Hi Mitch,

I have been focused on O3 cpu model running x86. Mainly, developing and testing 
things in SE mode, however FS should benefit from the modifications as well. 
Testing in FS will be something to be done in the near future.

The main goal of my modifications is to get SMT in O3 to work with x86 in SE 
mode for multiple programs and multi-threaded ones. This has been accomplished, 
currently the patches are still going through review internally. Some of these 
modification will benefit ARM ISA when running O3 as well. We are diligently 
working on getting other required patches (the ones that implement futex and 
clone) out before pushing the SMT ones.

It seems to me our work complements each other.

Best regards,
Alex
________________________________
From: gem5-users [[email protected]] on behalf of Mitch Hayenga 
[[email protected]]
Sent: Friday, April 24, 2015 9:55 AM
To: gem5 users mailing list
Subject: Re: [gem5-users] using SMT in gem5

Hi Alexander,

Just saw this thread and thought I'd contribute.  Are you focusing additional 
SMT support on the various CPU models or just x86-ISA/library side of things?  
I'm wondering how much overlap we have.

I've recently been working on extending gem5 SMT support by adding SMT to the 
Atomic, Timing, and MinorCPU models.   Atomic/Timing CPU SMT was simply done to 
enable fast-fowarding and restoring with a detailed (MinorCPU-based) 
multi-threaded model.  These patches will eventually become public, but need 
more refinement at the moment.

I've been focusing on getting SE mode up and running, but am about to move onto 
adding FS support.  The hope is that FS will be working sometime in the June 
timeframe for ARM under gem5.

Is there much overlap here with what's currently being done on your end of 
things?  I'm mostly focused on SE (currently just testing multiprogrammed) and 
FS support of the core models.


On Thu, Apr 16, 2015 at 10:12 PM, Dutu, Alexandru 
<[email protected]<mailto:[email protected]>> wrote:
Hi Sreejith,

We are planning on submitting some patches which fix SMT issues with X86. This 
will happen over the next couple of weeks. In terms of getting an openmp 
program to run in SE mode, there are some syscalls which need to be implemented 
(i.e. sched_getaffinity).

Best regards,
Alex
________________________________
From: gem5-users 
[[email protected]<mailto:[email protected]>] on behalf of 
Sreejith K M [[email protected]<mailto:[email protected]>]
Sent: Friday, April 03, 2015 1:33 AM
To: [email protected]<mailto:[email protected]>
Subject: [gem5-users] using SMT in gem5

Hi all,

Iam trying to run some simple openmp program in gem5 using SMT. When I ran it 
in SE mode, I got errors that many functions are unimplemented(set_tid_address 
etc). Now Iam trying to run it in FS mode in ARM linux ael image. But there 
also, Iam getting errors as follows,

gem5.opt: build/ARM/cpu/o3/fetch_impl.hh:702: void 
DefaultFetch<Impl>::finishTranslation(const Fault&, RequestPtr) [with Impl = 
O3CPUImpl, Fault = RefCountingPtr<FaultBase>, RequestPtr = Request*]: Assertion 
`!finishTranslationEvent.scheduled()' failed.
 Program aborted at tick 2315524304500
Aborted (core dumped)


Iam able to run the program in multiple CPUs but when I use multithreading in a 
single CPU, Iam getting errors.

I want to know whether SMT is implemented in SE/FS mode . Do I need to make any 
changes while doing SMT?

thanks and regards,

Sreejith K M

_______________________________________________
gem5-users mailing list
[email protected]<mailto:[email protected]>
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

_______________________________________________
gem5-users mailing list
[email protected]
http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users

Reply via email to