On Wed, 27 May 2015, Joel Hestness wrote:
On May 27, 2015, 12:22 a.m., Brad Beckmann wrote:
I just have two questions so far on this patch.
1. What does Nilay mean that the "we are dropping the functionality of connecting
buffers in python"? It wasn't too long ago when we added that functionality rather
than doing it in the controller constructors. Why is that functionality going away and
how was that decision determined?
2. Joel was it your intention to post this patch without moving the
MessageBuffers created within the network (not the endpoints) to python? Was
that to get early feedback before doing the possibly harder job of dealing with
those other MessageBuffers? Were you hoping that Mark Wilkening to pick this
patch up and add that support?
Thanks
1. I just made a quick decision to remove the buffer connection code in
pyobject.cc. It wasn't clear to me that it was actually doing anything, because
it just called into C++ code, which is generated to connect the buffers. I see
Nilay's point that the connections are printed in the config.ini. It would be
easy to add this back (though I still feel it is convoluted to connect a buffer
to a port). I'll be looking for ways to address this.
2. Yes, I created this patch quickly in an effort to get feedback and to
hopefully start us toward handling buffers in Python code. I have not yet
looked at what it would take to move the buffers from the Switch and
SimpleNetwork into Python code. I still prefer that we not introduce buffer
sizing inside cache controllers, and I'm hoping we can align on a
Python-configurable way to do it.
I'm willing to invest time in Python-configurability of MessageBuffers, because
it will very likely save me time later. Please let me know if you like the
direction of this patch (exposing MessageBuffers in Python), or if you think we
should do that another way.
Just realized that MessageBuffer.py file is missing from the posted patch.
--
Nilay
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev
