Interconnection Networks: Flow Control. Prof. Natalie Enright Jerger

Transcription

1 Interconnection Networks: Flow Control Prof. Natalie Enright Jerger

2 Switching/Flow Control Overview Topology: determines connectivity of network Routing: determines paths through network Flow Control: determine allocation of resources to messages as they traverse network Buffers and links Significant impact on throughput and latency of network 2

3 Flow Control Control State Channel Bandwidth Buffer capacity Control state records allocation of channels and buffers to packets current state of packet traversing node Channel bandwidth advances flits from this node to next Buffers hold flits waiting for channel bandwidth 3

4 Packets Messages: composed of one or more packets If message size is <= maximum packet size only one packet created Packets: composed of one or more flits Flit: flow control digit Phit: physical digit Subdivides flit into chunks = to link width 4

5 Packets (2) Message Header Payload Packet Route Seq# Head Flit Body Flit Tail Flit Flit Type VCID Head, Body, Tail, Head & Tail Phit Off-chip: channel width limited by pins Requires phits On-chip: abundant wiring means phit size == flit size 5

6 Packets(3) Cache line RC Type VCID Add r Bytes 0-15 Bytes Bytes Bytes Head Flit Body Flits Tail Flit Coherence Command RC Type VCID Add r Cmd Head & Tail Flit Packet contains destination/route information Flits may not all flits of a packet must take same route 6

7 Switching Different flow control techniques based on granularity Message-based: allocation made at message granularity (circuit-switching) Packet-based: allocation made to whole packets Flit-based: allocation made on a flit-by-flit basis 7

8 Message-Based Flow Control Coarsest granularity Circuit-switching Pre-allocates resources across multiple hops Source to destination Resources = links Buffers are not necessary Probe sent into network to reserve resources 8

9 Circuit Switching Once probe sets up circuit Message does not need to perform any routing or allocation at each network hop Good for transferring large amounts of data Can amortize circuit setup cost by sending data with very low perhop overheads No other message can use those resources until transfer is complete Throughput can suffer due setup and hold time for circuits Links are idle until setup is complete 9

10 Circuit Switching Example 0 Configuration Probe 5 Data Circuit Significant latency overhead prior to data transfer Data transfer does not pay per-hop overhead for routing and allocation Acknowledgement 10

11 Circuit Switching Example (2) Configuration Probe Data Circuit Acknowledgement When there is contention Significant wait time Message from 1 2 must wait 11

12 Location Time-Space Diagram: Circuit-Switching 0 S 08 A 08 D 08 D 08 D 08 D 08 T 08 1 S 08 S 08 A 08 D 08 D 08 D 08 D 08 T 08 2 S 28 A 08 D 08 D 08 D 08 D 08 T 08 S 28 5 S 08 S 08 A08 A 08 D 08 D 08 D 08 D 08 T 08 S 28 8 D 08 D 08 D 08 D 08 T Time to setup+ack circuit from 0 to 8 Time Time setup from 2 to 8 is blocked

13 Packet-based Flow Control Break messages into packets Interleave packets on links Better utilization Requires per-node buffering to store in-flight packets Two types of packet-based techniques 13

14 Store and Forward Links and buffers are allocated to entire packet Head flit waits at router until entire packet is received before being forwarded to the next hop Not suitable for on-chip Requires buffering at each router to hold entire packet Packet cannot traverse link until buffering allocated to entire packet Incurs high latencies (pays serialization latency at each hop) 14

15 Store and Forward Example 0 5 Total delay = 4 cycles per hop x 3 hops = 12 cycles High per-hop latency Serialization delay paid at each hop Larger buffering required 15

16 Location Time-Space Diagram: Store and Forward H B B B T H B B B T H B B B T H B B B T H B B B T Time

17 Packet-based: Virtual Cut Through Links and Buffers allocated to entire packets Flits can proceed to next hop before tail flit has been received by current router But only if next router has enough buffer space for entire packet Reduces the latency significantly compared to SAF But still requires large buffers Unsuitable for on-chip 17

18 Virtual Cut Through Allocate Example 4 flit-sized 0 Allocate 4 flit-sized buffers before buffers head before head proceeds proceeds 5 Total delay = 1 cycle per hop x 3 hops + serialization = 6 cycles Lower per-hop latency Large buffering required 18

