Re: [Gems-users] Odd EL_Timestamp behavior

[Jayaram Bobba <bobba@xxxxxxxxxxx>]

It is weird. Timestamp policy should not livelock... Can you get a more 
detailed output and see if it is really livelock
or if the oldest transaction died and the other transactions abort 
waiting for it?
The pathology numbers were computed from trace files output by 
XactVisualizer. However, I dont think the entire functionality
of the visualizer is included in the release. We didn't have the time to 
clean it up and put it out.
Jayaram

Jim Leek wrote:
> I've been running though the benchmarks with various logTM 
> configurations, when I hit one I find odd.  I set the following 
> variables (I can list the rest if anyone thinks it will help):
> XACT_EAGER_CD: true
> XACT_LAZY_VM: true
> XACT_CONFLICT_RES: TIMESTAMP
> XACT_NO_BACKOFF: false
>
>
> Which should mean EL_Timestamp.  When I ran the barnes benchmark it 
> livelocked.  I can't think of a case where EL_T should livelock, is 
> that no odd?
> Livelock evidence: 1 hour in:
>
> xact_size_dist: [binsize: 2 max: 68 count: 224 average: 7.95536 | 
> standard deviation: 10.643 | 0 51 0 145 0 0 15 5 0 0 0 0 0 0 0 0 0 0 0 
> 0 0 0 0 0 0 0 0 1 2 0 1 1 1 0 2 ]
> xact_aborts:        9164
>
> 9 hours in:
> xact_size_dist: [binsize: 2 max: 68 count: 224 average: 7.95536 | 
> standard deviation: 10.643 | 0 51 0 145 0 0 15 5 0 0 0 0 0 0 0 0 0 0 0 
> 0 0 0 0 0 0 0 0 1 2 0 1 1 1 0 2 ]
> xact_aborts:        14152
>
> BTW, how did you get the numbers in Table 2 of the pathologies paper?  
> (pathology % of total execution time.)
> Jim
> ------------------------------------------------------------------------
>
> _______________________________________________
> Gems-users mailing list
> Gems-users@xxxxxxxxxxx
> https://lists.cs.wisc.edu/mailman/listinfo/gems-users
> Use Google to search the GEMS Users mailing list by adding "site:https://lists.cs.wisc.edu/archive/gems-users/"; to your search.
>
>