1 Pipelining and Vector Processing

Computer Organization Computer Architectures Lab

PIPELINING AND VECTOR PROCESSING
Parallel Processing

Pipelining

Arithmetic Pipeline

Instruction Pipeline

RISC Pipeline

Vector Processing

Array Processors
2 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
PARALLEL PROCESSING
Levels of Parallel Processing
- Job or Program level

- Task or Procedure level

- Inter-Instruction level

- Intra-Instruction level
Execution of Concurrent Events in the computing
process to achieve faster Computational Speed
Parallel Processing
3 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
PARALLEL COMPUTERS
Architectural Classification
Number of Data Streams
Number of
Instruction
Streams
Single
Multiple
Single Multiple
SISD SIMD
MISD MIMD
Parallel Processing
Flynn's classification
Based on the multiplicity of Instruction Streams and
Data Streams
Instruction Stream
Sequence of Instructions read from memory
Data Stream
Operations performed on the data in the processor
4 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
COMPUTER ARCHITECTURES FOR PARALLEL
PROCESSING
Von-Neuman
based











Dataflow




Reduction
SISD






MISD


SIMD






MIMD
Superscalar processors

Superpipelined processors

VLIW


Nonexistence


Array processors

Systolic arrays

Associative processors


Shared-memory multiprocessors

Bus based
Crossbar switch based
Multistage IN based

Message-passing multicomputers

Hypercube
Mesh
Reconfigurable
Parallel Processing
5 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
SISD COMPUTER SYSTEMS
Control
Unit
Processor
Unit

Memory

Instruction stream
Data stream
Characteristics

- Standard von Neumann machine
- Instructions and data are stored in memory
- One operation at a time


Limitations

Von Neumann bottleneck

Maximum speed of the system is limited by the
Memory Bandwidth (bits/sec or bytes/sec)

- Limitation on Memory Bandwidth
- Memory is shared by CPU and I/O

Parallel Processing
6 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
SISD PERFORMANCE IMPROVEMENTS

Multiprogramming
Spooling
Multifunction processor
Pipelining
Exploiting instruction-level parallelism
- Superscalar
- Superpipelining
- VLIW (Very Long Instruction Word)

Parallel Processing
7 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
MISD COMPUTER SYSTEMS
M CU P
M CU P
M CU P






Memory
Instruction stream
Data stream
Characteristics

- There is no computer at present that can be
classified as MISD
Parallel Processing
8 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
SIMD COMPUTER SYSTEMS
Control Unit
Memory
Alignment network
P P P
M M M
Data bus
Instruction stream
Data stream
Processor units
Memory modules
Characteristics

- Only one copy of the program exists
- A single controller executes one instruction at a time
Parallel Processing
9 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
TYPES OF SIMD COMPUTERS

Array Processors

- The control unit broadcasts instructions to all PEs,
and all active PEs execute the same instructions
- ILLIAC IV, GF-11, Connection Machine, DAP, MPP


Systolic Arrays

- Regular arrangement of a large number of
very simple processors constructed on
VLSI circuits
- CMU Warp, Purdue CHiP


Associative Processors

- Content addressing
- Data transformation operations over many sets
of arguments with a single instruction
- STARAN, PEPE
Parallel Processing
10 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
MIMD COMPUTER SYSTEMS
Interconnection Network
P M P M P M

Shared Memory
Characteristics

- Multiple processing units

- Execution of multiple instructions on multiple data



Types of MIMD computer systems

- Shared memory multiprocessors

- Message-passing multicomputers
Parallel Processing
11 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
SHARED MEMORY MULTIPROCESSORS
Characteristics
All processors have equally direct access to
one large memory address space

Example systems
Bus and cache-based systems
- Sequent Balance, Encore Multimax
Multistage IN-based systems
- Ultracomputer, Butterfly, RP3, HEP
Crossbar switch-based systems
- C.mmp, Alliant FX/8

Limitations
Memory access latency
Hot spot problem
Interconnection Network(IN)


