Ah, that uint64 problem was relatively easy to fix. There are a couple
other things to fix too, and hopefully I'll send out a few patches later on
to get everything to build.

Gabe

On Fri, May 19, 2017 at 5:10 PM, Gabe Black <[email protected]> wrote:

> To clarify, I think this fixes the build problem when attempting to build
> gem5.opt, although we're running into other issues with pybind which I
> think might be specific to the version of Ubuntu we're trying to use. It
> "works" because while the systemc SConscript is processed, none of the
> build outputs are pulled into the dependency tree and so none are built.
> That's why my one change makes it possible to build things in ext/
> explicitly if you wanted to for, for example, testing the build.
>
> Gabe
>
> On Fri, May 19, 2017 at 5:03 PM, Gabe Black <[email protected]> wrote:
>
>> I made a pass at fixing this with these two CLs:
>>
>> https://gem5-review.googlesource.com/#/c/3460/
>> https://gem5-review.googlesource.com/#/c/3461/
>>
>> With those applied, I'm still running into compile errors, I think
>> because the underlying type for uint64 is unsigned long, and there are
>> operators which override for both types. I think g++ gets cranky because
>> those are equivalent or at least ambiguous, so you're not allowed to have
>> both. I'm not sure why that would blow up on 64 it ARM and not 64 bit x86,
>> but then again I'm not super familiar with this code base.
>>
>> Gabe
>>
>> On Fri, May 19, 2017 at 2:05 AM, Christian Menard <
>> [email protected]> wrote:
>>
>>> Hi Gabe,
>>>
>>> thanks for reporting this. I will look into it and hopefully post a patch
>>> soon. Maybe it would be best to leave QT threads out completely and use
>>> pthreads on all platforms. However, in the past, this caused some
>>> trouble on
>>> x86 when linking to gem5.
>>>
>>> Kind regards,
>>> Christian
>>>
>>> On Friday, 19 May 2017 09:25:51 CEST Gabe Black wrote:
>>> > Hi folks. We've just started to run into some problems building on ARM
>>> > because the systemc build fails out. While investigating, I found a
>>> couple
>>> > problems which are to blame.
>>> >
>>> > First, the SConscript seems to be assuming that it should be using the
>>> QT
>>> > threading implementation which requires architecture specific assembly
>>> > files. There is no file for ARM, so that will fail on ARM. Some
>>> Googling
>>> > suggests that systemc can in fact be built for ARM, but it needs to use
>>> > pthreads for its threading implementation.
>>> >
>>> > That led me to discover that the files which should be used with
>>> pthreads,
>>> > src/sysc/kernel/sc_cor_pthread.cpp, are already being included, and
>>> the
>>> > -pthread flag is being enabled unconditionally. I think those things
>>> should
>>> > be left out when using the QT threading library, and similarly the
>>> assembly
>>> > file shouldn't be required when building on ARM with pthreads.
>>> >
>>> > Gabe
>>> > _______________________________________________
>>> > gem5-dev mailing list
>>> > [email protected]
>>> > http://m5sim.org/mailman/listinfo/gem5-dev
>>>
>>>
>>> --
>>> Dipl.-Ing. Christian Menard
>>> Research Assistant
>>>
>>> TU Dresden
>>> Faculty of Computer Science
>>> Chair for Compiler Construction
>>> 01062 Dresden
>>>
>>> Phone: +49 351 463-42441
>>> e-Mail: [email protected]
>>>
>>> _______________________________________________
>>> gem5-dev mailing list
>>> [email protected]
>>> http://m5sim.org/mailman/listinfo/gem5-dev
>>
>>
>>
>
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to