Hi,
The tag and data access latency you are seeing in
\`src/mem/ruby/structures/RubyCache.py\` are actually not implemented by the
RubyCache itself. Instead, SLICC machines query these parameters using the
getTagLatency and getDataLatency member functions of the CacheMemory SimObject
(these
Regarding the rest of your question, I suggest getting familiar with the gem5
basics reading the “learning gem5” course. It will tell you what you need to
know to figure out how gem5 could help you in your research.
Unless I misunderstand what you are trying to achieve, you will likely have to
Hi Chetan,
se.py is no longer supported and will never be supported again.
configs/deprecated/example/se.py in its current shape is all you will get as
far as se.py is concerned.\
Hopefully, there’s the gem5 standard library! You can find many preconfigured
components and board configurations
Hi,
DES is a very well documented topic. You can find a lot of resources about this
topic searching for “discrete event simulation”. gem5 implements a simple
version of DES compared to, say, SystemC. It is explained in the *learning
gem5* tutorial. To put it simply : events are scheduled at
You need to add an attribute to the Cache python SimObject class:
> cctrl = Param.Center_Ctrl(“”)
You will need to import Center_Ctrl from m5.objects before as well.
You can find hundreds of examples of this construct in most python SimObject
classes.
Regards,
Gabriel
Hi Vijay,
1. *build/X86_MI_example/gem5.opt* is built with the MI protocol by default
while *configs/learning_gem5/part3/simple_ruby1.py* requires the MSI protocol.
I guess your simulation fails before it starts. You should see network related
stats in a well configured ruby-based simulation.
Hi Arth,
You can only have a single debug log destination. However, You can take
advantage of the FmtFlag special debug flag that will prepend the corresponding
debug flag to every logged line. You can then use grep to filter down the lines
of interest. If using bash, you can pipe the standard
Hi Arteen,
If you are not tied to the MESI_three_level protocol, then, as Gautam said, you
might want to switch to a different protocol that already supports the shared
state in the system level cache. CHI is the most advanced and configurable
protocol in that respect but others might do as
Hi Karim,
The short answer to “can I use a single vnet for MESI_Two_Level is “probably
not”, just as for any provided Ruby protocol.
The reason is that vnets provide traffic isolation between different message
classes (request, response, etc.). This isolation is required to avoid
deadlocks.
Hi Ghadeer,
I think this thread could help you with that:
https://gem5-workspace.slack.com/archives/C03K26BGTKL/p1694723886410439.
Regards,
Gabriel
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to
Hi,
For 1, I meant automatically: “lines will not get blocked automatically”.
As for 2, peek does not block anything unless you use the block_on mechanism.
You can peek as many queues as you want as well as dequeue and enqueue as many
as you want.
Regards,
Gabriel
Hi,
1. First, isLocked is part of the implementation of atomics based on
load-linked-store-conditional. When processing a request in ruby controllers,
lines will not get blocked manually. In order to avoid request hasards, several
mechanisms are available.
1. The block_on mechanism of
Hi Chunfeng,
Can you be more specific about “the buffer” you are referring to?
Also, “routing creating deadlock” sounds more like a routing and/or topology
issue than a micro-architecture issue.
Introducing a bubble as a deadlock avoidance countermeasure sounds like a
slippery option. If you
Hi,
With the current latest version of gem5, you must use python <= 3.10. Python
3.11 will be supported in the next release of gem5 coming very soon.
For details: https://gem5.atlassian.net/browse/GEM5-1295
Regards,
Gabriel
___
gem5-users mailing
Hi David,
If these “prerecorded traces” are not recorded using the standard gem5 cpu
tracer (elastic traces), you will likely need to roll your own trace-based
traffic generator.
TrafficGen is likely not the tool you are looking for. TraceCPU might be a
closer match. You can watch at the code
Hi Karim,
You can’t set the value of ruby_system.network.number_of_virtual_networks
arbitrarily. This value must match the actual virtual networks used by the
protocol it corresponds to. In the case of Garnet_standalone, only 3 vnets are
defined (c.f., network buffers defined by
Hi Karim,\
\
You can run simulation for a few thousand cycles using
m5.simulate(some_amount_of_ticks), then reset stats using m5.stats.reset() and
then run m5.simulate again. You will collect stats on a warmed up system.
Regards,
Gabriel
___
Hi Karim,
Can you be more specific about what you are trying to achieve for us to provide
accurate answers?
You seem to be really focused on the transport and not so much on the coherency
protocol. In that case, I recommend using the garnet standalone protocol:
*
Hi Karim,
To put it simply, virtual networks are logically independent transport
networks: messages traveling on a given vnet shall not be blocked by messages
on a different vnet.
Garnet is a transport model. As such, it does not decide on the required number
of vnets. The Ruby protocol used
Hi Eliot,
I can’t provide you with an assertive answer but I’ve also been looking at CXL
recently so here is what I understand so far.
>From a functional perspective, the classic cache system seems able to support
>the hierarchical coherency aspects just fine with the coherent Xbar of each
Hi Zhang,
Did you observe that behavior during a simulation or is your observation solely
based on your analysis of the protocol definition?
If you did observe that behavior, then I think it should be considered a bug
but not necessarily from the protocol.
First, it would be a bug in the
When getting a segfault, you want to run the simulation in a debugger to know
where it happens. Enabling the --with-asan --with-ubsan flags on scons can also
help a lot when the segmentation fault occurs at a different place than the bug
location in the code.
Gabriel
Hi Karim,
Where does the segfault happen? Have you tried your topology with the default
weight-based routing algorithm?
