Computer Organization and Design /fj^ The Hardware/Software Interface ^hi

instruction memory, fetch instruction Register numbers -> register file, read registers Depending on instruction class ■ Use ALU to calculate Arithmetic result Memory address for load/store Branch target address ■ Access data memory for load/store ■ PC <- target address or PC + 4 M 14 Chapter 4 — The Processor — 3 CPU Overview Add PC pH Address Instruction Instruction memory Add. Data Register # Registers Register # W Register # Address Data memory Data M 14 Chapter 4 — The Processor — 4 Multiplexers Add PC Address Instruction Instruction memory Add. Can't just join wires together ■ Use multiplexers Data Register # Registers Register # W Register # Address Data memory Data Chapter 4 — The Processor — 5 Control M u x Branch Add > pc pH Address Instruction Instruction memory Add, Data Register # Registers Register # Data Zero ALU ALU result RegWrite b. ALU Chapter 4 — The Processor — 14 Load/Store Instructions Read register operands Calculate address using 16-bit offset ■ Use ALU, but sign-extend offset Load: Read memory and update register Store: Write reaister value to memorv a. Data memory unit b. Sign extension unit M 14 ® Chapter 4 — The Processor — 15 Branch Instructions Read register operands Compare operands ■ Use ALU, subtract and check Zero output Calculate target address ■ Sign-extend displacement ■ Shift left 2 places (word displacement) - Add to PC + 4 Already calculated by instruction fetch M 14 Chapter 4 — The Processor — 16 Branch Instructions Just re-routes wires PC+4 from instruction datapath Instruction Read register 1 Read Read data 1 register 2 Write Registers register Read Write data 2 data < ALU operation Branch target To branch control logic Sign-bit wire replicated Chapter 4 — The Processor Composing the Elements First-cut data path does an instruction in one clock cycle ■ Each datapath element can only do one function at a time ■ Hence, we need separate instruction and data memories Use multiplexers where alternate data sources are used for different instructions M 14 Chapter 4 — The Processor — 18 R-Type/Load/Store Datapath Instruction Read register 1 Read Read data 1 register 2 Registers Write register Read data 2 Write data ALU operation Address MemWrite MemtoReg Read data Write data Data memory 1 M u x MemRead M 14 Chapter 4 — The Processor — 19 Full Datapath pc \Add PCSrc _^ ■ , , ALU Add result M u x Read address Instruction Instruction memory 11—** Read register 1 Read data 1 Read register 2 Write Re9jsters Read ALUSrc 4 J ALU operation Zero ALU ALU result MemWrite Address Read data Write Data data memory MemRead MemtoReg M u x W Chapter 4 — The Processor — 20 ALU Control ALU used for ■ Load/Store: F = add ■ Branch: F = subtract ■ R-type: F depends on funct field ALU control Function 0000 AND 0001 OR 0010 add 0110 subtract 0111 set-on-less-than 1100 NOR M 14 Chapter 4 — The Processor — 21 ALU Control Assume 2-bit ALUOp derived from opcode ■ Combinational logic derives ALU control opcode ALUOp Operation funct ALU function ALU control Iw 00 load word xxxxxx add 0010 sw 00 store word xxxxxx add 0010 beq 01 branch equal xxxxxx subtract 0110 R-type 10 add 100000 add 0010 subtract 100010 subtract 0110 AND 100100 AND 0000 OR 100101 OR 0001 set-on-less-than 101010 set-on-less-than 0111 M 14 Chapter 4 — The Processor — 22 The Main Control Unit Control signals derived from instruction R-type Load/ Store Branch opcode read, except for load 0 rs rt rd shamt funct 31:26 25:21 20:16 U 5:11 10:6 5:0 35 or 43 rs rt address 31:26 25:21 20:16 \ \ 15:0 4 rs rt \ address 31:26 v. j 25:21 20:16 \ V J 15:0 \ V V write for R-type and load sign-extend and add M 14 Chapter 4 — The Processor — 23 Datapath With Control > Add PC Read address Instruction [31-0] Instruction memory Instruction [31-26] RegDst Branch Mjeft2/^ Add ALU result MemRead MemtoReg ALUOp MemWrite ALUSrc Instruction [25-21] Instruction [20-16] Instruction [15-11] M u X Instruction [15-0] < RegWrite Read register 1 Read Read data 1 register 2 Write register Read data 2 Write data Registers Instruction [5-0] M u x > Zero I—I ALU ALU Address Read data Write Data data memorV M u X 0 Chapter 4 — The Processor — 24 R-Type Instruction > Add PC Read address Instruction [31-0] Instruction memory > Instruction [31-26] RegDst Branch * ALU /'Add result ^/shiftV Hjeft^T MemRead MemtoReg ALUOp MemWrite ALUSrc Instruction [25-21] Instruction [20-16] Instruction [15-11] M u X Instruction [15-0] < RegWrite ZD_ Read register 1 Read Read data 1 register 2 Write register Read data 2 Write data Registers Instruction [5-0] M u x > Zero I—I ALU ALU Address Read data Write Data data memory M u X 0 Chapter 4 — The Processor — 25 Load Instruction > Add PC Read address Instruction [31-0] Instruction memory Instruction [31-26] RegDst Branch ^left^T > Add MemRead MemtoReg ALUOp MemWrite ALUSrc Instruction [25-21] Instruction [20-16] Instruction [15-11] M u X Instruction [15-0] < RegWrite Read register 1 Read Read data 1 register 2 Write register Read data 2 Write data Registers Instruction [5-0] ALU result (0* M u x > Zero I—I ALU ALU Address Read data Write Data data memory M u X Chapter 4 — The Processor — 26 Branch-on-Equal Instruction > Add PC Read address Instruction [31-0] Instruction memory Instruction [31-26] RegDst Branch Mjeft2/^ \wd ALU result MemRead MemtoReg ALUOp MemWrite ALUSrc Instruction [25-21] Instruction [20-16] i-•- Instruction [15-11] M u X Instruction [15-0] < RegWrite ZD_ Read register 1 Read Read data 1 register 2 Write register Read data 2 Write data Registers Instruction [5-0] (0* M u x > Zero I—I ALU ALU Address Read data Write Data data memory M u X Chapter 4 — The Processor — 27 Implementing Jumps Jump address 31:26 25:0 Jump uses word address Update PC with concatenation of - Top 4 bits of old PC ■ 26-bit jump address - 00 Need an extra control signal decoded from opcode Chapter 4 — The Processor — 28 Datapath With Jumps Added Instruction [25-0] /gh-A Jump address [31-0] "Mien'2^ > Add PC Read address Instruction [31-0] Instruction memory 26 28 PC+ 4 [31-28] Instruction [31-26] RegDst Jump Branch Control MemRead MemtoReg ALUOp MemWrite ALUSrc Reg Write Instruction [25-21] Instruction [20-16] Instruction [15-11] Instruction [15-0] < Read register 1 Read Read data 1 register 2 ^ Write reg i Read ster data2 U Write data Registers 16 Instruction [5-0] M u X (0^ > ZeroM lualu result M u x Address Read data Write Data data memory M u X Chapter 4 — The Processor — 29 Performance Issues Longest delay determines clock period ■ Critical path: load instruction ■ Instruction memory -» register file -» ALU -> data memory ->• register file Not feasible to vary period for different instructions Violates design principle ■ Making the common case fast We will improve performance by pipelining M 14 Chapter 4 — The Processor — 30 Pipelining Analogy ■ Pipelined laundry: overlapping execution ■ Parallelism improves performance Time Task order A B C D Time - Task order A B C D 6PM 7 8 9 10 11 12 1 2AM Q Q □ ffl □ 6PM 7 8 9 10 11 12 1 2AM □ 5 □ 1 o Four loads: Speedup = 8/3.5 = 2.3 Non-stop: ■ Speedup = 2n/0.5n + 1.5*4 = number of stages M 14 Chapter 4 — The Processor — 31 MIPS Pipeline Five stages, one step per stage 1. IF: Instruction fetch from memory 2. ID: Instruction decode & register read 3. EX: Execute operation or calculate address 4. MEM: Access memory operand 5. WB: Write result back to register M 14 Chapter 4 — The Processor — 32 Pipeline Performance Assume time for stages is ■ 10Ops for register read or write ■ 200ps for other stages Compare pipelined datapath with single-cycle datapath Instr Instr fetch Register read ALU op Memory access Register write Total time Iw 200ps 100 ps 200ps 200ps 100 ps 800ps sw 200ps 100 ps 200ps 200ps 700ps R-format 200ps 100 ps 200ps 100 ps 600ps beq 200ps 100 ps 200ps 500ps M 14 Chapter 4 — The Processor — 33 Pipeline Performance Program execution order (in instructions) Single-cycle (Tc= 800ps) Time 200 —i— 400 —i— 600 800 —i— 1000 1200 1400 1600 1800 lw $1, 100($0) lw $2, 200($0) lw $3, 300($0) Instruction fetch Reg ALU Data access Reg 800 ps Instruction fetch Reg ALU Data access Reg 800 ps Instruction fetch Program execution order (in instructions) Pipelined (Tc= 200ps) 800 ps Time 200 400 600 800 1000 1200 1400 lw $1, 100($0) Instruction fetch Reg ALU Data access Reg lw $2, 200($0) 200 ps Instruction fetch Reg ALU Data access Reg lw $3, 300($0) 200 ps Instruction fetch Reg ALU Data access Reg < 200 ps 200 ps 200 ps 200 ps 200 ps Chapter 4 — The Processor — 34 Pipeline Speedup If all stages are balanced ■ i.e., all take the same time ■ Time between instructionspipe|ined = Time between instructionsnonpipe|ined Number of stages If not balanced, speedup is less Speedup due to increased throughput ■ Latency (time for each instruction) does not decrease M 14 Chapter 4 — The Processor — 35 Pipelining and ISA Design MIPS ISA designed for pipelining ■ All instructions are 32-bits Easier to fetch and decode in one cycle c.f. x86:1- to 17-byte instructions ■ Few and regular instruction formats Can decode and read registers in one step ■ Load/store addressing Can calculate address in 3rd stage, access memory in 4th stage ■ Alignment of memory operands Memory access takes only one cycle M 14 Chapter 4 — The Processor — 36 Hazards Situations that prevent starting the next instruction in the next cycle Structure hazards ■ A required resource is busy Data hazard ■ Need to wait for previous instruction to complete its data read/write Control hazard ■ Deciding on control action depends on previous instruction Chapter 4 — The Processor — 37 Structure Hazards Conflict for use of a resource In MIPS pipeline with a single memory ■ Load/store requires data access ■ Instruction fetch would have to stall'for that cycle Would cause a pipeline "bubble" Hence, pipelined datapaths require separate instruction/data memories ■ Or