19 Location Time-Space Diagram: VCT H B B B T H B B B T H B B B T 5 8 H B B B T H B B B T Time 19

20 Virtual Cut Through Cannot proceed because only 2 flit buffers available Throughput suffers from inefficient buffer allocation 20

21 Location Time-Space Diagram: VCT (2) 0 1 H B B B T H B B B T H B B B T Insufficient Buffers H B B B T H B B B T Time 21

22 Flit-Level Flow Control Help routers meet tight area/power constraints Flit can proceed to next router when there is buffer space available for that flit Improves over SAF and VCT by allocating buffers on a flit-by-flit basis 22

23 Wormhole Flow Control Pros More efficient buffer utilization (good for on-chip) Low latency Cons Poor link utilization: if head flit becomes blocked, all links spanning length of packet are idle Cannot be re-allocated to different packet Suffers from head of line (HOL) blocking 23

24 Wormhole Example Red holds this channel: channel remains idle until read proceeds Channel idle but red packet blocked behind blue Buffer full: blue cannot proceed 6 flit buffers/input port Blocked by other packets 24

25 Location Time-Space Diagram: Wormhole 0 1 H B B B T H B B B T 2 Contention H B B B T 5 H B B B T 8 H B B B T Time 25

26 Virtual Channels First proposed for deadlock avoidance We ll come back to this Can be applied to any flow control First proposed with wormhole 26

27 Virtual Channel Flow Control Virtual channels used to combat HOL blocking in wormhole Virtual channels: multiple flit queues per input port Share same physical link (channel) Link utilization improved Flits on different VC can pass blocked packet 27

28 Virtual Channel Flow Control (2) A (in) B (in) A (out) B (out) 28

29 Virtual Channel Flow Control (3) In1 AH A1 A2 A3 A4 A AT In2 BH B1 B2 B3 B4 B BT Out AH BH A1 B1 A2 B2 A3 B3 A4 B4 A5 B5 AT BT A Downstream B Downstream AH A1 A2 A3 A4 A5 AT BH B1 B2 B3 B4 B5 BT 29

30 Virtual Channel Flow Control (3) Packets compete for VC on flit by flit basis In example: on downstream links, flits of each packet are available every other cycle Upstream links throttle because of limited buffers Does not mean links are idle May be used by packet allocated to other VCs 30

31 Virtual Channel Example Buffer full: blue cannot proceed 6 flit buffers/input port 3 flit buffers/vc Blocked by other packets 31

32 Summary of techniques Circuit- Switching Store and Forward Virtual Cut Through Links Buffers Comments Messages N/A (buffer-less) Setup & Ack Packet Packet Head flit waits for tail Packet Packet Head can proceed Wormhole Packet Flit HOL Virtual Channel Flit Flit Interleave flits of different packets 32

33 Deadlock Using flow control to guarantee deadlock freedom give more flexible routing Recall: routing restrictions needed for deadlock freedom If routing algorithm is not deadlock free VCs can break resource cycle Each VC is time-multiplexed onto physical link Holding VC implies holding associated buffer queue Not tying up physical link resource Enforce order on VCs 33

34 Deadlock: Enforce Order C A0 A B A1 B1 B0 D D C D B 0 D 1 All message sent through VC 0 until cross dateline After dateline, assigned to VC 1 Cannot be allocated to VC 0 again 34 C1 A C0

35 Deadlock: Escape VCs Enforcing order lowers VC utilization Previous example: VC 1 underutilized Escape Virtual Channels Provide an escape path for every packet in a potential cycle Have 1 VC that uses a deadlock free routing subfunction Example: VC 0 uses DOR, other VCs use arbitrary routing function Access to VCs arbitrated fairly: packet always has chance of landing on escape VC Once in escape VC, must remain there (routing restrictions) Assign different message classes to different VCs to prevent protocol level deadlock Prevent req-ack message cycles 35

36 Virtual Channel Reallocation VC0 VC2 VC3 P3(H) P2(H) P3(B) P2(T) P3(T) R0 R1 R2 VC1 P1(H) P1(T) When can a VC be reallocated to a new packet? Two options: conservative and aggressive Conservative (atomic): only reallocate empty VC Aggressive (non-atomic): reallocate once tail flit has been received Implications for deadlock 36

