[gem5-users] Segmentation Fault in PARSEC Benchmark

[Riddhi Thakker]

Hi,

I am running PARSEC benchmark on X86 system. It is running fine without
checkpoints. But if I am using checkpoints, then after restoring checkpoint
I am getting segmentation fault in PARSEC benchmark. I have used
configs/boot/hack_back_ckpt.rcS to create checkpoint (cputype=timing).
After I restored checkpoint (restore-cpu-type=O3), it ran PARSEC benchmark
using the following script:

#!/bin/sh

# File to run the blackscholes benchmark

echo "Running blackscholes"

cd /parsec/install/bin

/sbin/m5 dumpstats

/sbin/m5 resetstats

./blackscholes 64 /parsec/install/inputs/blackscholes/in_4K.txt
/parsec/install/inputs/blackscholes/prices.txt

echo "Done :D"

/sbin/m5 exit

/sbin/m5 exit

This gives following output:
==== m5 slave terminal: Terminal 0 ====
Loading new script...
usage: /sbin/m5 exit [delay]
/sbin/m5 resetstats [delay [period]]
/sbin/m5 dumpstats [delay [period]]
/sbin/m5 dumpresetstats [delay [period]]
/sbin/m5 readfile [filename]
/sbin/m5 execfile <filename>
/sbin/m5 checkpoint [delay [period]]
/sbin/m5 loadsymbol <address> <symbol>
/sbin/m5 initparam
/sbin/m5 sw99param
/sbin/m5 pin <cpu> <program> [args ...]

All times in nanoseconds!
PARSEC Benchmark Suite Version 2.1
[HOOKS] PARSEC Hooks Version 1.2
Num of Options: 4096
Num of Runs: 100
Size of data: 163840
[HOOKS] Entering ROI
blackscholes[796]: segfault at fffffffffc973680 rip 00000000004048bc rsp
00002ae564f3bbf8 error 4
Segmentation fault
Done :D

Cmdline to restore: ./build/X86/gem5.opt configs/example/fs.py
--mem-type=DDR4_2400_8x8
--kernel=./x86parsec/binaries/x86_64-vmlinux-2.6.22.9.smp
--disk-image=../x86root-parsec.img --checkpoint-restore=1
--restore-with-cpu=DerivO3CPU
--script=./scripts/blackscholes_64c_simsmall.rcS

The disk image from
http://www.cs.utexas.edu/~parsec_m5/x86root-parsec.img.bz2. I tried to
search this error on the gem5 mail archive, but I couldn't find any
solution from there.

It would be great if anyone can help me sort out this problem. Thanks in
advance.

Regards,
Riddhi