Re: [Gems-users] ILOX/L1_PUTX in MOESI_CMP_directory

[Mike Marty <mikem@xxxxxxxxxxx>]

Thanks for letting me know.  I think either accepting or canceling the 
writeback will work in this case. 

--mike


Ricardo Fernández Pascual wrote:
> Hello,
>
>     I think I have found a bug in the MOESI_CMP_directory (GEMS 1.4).
> The problem is a missing ILOX/L1_PUTX transition in the L2. It has been
> reported before [1], but I am not sure that I understand it after
> reading that message. After some thinking and after reading [2] (which I
> think is a similar problem) I think that the correct fix is to add the
> following transition:
>
> transition(ILOX, L1_PUTX) {
>   ll_writebackNack;
>   o_popL1RequestQueue;
> }
>
> I am almost sure that the above transition is the correct one, however I
> wonder whether I should change the following transition (around line
> 1519 of the same file) to respond with a NAck too:
>
> // happened if we forwarded GETS to exclusive who tried to do writeback
> //  ?? should we just Nack these instead?  Could be a bugs here
>   transition(ILO, L1_PUTX, ILOW) {
>      l_writebackAckNeedData;
>      o_popL1RequestQueue;
>   }
>
> What do you think?
>
> I have put a tester trace at [3] which shows the problem, just in case
> you are interested.
>
> Thanks,
> Ricardo.
>
> [1]
> https://lists.cs.wisc.edu/archive/gems-users/2006-April/msg00010.shtml
> [2]
> https://lists.cs.wisc.edu/archive/gems-users/2005-July/msg00072.shtml
> [3] http://ditec.um.es/~rfernandez/debug.39.10.gz
>
>