1Lecture 3: Directory Protocol Implementations• Topics: coherence vs. msg-passing, corner cases indirectory protocols2Future Scalable Designs• Intel’s Single Cloud Computer (SCC): an example prototype• No support for hardware cache coherence• Programmer can write shared-memory apps by markingpages as uncacheable or L1-cacheable, but forcing memoryflushes to propagate results• Primarily intended for message-passing apps• Each core runs a version of Linux• Barrelfish-like OSes will likely soon be mainstream3Scalable Cache Coherence• Will future many-core chips forego hardware cache coherence in favor of message-passing or sw-managedcache coherence?• It’s the classic programmer-effort vs. hw-effort trade-off …traditionally, hardware has won (e.g. ILP extraction)• Two questions worth answering: will motivated programmersprefer message-passing?, is scalable hw cache coherencedo-able?4Message Passing• Message passing can be faster and more energy-efficient• Only required data is communicated: good for energy andreduces network contention• Data can be sent before it is required (push semantics;cache coherence is pull semantics and frequently requiresindirection to get data)• Downsides: more software stack layers and more memoryhierarchy layers must be traversed, and.. more programming effort5Scalable Directory Coherence• Note that the protocol itself need not be changed• If an application randomly accesses data with zero locality: long latencies for data communication also true for message-passing apps• If there is locality and page coloring is employed, the directoryand data-sharers will often be in close proximity• Does hardware overhead increase? See examples in last class… the overhead is ~2-10% and sharing can be tracked at coarse granularity… hierarchy can also be employed, with snooping-based coherence among a group of nodes6SGI Origin 2000• Flat memory-based directory protocol• Uses a bit vector directory representation• Two processors per node – combining multiple processorsin a node reduces costPL2CAM/DPL2Interconnect7Directory Structure• The system supports either a 16-bit or 64-bit directory(fixed cost); for small systems, the directory works as a full bit vector representation• Seven states, of which 3 are stable• For larger systems, a coarse vector is employed – eachbit represents p/64 nodes• State is maintained for each node, not each processor –the communication assist broadcasts requests to bothprocessors8Handling Reads• When the home receives a read request, it looks upmemory (speculative read) and directory in parallel• Actions taken for each directory state: shared or unowned: memory copy is clean, datais returned to requestor, state is changed to excl ifthere are no other sharers busy: a NACK is sent to the requestor exclusive: home is not the owner, request is fwdedto owner, owner sends data to requestor and home9Inner Details of Handling the Read• The block is in exclusive state – memory may or may nothave a clean copy – it is speculatively read anyway• The directory state is set to busy-exclusive and thepresence vector is updated• In addition to fwding the request to the owner, the memorycopy is speculatively forwarded to the requestor Case 1: excl-dirty: owner sends block to requestorand home, the speculatively sent data is over-written Case 2: excl-clean: owner sends an ack (without data)to requestor and home, requestor waits for this ackbefore it moves on with speculatively sent data10Inner Details II• Why did we send the block speculatively to the requestorif it does not save traffic or latency? the R10K cache controller is programmed to notrespond with data if it has a block in excl-clean state when an excl-clean block is replaced from the cache,the directory need not be updated – hence, directorycannot rely on the owner to provide data andspeculatively provides data on its own11Handling Write Requests• The home node must invalidate all sharers and allinvalidations must be acked (to the requestor), the requestor is informed of the number of invalidates to expect• Actions taken for each state: shared: invalidates are sent, state is changed toexcl, data and num-sharers are sent to requestor,the requestor cannot continue until it receives all acks(Note: the directory does not maintain busy state,subsequent requests will be fwded to new ownerand they must be buffered until the previous writehas completed)12Handling Writes II• Actions taken for each state: unowned: if the request was an upgrade and not aread-exclusive, is there a problem? exclusive: is there a problem if the request was anupgrade? In case of a read-exclusive: directory isset to busy, speculative reply is sent to requestor,invalidate is sent to owner, owner sends data torequestor (if dirty), and a “transfer of ownership”message (no data) to home to change out of busy busy: the request is NACKed and the requestormust try again13Handling Write-Back• When a dirty block is replaced, a writeback is generatedand the home sends back an ack • Can the directory state be shared when a writeback isreceived by the directory?• Actions taken for each directory state: exclusive: change directory state to unowned andsend an ack busy: a request and the writeback have crossedpaths: the writeback changes directory state toshared or excl (depending on the busy state),memory is updated, and home sends data torequestor, the intervention request is dropped14Writeback CasesP1 P2D3E: P1WbackThis is the “normal” caseD3 sends back an AckAck15Writeback CasesP1 P2D3E: P1 busyWbackIf someone else has the block in exclusive, D3 moves to busyIf Wback is received, D3 serves the requesterIf we didn’t use busy state when transitioning from E:P1 to E:P2, D3 may not have known who to service(since ownership may have been passed on to P3 and P4…)(although, this problem can be solved by NACKing the Wbackand having P1 buffer its “strange” intervention requests)FwdRd or Wr16Writeback CasesP1 P2D3E: P1 busyTransferownershipIf Wback is from new requester, D3 sends back a NACKFloating unresolved messages are a problemAlternatively, can accept the Wback and put D3 in some new busy stateConclusion: could have got rid of busy state between E:P1  E:P2, butwith Wback ACK/NACK and other bufferingcould have kept the busy state between E:P1  E:P2, couldhave got rid of