[gem5-users] Re: multithreaded program is SE mode using gem5 standard library

[bbruce--- via gem5-users]

Hey Ziyao,

Thank you for pointing this out. This comment is wrong. This was a restriction 
at some time, and even then only in the standard library. Right now you can run 
multi-threaded applications in se-mode via the \`set_se_binary_workload\` 
function. It’s supported.

I’ve submitted a pull-request here to remove this comment: 
https://github.com/gem5/gem5/pull/402.

Kind regards,

Bobby
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: Modifying X86Board to support multiple processors

[bbruce--- via gem5-users]

Hey James,

\
Perhaps I am very wrong in my assumptions here, but is there any difference in 
your idea between a "multiprocessor" setup and a multicore setup? In the gem5 
stdlib design each board has one processor, but each processor can have as many 
CPUs/Cores as you desire. I don't know the ins and outs of your research here, 
but do you really want to simulate separate processors or just have a system 
with many cores? The code you see in \`_setup_io_devices()\` sets up an 
\`X86IntelMPProcessor\` for each core in the processor and is our way of 
conforming to this standard (-ish... I think...).

Behind the scenes all a gem5 standard library "processor" is, is a list of 
Cores in which the processor object manages. While we could have more than one 
processor per-board I don't know how it would actually change what's being 
simulated.

I think you could have more than one "processor" per board but I think the way 
of handling them in a simulation would be to essentially to create a union of 
all the cores across all the processors and, effectively, treat them as one big 
processor anyway.

Again, perhaps I've completely misunderstood this idea, but I'd suggest looking 
at gem5 standard library processors and how they are constructed and used 
inside a simulation and see if you can make extensions/modifications there 
first: 
.

Kind regards,

Bobby
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: About using gem5 with RISCV arch.

[bbruce--- via gem5-users]

You’re trying to run a full-system simulation there and that requires 
specifying a disk image. The \`src/base/loader/image_file_data.cc:110: fatal: 
fatal condition fd < 0 occurred: Failed to open file None.\` error you are 
receiving is because you didn’t specify this.

I’d recommend using the gem5 standard library: 
https://www.gem5.org/documentation/gem5-stdlib/overview. If you go through this 
tutorial you can see how to setup a gem5 simulation and run some simple 
binaries in SE mode.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: m5.switchCpus for ARM starter_fs.py config

[bbruce--- via gem5-users]

What i dont understand here is why do you need cpu switching for starter_fs.py. 
Couldn't you juse use starter_fs.py to get the checkpoints you need then load 
these checkopoints into a simulation which uses the stdlib?

\
They'll be no easy way of introducing stdlib code like 
SimpleSwitchableProcessor into the stdlib. I'd say generally you either use the 
stdlib or use the starter_fs.py script.

\
I’d advise looking at the source-code for SimpleSwitchableProcessor which 
should contain Python code to help you understsand how to setup a system with 
switchable cores. 
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: Run multi-program workload in SE mode

[bbruce--- via gem5-users]

I don’t believe it’s possible to run multiple processes in SE mode. What you 
can do is have a single muli-threader process. You could construct a single 
binary which utilizes pthreads and have each thread it run something different, 
though this could prove difficult to do.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: SPEC2k17 V1.1.9 with gem5 gives errors in SE mode

[bbruce--- via gem5-users]

I think this is an annoying comment to here, but it means there’s an an 
instruction it doesn’t know how to process. In SE mode the error handling is 
very unhelpful in telling you what that instruction actually is. I’m almost 
certain in this case your compilation is including some ISA extension we don’t 
support. I’d strongly advise trying to use FS mode as you’ll get better error 
handling there.

In addition, i’d strongly recommend using the stdlib to do this kind of work. 
An example of using SPEC 2017 in FS mode with the stdlib can be found here: 
configs/example/gem5_library/x86-spec-cpu2017-benchmarks.py. The scripts you 
are using are deprecated and will be removed in a future release of gem5.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: Missing HDF5 libraries

[bbruce--- via gem5-users]

It can be resolved by installing the libraries that are missing. How you do so 
will different depending on your host system. However, these are just warnings, 
not errors. gem5 is still usable for the vast majority of cases without these 
libraries. These warnings simply exist to inform the users that certain niche 
features are disabled as libraries necessary for them to work cannot be found.

If gem5 is working for you, don’t worry about it.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org

[gem5-users] Re: RISCV build failure

[bbruce--- via gem5-users]

Getting this error is normally a side-effect of not having enough RAM. We 12GB 
or more to compile gem5 safely.
___
gem5-users mailing list -- gem5-users@gem5.org
To unsubscribe send an email to gem5-users-le...@gem5.org