separate instruction/data caches M 14 Chapter 4 — The Processor — 38 Data Hazards An instruction depends on completion of data access by a previous instruction - add $s0, $t0, $tl sub $t2, $s0, $t3 Time add $s0, $t0, $t1 sub $t2, $s0, $t3 200 —i— 400 10 >EX 600 800 1000 1200 1400 1600 i i I i i i * MEM WB ! LJ..J bubble/ C bubble/ C bubble/ ( bubble/ ( bubble bubble/ C bubble/ ( bubble/ ( bubble/ ( bubble ffl >EX MEM WB Chapter 4 — The Processor — 39 Forwarding (aka Bypassing) ■ Use result when it is computed ■ Don't wait for it to be stored in a register ■ Requires extra connections in the datapath Program execution order Time (in instructions) add $s0, $t0, $t1 sub $t2, $s0, $t3 M 14 Chapter 4 — The Processor — 40 Load-Use Data Hazard Can't always avoid stalls by forwarding ■ If value not computed when needed ■ Can't forward backward in time! Program execution order Time (in instructions) Iw $s0, 20($t1) 200 400 600 800 1000 1200 1400 —i-1-1-1-1-1-1— ID >EX MEM WB Cbubble) Cbubble) Cbubble)\ (/bubble) (/bubble) sub $t2, $s0, $t3 MEM WB Chapter 4 — The Processor — 41 Code Scheduling to Avoid Stalls Reorder code to avoid use of load result in the next instruction stall stall e for A = B + e; C B + F lw $tl, 0($t0) lw $tl, 0($t0) lw 4($t0) lw 4($t0) add $t3, $tl, rlw 8($t0) sw $t3, 12($t0) add $t3, $tl, lw 8($to;r sw $t3, 12($t0) add $t5, $ti, add $t5, $tl, sw $t5, 16($t0) sw $t5, 16($t0) 13 cycles 11 cycles Chapter 4 — The Processor — 42 Control Hazards Branch determines flow of control ■ Fetching next instruction depends on branch outcome ■ Pipeline can't always fetch correct instruction Still working on ID stage of branch In MIPS pipeline ■ Need to compare registers and compute target early in the pipeline ■ Add hardware to do it in ID stage M 14 Chapter 4 — The Processor — 43 Stall on Branch Wait until branch outcome determined before fetching next instruction Program execution order (in instructions) Time 200 —I— 400 600 800 1000 1200 1400 add $4, $5, $6 beq $1, $2, 40 or $7, $8, $9 Instruction fetch Reg ALU Data access Reg Instruction fetch Reg ALU Data access Reg -► 200 ps 400 ps T T Reg ALU Data access Reg Chapter 4 — The Processor Branch Prediction Longer pipelines can't readily determine branch outcome early ■ Stall penalty becomes unacceptable Predict outcome of branch ■ Only stall if prediction is wrong In MIPS pipeline ■ Can predict branches not taken ■ Fetch instruction after branch, with no delay M 14 Chapter 4 — The Processor — 45 MIPS with Predict Not Taken Prediction correct Program execution order (in instructions) Time 200 400 600 800 1000 1200 1400 add $4, $5, $6 Instruction fetch Reg ALU Data access Reg beq $1, $2, 40 Instruction Reg ALU Data Reg 200 ps fetch access Instruction fetch Data access Iw $3, 300{$0) 200 ps Reg ALU Reg Prediction incorrect Program execution order (in instructions) Time 200 400 600 800 1000 1200 1400 add $4, $5, $6 beq $1, $2, 40 Instruction fetch Reg ALU Data access Reg Instruction Reg ALU Data Reg 200 ps fetch access or $7, $8, $9 400 ps Instruction fetch Reg ALU Data access Reg Chapter 4 — The Processor — 46 More-Realistic Branch Prediction Static branch prediction ■ Based on typical branch behavior ■ Example: loop and if-statement branches Predict backward branches taken Predict forward branches not taken Dynamic branch prediction ■ Hardware measures actual branch behavior e.g., record recent history of each branch ■ Assume future behavior will continue the trend When wrong, stall while re-fetching, and update history M 14 Chapter 4 — The Processor — 47 Pipeline Summary The BIG Picture Pipelining improves performance by increasing instruction throughput ■ Executes multiple instructions in parallel ■ Each instruction has the same latency Subject to hazards ■ Structure, data, control Instruction set design affects complexity of pipeline implementation Chapter 4 — The Processor — 48 MIPS Pipelined Datapath WB: Write back Right-to-left flow leads to hazards Chapter 4 — The Processor — 49 Pipeline registers Need registers between stages ■ To hold information produced in previous cycle 0 i M u PC x Address Instruction memory Read register 1 Read data 1 Read register 2 Registers Read Write data 2 register Write data I Shift \ I left 2 I-*~ result 16 ^ f Sign- \ 3f \ * í extend ^~ M u x Salu ALU result Read Address data Data memory Write data (A M M 14 Chapter 4 — The Processor — 50 Pipeline Operation Cycle-by-cycle flow of instructions through the pipelined datapath ■ "Single-clock-cycle" pipeline diagram Shows pipeline usage in a single cycle Highlight resources used ■ c.f. "multi-clock-cycle" diagram Graph of operation over time We'll look at "single-clock-cycle" diagrams for load & store M 14 Chapter 4 — The Processor — 51 IF for Load, Store, lw Instruction fetch M 14 Chapter 4 — The Processor — 52 ID for Load, Store, ... M 14 Chapter 4 — The Processor — 53 EX for Load lw Execution M 14 Chapter 4 — The Processor — 54 MEM for Load lw Memory M 14 