinfo: ignoring ieee_set_fp_control() arg 0x11ff91f98
warn: ignoring syscall setrlimit(3, 4831387592, ...)
warn: Increasing stack size by one page.
does increasing the stack size really take that long? or is this
indicative of an error?
It doesn't take long at all, the stack size was increased and the
execution of the program was immediately continued. Is completely
expected that the stack with grow, the warning is simply there to let
you know that another page of memory was consumed by the stack.
also, should i be concerned about the syscalls that are ignored in
this and the other benchmarks?
Generally no, if we set them to be ignored (as opposed to causing a
panic), we're pretty certain that they won't alter the behavior of
most programs. The only time they would is if the program needed to
take something like FP trap to be correct (while it's possible to
write such a program, it's not clear why one would exist) or a program
that needs to send signals to itself.
The runtime of the spec benchmarks is dependent on the input set that
you're using. If you're using the standard input for spec benchmarks
starting from the beginning and running to completion you should
expect running the benchmarks in M5 to takes days if not weeks.
Ali
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
