Re: [Gems-users] Warm-up caches problem

[Philip Garcia <pcgarcia@xxxxxxxx>]

i've never really done this myself, but it sounds like you should only remove the timing_model: ruby0 from the file.  Removing all of phys_mem will really screw up the system, as it needs the IO and memory mapping of the machine to operate.

Phil

On Mar 13, 2009, at 11:46 PM, Edward Lee wrote:

I am still stuck at this point.

After step2, if I do not remove the  ruby0 objects I get this error.

/***************************************************************************\
 > Physical Memory object cannot be found. If you are NOT compiling Ruby and <
 > you see this message, something is wrong.                                 <
 > This message is part of the normal compilation process.                   <
 \***************************************************************************/

I know this is expected thats why I am removing the following objects, after grepping ruby in the warm.conf file:

OBJECT phys_mem TYPE memory-space {
        queue: cpu0
        map: ((0, memory, 0, 0, 0x40000000, NIL, 0, 0, 0),
              (0x800000000, memory, 0, 0x40000000, 0x40000000, NIL, 0, 0, 0),
              (0x7fff07ffff0, simicsfs, 0, 0, 16, NIL, 0, 0, 0))
        timing_model: ruby0
}

...
...
...

OBJECT ruby0 TYPE ruby {

}

And when I try to read the configuration after removing these lines I am getting the following error:

Segmentation fault (SIGSEGV) in main thread
The simulation state has been corrupted. Simulation cannot continue.
Please restart Simics.

So, am I doing anything wrong here?

Also, an interesting stuff with the magic breaks. Maybe I shouldn't have enabled them at all. When I only run SIMICS at step1, magic breaks are recognized fine and I get no warnings but I when I run the benchmark with RUBY I am getting the following warnings. Can these be related?

Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:476: val is 1
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:476: val is 1
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:477: SIMICS_get_program_counter(proc_num) is [0x12fe8, line 0x12fc0]
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:477: SIMICS_get_program_counter(proc_num) is [0x12fe8, line 0x12fc0]
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:478: Unexpected magic call
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:478: Unexpected magic call
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:476: val is 2
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:476: val is 2
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:477: SIMICS_get_program_counter(proc_num) is [0x1308c, line 0x13080]
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:477: SIMICS_get_program_counter(proc_num) is [0x1308c, line 0x13080]
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:478: Unexpected magic call
Warning: in fn void magic_instruction_callback(void*, void*, integer_t) in simics/commands.C:478: Unexpected magic call

I really appreciate any input.

Thanks in advance,

Ed

On Fri, Mar 13, 2009 at 12:01 AM, Edward Lee <edwl202@xxxxxxxxx> wrote:

Hi,

I am a newbie to GEMS and I am trying to run some benchmarks using SIMICS+RUBY. I followed the guidelines in the ISCA tutorial slides and I also checked the earlier posts.
Here is what I am doing:

Step1:
Only SIMICS
./simics -stall
configure simics (itsc/dstc-disable ... etc.)
Run FFT with small input
write-configuration cold.conf
exit simics

Step2:
SIMICS+RUBY
./simics -stall
read-configuration cold.conf
configure simics
setup ruby
Run FFT with small input
ruby0.save-caches fft.caches_8p.gz
write-configuration warm.conf
exit simics

Step3:
./simics -stall
Edit warm.conf to delete ruby object and ruby timing model memory space object, save the file
read-configuration warm.conf

And at this stage I am getting the following error:

Segmentation fault (SIGSEGV) in main thread
The simulation state has been corrupted. Simulation cannot continue.
Please restart Simics.

Any ideas on what I am doing wrong?

Regards,

Ed

_______________________________________________
Gems-users mailing list
Gems-users@xxxxxxxxxxx
https://lists.cs.wisc.edu/mailman/listinfo/gems-users
Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/" to your search.