[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

lt/cached kernels are handled >>>>> in MIOpen (stored in database vs directories), which we can also change >>>>> when compiling from source. >>>>> >>>>> I've also had issues with trying to install parts of ROCm from source >>>>>

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

had issues with trying to install parts of ROCm from source >>>> and then subsequent parts of ROCm from .deb packages or apt. That's one of >>>> the reasons for the complexity in the GCN dockerfile. If that issue has >>>> gone away with ROCm 4 (I haven't tried yet

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

>> and then subsequent parts of ROCm from .deb packages or apt. That's one of >>> the reasons for the complexity in the GCN dockerfile. If that issue has >>> gone away with ROCm 4 (I haven't tried yet) that would simplify the >>> dockerfile greatly as we would only need to build ROCcl

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

s we would only need to build ROCclr from source, assuming >> memfd_create is implemented, and everything else could be installed from >> apt. >> >> Kyle >> ---------- >> *From:* gem5-dev@gem5.org >> *Sent:* Friday, May 14, 2021 1:07

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

led from > apt. > > Kyle > -- > *From:* gem5-dev@gem5.org > *Sent:* Friday, May 14, 2021 1:07 PM > *To:* gem5 Developer List > *Cc:* Matt Sinclair ; Poremba, Matthew < > matthew.pore...@amd.com>; Gabe Black ; Bobby Bruce < > bbr

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

Re: Build failed in Jenkins: compiler-checks #72 To clarify the docker situation, here's a quick rundown of how docker is used with gem5: Docker is the fallback for people who are using outdated/weird/unsupported systems and want to use gem5. It's also handy for testing as we can quickly spin up

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

To clarify the docker situation, here's a quick rundown of how docker is used with gem5: Docker is the fallback for people who are using outdated/weird/unsupported systems and want to use gem5. It's also handy for testing as we can quickly spin up different environments (different OS's, different

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

So, is this something *inside* the simulation which needs to be compiled with a particular version of gcc, or is it a part of the simulator itself? I was imagining the former, but if it's the later I see why it's problematic. How tightly coupled is this ROCm code base? Is it just a matter of

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

I'm pretty clueless as far as how the GPU code is built, but would it be possible to build it in a docker or something with a separate older compiler? Would it be reasonable to provide a docker for gem5 building in general, so we can decouple from ye-olde versions of RedHat? In the past Jason

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

I think the main question for (not) dropping support is the LTS for Ubuntu and RHEL. It looks like Ubuntu 16.04 just dropped out of standard support, so we can probably drop support for the default there, now. https://wiki.ubuntu.com/Releases TBH, I can't really tell what's currently supported

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

+1 to raising the minimum GCC version to 7 (not sure about 7.3; 2018 may be too recent) and enabling C++17. Em quinta-feira, 13 de maio de 2021 19:58:49 BRT, Gabe Black via gem5-dev escreveu: I have no objection to moving the compiler versions up. I don't really know what benchmark

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

If we were willing to bump clang support all the way up to version 9 (a big leap, released 19 September 2019) we would get support for __builtin_LINE and __builtin_FILE, which are there to support std::source_location which is a c++20 feature but which can be used without enabling c++20. That

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

I have no objection to moving the compiler versions up. I don't really know what benchmark we use to decide when that's ok to do. If we do move up, it would be nice to move to a version which would let us use c++17. For gcc, the oldest version with any support is 5, there seems to be pretty solid

[gem5-dev] Re: Build failed in Jenkins: compiler-checks #72

These two patchset should fix most of this: https://gem5-review.googlesource.coThism/c/public/gem5/+/45479 , https://gem5-review.googlesource.com/c/public/gem5/+/45481 Unfortunately, we currently can't compile with GCC 5 as deprecation