On Apr 23, 2011, at 1:01 PM, Lisa Hsu wrote: > >> On 2011-04-22 18:51:32, Nathan Binkert wrote: > >> On 2011-04-22 18:51:32, Nathan Binkert wrote: >>> src/mem/cache/cache_impl.hh, line 1087 >>> <http://reviews.m5sim.org/r/654/diff/1/?file=11762#file11762line1087> >>> >>> I think we should force context IDs to be set even by devices. >> >> Lisa Hsu wrote: >> not it. >> >> Nathan Binkert wrote: >> Is is that hard? > > I don't know. Is it? I haven't opened a file in dev/ in years. Does this > involve changing all device models? Changing something in a port interface? > Which? I don't know. I have not looked into it at all. I tried to page out > as little of what I was doing to fix this vector length matching thing, and I > don't want to delay getting back to other crap that I have to do. I consider > this a difference between "should" and "must". I consider the vector thing > a "must" - I introduced the code that broke it and it currently causes > problems. This is definitely a "should", IMHO. > > I've got to get to Crossfit now, I'm going to be late! It should be a one line change to DmaPort::dmaAction() in io_device.cc.
Ali _______________________________________________ m5-dev mailing list m5-dev@m5sim.org http://m5sim.org/mailman/listinfo/m5-dev
