1. \course\cpeg323-05F\Topic6b-323
Pipeline Control
2018/12/6
\course\cpeg323-05F\Topic6b-323

2. \course\cpeg323-05F\Topic6b-323
Assumptions
PC is updated at each cycle
Pipeline registers are also updated at each cycle
Only need to set control values during each pipeline stage
2018/12/6
\course\cpeg323-05F\Topic6b-323

3. Five Groups of Control Lines
1. Instruction Fetch: The control signals to read instruction memory and to write the PC are always asserted, so there is nothing special to control in this pipeline stage.
2. Instruction Decode/Register Fetch: As in the previous stage, the same thing happens at every clock cycle, so there are no optional control lines to set.
3. Execution: The signals to be set are RegDst, ALUop, and ALUSrc. The signals select the Result register, the ALU operation, and either a register or a sign-extended immediate for the ALU.
2018/12/6
\course\cpeg323-05F\Topic6b-323

4. Five Groups of Control Lines
(Cont’d)
4. Memory Stage: The control lines set in this stage are Branch, MemRead, and MemWrite. These signals are set by the branch equal, load, and store instructions, respectively.
5. Write Back: The two control lines are MemtoReg, which decides between sending the ALU result or the memory value to the registers, and Reg-Write, which writes the chosen value.
2018/12/6
\course\cpeg323-05F\Topic6b-323

5. \course\cpeg323-05F\Topic6b-323
A copy of Figure 6.26 from page 468
2018/12/6
\course\cpeg323-05F\Topic6b-323

6. \course\cpeg323-05F\Topic6b-323
A copy of Figure 6.27 from page 468
2018/12/6
\course\cpeg323-05F\Topic6b-323

7. \course\cpeg323-05F\Topic6b-323
The values of the control lines are the same as in Figure5.20, but they have been shuffled into three group corresponding to the last three pipeline stages.
2018/12/6
\course\cpeg323-05F\Topic6b-323

8. \course\cpeg323-05F\Topic6b-323
The Control Lines for the Final Three Stages
2018/12/6
\course\cpeg323-05F\Topic6b-323

9. \course\cpeg323-05F\Topic6b-323
The pipelined datapath, with the control signals connected to the control portions of the
pipeline registers. The control values for the last three stages are created during the instruction
decode stage and then placed in the ID/EX pipeline register. The control lines for each pipe stage
are used, and remaining control lines are then passed to the next pipeline stage.
2018/12/6
\course\cpeg323-05F\Topic6b-323

10. Hazards and Hazards Resolution
Hazards detection
Hazards resolution
2018/12/6
\course\cpeg323-05F\Topic6b-323

11. \course\cpeg323-05F\Topic6b-323
(cont’d)
Hazards:
situations that prevent the next instruction in the instruction stream to be executed at the designated clock cycle.
2018/12/6
\course\cpeg323-05F\Topic6b-323

12. Hazards (structral and data)
(cont’d)
Hazards (structral and data)
Structural: Resource conflicts when hardware cannot support all possible combinations of instructions in simultaneous overlapped execution
Data: Instructions depend on the results of a previous instruction in a way that is exposed by the overlapping execution.
2018/12/6
\course\cpeg323-05F\Topic6b-323

13. \course\cpeg323-05F\Topic6b-323
(cont’d)
Hazards (control)
Control: Due to branches and other instructions that change PC
Hazard will cause “stall”, but in pipeline “stall” is serious - it will hold multiple instructions.
It stalls later instructions, not earlier instructions.
2018/12/6
\course\cpeg323-05F\Topic6b-323

14. \course\cpeg323-05F\Topic6b-323
Hazards (cont’d)
Structural Hazards
Non-pipelined FUs
One port of a R-file
One port of M.
Data hazards
for some data hazards ( e.g. ALU/ALU ops):
forwards (bypass)
2018/12/6
\course\cpeg323-05F\Topic6b-323

15. \course\cpeg323-05F\Topic6b-323
Data Hazards (cont’d)
for others:
pipeline interlock + pipeline stall
(bypass cannot do on time)
LD R1 A
+ R4 R1 R7
this may need a “stall” or bubble
2018/12/6
\course\cpeg323-05F\Topic6b-323

16. Example of Structural Hazard
Instruction Clock cycle number
Load instruction IF ID EX MEM WB
Instruction i IF ID EX MEM WB
Instruction i IF ID EX MEM WB
Instruction i stall IF ID EX MEM WB
Instruction i IF ID EX MEM
A pipeline stalled for a structural hazard-a load with one memory port. With only one memory port, the pipeline cannot initiate a data fetch and instruction fetch in the same cycle. A load instruction effectively steals an instructin-fetch cycle, causing the pipeline to stall - no instruction is initiated on clock cycle 4 (which normally would be instruction i+3). Because the instruction being fetched is stalled, all other instructions in the pipeline can proceed normally. The stall cycle will continue to pass through the pipeline.
2018/12/6
\course\cpeg323-05F\Topic6b-323

