Writebacks consume memory bus bandwidth, but that's only a function of block size and available bus bandwidth, unrelated to memory write latency. Basically the memory latency only determines the delay between when a request is received and when the corresponding response is generated, and since writebacks don't generate responses (in the current protocol), the write latency doesn't have an effect.
Regarding your other email, we have Ruby FS working, but the code may not be committed yet. It should be coming soon though. Brad Beckmann has done most of the work on this. Steve On Mon, Jul 5, 2010 at 3:08 PM, sheng qiu <[email protected]> wrote: > another thing is does that mean in the current physicalmemory model, reads > won't get delayed because of a longer latency of writeback request to > memory, and memory bus bandwidth won't tied up because of longer latency > writes? So writeback won't consume memory bus bandwidth and only read > request consume memory bus bandwidth? > > Thanks, > Sheng > > > > _______________________________________________ > 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