Chapter 4 — The Processor — 55 WB for Load lw Write back 0 1 M u PC X Instruction memory Wrong register number Read register 1 Read data 1 Head r°qister 2 Registers Read Write data 2 regis+jr Write data 16 / Sign- I \ ^ 1 extend 32 -V- Read Address data Data memory Write data M 14 Chapter 4 — The Processor — 56 Corrected Datapath for Load M u ^ X Address Instruction memory Head register 1 Read data 1 Read register 2 Registers Read Write data 2 register Write data 1*/ Sign- \ 32 \ extend /~\ >Add (am) f ,esu~ I left 2 J- MEM/WB Read Address data Data memory Write data M 14 Chapter 4 — The Processor — 57 EX for Store sw Execution MEM/WB Chapter 4 — The Processor — 58 MEM for Store sw Memory r0 1 M u PC X Instruction memory Read register 1 Read data 1 Read register 2 Registers Read Write data 2 register Write data '6 / Sign- 1 -X * I extend Read Address data Data memory Write data Chapter 4 — The Processor — 59 WB for Store sw Write-back IF/ID ID/EX EX/MEM 1^ M U H Instruction memory Read register 1 Read data 1 Read register 2 Registers Head Write data 2 register Write data 16 Sign- extend MEM/WB Read Address data □ate memory Write data M M 14 Chapter 4 — The Processor — 60 Multi-Cycle Pipeline Diagram Form showing resource usage Time (in clock cycles) ->■ CC1 CC2 CC3 CC4 CCS CC 6 CC 7 CC 8 CC 9 Program execution order (in instructions) lw$10, 20($1) sub $11, $2, $3 add $12, $3, $4 lw$13, 24($1) add $14, $5, $6 IM < w -■nReg _>ALU^ -jJDM — im I--^Regp >ALU^ IM [—I pReg_ -Regi DM 1m]—4 Regp -J- UM " -■fieg —^egi r DM r n ■DM ^egj Chapter 4 — The Processor — 61 Cycle Pipeline Diagram Traditional form Time (in clock cycles) -» CC1 CC2 CC3 CC4 CC5 CC6 CC7 CC8 CC9 Program execution order (in instructions) lw$10, 20($1) Instruction fetch Instruction decode Execution Data access Write back sub $11, $2, $3 Instruction fetch Instruction decode Execution Data access Write back add $12, $3, $4 Instruction fetch Instruction decode Execution Data access Write back lw$13, 24($1) Instruction fetch Instruction decode Execution Data access Write back add $14, $5, $6 Instruction fetch Instruction decode Execution Data access Write back Chapter 4 — The Processor — 62 90 Single-Cycle Pipeline Diagram State of pipeline in a given cycle add $14, $5, $6 Iw $13, 24 ($1) add $12, $3, $4 sub $11, $2, $3 lw$10, 20($1) Instruction fetch Instruction decode Execution Memory Write-back ro 1 M u PC X KU Instruction memory IF/ID Read register 1 Read datal Read register 2 Registers Head Write data 2 register Write data id/ex I Shift \ I left 2 I-*■ >Add A* ' result ■~^U EX/MEM . Zero >LU ALU mem/wb Read Address data Data memory Write data \U M 14 Chapter 4 — The Processor — 63 Pipelined Control (Simplified) MEM/WB M 14 Chapter 4 — The Processor — 64 Pipelined Control ■ Control signals derived from instruction ■ As in single-cycle implementation Instruction WB IF/ID ID/EX EX/MEM MEIWWB M 14 Chapter 4 — The Processor — 65 Pipelined Control PCSrc (0s M u PC X Address Instruction memory ID/EX EX/MEM Instruction [15-0] 1 Instruction [20-16] Instruction [15-11] Read reqister 1 Read data 1 Read register 2 w.te Registers register aa,a d Write data Branch MEM/WB Read Address data Data memory Write data Mem Read /0\ Chapter 4 — The Processor — 66 Data Hazards in ALU Instructions Consider this sequence: sub $2, $1,$3 and $12,$2,$5 or $13,$6,$2 add $14,$2,$2 sw $15,100($2) We can resolve hazards with forwarding ■ How do we detect when to forward? M 14 Chapter 4 — The Processor — Dependencies & Forwarding Time (in clock cycles) - Value of cc 1 cc 2 cc 3 cc 4 cc 5 cc 6 cc 7 cc 8 cc 9 register $2: 10 10 10 10 10/-20 -20 -20 -20 -20 Program execution order Detecting the Need to Forward Pass register numbers along pipeline ■ e.g., ID/EX.RegisterRs = register number for Rs sitting in ID/EX pipeline register ALU operand register numbers in EX stage are given by - ID/EX.RegisterRs, ID/EX.RegisterRt Data hazards when 1a. EX/MEM.RegisterRd = ID/EX.RegisterRs 1b. EX/MEM.RegisterRd = ID/EX.RegisterRt 2a. MEM/WB.RegisterRd = ID/EX.RegisterRs 2b. MEM/WB.RegisterRd = ID/EX.RegisterRt Fwd from EX/MEM pipeline reg Fwd from MEM/WB pipeline reg < Chapter 4 — The Processor — 69 Detecting the Need to Forward But only if forwarding instruction will write to a register! . EX/MEM.RegWrite, MEM/WB. Reg Write And only if Rd for that instruction is not $zero . EX/MEM.RegisterRd t 0, MEM/WB.RegisterRd#0 M 14 Chapter 4 — The Processor — 70 Forwarding Paths ID/EX EX/MEM MEM/WB Registers Rs M u X ForwardA M u X Y ForwardB M u X ALU Forwarding V«_ unit /*■ Data memory EX/MEM. RegisterRd MEM/WB.RegisterRd M u X b. With forwarding < Chapter 4 — The Processor — 71 Forwarding Conditions EX hazard . if (EX/MEM.RegWrite and (EX/MEM.RegisterRd * 0) and (EX/MEM.RegisterRd = ID/EX.RegisterRs)) ForwardA = 10 . if (EX/MEM.RegWrite and (EX/MEM.RegisterRd * 0) and (EX/MEM.RegisterRd = ID/EX.RegisterRt)) Forward B = 10 MEM hazard . if (MEM/WB.RegWrite and (MEM/WB.RegisterRd * 0) and (MEM/WB.RegisterRd = ID/EX.RegisterRs)) ForwardA = 01 . if (MEM/WB.RegWrite and (MEM/WB.RegisterRd * 0) and (MEM/WB.RegisterRd = ID/EX.RegisterRt)) ForwardB = 01 M 14 Chapter 4 — The Processor — 72 Double Data Hazard Consider the sequence: add $l,$l,$2 add $1,$1,$3 add $1,$1,$4 Both hazards occur ■ Want to use the most recent Revise MEM hazard condition ■ Only fwd if EX hazard condition isn't true M 14 Chapter 4 — The Processor — 73 Revised Forwarding Condition MEM hazard . if (MEM/WB.RegWrite and (MEM/WB.RegisterRd * 0) and not (EX/MEM.RegWrite and (EX/MEM.RegisterRd * 0) and (EX/MEM.RegisterRd = ID/EX.RegisterRs)) and (MEM/WB.RegisterRd = ID/EX.RegisterRs)) Forward A = 01 . if (MEM/WB.RegWrite and (MEM/WB.RegisterRd * 0) and not (EX/MEM.RegWrite and (EX/MEM.RegisterRd * 0) and (EX/MEM.RegisterRd = ID/EX.RegisterRt)) and (MEM/WB.RegisterRd = ID/EX.RegisterRt)) ForwardB = 01 M 14 Chapter 4 — The Processor — 74 Datapath with Forwarding ID/EX MEM/WB PC- Instruction memory < Chapter 4 — The Processor — 75 Load-Use Data Hazard Time (in clock cycles) CC1 CC2 CC3 CC4 CC 5 CC6 CC 7 CC 8 CC 9 Program execution order (in instructions) Iw $2, 20($1) and $4, $2, $5 or $8, $2, $6 add $9, $4, $2 sit $1, $6, $7 < IM n ^Reg Need to stall for one cycle MRegi Chapter 4 — The Processor — 76 Load-Use Hazard Detection Check when using instruction is decoded in ID stage ALU operand register numbers in ID stage are given by . IF/ID.RegisterRs, IF/ID.RegisterRt Load-use hazard when - ID/EX. Mem Read and ((ID/EX.RegisterRt = IF/ID.RegisterRs) or (ID/EX.RegisterRt = IF/ID.RegisterRt)) If detected, stall and insert bubble M 14 Chapter 4 — The Processor — 77 How to Stall the Pipeline Force control values in ID/EX register toO ■ EX, MEM and WB do nop (no-operation) Prevent update of PC and IF/ID register ■ Using instruction is decoded again ■ Following instruction is fetched again ■ 1-cycle stall allows MEM to read data for lw Can subsequently forward to EX stage M 14 Chapter 4 — The Processor — 78 Stall/Bubble in the Pipeline Time {in clock cycles)-+~ CC1 CC2 CC 3 CC4 CC 5 CC 6 CC 7 CC 8 CC 9 CC10 Program execution order (in instructions) lw$2, 20($1) and becomes nop and $4, $2, $5 or $8, $2, $6 add $9, $4, $2 IM ^Reg IM Stall inserted here DM Reg Chapter 4 — The Processor — 79 Stall/Bubble in the Pipeline Time (in clock cycles)- CC 1 CC2 CC3 CC4 CC 5 CC 6 CC 7 CC 8 CC 9 CC 10 Program execution order (in instructions) Iw $2, 20($1) IM - and becomes nop IM--RRe> tJRe^ and $4, $2, $5 stalled in ID or $8, $2, $6 stalled in IF r-i-|DM|—I ^Reg IM - Reg, n bubble U ^Reg L _ add $9, $4, $2 Or, more accurately. IM - Chapter 4 — The Processor — Datapath with Hazard Detection Q. PC Instruction memory IR ID J Hazard detection A unit ID/EX.MemRead Registers IF/ID.RegisterRs IF/ID.RegisterRt IF/ID.RegisterRt IF/ID.RegisterRd ID/EX. RegisterRt EX/MEM M u X M u BL Rs V M u X WB M Rt Forwarding unit MEM/WB Data memory WB M u X Chapter 4 — The Processor — 81 Stalls and Performance The BIG Picture Stalls reduce performance ■ But are required to get correct results Compiler can arrange code to avoid hazards and stalls ■ Requires knowledge of the pipeline structure M 14 Chapter 4 — The Processor — 82 Branch Hazards If branch outcome determined in MEM Time (in clock cycles) CC1 CC 2 CC 3 CC 4 CC 5 CC 6 CC 7 CC 8 CC 9 Program execution order (in instructions) 40 beq $1, $3, 28 w 44 and $12, $2, $5 48 orSl3, $6, $2 52 add $14, $2, $2 72 Iw $4, 50($7) 3Reg r-p IM 3Reg |-|-|dm"—i ~Jm]--°^eg^ im — -Reg' Dm[— —Reg Reg__> im PC 3 Reg Jjm|--Reg| DM — Flush these instructions (Set control values to 0) J Regj Chapter 4 — The Processor — 83 Reducing Branch Delay Move hardware to determine outcome to ID stage ■ Target address adder ■ Register comparator Example: branch taken 36: sub $10, $4, $8 40: beq $1, $3, 7 44: and $12, $2, $5 48: or $13, $2, $6 52: : add $14, $4, $2 56: : sit $15, $6, $7 72: ■ ■ ■ : lw $4, 50($7) M 14 Chapter 4 — The Processor — 84 Example: Branch Taken before<2> Chapter 4 — The Processor — 85 Example: Branch Taken if.FIush Iw $4, 50{$7) Bubble (nop) M u PC X . J 76 IF, 76 72 Instruction ^mer^^l ,ID IH- Control L- 73 V___> Hazard detection unit Clock 4 Hegisters beq $1, $3, 7 sub $10,... I before<1> M ID/EX wb|- EX M M EX/^EM .11 \ )ALU $3 10 Forwarding unit MEM/WB Data memory Chapter 4 — The Processor — 86 Data Hazards for Branches If a comparison register is a destination of 2nd or 3rd preceding ALU instruction add $1, $2, $3 add $4, $5, $6 IF ID EX MEM WB IF ID EX rVIEM WB IF ID beq $1, $4, target IF ID> MEM WB EX MEM WB Can resolve using forwarding Chapter 4 — The Processor — 87 Data Hazards for Branches If a comparison register is a destination of preceding ALU instruction or 2nd preceding load instruction ■ Need 1 stall cycle lw L, addr |_f add $4, $5, $6 beq stalled beq $1, $4, target ID IF EX ID IF MEM EX ID WB MEM npj| WB \o\ o pi EX MEM WB Chapter 4 — The Processor — 88 Data Hazards for Branches If a comparison register