P P P
M M M
Buses,
Multistage IN,
Crossbar Switch
Parallel Processing
12 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
MESSAGE-PASSING MULTICOMPUTER
Characteristics

- Interconnected computers
- Each processor has its own memory, and
communicate via message-passing

Example systems

- Tree structure: Teradata, DADO
- Mesh-connected: Rediflow, Series 2010, J-Machine
- Hypercube: Cosmic Cube, iPSC, NCUBE, FPS T Series, Mark III

Limitations

- Communication overhead
- Hard to programming
Message-Passing Network

P P
P
M
M M

Point-to-point connections
Parallel Processing
13 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
PIPELINING
R1 A
i
, R2 B
i
Load A
i
and B
i
R3 R1 * R2, R4 C
i
Multiply and load C
i
R5 R3 + R4 Add
A technique of decomposing a sequential process
into suboperations, with each subprocess being
executed in a partial dedicated segment that
operates concurrently with all other segments.
A
i
* B
i
+ C
i
for i = 1, 2, 3, ... , 7
A
i
R1 R2
Multiplier
R3 R4
Adder
R5
Memory
Pipelining
B
i
C
i
Segment 1
Segment 2
Segment 3
14 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
OPERATIONS IN EACH PIPELINE STAGE
Clock
Pulse
Segment 1 Segment 2 Segment 3
Number R1 R2 R3 R4 R5
1 A1 B1
2 A2 B2 A1 * B1 C1
3 A3 B3 A2 * B2 C2 A1 * B1 + C1
4 A4 B4 A3 * B3 C3 A2 * B2 + C2
5 A5 B5 A4 * B4 C4 A3 * B3 + C3
6 A6 B6 A5 * B5 C5 A4 * B4 + C4
7 A7 B7 A6 * B6 C6 A5 * B5 + C5
8 A7 * B7 C7 A6 * B6 + C6
9 A7 * B7 + C7

Pipelining
15 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
GENERAL PIPELINE
General Structure of a 4-Segment Pipeline
S R
1 1
S R
2 2
S R
3 3
S R
4 4
Input
Clock
Space-Time Diagram
1 2 3 4 5 6 7 8 9
T1
T1
T1
T1
T2
T2
T2
T2
T3
T3
T3
T3 T4
T4
T4
T4 T5
T5
T5
T5 T6
T6
T6
T6
Clock cycles
Segment 1
2
3
4
Pipelining
16 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
PIPELINE SPEEDUP
n: Number of tasks to be performed

Conventional Machine (Non-Pipelined)
t
n
: Clock cycle
t
1
: Time required to complete the n tasks
t
1
= n * t
n

Pipelined Machine (k stages)
t
p
: Clock cycle (time to complete each suboperation)
t
k
: Time required to complete the n tasks
t
k
= (k + n - 1) * t
p

Speedup
S
k
: Speedup

S
k
= n*t
n
/ (k + n - 1)*t
p





n
S
k
=
t
n
t
p
( = k, if t
n
= k * t
p
) lim
Pipelining
17 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
PIPELINE AND MULTIPLE FUNCTION UNITS
P
1
I
i
P
2
I
i+1
P
3
I
i+2
P
4
I
i+3
Multiple Functional Units
Example
- 4-stage pipeline
- subopertion in each stage; t
p
= 20nS
- 100 tasks to be executed
- 1 task in non-pipelined system; 20*4 = 80nS

Pipelined System
(k + n - 1)*t
p
= (4 + 99) * 20 = 2060nS

Non-Pipelined System
n*k*t
p
= 100 * 80 = 8000nS

Speedup
S
k
= 8000 / 2060 = 3.88

4-Stage Pipeline is basically identical to the system
with 4 identical function units
Pipelining
18 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
ARITHMETIC PIPELINE
Floating-point adder
[1] Compare the exponents
[2] Align the mantissa
[3] Add/sub the mantissa
[4] Normalize the result
X = A x 2
a
Y = B x 2
b

