Hello, Hmm... this is surprising since that commit only adds functionality (it actually implements those instructions).
This could be something to do with the classic memory system. In Ruby, all DMAs are coherent. I'm not sure that's true in the classic caches. >From what I can tell, the working and not working commits are quite far apart. It would be helpful to narrow that down. I suggest using git bisect to find the problem. I'm currently working to verify which kernel versions, gem5 versions, CPUs, etc. work with x86. I'll post something soon about this :). Jason On Mon, Sep 16, 2019 at 2:03 AM Eleanor <[email protected]> wrote: > Hi Gabe, > > I get this warning in stdout: > "warn: MOVNTI: Ignoring non-temporal hint, modeling as cacheable!" > At this point the gem5 gets stuck (at mounting filesystems). > Coincidentally, the simulation also gets stuck at that warning when my > colleague runs this command : > $GEM5_DIR/build/X86/gem5.opt $GEM5_DIR/configs/example/se.py --caches > --l2cache --cpu-type=AtomicSimpleCPU -c > benchmarks/benchspec/CPU2006/462.libquantum/exe/libquantum_base.gcc41-64bit > -o '33 5'. > I didn't have this issue earlier. > gem5 version with issue: 904784fb1e15f0c090fb1f1e5c5057e74b0b4ea8 > It seems this issue does not show up on gem5 version: > 08e7849f2710e6b62aa07db86e356edb15a99f9c. > > I think the issue can be correlated to this commit > https://gem5.googlesource.com/public/gem5/+/8d439c29ed86f55c454396fdd03d606cfa13a95. > > _______________________________________________ > gem5-users mailing list > [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
