On Nov. 20, 2015, 1:33 p.m., Tony Gutierrez wrote:
> > I am still not really understanding the request and the packet part. How 
> > are the two related (are they)?
> 
> Tony Gutierrez wrote:
>     I have a patch to undo the packet ACQ/REL changes, and will post it soon. 
> Honestly I think the original author of this patch may have only added the 
> Command for ACQ/REL because they needed something to pass to the Packet() 
> ctor and no other Commands fit, at least that is my best guess.
> 
> Andreas Hansson wrote:
>     What is the big picture here? They are added as request flags (although 
> they are mutually exclusive with the other request flags)? They are not 
> appearing as packets, but there will be read/write packets that have requests 
> that are acquire/release? Please don't commit this until we have sorted out 
> all this confusion.
> 
> Tony Gutierrez wrote:
>     Yes, they are mutually exclusive with other request flags, i.e., all code 
> for ACQ/REQ requests in the memory system is unique to only requests with 
> isAcquire() or isRelease(). After creating the patch to remove ACQ/REL stuff 
> from the packet our code works ok with regular Read/Write Cmds, but I was 
> thinking it would make sense to have something like SerializingReq or 
> SynchronizingReq or FenceReq in the packet with its own unique attributes, 
> which could potentially be used for all sync/serializing requests across 
> ISAs. This could even potentially clean up some of this code, because as 
> noted before the Command/Attribute/Request Flags interface isn't very clearly 
> defined and everyone seems to have their own ideas as to what they mean.

That sounds great. I agree that it is a bit of a mess at the moment, and 
finding one common abstraction for all the ISAs would be a massive step in the 
right direction. Will you post a proposal once you've got something working?


- Andreas


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviews.gem5.org/r/3180/#review7622
-----------------------------------------------------------


On Nov. 20, 2015, 9:52 p.m., Tony Gutierrez wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://reviews.gem5.org/r/3180/
> -----------------------------------------------------------
> 
> (Updated Nov. 20, 2015, 9:52 p.m.)
> 
> 
> Review request for Default.
> 
> 
> Repository: gem5
> 
> 
> Description
> -------
> 
> Changeset 11225:89acf0bcf147
> ---------------------------
> mem: add request types for acquire and release
> 
> Add support for acquire and release requests.  These synchronization 
> operations
> are commonly supported by several modern instruction sets.
> 
> 
> Diffs
> -----
> 
>   src/mem/request.hh c0ea80fed78fef29ad2829b9d93e7bd568c46665 
> 
> Diff: http://reviews.gem5.org/r/3180/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Tony Gutierrez
> 
>

_______________________________________________
gem5-dev mailing list
gem5-dev@gem5.org
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to