R
Compare
exponents
by subtraction
a b
R
Choose exponent
Exponents
R
A B
Align mantissa
Mantissas
Difference
R
Add or subtract
mantissas
R
Normalize
result
R
R
Adjust
exponent
R
Segment 1:
Segment 2:
Segment 3:
Segment 4:
Arithmetic Pipeline
19 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
4-STAGE FLOATING POINT ADDER
A = a x 2
p
B = b x 2
q
p a q b
Exponent

subtractor
Fraction

selector
Fraction with min(p,q)
Right shifter
Other
fraction
t = |p - q|
r = max(p,q)
Fraction

adder
Leading zero

counter
r c
Left shifter
c
Exponent

adder
r
s d
d
Stages:
S1
S2
S3
S4
C = A + B = c x 2 = d x 2
r s
(r = max (p,q), 0.5 s d < 1)
Arithmetic Pipeline
20 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
INSTRUCTION CYCLE
Six Phases* in an Instruction Cycle

[1] Fetch an instruction from memory
[2] Decode the instruction
[3] Calculate the effective address of the operand
[4] Fetch the operands from memory
[5] Execute the operation
[6] Store the result in the proper place

* Some instructions skip some phases
* Effective address calculation can be done in
the part of the decoding phase
* Storage of the operation result into a register
is done automatically in the execution phase

==> 4-Stage Pipeline

[1] FI: Fetch an instruction from memory
[2] DA: Decode the instruction and calculate
the effective address of the operand
[3] FO: Fetch the operand
[4] EX: Execute the operation
Instruction Pipeline
21 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
INSTRUCTION PIPELINE
Execution of Three Instructions in a 4-Stage Pipeline

Instruction Pipeline
FI DA FO EX
FI DA FO EX
FI DA FO EX
i
i+1
i+2
Conventional
Pipelined
FI DA FO EX
FI DA FO EX
FI DA FO EX
i
i+1
i+2
22 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
INSTRUCTION EXECUTION IN A 4-STAGE PIPELINE
1 2 3 4 5 6 7 8 9 10 12 13 11
FI DA FO EX 1
FI DA FO EX
FI DA FO EX
FI DA FO EX
FI DA FO EX
FI DA FO EX
FI DA FO EX
2
3
4
5
6
7
FI
Step:
Instruction
(Branch)
Instruction Pipeline
Fetch instruction

from memory
Decode instruction

and calculate

effective address
Branch?
Fetch operand

from memory
Execute instruction
Interrupt?
Interrupt

handling
Update PC
Empty pipe
no
yes
yes
no
Segment1:
Segment2:
Segment3:
Segment4:
23 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
MAJOR HAZARDS IN PIPELINED EXECUTION
Structural hazards(Resource Conflicts)

Hardware Resources required by the instructions in
simultaneous overlapped execution cannot be met

Data hazards (Data Dependency Conflicts)

An instruction scheduled to be executed in the pipeline requires the
result of a previous instruction, which is not yet available
JMP ID PC + PC
bubble IF ID OF OE OS
Branch address dependency
Hazards in pipelines may make it
necessary to stall the pipeline
Pipeline Interlock:
Detect Hazards Stall until it is cleared

Instruction Pipeline
ADD DA B,C +
INC DA +1 R1 bubble
Data dependency
R1 <- B + C
R1 <- R1 + 1
Control hazards

Branches and other instructions that change the PC
make the fetch of the next instruction to be delayed
24 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
STRUCTURAL HAZARDS
Structural Hazards

Occur when some resource has not been
duplicated enough to allow all combinations
of instructions in the pipeline to execute

Example: With one memory-port, a data and an instruction fetch
cannot be initiated in the same clock
The Pipeline is stalled for a structural hazard
<- Two Loads with one port memory
-> Two-port memory will serve without stall
Instruction Pipeline
FI DA FO EX
i
i+1
i+2
FI DA FO EX
FI DA FO EX stall stall
25 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
DATA HAZARDS
Data Hazards