is a destination of immediately preceding load instruction ■ Need 2 stall cycles lw $1, addr if beq stalled beq stalled beq $1, $0, target id if EX MEM u WB id id CP o o id EX MEM WB Chapter 4 — The Processor — 89 Dynamic Branch Prediction In deeper and superscalar pipelines, branch penalty is more significant Use dynamic prediction ■ Branch prediction buffer (aka branch history table) ■ Indexed by recent branch instruction addresses ■ Stores outcome (taken/not taken) ■ To execute a branch Check table, expect the same outcome Start fetching from fall-through or target If wrong, flush pipeline and flip prediction M 14 Chapter 4 — The Processor — 90 1-Bit Predictor: Shortcoming Inner loop branches mispredicted twice! outer: inner: beq ..., ..., inner beq ..., ..., outer Mispredict as taken on last iteration of inner loop Then mispredict as not taken on first iteration of inner loop next time around Chapter 4 — The Processor — 91 2-Bit Predictor Only change prediction on two successive mispredictions Calculating the Branch Target Even with predictor, still need to calculate the target address ■ 1 -cycle penalty for a taken branch ' Branch target buffer ■ Cache of target addresses ■ Indexed by PC when instruction fetched If hit and instruction is branch predicted taken, can fetch target immediately M 14 Chapter 4 — The Processor — 93 Exceptions and Interrupts "Unexpected" events requiring change in flow of control ■ Different ISAs use the terms differently Exception ■ Arises within the CPU e.g., undefined opcode, overflow, syscall, ... Interrupt ■ From an external I/O controller Dealing with them without sacrificing performance is hard M 14 Chapter 4 — The Processor — Handling Exceptions In MIPS, exceptions managed by a System Control Coprocessor (CPO) Save PC of offending (or interrupted) instruction ■ In MIPS: Exception Program Counter (EPC) Save indication of the problem ■ In MIPS: Cause register ■ We'll assume 1-bit 0 for undefined opcode, 1 for overflow Jump to handler at 8000 00180 M 14 Chapter 4 — The Processor — 95 An Alternate Mechanism Vectored Interrupts ■ Handler address determined by the cause Example: - Undefined opcode: C000 0000 . Overflow: C000 0020 ■ C000 0040 Instructions either ■ Deal with the interrupt, or ■ Jump to real handler M 14 Chapter 4 — The Processor — 96 Handler Actions Read cause, and transfer to relevant handler Determine action required If restartable ■ Take corrective action ■ use EPC to return to program Otherwise ■ Terminate program ■ Report error using EPC, cause, ... M 14 Chapter 4 — The Processor — 97 Exceptions in a Pipeline Another form of control hazard Consider overflow on add in EX stage add $1, $2, $1 ■ Prevent $1 from being clobbered ■ Complete previous instructions ■ Flush add and subsequent instructions ■ Set Cause and EPC register values ■ Transfer control to handler Similar to mispredicted branch ■ Use much of the same hardware M 14 Chapter 4 — The Processor — 98 Pipeline with Exceptions Chapter 4 — The Processor — 99 Exception Properties Restartable exceptions ■ Pipeline can flush the instruction ■ Handler executes, then returns to the instruction Refetched and executed from scratch PC saved in EPC register ■ Identifies causing instruction ■ Actually PC + 4 is saved Handler must adjust M 14 Chapter 4 — The Processor — 100 Exception Example Exception on add in 40 sub $11, $2, $4 44 and $12, $2, $5 48 or $13, $2, $6 4C add $1, $2, $1 50 sit $15, $6, $7 54 lw $16, 50($7) Handler 80000180 sw $25, 1000($0) 80000184 sw $26, 1004($0) M 14 Chapter 4 — The Processor — 101 Exception Example lw $16, 50($7) IF. Flush Sit $15, $6, $7 add $1, $2, $1 EX.FIush Clock 6 Chapter 4 — The Processor — 102 Exception Example sw$25, 1000($0) .Flush bubble (nop) bubble EX.FIush bubble or $13, . Clock 7 Chapter 4 — The Processor — 103 Multiple Exceptions Pipelining overlaps multiple instructions ■ Could have multiple exceptions at once Simple approach: deal with exception from earliest instruction ■ Flush subsequent instructions ■ "Precise" exceptions In complex pipelines ■ Multiple instructions issued per cycle ■ Out-of-order completion ■ Maintaining precise exceptions is difficult! M 14 Chapter 4 — The Processor — 104 Imprecise Exceptions Just stop pipeline and save state ■ Including exception cause(s) Let the handler work out ■ Which instruction(s) had exceptions ■ Which to complete or flush May require "manual" completion Simplifies hardware, but more complex handler software Not feasible for complex multiple-issue out-of-order pipelines M 14 Chapter 4 — The Processor — 105 Instruction-Level Parallelism (ILP) Pipelining: executing multiple instructions in parallel To increase ILP ■ Deeper pipeline Less work per stage =^> shorter clock cycle ■ Multiple issue Replicate pipeline stages =^> multiple pipelines Start multiple instructions per clock cycle CPI < 1, so use Instructions Per Cycle (IPC) E.g., 4GHz 4-way multiple-issue 16 BIPS, peak CPI = 0.25, peak IPC = 4 