sandsifter-tests
sandsifter-tests copied to clipboard
Add new CPUs
based on https://github.com/rigred/sandsifter
Thanks for the submission.
I've noticed something strange in the logs, they seem to end far before the usual runs.
660f8fff000000
in both files as opposed to f3dfc7
. This means your logs fall several thousand instructions short of the expected result.
What kernel did you test this with? I haven't yet checked for what effect spectre and meltdown mitigations have on sandsifter testing, but it may be the case here. Otherwise it might be system stability related, I have previously noted that certain overclocked systems fail to run sandsifter correctly.
-- Rigo
For the time being I have pushed this to the evaluation_data
branch.
I'm using stable debian stretch with the current kernel updates: 4.9.0-6-amd64 #1 SMP Debian 4.9.88-1+deb9u1 (2018-05-07) x86_64 GNU/Linux
. It's a minimal (main only) installation with default bios settings so there is no much space for instabilities here.
$ cat /proc/cpuinfo | grep bugs
bugs : cpu_meltdown spectre_v1 spectre_v2
compiling with just a -no-pie flag based on the original repo: https://github.com/xoreaxeaxeax/sandsifter my sandsifter is crashing after 7min on 0fbafc0b
later on 260fbaf405
, 2e0fbaf407
, 360fbaf406
, ... (I will try to make a full list of this problematic instructions)
If you have any idea what is the reason of this misbehavior and how to workaround it I can correct my results. btw. soon I will upload 2more AMD cpus