Occurs when the execution of an instruction
depends on the results of a previous instruction
ADD R1, R2, R3
SUB R4, R1, R5
Hardware Technique

Interlock
- hardware detects the data dependencies and delays the scheduling
of the dependent instruction by stalling enough clock cycles
Forwarding (bypassing, short-circuiting)
- Accomplished by a data path that routes a value from a source
(usually an ALU) to a user, bypassing a designated register. This
allows the value to be produced to be used at an earlier stage in the
pipeline than would otherwise be possible

Software Technique
Instruction Scheduling(compiler) for delayed load
Data hazard can be dealt with either hardware
techniques or software technique

Instruction Pipeline
26 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
FORWARDING HARDWARE
Register
file
Result
write bus
Bypass
path
ALU result buffer
MUX
ALU
R4
MUX
Instruction Pipeline

Example:

ADD R1, R2, R3
SUB R4, R1, R5

3-stage Pipeline

I: Instruction Fetch
A: Decode, Read Registers,
ALU Operations
E: Write the result to the
destination register

I A E
ADD
SUB
I A E
Without Bypassing
I A E
SUB With Bypassing
27 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
INSTRUCTION SCHEDULING
a = b + c;
d = e - f;
Unscheduled code:
Delayed Load

A load requiring that the following instruction not use its result
Scheduled Code:
LW Rb, b
LW Rc, c
LW Re, e
ADD Ra, Rb, Rc
LW Rf, f
SW a, Ra
SUB Rd, Re, Rf
SW d, Rd
LW Rb, b
LW Rc, c
ADD Ra, Rb, Rc
SW a, Ra
LW Re, e
LW Rf, f
SUB Rd, Re, Rf
SW d, Rd
Instruction Pipeline
28 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
CONTROL HAZARDS
Branch Instructions

- Branch target address is not known until
the branch instruction is completed








- Stall -> waste of cycle times
FI DA FO EX
FI DA FO EX
Branch
Instruction
Next
Instruction
Target address available
Dealing with Control Hazards

* Prefetch Target Instruction
* Branch Target Buffer
* Loop Buffer
* Branch Prediction
* Delayed Branch

Instruction Pipeline
29 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
CONTROL HAZARDS
Instruction Pipeline
Prefetch Target Instruction
Fetch instructions in both streams, branch not taken and branch taken
Both are saved until branch branch is executed. Then, select the right
instruction stream and discard the wrong stream
Branch Target Buffer(BTB; Associative Memory)
Entry: Addr of previously executed branches; Target instruction
and the next few instructions
When fetching an instruction, search BTB.
If found, fetch the instruction stream in BTB;
If not, new stream is fetched and update BTB
Loop Buffer(High Speed Register file)
Storage of entire loop that allows to execute a loop without accessing
memory
Branch Prediction
Guessing the branch condition, and fetch an instruction stream based
on
the guess. Correct guess eliminates the branch penalty
Delayed Branch
Compiler detects the branch and rearranges the instruction sequence
by inserting useful instructions that keep the pipeline busy
in the presence of a branch instruction

30 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
RISC PIPELINE
Instruction Cycles of Three-Stage Instruction Pipeline
RISC Pipeline
RISC
- Machine with a very fast clock cycle that
executes at the rate of one instruction per cycle
<- Simple Instruction Set
Fixed Length Instruction Format
Register-to-Register Operations

Data Manipulation Instructions
I: Instruction Fetch
A: Decode, Read Registers, ALU Operations
E: Write a Register

Load and Store Instructions
I: Instruction Fetch
A: Decode, Evaluate Effective Address
E: Register-to-Memory or Memory-to-Register

Program Control Instructions
I: Instruction Fetch
A: Decode, Evaluate Branch Address
E: Write Register(PC)
31 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
DELAYED LOAD
Three-segment pipeline timing
Pipeline timing with data conflict

clock cycle 1 2 3 4 5 6
Load R1 I A E
Load R2 I A E
Add R1+R2 I A E
Store R3 I A E

