Yes, a good observation of a flaw that will be fixed shortly. The LogTM
requires the fast path to be disabled so that R/W bits are set. But since
the released LogTM protocol is an SMP with combined L1/L2 controllers, by
disabling fast path, the L1 timing is disabled as well.
We will release a fix to this shortly (along with a LogTM CMP protocol).
If you need a faster fix or workaround, one of the LogTM people might be
able to chime in and help you.
--Mike
> When an ld_xact/st_xact hits in a fast path, the request never goes to Cache
> Controller. The question is that, in such a situation, how the simulator
> sets up the R/W bit for the corresponding cache line? It seems that the
> simulator leaves the flag unchanged--be a big bug? If the problem do exists,
> how to get a quick fix?
>
> G.R.
>
> _______________________________________________
> 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.
>
|