On Mon, 18 May 2015, Brad Beckmann wrote:
On May 12, 2015, 3:48 a.m., Nilay Vaish wrote:
This patch is incorrect on a couple of counts. We are not in 32-bit world.
Secondly, you should expose the getAddress() function as we expose functions
related to other classes.
Brad Beckmann wrote:
This comment is dissappointing and unprofessional. I'm well aware of
address sizes supported in modern systems. Also I would like to point out this
patch adds back in a function verbatim that was previously removed. I'm not
introducing something new here.
As the description specifies, this funciton is used to convert an address
offset to an integer, not a 64-bit address. An integer is needed because that
is required to create a NodeID. Also there are plenty of uses of 32-bit
integers in this file and across gem5. I do not think anyone has expectations
to instantiate more than 2 B ruby controllers.
Finally what is the benefit of exposing the getAddress function and doing
the casting within SLICC? It is not how these sorts of conversions have been
handled in the past. There doesn't seem to be any real benefit other than
adding more work for us.
Would it be sufficient if we added an assert to this function that ensures the
top 32 bits in the address are masked off?
OK.
--
Nilay
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