37 VCT + non-atomic allocation Typically uses non-atomic VC allocation VC allocated only if there is enough free buffer space to hold entire packet Guarantees that whole packet can be received by VC Will entirely vacate upstream VC New packet at head of VC can bid for resources 37

38 VC Reallocation: Wormhole + Aggressive Turn-model and deterministic routing No cyclic channel dependence Can use aggressive reallocation VC0 VC2 VC3 P3(B) P2(H) R0 R1 R2 P1(T) P1(B) VC1 P1(H) P1(T) P1(B) P1(H) P3(T) P2(T) P1 can be forwarded immediately High VC utilization but low adaptivity 38

39 VC Reallocation: Wormhole + Adaptive Routing Adaptive routing Cyclic channel dependence With aggressive reallocation Packet spans 2 VCs and head flit not at head of one VC Cannot reach escape VC Leads to deadlock Requires conservative reallocation Can only reallocate an empty VC Low VC utilization R0 EVC 1 AVC 1 P 1 (B) P 1 (T) P 0 (H) P 0 (B) R2 R4 P 4 (B) P 4 (H) P 5 (T) P 5 (B) P2(B) P2(T) P1(H) P1(B) Dest(P 2 )=R3 Dest(P1)=R1 Dest(P5)=R5 EVC2 AVC2 R1 P 2 (B) P 2 (H) P 3 (T) P 3 (H) AVC 3 EVC 3 Dest(P 0 )=R0 AVC0 P0(B) EVC0 P4(B) P0(B) P0(B) P4(B) P4(B) EVC 4 AVC 4 Dest(P 6 )=R3 P5(H) P5(B) R3 Dest(P 4 )=R4 AVC 6 EVC 6 P 7 (H) P 7 (T) P6(B) P6(T) AVC5 EVC5 Dest(P3)=R2 P0(T) P4(T) Dest(P7)=R2 P 6 (H) P 6 (B) R5 39

40 Buffer Backpressure Need mechanism to prevent buffer overflow Avoid dropping packets Upstream nodes need to know buffer availability at downstream routers Significant impact on throughput achieved by flow control Two common mechanisms Credits On-off 40

41 Credit-Based Flow Control Upstream router stores credit counts for each downstream VC Upstream router When flit forwarded Decrement credit count Count == 0, buffer full, stop sending Downstream router When flit forwarded and buffer freed Send credit to upstream router Upstream increments credit count 41

42 Credit Timeline t1 t2 t3 Process Node 1 Node 2 Flit departs router Credit round trip delay t4 t5 Process Round-trip credit delay: Time between when buffer empties and when next flit can be processed from that buffer entry If only single entry buffer, would result in significant throughput degradation Important to size buffers to tolerate credit turn-around 42

43 On-Off Flow Control Credit: requires upstream signaling for every flit On-off: decreases upstream signaling Off signal Sent when number of free buffers falls below threshold F off On signal Sent when number of free buffers rises above threshold F on 43

44 t1 On-Off Timeline Node 1 Node 2 F off threshold reached F off set to prevent flits arriving before t4 from overflowing t2 t3 t4 Process F on set so that Node 2 does not run out of flits between t5 and t8 t5 t6 t7 t8 Process Less signaling but more buffering Fall 2014 On-chip buffers ECE 1749H: more Interconnection expensive Networks (Enright than Jerger) wires F on threshold reached 44

45 Buffer Sizing Prevent backpressure from limiting throughput Buffers must hold flits >= turnaround time Assume: 1 cycle propagation delay for data and credits 1 cycle credit processing delay 3 cycle router pipeline At least 6 flit buffers 45

46 Actual Buffer Usage & Turnaround Delay Actual buffer usage Credit propagation delay Credit pipeline delay flit pipeline delay flit propagation delay Flit arrives at node 1 and uses buffer Flit leaves node 1 and credit is sent to node 0 Node 0 receives credit Node 0 processes credit, freed buffer reallocated to new flit New flit leaves Node 0 for Node 1 New flit arrives at Node 1 and reuses buffer 46

47 Flow Control Summary On-chip networks require techniques with lower buffering requirements Wormhole or Virtual Channel flow control Avoid dropping packets in on-chip environment Requires buffer backpressure mechanism Complexity of flow control impacts router microarchitecture (next) 47