17. Why ALUOUTPUT has a “1” in IF?PC IF ID EX
MEM WB M R
a ALU op
IF ID EX MEM WB
PC Unit R
ALU
ALU OUTPUT M
Why ALUOUTPUT has a “1” in IF?
2018/12/6
\course\cpeg323-05F\Topic6b-323

18. \course\cpeg323-05F\Topic6b-323
Pipelined dependencies in a five-instruction sequence using simplified datapaths to show the dependencies.
2018/12/6
\course\cpeg323-05F\Topic6b-323

19. \course\cpeg323-05F\Topic6b-323
The dependencies between the pipeline registers move forward in time, so it is possible to supply the inputs to the ALU needed by the and instruction and or instruction by forwarding the results found in the pipeline registers.
2018/12/6
\course\cpeg323-05F\Topic6b-323

20. Data Hazard Resolution
Software method
Architecture features
2018/12/6
\course\cpeg323-05F\Topic6b-323

21. \course\cpeg323-05F\Topic6b-323
A B + C .
E A + D
Flow-dependency
(R/W conflicts)
2018/12/6
\course\cpeg323-05F\Topic6b-323

22. \course\cpeg323-05F\Topic6b-323
A B + C .
A B - C
Output dependency
(W/W conflicts)
Leave A in wrong state if order is changed
2018/12/6
\course\cpeg323-05F\Topic6b-323

23. \course\cpeg323-05F\Topic6b-323
X A + B .
A C + D
anti-dependency
(W/R conflicts)
2018/12/6
\course\cpeg323-05F\Topic6b-323

24. \course\cpeg323-05F\Topic6b-323
How about arrays? .
A [i] =
= A[i-1]+..
RAW (read after write) - j tries to read a source before i writes it, so j incorrectly gets the old value. This is the most common type of hazard. WAR (write after read) - j tries to write a destination before it is read by i, so i incorrectly gets the new value. This cannot happen in our example pipeline because all reads are early (in ID) and all writes are late (in WB). This hazard occurs when there are some instructions that write results early in the instruction pipeline, and other instructions that read a source after a write of an instruction later in the pipeline. For example, autoincrement addressing can create a WAR hazard.
WAW (write after write) - j tries to write an operand before it is written by i. The writes end up being performed in the wrong order, leaving the value written by i rather than the value written by j in the destination. This hazard is present only in pipelines that write in more than one pipe stage (or allow an instruction to proceed even when a previous instruction is stalled). The DLX pipeline writes a register only in WB and avoids this class of hazards.
2018/12/6
\course\cpeg323-05F\Topic6b-323

25. \course\cpeg323-05F\Topic6b-323
j i MIPS
Read / Read
Write / Write no
Read / Write yes
Write /Read no
“Shared Datum” conflicts
This case is different from the situation with back-to-back ALU operations. The LW instructioan does not have the data until the end of the MEM cycle., while the ADD instruction needs to have the data by the beginning of that clock cycle. Thus, the data hazard from using the result of a load instruction cannot be completely eliminated with simple hardware. We can forward the result immediately to the ALU from the MDR and for the SUB instruction - which begins two clock cycles after the load - the result arrives in time, as shown in ?. However, for the ADD instruction, the forwarded result arrives too late - at the end of a clock cycle though it is needed at the beginning.
LW R1, 32 (R6) IF ID EX MEM WB
ADD R4, R1, R IF ID EX MEM
SUB R5, R1, R IF ID EX
AND R6, R1, R IF ID
Pipeline hazard occurring when the result of a load instruction is used by the next instruction as a source operand and is forwarded. The value is avaiable when it returns from memory at the end of the load instructions’s MEM cycle. However, it is needed at the beginning of that clock cycle for the ADD (the EX stage of the add). The load value can be forwarded to the SUB instruction and will arrive in time for that instruction (EX). The AND can simply read the value during ID since it reads the registers in the second half of the cycle and the value is written in the first half.
2018/12/6
\course\cpeg323-05F\Topic6b-323

26. \course\cpeg323-05F\Topic6b-323
Data Dependence S1
S1: $R1 = $R2 + 1
S2: $R3 = $R1 + $R4
S3: $R4 = $R1 – 2
S4: $R3 = 2 S2 S3
Dependences:
S1, S2 -> flow-dependence
S1, S3 -> flow-dependence
S2, S3 -> anti-dependence
S2, S4 -> output-dependence S4
2018/12/6
\course\cpeg323-05F\Topic6b-323