On 07/24/11 10:21, nathan binkert wrote:
>> What about all the declarations that appear in
>> src/mem/protocol/RubySlicc_Exports.sm? Those are really common to all the
>> protocols and should be generated only once. Are you suggesting that these
>> be duplicated as well?
> They are now.  I'd love it if we could figure out how to not do that.
>
> Just so we're clear about what's going on right now.  Every single
> file is compiled for every single version of gem5
> (ALPHA_SE_MESI_CMP_directory, SPARC_FS, X86_SE).  There are currently
> 16 versions of gem5 in build_opts and that isn't an exhaustive list of
> what we should be testing.  Now just about all files in base and sim
> could care less about any of that stuff and should really only be
> compiled once.  The memory system (both classic and ruby) shouldn't
> care about the CPU model or the ISA and should only be compiled once.
> The CPU models shouldn't care about the ruby protocol.  In all of
> these instances, we're compiling lots of extra stuff.  The easiest
> thing to get rid of is the protocol thing.  Then FS vs. SE.  The ISA
> thing will probably be a lot of work to fix.  Compiling would be far
> faster if we got rid of all of this duplication.
>
>   Nate
> _______________________________________________
> gem5-dev mailing list
> [email protected]
> http://m5sim.org/mailman/listinfo/gem5-dev

I'm planning on making some headway on the FS vs. SE part. I'm hoping to
do a lot of integration instead of duplication, or in other words make
things flexible enough to do both, not to have two copies of everything
that can be compiled together.

Gabe
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to