Re: [gem5-dev] Memory fence neglected/obviated after squashing a younger/later fence

2020-01-18 Thread Jason Lowe-Power
Hey Eduardo, From a quick skim, I think what you are describing makes sense, and your solution sounds like it's on the right track. Instead of attaching to email (which is stripped on this mailing list), could you please create a changeset on gerrit. See http://new.gem5.org/contributing and

[gem5-dev] Memory fence neglected/obviated after squashing a younger/later fence

2020-01-15 Thread Eduardo Jose Gomez Hernandez
Hi everybody, During  running  some  benchmarks  (actually, a  modified version  of ocean-contiguous-partitions    from Splash-3 "https://github.com/SakalisC/Splash-3;), we encountered a deadlock. After diving into the trace files, we found that an atomic instruction was locking the