On Thu, Jan 5, 2012 at 1:50 PM, nathan binkert <[email protected]> wrote:

> > So unless someone can explain why we really need it, I propose just>
> deleting the " & ~0x03" from both of the places it appears in doService()>
> and moving on.
>
> PAL addresses as listed in the palcode file (i.e. the output of nm)
> don't have the final bit set, so when you set a breakpoint in the
> remote debugger, the breakpoint that is set is on the address without
> the final bit set, so if we did this, we'd break the ability to break
> and step through palcode in the remote debugger.  Probably nothing
> that anybody other than me has ever done and I haven't done it in >5
> years, but it would be functionality that we'd break.
>
> I'm not sure if Ali's bottleneck stuff traced through palcode or if it
> used PCEvents.
>
> Anyway, there's a counter example.
>

Yea, that's the counter example I already brought up ;-).  Thanks for
confirming it.  You might be able to work around it by manually adding 1 to
the symbol value on the command line, or if you wanted to get really fancy
you could add the 1 in the Alpha remotegdb support if it looked like a PAL
address.  If that's the only case though, I'm inclined to just drop the
mask.  We could also put a warning in the Alpha remotegdb code in case
anyone tries to do this in the future.

Steve
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to