Panic on --fast-forward flag

[Arth Patel]

Hi,

I am trying to run a fast-forward simulation, and the simulation panics with following error.

Switch at instruction count:2000000000

info: Entering event queue @ 0.  Starting simulation...

panic: RubyPort::MemSlavePort::recvAtomic() not implemented!

 @ tick 0

[recvAtomic:build/X86_VI_hammer_GPU/mem/ruby/system/RubyPort.hh, line 91]

Memory Usage: 2401440 KBytes

Program aborted at tick 0

I don't know what this mean (I'm not very familiar with Ruby/gem5).

Is there a way I can fix this?

When --fast-forward flag is not supplied, the simulation works just fine.

Thank you.. :)

--

Arth (PioneerAxon)

[Jason Power]

Hi Arth,

You can't use the atomic CPU with Ruby.

Although I've never tried it, you may be able to fast forward with the classic memory system, take a checkpoint, then restore the checkpoint into a system with Ruby. There's no guarantee this would work, but it's something to try.

Jason

[15053...@mail.nwpu.edu.cn]

Hi Jason,

I run 4 spec cpu2006 benchmarks and a Rodinia benchmark  with --fast-forward 1000000000,-I 1000000000 and input data size is 16384,but the value of  CPU's instpercycle  is non in  stats.txt

nohup build/X86_VI_hammer_GPU/gem5.opt -d ./m5out/444-456-483-403-bp-open ../gem5-gpu/configs/se_fusion_v2.py --mem-size=4GB --cpu-type=detailed -n 5 -b 444-456-483-403-bp -F 1000000000 -I 1000000000 &

How to do ?

Thanks!

Xiaofeng Li

在 2015年3月19日星期四 UTC+8下午11:54:49，Jason Lowe-Power写道：

[Jason Lowe-Power]

When you run in fast-forward mode, gem5 uses the Atomic CPU, which does not count many statistics. The atomic CPU does not do timing accesses, but tries to complete every action "atomically". Your application may be ending before the 1000000000 instructions.

Jason

--

Jason