But dependencies reduce this in practice M 14 Chapter 4 — The Processor Multiple Issue Static multiple issue ■ Compiler groups instructions to be issued together ■ Packages them into "issue slots" ■ Compiler detects and avoids hazards Dynamic multiple issue ■ CPU examines instruction stream and chooses instructions to issue each cycle ■ Compiler can help by reordering instructions ■ CPU resolves hazards using advanced techniques at runtime M 14 Chapter 4 — The Processor — 107 Speculation "Guess" what to do with an instruction ■ Start operation as soon as possible ■ Check whether guess was right If so, complete the operation If not, roll-back and do the right thing Common to static and dynamic multiple issue Examples ■ Speculate on branch outcome Roll back if path taken is different ■ Speculate on load Roll back if location is updated M 14 Chapter 4 — The Processor — 108 Compiler/Hardware Speculation Compiler can reorder instructions ■ e.g., move load before branch ■ Can include "fix-up" instructions to recover from incorrect guess Hardware can look ahead for instructions to execute ■ Buffer results until it determines they are actually needed ■ Flush buffers on incorrect speculation M 14 Chapter 4 — The Processor — 109 Speculation and Exceptions What if exception occurs on a speculatively executed instruction? ■ e.g., speculative load before null-pointer check Static speculation ■ Can add ISA support for deferring exceptions Dynamic speculation ■ Can buffer exceptions until instruction completion (which may not occur) M 14 Chapter 4 — The Processor — 110 Static Multiple Issue Compiler groups instructions into "issue packets" ■ Group of instructions that can be issued on a single cycle ■ Determined by pipeline resources required Think of an issue packet as a very long instruction ■ Specifies multiple concurrent operations ■ => Very Long Instruction Word (VLIW) M 14 Chapter 4 — The Processor — 111 Scheduling Static Multiple Issue Compiler must remove some/all hazards ■ Reorder instructions into issue packets ■ No dependencies with a packet ■ Possibly some dependencies between packets Varies between ISAs; compiler must know! ■ Pad with nop if necessary M 14 Chapter 4 — The Processor — 112 MIPS with Static Dual Issue Two-issue packets ■ One ALU/branch instruction ■ One load/store instruction ■ 64-bit aligned ALU/branch, then load/store Pad an unused instruction with nop Address Instruction type Pipeline Stages n ALU/branch IF ID EX MEM WB n + 4 Load/store IF ID EX MEM WB n + 8 ALU/branch IF ID EX MEM WB n + 12 Load/store IF ID EX MEM WB n + 16 ALU/branch IF ID EX MEM WB n + 20 Load/store IF ID EX MEM WB M 14 Chapter 4 — The Processor — 113 MIPS with Static Dual Issue 80000180 M u PC X Chapter 4 — The Processor — 114 Hazards in the Dual-Issue MIPS More instructions executing in parallel EX data hazard ■ Forwarding avoided stalls with single-issue ■ Now can't use ALU result in load/store in same packet add $t0, $s0, $sl load $s2, 0($t0) Split into two packets, effectively a stall Load-use hazard ■ Still one cycle use latency, but now two instructions More aggressive scheduling required Chapter 4 — The Processor — 115 Scheduling Example Schedule this for dual-issue MIPS Loop: lw $t0, 0($sl) addu $t0, $t0, $s2 sw $t0, 0($sl) addi $sl, $sl,-4 # $tO=array element # add scalar in $s2 # store result # decrement pointer bne L, $zero, Loop # branch $sl!=0 ALU/branch Load/store cycle Loop: noD lw $t0, 0($sl) 1 addi $sl, $sl,-4 nop 2 addu $t0, $t0, $s2 nop 3 bne L, $zero, Loop sw $t0, 4($sl) 4 IPC = 5/4 = 1.25 (c.f. peak IPC = 2) < Chapter 4 — The Processor — 116 Loop Unrolling Replicate loop body to expose more parallelism ■ Reduces loop-control overhead Use different registers per replication ■ Called "register renaming" ■ Avoid loop-carried "anti-dependencies" Store followed by a load of the same register Aka "name dependence" Reuse of a register name M 14 Chapter 4 — The Processor — 117 Loop Unrolling Example ALU/branch Load/store cycle Loop: addi $sl, $sl,-16 lw $t0, 0($sl) 1 nop lw $tl, 12($sl) 2 addu $t0, $t0, $s2 lw $t2, 8($sl) 3 addu $tl, $tl, $s2 lw $t3, 4($sl) 4 addu $t2, $t2, $s2 sw $t0, 16($sl) 5 addu $t3, $t4, $s2 sw $tl, 12($sl) 6 sw $t2, 8($sl) 7 bne L, $zero, Loop sw $t3, 4($sl) 8 IPC = 14/8 = 1.75 ■ Closer to 2, but at cost of registers and code size M 14 Chapter 4 — The Processor — 118 Dynamic Multiple Issue "Superscalar" processors CPU decides whether to issue 0, 1,2, ... each cycle ■ Avoiding structural and data hazards Avoids the need for compiler scheduling ■ Though it may still help ■ Code semantics ensured by the CPU M 14 Chapter 4 — The Processor — 119 Dynamic Pipeline Scheduling Allow the CPU to execute instructions out of order to avoid stalls ■ But commit result to registers in order Example lw $t0, 20($s2) addu $tl, $t0, $t2 sub $s4, $s4, $t3 slti $t5, $s4, 20 ■ Can start sub while addu is waiting for lw M 14 Chapter 4 — The Processor — 120 Dynamically Scheduled CPU Instruction