Seyed,
It's not unreasonable; the other cores were just halted, and there were
still an outstanding number of instructions remaining. So, the remaining
instructions were given to the unhalted vCPU. If you look at the code, you
will see:
1600 if (others_halted && insns_remaining > 100) {
1601 /* This happens because all other CPUs are either halted
1602 * or stopped and if we still have some instructions
1603 * remaining then we allocate more instructions to this core
1604 * and let it run in emulation mode untill our instruciton
1605 * count reaches to near zero. */
1606 adjust_fwd_insts(ctx);
...
...
fast-forwarding is done using qemu. The number of iterations, simulation,
logging, etc. do not get counted or occur during qemu operation.
Therefore, the second two options will be considered wrt the instance in
time after fast-forwarding has occured.
Tyler
> Hi all,
>
>
> I'm trying to use "-fast-fwd-insns" option in my simulations.
> I'm getting the following warning every time I run simulations.
>
>
> All CPU context will be fast-forwared to 5000000 instructions.
> WARNING: Early switching to simulation mode. Instrucitons remaining in
> each
> CPU context to fast-forward are:
> CPU 0: 4920101
> CPU 1: 3197246
> CPU 2: 4578852
> CPU 3: 12696199
>
> The remaining instructions for CPU 3 is not reasonable at all. I was also
> wondering why fast-forwarding couldn't be completed in any of my
> simulations.
>
> For the record, this is my simconfig:
> -machine shared_l2
> -corefreq 2G
> -fast-fwd-insns 20M
> -stats %(out_dir)s/%(bench)s.yml
> -dramsim-device-ini-file ini/sg125.ini
> -dramsim-results-dir-name %(out_dir)s_%(bench)s
> -startlog 1M
> -loglevel 1
> -stopinsns 20M
>
> I have another question. What does happen if we have "-stopinsns X",
> "-fast-fwd-insns Y", and "-startlog Z". Does it mean that after Y+Z
> instructions MARSS starts collecting stats and after Y+Z instructions it
> terminates (rather than starting stat collection after min(Y,Z) and
> terminating after max(Y,X))?
>
> Thank you in advance for your time.
>
> Best,
> Seyed
>
>
> _______________________________________________
> http://www.marss86.org
> Marss86-Devel mailing list
> [email protected]
> https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
>
_______________________________________________
http://www.marss86.org
Marss86-Devel mailing list
[email protected]
https://www.cs.binghamton.edu/mailman/listinfo/marss86-devel