Pipeline timing with delayed load

clock cycle 1 2 3 4 5 6 7
Load R1 I A E
Load R2 I A E
NOP I A E
Add R1+R2 I A E
Store R3 I A E

LOAD: R1 M[address 1]
LOAD: R2 M[address 2]
ADD: R3 R1 + R2
STORE: M[address 3] R3
RISC Pipeline
The data dependency is taken
care by the compiler rather
than the hardware
32 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
DELAYED BRANCH
1
I
3 4 6 5 2 Clock cycles:
1. Load A
2. Increment
4. Subtract
5. Branch to X
7
3. Add
8
6. NOP
E
I A E
I A E
I A E
I A E
I A E
9 10
7. NOP
8. Instr. in X
I A E
I A E
1
I
3 4 6 5 2 Clock cycles:
1. Load A
2. Increment
4. Add
5. Subtract
7
3. Branch to X
8
6. Instr. in X
E
I A E
I A E
I A E
I A E
I A E
Compiler analyzes the instructions before and after
the branch and rearranges the program sequence by
inserting useful instructions in the delay steps
Using no-operation instructions
Rearranging the instructions
RISC Pipeline
33 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
VECTOR PROCESSING
Vector Processing
Vector Processing Applications
Problems that can be efficiently formulated in terms of vectors
Long-range weather forecasting
Petroleum explorations
Seismic data analysis
Medical diagnosis
Aerodynamics and space flight simulations
Artificial intelligence and expert systems
Mapping the human genome
Image processing

Vector Processor (computer)
Ability to process vectors, and related data structures such as
matrices
and multi-dimensional arrays, much faster than conventional
computers

Vector Processors may also be pipelined

34 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
VECTOR PROGRAMMING
DO 20 I = 1, 100
20 C(I) = B(I) + A(I)
Conventional computer
Initialize I = 0
20 Read A(I)
Read B(I)
Store C(I) = A(I) + B(I)
Increment I = i + 1
If I s 100 goto 20

Vector computer
C(1:100) = A(1:100) + B(1:100)
Vector Processing
35 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
VECTOR INSTRUCTIONS
f1: V - V
f2: V - S
f3: V x V - V
f4: V x S - V
V: Vector operand
S: Scalar operand
Type Mnemonic Description (I = 1, ..., n)
Vector Processing
f1 VSQR Vector square root B(I) - SQR(A(I))
VSIN Vector sine B(I) - sin(A(I))
VCOM Vector complement A(I) - A(I)
f2 VSUM Vector summation S - E A(I)
VMAX Vector maximum S - max{A(I)}
f3 VADD Vector add C(I) - A(I) + B(I)
VMPY Vector multiply C(I) - A(I) * B(I)
VAND Vector AND C(I) - A(I) . B(I)
VLAR Vector larger C(I) - max(A(I),B(I))
VTGE Vector test > C(I) - 0 if A(I) < B(I)
C(I) - 1 if A(I) > B(I)
f4 SADD Vector-scalar add B(I) - S + A(I)
SDIV Vector-scalar divide B(I) - A(I) / S
36 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
VECTOR INSTRUCTION FORMAT
Operation
code
Base address
source 1
Base address
source 2
Base address
destination
Vector
length
Vector Processing
Vector Instruction Format
Source
A
Source
B
Multiplier
pipeline
Adder
pipeline
Pipeline for Inner Product
37 Pipelining and Vector Processing
Computer Organization Computer Architectures Lab
MULTIPLE MEMORY MODULE AND INTERLEAVING
Vector Processing
Multiple Module Memory
Address Interleaving

Different sets of addresses are assigned to
different memory modules
AR
Memory

array
DR
AR
Memory

array
DR
AR
Memory

array
DR
AR
Memory

array
DR
Address bus
Data bus
M0 M1 M2 M3