[Qemu-devel] QEMU: DBT vs. Interpretation

[Qemu-devel] QEMU: DBT vs. Interpretation

[Javier Picorel]

Hi,

We are trying to make QEMU deterministic for 
architectural simulation. In the absence of I/O, 
let’s say only executing user code or exceptions, 
is there any source of indeterminism (e.g., non 
deterministic compiler optimizations, TB indeterminism) 
of QEMU’s DBT versus a canonical interpreter? Thanks!

Regards,
Javier

Re: [Qemu-devel] QEMU: DBT vs. Interpretation

[Paolo Bonzini]

On 26/01/2015 21:54, Javier Picorel wrote:
> Hi,
> 
> We are trying to make QEMU deterministic for 
> architectural simulation. In the absence of I/O, 
> let’s say only executing user code or exceptions, 
> is there any source of indeterminism (e.g., non 
> deterministic compiler optimizations, TB indeterminism) 
> of QEMU’s DBT versus a canonical interpreter? Thanks!

No, there isn't.

Paolo

Re: [Qemu-devel] QEMU: DBT vs. Interpretation

[Laurent Desnogues]

On Mon, Jan 26, 2015 at 9:54 PM, Javier Picorel <javier.picorel@epfl.ch> wrote:
> Hi,
>
> We are trying to make QEMU deterministic for
> architectural simulation. In the absence of I/O,
> let’s say only executing user code or exceptions,
> is there any source of indeterminism (e.g., non
> deterministic compiler optimizations, TB indeterminism)
> of QEMU’s DBT versus a canonical interpreter? Thanks!

I have only hit two sources of variations:

- time;  for instance SPEC 2000 173.gcc dumps time at
  the end, so you'll get slightly different results each run

- environment variables which are parsed by the runtime
  at process start;  to get around that issue you'll have
  to reintroduce -ignore-environment which was removed
  some time ago (see commit
  fc9c54124d134dbd76338a92a91804dab2df8166)

A "canonical interpreter" would have the same issues.

HTH,

Laurent

Re: [Qemu-devel] QEMU: DBT vs. Interpretation

[Peter Maydell]

On 28 January 2015 at 15:10, Laurent Desnogues
<laurent.desnogues@gmail.com> wrote:
> - environment variables which are parsed by the runtime
>   at process start;  to get around that issue you'll have
>   to reintroduce -ignore-environment which was removed
>   some time ago (see commit
>   fc9c54124d134dbd76338a92a91804dab2df8166)

Wouldn't just running QEMU under 'env -i' have the same
effect?

thanks
-- PMM

Re: [Qemu-devel] QEMU: DBT vs. Interpretation

[Laurent Desnogues]

On Wed, Jan 28, 2015 at 4:39 PM, Peter Maydell <peter.maydell@linaro.org> wrote:
> On 28 January 2015 at 15:10, Laurent Desnogues
> <laurent.desnogues@gmail.com> wrote:
>> - environment variables which are parsed by the runtime
>>   at process start;  to get around that issue you'll have
>>   to reintroduce -ignore-environment which was removed
>>   some time ago (see commit
>>   fc9c54124d134dbd76338a92a91804dab2df8166)
>
> Wouldn't just running QEMU under 'env -i' have the same
> effect?

I didn't know about env -i and it indeed works fine!

Thanks,

Laurent