In b4 the numbers are all set to ns, so they're closer to reality in that case. Either way you should understand what you're simulating and all the parameters and the assumptions based in the models. I don't believe one default parameter is any more or less correct than another one, if it matters to you, you should understand what the parameter does and change it to the appropriate value for your work. Like the warning on the DRAM model says, it hasn't been tested or validated since the 1.X M5 releases. If you want to spend the time testing it, validating it, and coming up with reasonably parameters, please do we will gladly accept the patches. Even better would be to have the many of the dram parameters automatically set based on the the other simulated parameters.

Ali


On Nov 12, 2007, at 6:09 PM, Sujay Phadke wrote:

HIi ALi,
As you replied, we have to set the DRAM parameters independently of the simulation setup. But I feel that these default DRAM parameters are wrong, and will give wrong results for people doing timing simulation. Correct me if I am wrong.

- Sujay

----- Original Message ----- From: "Ali Saidi" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>; "M5 users mailing list" <m5- [EMAIL PROTECTED]>
Sent: Monday, November 12, 2007 6:08 PM
Subject: Re: [m5-users] How to determine phyMem latency


No it means it's one tick of simulator time which is 1ps.
Ali

On Nov 12, 2007, at 5:40 PM, Rick Strong wrote:

Hi,

I noticed that the PhysicalMemory class sets memory latency to 1t which I assume is 1 tick from looking at the python. Does this imply that my memory latency is 1 tick of the system.membus clock @1GHZ?

class PhysicalMemory(MemObject):
 type = 'PhysicalMemory'
 port = VectorPort("the access port")
 range = Param.AddrRange(AddrRange('128MB'), "Device Address")
 file = Param.String('', "memory mapped file")
 latency = Param.Latency('1t', "latency of an access")
 zero = Param.Bool(False, "zero initialize memory")

-Richard
_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users


_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users


_______________________________________________
m5-users mailing list
m5-users@m5sim.org
http://m5sim.org/cgi-bin/mailman/listinfo/m5-users

Reply via email to