Hi Giacomo. Some of the register accessor functions (maybe all of them?)
used to return Fault objects, but I changed them to be like they are now a
long time ago. I don't remember exactly what the reason was, just that I
had one and made the switch. I can try to find some info about why. Please
wait until I've had a chance to remind myself so I can comment
intelligently about whether it makes sense to switch back.
Gabe
On Thu, Jun 7, 2018 at 5:55 AM Giacomo Travaglini <
giacomo.travagl...@arm.com> wrote:
> Hi all,
>
>
>
> It's been a while that I am lingering about the idea of proposing a
> signature change for the ISAs MiscReg setter:
>
>
>
> What it currently is:
>
>
>
> * void setMiscReg(int misc_reg, const MiscReg &val);
>
>
>
> What I would like to have:
>
>
>
> * Fault setMiscReg(int misc_reg, const MiscReg &val);
>
>
>
> IMO this will improve system register access design, and will make it
> easier to cover cases where the access to a System register produces a a
> fault as a side effect.
>
> This could be for instance a register access that is trapped or the
> setting of a particular Bit in a system register which triggers an
> exception.
>
>
>
> At the moment the current signature imposes us to have multiple table
> walks for a specific
>
> MiscReg number: when writing a system register I have to do the following
> calls (this is ISA agnostic pseudocode)
>
>
>
> 1) checkIfTraps(miscreg_number);
>
>
>
> Which will be implemented more or less in this way:
>
>
>
> Switch(miscreg_number) {
>
> Case MISCREG_1:
>
> // return boolean or fault if trapping
>
> Case MISCREG_2:
>
> […]
>
> }
>
>
> 2) setMiscReg(miscreg_number, val);
>
>
>
> Which will be implemented more or less in this way:
>
>
>
> Switch(miscreg_number) {
>
> Case MISCREG_1:
>
> // Here I do the register masking for writing
>
> Case MISCREG_2:
>
> […]
>
> }
>
>
> This could be inefficient because you have to traverse switch cases twice.
> This is even more
>
> True if the exception is generated by the write of a specific value in a
> bitfield: other problems arise
>
> If you have to generate the fault but still perform a register write.
>
>
> Switching to
>
>
>
> Fault setMiscReg(int misc_reg, const MiscReg &val);
>
>
>
> Will be more callback oriented and solve the issue:
>
>
>
> Do you have any thoughts about this? I'd like to have your opinion before
>
> I start working on this big change.
>
>
> Thanks
>
>
> Giacomo Travaglini
>
> IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
> _______________________________________________
> gem5-dev mailing list
> gem5-dev@gem5.org
> http://m5sim.org/mailman/listinfo/gem5-dev
_______________________________________________
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev
