Re: [gem5-users] Stats in garnet 2.0

To get utilization of individual links, go to GarnetNetwork.cc … you’ll see that it gets the link utilization of each link and aggregates it. You can print each of them out instead. > On Jul 29, 2017, at 2:42 PM, SHARJEEL KHILJI > wrote: > > Hi, > I am running ARM

Re: [gem5-users] injection rate

For avg rate just divide total packets injected by total cycles (sim_ticks) Different values are per virtual network. Their sum will add up to the total. > On Aug 12, 2017, at 3:22 AM, matild breo wrote: > > Hi Tushar > > > yes I saw the packet and flit injection like bellow in my stas.txt:

Re: [gem5-users] Regarding garnet2.0

Take a look at http://www.gem5.org/Garnet2.0#Configuration control_msg_size, data_msg_size, and ni_flit_size are the parameters you an play with for what you want to do. Cheers Tushar > On Aug 17, 2017, at 1:22 PM, evania.t...@gmail.com wrote: >

Re: [gem5-users] Regarding garnet2.0

sg_size differently when > generating packet each time? > And is ni_flit_size equal to rubysystem::getBlockSizeBytes()? > Thanks for your kind help. > Best, > Evania > > 2017-08-18 3:36 GMT+08:00 Tushar Krishna <mailto:tus...@ece.gatech.edu>>: > Take a look at > http://

Re: [gem5-users] vcs per vnet and ni_flit_size and link width in garnet2

You can find the answers here: http://www.gem5.org/Garnet2.0 > On Aug 19, 2017, at 1:49 AM, matild breo wrote: > > Hi every body > I have some questions: > > 1- what is default link width in garnet 2 in bytes? > > 2- if the ni_flit_size is 16 byte, does it mean

Re: [gem5-users] vcs per vnet and ni_flit_size and link width in garnet2

Depending on the coherence protocol, there are certain number of “virtual networks" or “message classes”, such as request, response, forward etc. This to carry coherence messages. Within each vnet, the NoC can have one or more VCs. So together, num_vnets x vcs_per_vnet is the total number of VCs *

Re: [gem5-users] Regarding garnet2.0

not? Can I set mesh_row with 9 cpus > and 16 dirs? Because I encountered the errors as assert(node.type == > 'DMA_Controller') > assert(num_columns * num_rows == num_routers) > > Or can I comment out these assertion? > Thanks in advance. > Best, > Evania > >

Re: [gem5-users] Garnet 2.0 + How the dir_nodes are connected in Mesh_XY network

Hi Faisal, > On Aug 31, 2017, at 2:17 AM, F. A. Faisal wrote: > > Dear All, > > 1. I cross-checked the code for Mesh_XY.py and MeshDirCorners_XY.py file. > Where, in the MeshDirCorners_XY.py has the proper implementation of the > connected dir_nodes. > However, I couldn't find any code that c

Re: [gem5-users] DMA controller

You can remove that assertion if you are sure what controllers are getting connected. The topology file makes an assumption about the number of L2s being equal to the number of CPUs. > On Sep 7, 2017, at 9:17 AM, Hagai David wrote: > > Hi, > > What is the best way to add DMA controllers to

Re: [gem5-users] Garnet 2.0 + How the dir_nodes are connected in Mesh_XY network

ort) 0 -> 2 (inport) Router_1 > ROTUER_0 --> (outport) 0 -> 3 (inport) Router_1 > > Thanks again. > > Best Regards, > > F. A. Faisal > > > On Thu, Aug 31, 2017 at 3:44 PM, Tushar Krishna <mailto:tus...@ece.gatech.edu>> wrote: > Hi Faisal, >

Re: [gem5-users] Fault Injection in garnet Network on chip Router

The codebase has a fault model based on http://projects.csail.mit.edu/wiki/pub/LSPgroup/PublicationList/FaultModel_DAC2011.pdf I believe it generates probabilities of failures for a given NoC configuration

Re: [gem5-users] Question about Garnet network

Hi Xi, Hmm one way to make the entire network (or parts of it) stall for a certain time could be to use the event queue: every time a wakeup is called for any network event, you could re-insert that back into the event queue with wakeup time such that it wakes up after the voltage transition is

Re: [gem5-users] Decouple Garnet

Hi Daniel, Firstly regarding the mapping of input port numbers to port type: the network models in Ruby can model any topology, not just a mesh. Even in a mesh, some routers (edges/corners) have fewer ports, while others have more etc. Hence the mapping between input port id and direction is arbi

Re: [gem5-users] Decouple Garnet

Hi Daniel, As long as you make sure a flit that comes in sits in a correct VC and gets to use the switch, it should all work fine. Both units should access all flits at all input ports, there are no other dependencies I can immediately think of. - Tushar On Nov 27, 2012, at 2:08 PM, Bueso, Da

Re: [gem5-users] Default network test

Hi Vineeth, Yeah it should be ruby_random_test.py instead of run_random_test.py .. I will edit that on the wiki. But you seem to indicate that "it does not work" with ruby_random_test.py? What do you mean by that? Did it give you some error? - Tushar On Dec 23, 2012, at 12:29 AM, Vineeth Mohan

Re: [gem5-users] Default network test

n. > > Best, > Vineeth > > > On Sun, Dec 23, 2012 at 1:34 PM, Tushar Krishna wrote: > Hi Vineeth, > Yeah it should be ruby_random_test.py instead of run_random_test.py .. I > will edit that on the wiki. > But you seem to indicate that "it does not work&quo

Re: [gem5-users] Concentrated 2D Mesh

Hi Ni, You can create your own topology file for the concentrated mesh. Currently each cache controller connects to one router. You can instead connect 4 controllers to each router. Take a look at Mesh.py and MeshDirCorner.py for reference. - Tushar Rx Ni wrote: >Hi, all, > >I am running

Re: [gem5-users] how to run interconnect network in gems

Make sure you have the correct python and C++ libraries. Look at the website to see which versions are compatible. - Tushar On Jan 28, 2013, at 1:22 PM, abbas dehghani wrote: > hello > I downloaded the gem5.I used the command line > scons build/ALPHA_Network_test/gem5. debug > > It is built gem

Re: [gem5-users] how to run interconnect network in gems

Hi Dehghani, I just pulled out the latest repo and ran the same simulation and it worked fine. The error itself is due to older python/c++ libraries. Note sure why it did not show up without garnet. Can you double check if all your libraries match the version numbers listed here: http://gem5.org/

Re: [gem5-users] The problem with running interconnct network simulation

Hi Dehghani, Sorry I am super swamped with work so can't help out too much on this. Anyone else seen this error earlier and know what the problem is? A few quick suggestions from the top of my head: - check if you have python-dev installed - am surprised why it would run without garnet, but break

Re: [gem5-users] Can anyone explain the items in ruby.stats?

Hi Lu, The routers on the edges will have fewer ports (links). I think Router 0 gets connected to DMA and hence has more links. You can take a look at src/mem/ruby/network/topologies/Mesh.py to see how the topology gets constructed. 550 4400 […] means that 550 messages were sent, with size 550*8

Re: [gem5-users] Can anyone explain the items in ruby.stats?

The number outside is the count of the number of messages of type Request_Control on that link, which is 550. The number of bytes transferred is 4400 (since each Request_Control message is 8 byte long). Note sure why there are 10 numbers inside. It is some kind of histogram. You can take a look

Re: [gem5-users] What does this warning mean?

You can ignore the warnings. -Tushar On Mar 7, 2013, at 5:44 PM, Lu Hang wrote: > Dear all, I use ruby with garnet network to run parsec benchmark. The > simulation goes smoothly without any error, except that some warnings are > printed on the screen, as the last few lines below: "warning:

Re: [gem5-users] how to declare a MessageBuffer transporting MemoryMsg

The vnet_type field is used by the garnet on-chip network to decide whether the message only carries control (= request) information or data (i.e. the cache line) (= response), and size the number of buffers required within each VC appropriately (e.g. vnets carrying control have 1-flit buffers,

Re: [gem5-users] About Network Interfaces Garnet flexible pipeline in Ruby

Each coherence controller (L1, L2, Directory and DMA) creates an NI. - Tushar On Mar 19, 2013, at 2:54 PM, Lu Hang wrote: > Hi all, I have a question in terms of the network interfaces (NI) implemented > in Garnet flexible pipeline in ruby. I used 16 cores for simulation, so the > routers and

Re: [gem5-users] SE mode: 2DMesh topology

The simulation itself worked fine. The error happened after simulation when Orion is trying to compute network power using activity stats. The assertion seems to indicate that a router with zero input ports was created, which is weird. I know the Mesh topology works if you use equal number of CPU

Re: [gem5-users] gem5-users Digest, Vol 81, Issue 65

> ---------- >> >> Message: 4 >> Date: Thu, 25 Apr 2013 12:01:12 -0400 >> From: Tushar Krishna >> To: gem5 users mailing list >> Subject: Re: [gem5-users] SE mode: 2DMesh topology >> Message-ID: >> Content-Type: text/plain; char

Re: [gem5-users] virtual_network : 0 / 1 / 2 meaning?

Protocol messages are sent in different virtual networks (message classes) to avoid deadlock .. example requests and responses go on separate virtual networks which use separate queues... Mann wrote: >hi all > >what virtual_network : 0 / 1 / 2 refers to/mean in protocol (.sm) >files ? > >eg.

Re: [gem5-users] how .sm files communicate?

They follow a producer-consumer relationship. The slicc parser makes some module the consumer for this queue, and that module will call the dequeue for its producer. Take a look at build/X86/mem/protocol/L1CacheController.cc for instance, after you build your protocol. You will see lines like thi

Re: [gem5-users] how .sm files communicate?

cation between these or is >there >other way? > > >On Tue, Jun 4, 2013 at 5:09 AM, Tushar Krishna >wrote: > >> They follow a producer-consumer relationship. >> The slicc parser makes some module the consumer for this queue, and >that >> module will call

Re: [gem5-users] MessageBuffer Usage

Hi Mann, Many of the SLICC experts are perhaps not actively monitoring this list currently, hence you have not gotten a reply. While I have not used MessageBuffers much, here are my 2 cents. Try to read up on SLICC (http://gem5.org/SLICC) and Life of a Memory Request in Ruby (http://gem5.org/Rub

Re: [gem5-users] buffers in the router

The organization is as follows: Each Input Unit has a vector of Virtual Channels Each Virtual Channel has a member called flitBuffer_d Each flitBuffer_d is a vector of flits (flit_d) Each flit_d carries a actual message (MsrPtr) in addition to network related information (route, vc id etc etc). If

Re: [gem5-users] Perfect Network

Hi Mahshid, I have often done the following to model a 0/fixed-delay network: Magically send the message from the source NetworkInterface to the destination NetworkInterface without sending it through the routers. Create some function in the global network class (GarnetNetwork_d.cc) to receive a

Re: [gem5-users] Network latency breakdown for Garnet

That breakdown is not tracked. You will have to add your own counters. - Tushar Mahshid Sedghi wrote: >Hi all, > >I am using Garnet and would like to know the breakdown of the network >latency. To be more exact, I want to know the queuing latency as well >as >transfer latency. It seems that th

Re: [gem5-users] Change garnet to 3D

Hi Yuhang, Once you set up a 3D topology, fundamentally nothing has to change in garnet for it to work - each router simply has more ports now. If you want the Z links to have different latencies, I think you can set that in the topology file as well. But if you actually want to model some differ

Re: [gem5-users] routing table and weight table

Hi Yuhang, msg_destination is basically a bit vector for the destination(s). 0 0 0 0 - 1 0 0 0 - 0 0 0 0 This represents the 4 directories, 4 L2s and 4 L1s. In your example, this means that the destination is L2 #3, which will be connected to Router #3. Now, with regard to the code, the actual co

Re: [gem5-users] odd even turn model

Hi Yuhang, There is no way to identify the directions of the ports since the current setup allows you to potentially declare any topology, not just a mesh. The way I typically do this is the following: In GarnetNetwork_d.cc, when the router ports are being added (makeInternalLink etc), you can se

Re: [gem5-users] About OrionLink and wire

Hi Lyh, The "total router power" in the stats is the sum of the power at all network routers. If you look at src/mem/ruby/network/garnet/fixed-pipeline/GarnetNetwork_d.cc, you will see that calculate_power() is called for each router, and the power summed up and displayed. calculate_power()

Re: [gem5-users] Question about Mesh Topology

Hi Tejasi, In general there is *no fixed rule* for these, and depends on the coherence protocol. For most protocols currently in gem5, each "Directory" Controller implies both a Directory (i.e. state) and a Memory Controller. The memory controller models the delay to main memory. We don't defin

Re: [gem5-users] Problems about Mesh in FS mode

Hi Xing, How many directories did you declare? A few things to note: The Mesh topology by default works with equal cpus (= l1 caches) and directories, connecting them all to one router each. If you use only 4 directories, MeshDirCorners might be a better topology as it connects each cache to a r

Re: [gem5-users] Question about Mesh Topology

where does the routing_table_entry variable in routing_unit.cc gets its value from? I am unable to figure these things out by looking at the code and hence stuck at one point. Hope to hear from you soon. On Thu, Oct 27, 2011 at 9:11 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> w

Re: [gem5-users] Problems about Mesh in FS mode

Xing 2011/11/17 15:18, By Tushar Krishna: Hi Xing, How many directories did you declare? A few things to note: The Mesh topology by default works with equal cpus (= l1 caches) and directories, connecting them all to one router each. If you use only 4 directories, MeshDirCorners might be a better

Re: [gem5-users] Question about Mesh Topology

in makeTopology assert(len(dir_nodes) == 1) AssertionError I am not sure why is this error? Also, how does it know the node type in line 48? Could you please guide me on this? On Sat, Nov 19, 2011 at 11:52 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: H

Re: [gem5-users] Question about Mesh Topology

, 2011 at 8:33 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: Hi Tejasi, The assertion for number of directories being 1 failed. The node type is being set in line 52: elif node.type == 'Directory_Controller': dir_nodes.append

Re: [gem5-users] Question about Mesh Topology

= 0) > AssertionError > > I am not getting what is the remainder value coming out to be because I just > changed the num_dirs and its connections. Could you please help me out with > this? I really appreciate your prompt replies. > > On Sat, Nov 26, 2011 at 11:41 PM, Tusha

Re: [gem5-users] Issue Defining new Ruby Topology/ Possible Bug in Topology.cc

Hi Malek, Hmm yeah checking the condition once should be enough, not sure why it is checked twice. But in any case, the condition being checked is correct, even for your topology. Each "ext_link" is the link connecting a controller (L1, L2, dir etc) to a router, while "int_link" is for links

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

Hi Pavan, As you probably noticed, a patch has just been approved on the gem5-dev list to fix this issue... Keep a look out on that list for the patch to be pushed, and then try it .. cheers, Tushar On 1/27/12 1:11 AM, Pavan Poluri wrote: Hello, I am trying to invoke garnet module in gem5.

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

our command line ran just fine over here both in the debug and opt versions (provided the missing space after --fixed-pkts was a typo) If you still have no success, I'd do what Tushar said and wait for the patch to be pushed. Regards, Alexandre Yasuo Yamamoto On Fri, Jan 27, 2012 at

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

of ruby_network_test.py or changing the topology? Thanks, Pavan On Jan 27, 2012, at 11:28 AM, Tushar Krishna wrote: Yeah the warnings are always there but they are not a problem. The specific error with regard to the system port are specific to the ruby network tester which the patch fixes

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

subscribe to the gem5-dev list to receive the emails regarding it or do those emails appear on gem5-users list too? Thanks, Pavan On Fri, Jan 27, 2012 at 11:57 AM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: Hi Pavan, I saw Nilay approve the patch on the gem5-dev mailing l

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

39 PM, Tushar Krishna wrote: No emails related to code reviews and changes only come on the dev list. On 1/27/12 2:18 PM, Pavan Poluri wrote: Hi Tushar, I am currently building gem5 binary with MOESI_hammer. I will try to invoke garnet with that binary as you suggested after it is done. As pe

Re: [gem5-users] Problem trying to invoke Garnet on gem5 simulator

. There are two fields which I am not able to understand as to what do they mean. 1. m_virtual_networks -> number of virtual networks 2. m_vc_per_vnet -> number of vcs per virtual network What does number of virtual networks mean? Thanks, Pavan On Jan 28, 2012, at 12:55 PM, Tushar Krishna

Re: [gem5-users] Ruby Alone

You can run garnet (+ orion) standalone like this: http://gem5.org/Networktest You can read more about the network implementation and how to invoke garnet here: http://gem5.org/Interconnection_Network cheers, Tushar On Jan 31, 2012, at 8:14 PM, Monobrata Debnath wrote: > Hello Gem5, > > Can

Re: [gem5-users] Interconnection_Network interface

Hi Udit, Do you want to add a new network model in place of garnet or just edit garnet? Replacing garnet will require you to make sure your model follows the same interface as garnet and follows the event-driven nature of ruby. Will be easier for you to just edit garnet. Some useful files within

Re: [gem5-users] Garnet trace file

Hi Andreas, As Ankit mentioned, one way is to hack the NetworkInterface_d.cc file which reads messages from the protocol queues, converts them into flits and injects them into the router connected to it. But you will still need a way for the simulation to call the wakeup of network interface eve

Re: [gem5-users] Garnet trace file

can return the favor some day, > Andreas Prodromou > > On Mon, Feb 20, 2012 at 5:32 AM, Tushar Krishna wrote: > Hi Andreas, > As Ankit mentioned, one way is to hack the NetworkInterface_d.cc file which > reads messages from the protocol queues, converts them into flits and injec

Re: [gem5-users] Problem with GARNET and Network Tester

Hi Nilay, Seems like the problem is deadlock in the Torus topology. The code enforces XY routing in the Torus (similar to a Mesh) but since there are rings in every row/column, it leads to a cyclic dependency within a dimension at high loads. Each response message becomes 5 flits increasing load

Re: [gem5-users] Problem with GARNET and Network Tester

n waiting. I think this time > should be part of the queueing delay stat that is already present. What do > you think? > > -- > Nilay > > > On Mon, 19 Mar 2012, Tushar Krishna wrote: > >> Hi Nilay, >> Seems like the problem is deadlock in the Torus topolo

Re: [gem5-users] bug at NetworkLink_d pointers??

Hi Dean, Looking at the code, I can see that NetworkLink_d has 2 constructors: one where just the id is passed, and is used when CreditLink_d links are created, and one where the network pointer is passed when NetworkLink_d is created. In the latter case, GarnetNetwork_d passes the "this" pointer

Re: [gem5-users] Problem with GARNET and Network Tester

the network. The > queueing delay for these packets will be very high in case of a deadlock in > the routing protocol. > > -- > Nilay > > On Tue, 20 Mar 2012, Tushar Krishna wrote: > >> Hi Nilay, Yes this is precisely the queueing delay stat. If the protocol

Re: [gem5-users] (garnet question) deflection routing

Hi Dean, The OutputUnit does not wake up when the flit has to leave the router (it only wakes up to receive credits). It is the switch that wakes up and schedules the output link. The SWallocator_d removes winning flits from their respective input VCs and places them in the switch buffer. The

Re: [gem5-users] (garnet question) deflection routing

ome back to the same router. cheers! dean On Thu, Mar 22, 2012 at 12:40 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: Hi Dean, The OutputUnit does not wake up when the flit has to leave the router (it only wakes up to receive credits). It is the switch th

Re: [gem5-users] (garnet question) deflection routing

I am not too familiar with the current state of the debug flags. The ruby README might be outdated. Try to search for how the debug flags work within the other modules. On 03/22/2012 05:57 PM, Dean Michael Ancajas wrote: On Thu, Mar 22, 2012 at 3:41 PM, Tushar Krishna <mailto:

Re: [gem5-users] Issue Defining new Ruby Topology/ On-chip versus Off-chip

single > router through "ext_link" correct? > > However, the code for Crossbar seems to contradict this, because there > are N = len(nodes) int_links and ext_links created, where the Crossbar > simulates a pure CMP. > > This is all assuming I am using the MOESI_CMP_

Re: [gem5-users] (garnet question) tester generates load only on 1st virtual network

Hi Dean, The number of vnets that get traffic ultimately depend on the coherence protocol. When you ran Simics, the coherence protocol you built will determine how many vnets are active and what kinds of traffic they carry. If you are using the GEMS network tester, the traffic injector code in T

Re: [gem5-users] (garnet question) tester generates load only on 1st virtual network

using PARSEC > benchmarks) config so this got me a bit puzzled for a while now. > > thanks, > dean > > > > On Tue, Apr 3, 2012 at 5:38 PM, Tushar Krishna wrote: > Hi Dean, > The number of vnets that get traffic ultimately depend on the coherence > protocol. Wh

Re: [gem5-users] Ruby Network Tester

Hi Nilay, Ah yes, I don't think that memory is being deallocated. Looking at memtest/memtest.cc after which networktest was modeled, I believe this needs to be done in the completeRequest(PacketPtr pkt) function right? -- Request *req = pkt->req; Ne

Re: [gem5-users] Issue Defining new Ruby Topology/ On-chip versus Off-chip

there are references to multiple chips, and simulationg SMP/SCMPs, I was trying to convince myself otherwise. Malek On Sun, Apr 1, 2012 at 7:19 PM, Tushar Krishna wrote: Hi Malek, The "int" links just mean links internal to the network, and "ext" links just mean links connecting the

Re: [gem5-users] NoC in ruby

cc:ing the users list as this might be useful info... Hi Xuhao, The constructor for SimpleNetwork is actually being called from python in configs/ruby/Ruby.py Note that I am not very familiar with the python-C++ interface … The total number of vnets are defined in src/mem/ruby/network/Network.py

Re: [gem5-users] #VNs in MOESI_Directory

Hi Monobrata, To see the number of VNs required by each protocol, you should take a look at its .sm files. Take a look at src/mem/protocol/MOESI_CMP_directory-L1cache.sm, src/mem/protocol/MOESI_CMP_directory-L2cache.sm, src/mem/protocol/MOESI_CMP_directory-dir.sm and src/mem/protocol/MOESI_CMP_

Re: [gem5-users] Replacing Garnet

Take a look at src/mem/ruby/network/garnet (where all the garnet source files exist) and src/configs/ruby/Ruby.py (where garnet is instantiated). You should edit the SConscript file in the src/mem/ruby directory to make sure your simulator gets built. - Tushar On Apr 25, 2012, at 12:24 AM, Mono

Re: [gem5-users] Question on Network_test (networktest.cc) to test GARNET

Hi Pavan, A couple of things to note: 1) 0.01 is an average injection rate. So a node *could* inject more than 10 packets (but not a lot more). The implementation is basically a bernoulli trial, where every cycle you generate a random number and decide to inject or not based on whether it is les

Re: [gem5-users] Question on Network_test (networktest.cc) to test GARNET

anks for your time. > > Thanks, > Pavan > > On Sun, Apr 29, 2012 at 1:05 PM, Tushar Krishna wrote: > Hi Pavan, > A couple of things to note: > 1) 0.01 is an average injection rate. So a node *could* inject more than 10 > packets (but not a lot more). > The implement

Re: [gem5-users] Question on Network_test (networktest.cc) to test GARNET

the same thing it reported when simcycles was of type Tick. Thanks for your time. Thanks, Pavan On Sun, Apr 29, 2012 at 1:57 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: You can specify --random_seed=xxx in the command line. Yes numPacketsSent in networktest.cc

Re: [gem5-users] Question on Network_test (networktest.cc) to test GARNET

another question. When injecting synthetic traffic we control the injection rate via the --injectionrate argument. Is it possible to control the injection rate of benchmark traffic too? Thanks, Pavan On Apr 30, 2012, at 12:02 PM, Tushar Krishna wrote: Hmm, you probably need to take a look at src

Re: [gem5-users] Run a new topology in Gem5

I assume in Mytopology.py, you start with class Mytopology(Topology): … right? If yes, what was the command you ran? - Tushar On May 8, 2012, at 5:05 AM, Haera Chung wrote: > Hi Gem5, > > I created my network topology and copied the ‘Mytopology.py’ into > “~/gem5-stable/src/mem/ruby/networ

Re: [gem5-users] [gem5-dev] Review Request: Garnet: Stats at vnet granularity + code cleanup

Hi Juyoung, In future pls send the emails directly on gem5-users so that others who experienced similar problems can help out as well? I am cc:ing gem5-users I just pulled out the latest version of gem5 and ran this exact command and it worked fine. Are you getting this error only with ruby and

Re: [gem5-users] Run a new topology in Gem5

> --timing --garnet-network=fixed ……… --topology=Mytopology --mesh-rows=8 -b fft > > Thanks, > HR > > From: gem5-users-boun...@gem5.org [mailto:gem5-users-boun...@gem5.org] On > Behalf Of Tushar Krishna > Sent: Tuesday, May 08, 2012 12:01 PM > To: gem5 users mailing list &

Re: [gem5-users] [gem5-dev] Review Request: Garnet: Stats at vnet granularity + code cleanup

version of c++ you are > using on your platform? (including other packages' version such as > swig/python/zlib/m4 if possible). > > Can you give me more comment on this garnet invocation issue? > Thanks. > > > - Juyoung > > > > > > > &g

Re: [gem5-users] Run a new topology in Gem5

kill the simulation? > > > Regards, > HR > > From: gem5-users-boun...@gem5.org [mailto:gem5-users-boun...@gem5.org] On > Behalf Of Tushar Krishna > Sent: Tuesday, May 08, 2012 1:16 PM > To: gem5 users mailing list > Subject: Re: [gem5-users] Run a new topology in Ge

Re: [gem5-users] [gem5-dev] Review Request: Garnet: Stats at vnet granularity + code cleanup

imulation even > though I am able to invoke a garnet object without error in this time. I am > not sure if this deadlock or invalid transition problem is also caused by > slightly different development tool version. What would be the real cause of > the deadlock/invalid transition be

Re: [gem5-users] [gem5-dev] Review Request: Garnet: Stats at vnet granularity + code cleanup

the problem is only related to gem5-stable version, because I > am able to run the same command with gem5 (development version). Can you > perform the regression test for the protocol with gem5-stable in your side? > Thanks. > > Regards, > > Juyoung > > On Wed, May

Re: [gem5-users] Network Mesh topology not working

Hi Xi, The Network_test coherence protocol and ruby_network_test.py work in tandem, and is for simulating synthetic network-only traffic. With other coherence protocols (such as MESI_CMP_directory) as you are using currently, you should use other testers (such as ruby_random_test.py), or run in

Re: [gem5-users] Network Mesh topology not working

Ah ... I think the problem is that the number of L2's (required by MESI_CMP_directory), has not been defined as 16 (default is 1). [Note that for private L2 coherence protocols like MOESI_hammer the number of L2's is equal to the number of L1's by default, while for shared L2 protocols like MES

Re: [gem5-users] NetworkTester and GarnetNetwork_d

Hi Kypro, NetworkTest is simply the dummy coherence protocol that you can run to simulate network-only traffic. GarnetNetwork_d is the actual network. The messages to/from the NetworkInterface_d come from/to the cache/dir controllers in NetworkTest. Could you give me more details about what is it

Re: [gem5-users] NetworkTester and GarnetNetwork_d

Hi Kypros, The hierarchy is something like this: src/cpu/testers/networktest.cc (~CPU, issues reads and writes) --> NetworkTest protocol (~cache coherence protocol which creates cache/dir controllers) --> garnet/fixed-pipeline/NetworkInterface_d (receives msgs from cache controller and converts

Re: [gem5-users] Modify MOESI_hammer protocol

The error has nothing to do with the virtual network being ordered or not. If you look at MessageBuffer.hh, line 131, a setOrdering function needs to be called for each MessageBuffer. This is called from the coherence protocol's generated files. [The ordering itself might be true or false dependi

Re: [gem5-users] Modify MOESI_hammer protocol

ight way to do or there is some other better method? > Thank you very much for your help. > > Best, > > Jinzhu > > > > On Thu, Jul 26, 2012 at 12:53 PM, Tushar Krishna wrote: > The error has nothing to do with the virtual network being ordered or not. > If you