Regards,
Gabriel
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org
Hi Ziyao,
Even though protocols are often implemented that way, SLICC does not require
that each request type corresponds to an event type. For instance, the event
can be a function of both an incoming message type and of the internal state of
the controller. The only true way to understand
Hi Karim,
You could indeed go for 16 directories and connect the 7 remaining directories
to a dummy router and then ensure that traffic generators never send requests
to these directory. This can be difficult depending on the traffic generators
logic so here is another approach if that one
Hi Ali,
It is difficult to tell where is the problem based on the information available
(I can’t open .py attachments) but here is a list of checks that you can
perform.
First, ensure that your modules are present and configured in m5out/config.ini.
Especially, you want to check that the
Hi,
Just replace options.voltage_value with, say, “1V” in your script.
options.voltage_value tries to read the value from the command line options but
that option does not seem to exist anymore.
Regards,
Gabriel
___
gem5-users mailing list --
Hi,
If by page you mean an aligned 4k memory region, then no, packets address are
not page addresses. They are the address of individual requests that target
arbitrary addresses with accesses of various sizes. Mask with \~0xFFF and you
will get the page address you are looking for, assuming
Hi,
If you are running regular programs under an OS, two different processes will
never access the same physical memory unless explicitly asked to the OS through
mechanisms like mmap. In particular, malloc will return virtual addresses that
are mapped to physical addresses that are specific to
Hi Wilson,\
\
I’ve never tried that myself but I believe that the issue comes from
RubyPort::MemResponsePort::getAddrRanges() always returning an empty list,
making the XBar believe that the ruby_system is not responding to any address.
That would require a fix likely going through RubySystem
Hi,
I’m not aware of a way to filter stats before dumping. Stats file size is
usually not a big concern.
If you still want to filter stats before writing them to disk, you can filter
the dump on the flight using process substitution:
```
gem5.opt --stat-file >(grep -e numCycles -e miss >
Hi Vipin,\
\
You can use the comm_monitor SimObject. It can be placed in between the
processors and the L1 caches in the python configuration script. You can set
the address mask parameters of the comm_monitor to record cacheline-aligned
addresses.\
\
However, I am cautious about the accuracy
Hi Prakhar,
A standard solution to 1 would be to use M5ops:
https://www.gem5.org/documentation/general_docs/m5ops/. That small library
enables you to call magic functions from your guest binary that will trigger
events in the host simulation environment. Namely, the dumpresetstats op will
NI decided once the tail flit reaches its destination router?
This I don't know for sure. I would suggest looking at
RoutingUnit::outportCompute() and around.
>
> Thank you,
> Vedika
>
> Get Outlook for Android<https://aka.ms/AAb9ysg>
>
> _______
Hi,
This is based on the virtual network ID specified at the top of the machine
declaration with the MessageBuffer* declaration.
E.g.: MessageBuffer * L1RequestToL2Cache, network="From", virtual_network="0",
vnet_type="request";
Gabriel
___
Response in the original thread.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org
%(web_page_url)slistinfo%(cgiext)s/%(_internal_name)s
You responded in a new thread instead of this one ;) I'm answering here.
Go to the gem5 directory, run the following commands and attach the output in
your answer:
$ file tests/test-progs/hello/bin/arm/linux/hello
$ file tests/test-progs/hello/bin/arm/linux/myfile_name
For instance on the hello
Hi,
I would suggest you to double-check that your binary is compatible with
gem5-arm. I get the same error when I attempt to run an X86 binary on an ARM
platform for instance. This is a result of the loader failing to load your
binary.
You can compare the output of the command "file" on both
Hi Gabe,
Such configuration tool would be great, indeed!
I buy the performance argument for using static linking in release builds.
But for debug builds, it is slightly less convincing to me ;)
In the mean time, if you have any tip you are using to speed up the Edit,
Compile, Debug cycle, I
Hi all,
I would like to know what is your favorite way to run the ECD cycle with this
gem5.debug beast of a binary.
I am currently developing a pretty large Ruby protocol and use 2 build
configurations: debug and fasta for "fast with asserts". I've defined the fasta
build target myself and
Hi John,
Short answer : no, you can only run several simulations in parallel, but not a
single simulation using one thread per CPU.
Gem5 relies on Discrete Event Simulation (DES) to simulate the concurrent
behavior of HW.
DES is intrinsically sequential in its execution as it relies on
Hi Jason,
Thanks for your reply, it should be an adequate solution, even if it looks kind
of awkward to me right now ;)
While I wanted the downstream controller (B) to have a reference to the
upstream (proxied) controller (A), I must actually do the opposite and have (A)
store a reference to
I am currently hacking my way through it using the NodeID (e.g, (A).version on
python side) as a parameter of controller (B) so that I can basically write in
(B)'s state machine :
MachineID proxiedController :=
createMachineID(MachineType:, proxiedNodeID);
It seems to work for now but I
Hi,
I need to reference an upstream controller (A) in the parameters of another
ruby controller (B).
(B) is a downstream_destination for (A) in the ruby system.
However, using a Param.RubyController in (B) to reference (A) obviously creates
a cycle in the hierarchy.
How can I get the MachineID
Hi,
I want to model a coherent interconnect in ruby.
The protocol inside the interconnect defers from the protocol outside of it.
For instance, caches use a subset of the CHI protocol while the interconnect
internally uses an internal protocol.
To interface the caches and the interconnect, an
Hi,
In the file SimpleNetwork.py at line 67 (version v20.1.0.0), it is checked that
the internal side of the network link is connected to ONE OF the routers before
instantiating a buffer for that link:
if link.int_node in self.routers:
This action is done in a loop on the
Hi Gem5 community,
I want to introduce myself as a new Gem5 user.
My name is Gabriel Busnot and I work at Arteris IP, a major network-on-chip IP
provider.
I am working on a new model for performance evaluation of architectural
changes, coherence protocol testing and quality-of-service
47 matches
Mail list logo