fetch and decode unit Ii 1 Reservation station Reservation station Reservation station Reservation station In-order issue Functional units Integer Integer J Floating point Out-of-order execute Reorders buffer for register writes Commit unit In-order commit Can supply operands for issued instructions Preserves dependencies Hold pending operands Results also sent to any waiting reservation stations Chapter 4 — The Processor — 121 Register Renaming Reservation stations and reorder buffer effectively provide register renaming On instruction issue to reservation station ■ If operand is available in register file or reorder buffer Copied to reservation station No longer required in the register; can be overwritten ■ If operand is not yet available It will be provided to the reservation station by a function unit Register update may not be required Chapter 4 — The Processor — 122 Speculation Predict branch and continue issuing ■ Don't commit until branch outcome determined Load speculation ■ Avoid load and cache miss delay Predict the effective address Predict loaded value Load before completing outstanding stores Bypass stored values to load unit ■ Don't commit load until speculation cleared M 14 Chapter 4 — The Processor — 123 Why Do Dynamic Scheduling? Why not just let the compiler schedule code? Not all stalls are predicable ■ e.g., cache misses Can't always schedule around branches ■ Branch outcome is dynamically determined Different implementations of an ISA have different latencies and hazards M 14 Chapter 4 — The Processor — 124 Does Multiple Issue Work? The BIG Picture Yes, but not as much as we'd like Programs have real dependencies that limit ILP Some dependencies are hard to eliminate ■ e.g., pointer aliasing Some parallelism is hard to expose ■ Limited window size during instruction issue Memory delays and limited bandwidth ■ Hard to keep pipelines full Speculation can help if done well Chapter 4 — The Processor — 125 Power Efficiency Complexity of dynamic scheduling and speculations requires power Multiple simpler cores may be better Microprocessor Year Clock Rate Pipeline Stages Issue width Out-of-order/ Speculation Cores Power i486 1989 25MHz 5 1 No 1 5W Pentium 1993 66MHz 5 2 No 1 10W Pentium Pro 1997 200MHz 10 3 Yes 1 29W P4 Willamette 2001 2000MHz 22 3 Yes 1 75W P4 Prescott 2004 3600MHz 31 3 Yes 1 103W Core 2006 2930MHz 14 4 Yes 2 75W UltraSparc III 2003 1950MHz 14 4 No 1 90W UltraSparcTI 2005 1200MHz 6 1 No 8 70W M 14 Chapter 4 — The Processor — 126 The Opteron X4 Microarchitecture Instruction cache Branch prediction Instruction prefetch and decode T RISC-operation queue I Dispatch and register remaining Integer and floating-point operation queue Integer ALU. Multiplier Integer ALU Integer ALU Floating point Adder /SSE Floating point Multiplier /SSE Load/Store queue Data cache < Commit unit 72 physical registers Chapter 4 — The Processor — 127 The Opteron X4 Pipeline Flow For integer operations Instruction Fetch II Number of clock cycles Decode and translate queue Reorder f buffer allocation + register renaming 1 buffer — —► 1- Scheduling + dispatch unit Execution Data Cache/I Commit ■ FP is 5 stages longer ■ Up to 106 RISC-ops in progress Bottlenecks ■ Complex instructions with long dependencies ■ Branch mispredictions ■ Memory access delays Chapter 4 — The Processor — 128 Fallacies Pipelining is easy (!) ■ The basic idea is easy ■ The devil is in the details e.g., detecting data hazards Pipelining is independent of technology ■ So why haven't we always done pipelining? ■ More transistors make more advanced techniques feasible ■ Pipeline-related ISA design needs to take account of technology trends e.g., predicated instructions M 14 Chapter 4 — The Processor — 1 Pitfalls Poor ISA design can make pipelining harder ■ e.g., complex instruction sets (VAX, IA-32) Significant overhead to make pipelining work IA-32 micro-op approach ■ e.g., complex addressing modes Register update side effects, memory indirection ■ e.g., delayed branches Advanced pipelines have long delay slots M 14 Chapter 4 — The Processor — 130 Concluding Remarks ISA influences design of datapath and control Datapath and control influence design of ISA Pipelining improves instruction throughput using parallelism ■ More instructions completed per second ■ Latency for each instruction not reduced Hazards: structural, data, control Multiple issue and dynamic scheduling (ILP) ■ Dependencies limit achievable parallelism ■ Complexity leads to the power wall M 14 Chapter 4 — The Processor — Exercises Answer the following exercises, and send your answers as a PDF attachment to the email address listed below xamiri@fi.muni.cz I Leave body of the email blank Deadline is April 28th Exercises 4.1.1(b), 4.2.4(a), 4.7.3(a), 4.9.2(a)~4.9.4(a), 4.10.4(a), 4.11.1(a)~4.11.4(a), 4.13.1(a), 4.13.5(a), 4.16.4(b), 4.18.1(b), 4.21.2(a), 4.21.4(a), 4.22.1(b), 4.23.2(a), 4.24.2(b) M 14 Chapter 1 — Computer Abstractions and Technology — 132