Re: [gem5-users] Modify MOESI_hammer protocol

ar is > very very close to what I wanna do. Thank you very much! > > Best, > > Jinzhu > > On Thu, Jul 26, 2012 at 7:30 PM, Tushar Krishna wrote: > Hi Jinzhu, > Why do you want to broadcast to L1s and the directories? Just to add more > messages in the network? MOE

Re: [gem5-users] Modify MOESI_hammer protocol

ection to go? Thanks a lot again! > > Best, > > Jinzhu > > On Fri, Jul 27, 2012 at 12:44 AM, Tushar Krishna wrote: > Hi Jinzhu, > Ok so I see two components in your design: > (1) broadcast-based coherence protocol > (2) shared-bus based crossbar > > For (1) y

Re: [gem5-users] Modify MOESI_hammer protocol

For instance, you could think of the "directory" in the MOESI_hammer protocol as something analogous to the bus arbiter, which receives unicast requests from all caches, and then sends them out on the bus (broadcast) in some order… - Tushar On Jul 27, 2012, at 6:10 PM, Tushar Kri

Re: [gem5-users] Modify MOESI_hammer protocol

nd then modify the MOSI_SMP_bcast? I > should be able to create a hierarchical switch with gem5 right? > > Thank you! It's so nice of you! > > Best, > > Jinzhu > > On Fri, Jul 27, 2012 at 6:24 PM, Tushar Krishna wrote: > For instance, you could think of the &qu

Re: [gem5-users] NetworkTest: Packets generated but not actually injected?

Hi Kypros, The network tester generates (InjectionRate * num-dirs * sim-cycles) number of flits, and these are sent to the NetworkInterface (NI) in Garnet. The NI now tries to inject them into the network. That depends on whether there are free VCs at the input port of the connected router, which

Re: [gem5-users] NetworkTest: Packets generated but not actually injected?

Hi Kypros, Yes they should be injected at some point, they will not get thrown away. But you defined your simulation's simCycles to be 100,000 cycles at which point the simulation ended. If you use the --fixed-pkts option to inject a fixed number of packets or flits say 640,000 with injectionRate

Re: [gem5-users] Questions with moving broadcast protocol MOSI_SMP_bcast from GEMS to GEM5

Hi Jinzhu, The error is that your protocol found an invalid transition when in state "I". If you grep for "Invalid transition", you will see that this error message was most likely printed out from build/X86_SE/mem/protocol/L1Cache_Transitions.cc You can print out a protocol trace in such cases t

Re: [gem5-users] Questions with moving broadcast protocol MOSI_SMP_bcast from GEMS to GEM5

ing(the root node is the serialization point). Do I misunderstand something here? Can the protocol cause the violation if it's not the topology? Thank you! Best, Jinzhu On Fri, Aug 3, 2012 at 8:12 PM, Tushar Krishna <mailto:tus...@csail.mit.edu>> wrote: Hi Jinzhu, The er

Re: [gem5-users] Ruby Message.clone() -- multicast

Hi Mark, No sending the same MsgPointer around can cause problems because the NI that receives it earlier will send it up to the cache/dir controller which will delete it after consuming it. You could perhaps add some extra bcast_id in the message at the source and identify it based on the {sou

Re: [gem5-users] Set total order among multiple logic buses

Hi Jinzhu, Hmm if I understand your topology correctly, each controller connects to multiple routers: each router corresponding to one bus. The problem is that all these routers connect to the same MessageBuffer in the controller... ? And I guess the problem occurs because the routers on each bu

Re: [gem5-users] Set total order among multiple logic buses

router arbitrates/allocates everything following the same round robin, but > maybe that's not the case... > > Thanks for your suggestion. I will look at the simple router first and then > check the event queue if necessary. > > Best, > > Jinzhu > > > On Fri, Aug

Re: [gem5-users] Question on GARNET + Ruby

That is the combined power for entire NoC. If you want individual router power, and/or breakdown of individual components, look at network/orion/NetworkPower.cc - Tushar On Aug 28, 2012, at 12:30 PM, Pavan Poluri wrote: > Hello, > > I have done a full system simulation with 8 cores, 8 routers,

  1   2   >