Re: [Gems-users] Generated files in ruby

["Lei Yang" <lya755@xxxxxxxxxxxxxxxxxxxx>]

Hi Dan,

Thanks for the fast reply!

> That is approximately correct. A simple CMP protocol would be very 
> illustrative for you...I suggest MOESI_CMP_token, especially 
> MOESI_CMP_token-L1cache.sm and MOESI_CMP_token-L2cache.sm.
I am looking for a protocol where each processor has split L1 and unified 
L2.
I think the CMP protocols GEMS provides has a shared L2 for all processors.
Is this a fact?

> SLICC makes some assumptions about what caches "do". One of these 
> assumptions is that they contain data and can return data when 
> requested...thats what getDataBlk() is doing. You should be aware that the 
> dataBlk functionality of ruby is deprecated and no longer functions 
> correctly.
> Essentially, the behaviour of the coherence controllers is specified in 
> the SLICC code, but some functionality is thrown in automatically.
This is now clear. So I guess I can't change those assumptions of SLICC. Say 
if
I want add more public methods for L1Cache_Entry, I can't change it in 
generated
file. But I should be able change the AbstractCacheEntry class. Also, after 
I
make the changes, I should recompile ruby like this, right?

make PROTOCOL=MOSI_SMP_bcast DESTINATION=MOSI_SMP_bcast

BTW, any idea how I can make dataBlk function correctly? Or is
there a release of GEMS that supports dataBlk? In my project I do need the
cache to contain actual data.

Thanks a lot,
Lei

> > ----- Original Message ----- From: "Dan Gibson" <degibson@xxxxxxxx>
> > To: "Lei Yang" <lya755@xxxxxxxxxxxxxxxxxxxx>; "Gems Users" 
> > <gems-users@xxxxxxxxxxx>
> > Sent: Thursday, September 07, 2006 11:55 AM
> > Subject: Re: [Gems-users] Generated files in ruby
> >
> >
> > > These files are generated directly from the protocol specification via 
> > > the SLICC utility.
> > > The files in $GEMS/protocols are used to specify protocol behaviour. The 
> > > SLICC compiler reads these specifications and produces the 
> > > generated/[protocol] directory which is used in the actual simulation.
> > > The reason why you're not finding separate files for the L1's and L2's 
> > > is that under the "SMP" protocols, the caches use a single controller 
> > > for coherence. Try one of the "CMP" protocols and you will find separate 
> > > L1 and L2 controllers.
> > > Regards,
> > > Dan Gibson
> > >
> > >
> > > Lei Yang wrote:
> > > > Dear list,
> > > >  In gems/ruby/generated, there are a lot of generated C and h files for 
> > > > a certain protocol, say MOSI_SMP_bcast. I wonder how these files are 
> > > > generated and what are the source files. For example, in 
> > > > L1Cache_Entry.C, class 'L1Cache_Entry' inherits from 
> > > > 'AbstractCacheEntry', which is in turn, defined in 
> > > > ruby/slicc_interface/AbstractCacheEntry.h . However, as I understand 
> > > > the MOSI_SMP_bcast protocol, in addition to L1I, L1D cache, each 
> > > > processor also owns a unified L2 cache. But there is no 
> > > > L2Cache_Entry.C/h file in ruby/generated. Since I want to do some 
> > > > modification to the L2 cache architecture, I need to modify the 
> > > > L2Cache_Entry class, if presents. More importantly I want to understand 
> > > > how these files are automatically generated, because in 
> > > > MOSI_SMP_bcast-cache.sm I can only see these lines related to 
> > > > L1Cache_Entry and L2Cache_Entry:
> > > >  CacheMemory L1IcacheMemory, template_hack="<L1Cache_Entry>", 
> > > > constructor_hack='L1_CACHE_NUM_SETS_BITS,L1_CACHE_ASSOC,MachineType_L1Cache,int_to_string(i)+"_L1I"', 
> > > > abstract_chip_ptr="true";
> > > >   CacheMemory L1DcacheMemory, template_hack="<L1Cache_Entry>", 
> > > > constructor_hack='L1_CACHE_NUM_SETS_BITS,L1_CACHE_ASSOC,MachineType_L1Cache,int_to_string(i)+"_L1D"', 
> > > > abstract_chip_ptr="true";
> > > >   CacheMemory L2cacheMemory,  template_hack="<L1Cache_Entry>", 
> > > > constructor_hack='L2_CACHE_NUM_SETS_BITS,L2_CACHE_ASSOC,MachineType_L1Cache,int_to_string(i)+"_L2"', 
> > > > abstract_chip_ptr="true";
> > > >  Your help will be greatly appreciated!!
> > > >  Thanks,
> > > > Lei
> > > > ------------------------------------------------------------------------
> > > >
> > > >
> > > > _______________________________________________
> > > > 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.
> > > --
> > > http://www.cs.wisc.edu/~gibson [esc]:wq!
> --
> http://www.cs.wisc.edu/~gibson [esc]:wq!