Date: | Sun, 16 Nov 2008 13:25:52 -0600 |
---|---|
From: | "Mike Marty" <mike.marty@xxxxxxxxx> |
Subject: | Re: [Gems-users] Couple of questions |
On Wed, Nov 12, 2008 at 7:23 PM, Arkaprava Basu <basu@xxxxxxxxxxx> wrote:
Hi All, see the -dir.sm files in $GEMS/slicc for where the naive memory access is performed. The memory latency is controller by a parameter in $GEMS/config/rubyconfig.defaults
The bandwidth can be constrained by setting the bandwidth of the appropriate link that connects the DirectoryMemory controller (specified in the -dir.sm file) I believe there has been work done on a more realistic memory controller implementation since I left the university. Thus the bandwidth might be more directly controlled.
You will need to add an invalidate type to CacheRequest. These are initialized in the per-processor Ruby sequencer. --Mike
|
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | Re: [Gems-users] gems suitability, Dan Gibson |
---|---|
Next by Date: | Re: [Gems-users] tracer-output-file, Mike Marty |
Previous by Thread: | [Gems-users] Couple of questions, Arkaprava Basu |
Next by Thread: | [Gems-users] Determining capacity and cold start misses using RUBY., Ashwath Narasimhan |
Indexes: | [Date] [Thread] |