Hello everyone,
I'm trying to execute the parsec benchmarks (version 2.1 in the pre-compiled
disk image) on an ALPHA MESI CMP directory architecture. My parameters for 64
cores are (everything else is default):
./build/ALPHA_MESI_CMP_directory/gem5.opt --remote-gdb-port=0 --outdir=$1_dir
--dump-config=$1_config --redirect-stdout --redirect-stderr
--stdout-file=$1_out --stderr-file=$1_err --stats-file=$1_stats
configs/example/fs.py --script=./runscripts/$1
--disk-image=linux-parsec-2-1-m5-with-test-inputs.img
--kernel=vmlinux_2.6.27-gcc_4.3.4 --cpu-type=atomic --caches --l2cache
--clock=2GHz --num-dirs=16 --num-l2caches=64 --num-l3caches=16 --l2_size=2MB
--l3_size=16MB --l1d_assoc=2 --l1i_assoc=2 --l2_assoc=8 --l3_assoc=16
--cacheline_size=64 --ruby --num-cpus=64
and the error:
em5.opt: build/ALPHA_MESI_CMP_directory/cpu/simple/atomic.cc:313: Fault
AtomicSimpleCPU::readMem(Addr, uint8_t*, unsigned int, unsigned int): Assertion
`!pkt.isError()' failed.
Program aborted at cycle 469873597000
If I run the same parameters except 128 CPUs and 128 L2 caches:
fatal: Unable to find destination for addr 0x80200000468 on bus system.iobus
@ cycle 196665500
[findPort:build/ALPHA_MESI_CMP_directory/mem/bus.cc, line 374]
Are there some restrictions to how many caches, cpus and such the system can
have?
By the way, is running a 512 or even 1024 core simulation in the above
configuration feasible? I briefly tried but the memory demands were too high.
Thank you!
George M
_______________________________________________
gem5-users mailing list
gem5-users@gem5.org
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
