I don't know the details, but a key thing to realize is that originally
SimObject developers had to write their own C++ Params structs (such as what
you see in cpu/o3/params.hh), while more recently we've finished our
long-held goal of autogenerating the C++ structs from the Param.Foo()
attributes in the Python object classes (which is what you find in
build/*/params).

I'm guessing the O3 model is stuck in the old world, or perhaps is in some
intermediate limbo state, due to the complexity of converting all the
template-handling cruft into the new model, and you're getting the structs
from the two worlds confused.  Probably Kevin or Nate would have to provide
more details on just what the current status is, if that's not enough of a
clue to get you going.

If you continue to have problems, providing specific compiler error messages
would also be helpful.

Steve

On Wed, Jul 30, 2008 at 11:35 PM, Rick Strong <[EMAIL PROTECTED]> wrote:

> Hi,
>
> I am adding support to dynamically set structure parameters sizes in
> wattch and want to pass the struct DerivO3CPUParams to the wattch
> initialization section. The easiest way to do this for every cpu  to
> initialize its wattch structure during initialization of the cpu class.
>
> I decided to put the initialization call in
> src/cpu/cpu.cc::FullO3CPU(...) constructor  and pass params
> (Impl::Params) to the wattch initialization function. I then included
> params/DerivO3CPU.hh in the wattch file, but the compiler does not like
> this. I also tried to do a dynamic_cast<DerivO3CPUParams>(params) where
> params is Impl::Params of src/cpu/cpu.hh::FullO3CPU class. I have
> resigned that I have no idea how the params system is working ... lol.
>
> So my question is, what file do I have to include in wattch so I can
> access the Impl::params parameter of FullO3CPU after I pass it to the
> function. Also, do I need to do a dynamic_cast? If you could point me to
> an example, that would be equally helpful.
>
> -Rick
>
> _______________________________________________
> m5-users mailing list
> [email protected]
> http://m5sim.org/cgi-bin/mailman/listinfo/m5-users
>
_______________________________________________
m5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to