Revision Slide

Dr. Bryan Raj

Department of Computer System & Technology
Faculty of Computer Science & Information Technology
University of Malaya
[email protected]
Pipelining
Pipelining example (1)

Add the floating point

numbers 9.87×104 and
6.54×103
Pipelining example (2)

 Assume each
operation takes one
nanosecond (10-9
seconds).

 This for loop takes

about 7000
nanoseconds.
 Pipelining (3)
 Divide the floating point adder into 7 separate
pieces of hardware or functional units.
 First unit fetches two operands, second unit
compares exponents, etc.
 Output of one functional unit is input to the next.
 Pipelining (4)

Table 2.3: Pipelined Addition.

Numbers in the table are subscripts of
operands/results.
 Pipelining (5)
 One floating point addition still takes
7 nanoseconds.

 But 1000 floating point additions

now takes 1006 nanoseconds!
PERFORMANCE

8
 Speedup
 Number of cores = p
 Serial run-time = Tserial
 Parallel run-time = Tparallel

u p
ed
s pe
r
lin
ea
Tparallel = Tserial / p

9
Speedup of a parallel program

Tserial
S=
Tparallel

10
Efficiency of a parallel program

Tserial

S Tparallel Tserial

E= = =
p p
.
p Tparallel

11
 Speedups and
efficiencies of a parallel
program

12
 Example
 We can parallelize 90% of a serial
program.
 Parallelization is “perfect”
regardless of the number of cores p
we use.
 Tserial = 20 seconds
 Runtime
0.9 x T of
/ p parallelizable
serial = 18 / p part is

13
 Example (cont.)
 Runtime of “unparallelizable” part is
0.1 x Tserial = 2

 Overall parallel run-time is

Tparallel = 0.9 x Tserial / p + 0.1 x Tserial = 18 / p + 2

14
 Example (cont.)
 Speed up

Tserial 20
S= =
0.9 x Tserial / p + 0.1 x Tserial 18 / p + 2

15
PARALLEL PROGRAM
DESIGN

16
 Foster’s methodology

1. Partitioning: divide the computation to be

performed and the data operated on by the
computation into small tasks.

The focus here should be on identifying tasks

that can be executed in parallel.

17
 Foster’s methodology
2. Communication: determine what communication
needs to be carried out among the tasks
identified in the previous step.

18
 Foster’s methodology
3. Agglomeration or aggregation: combine tasks
and communications identified in the first step
into larger tasks.

For example, if task A must be executed before

task B can be executed, it may make sense to
aggregate them into a single composite task.

19
 Foster’s methodology
4. Mapping: assign the composite tasks identified in
the previous step to processes/threads.

This should be done so that communication is

minimized, and each process/thread gets roughly
the same amount of work.

20
Concurrency VS Parallelism
 Concurrency vs.
Parallelism
■ Concurrency means dealing with several things at once
□ Programming concept for the developer
□ In shared-memory systems, implemented by time sharing
■ Parallelism means doing several things at once
□ Demands parallel hardware
■ Parallel programming is a misnomer
□ Concurrent programming aiming at parallel execution
■ Any parallel software is concurrent software
□ Note: Some researchers disagree, most practitioners agree
■ Concurrent software is not always parallel software
Concurrency
□ Many server applications achieve scalability
by optimizing concurrency only (web server)
Parallelism
 Server Example:
No Concurrency, No
Parallelism

Client Core 1 Storage

Receive request A

Fetch A data

Return A data

R
e
t
Client Core 1u Storage
r
n

r
e
s
u
Server Example:
Concurrency for
Throughput
Client 1 Client 2 Core 1 Storage

Receive request A

Fetch A data

R
e
c
e
i
v
e

r
e
q
u
e
s
t
Client 1 Client 2 Core 1 Storage
B
 Server Example:
Parallelism for Throughput
Client 1 Client 2 Core 1 Core 2 Storage

Receive request A

Receive quest B
re
Fe tch A data

Fetch B data

Return B data

Return A data

Return re sult B'

Return result A'

Client 1 Client 2 Core 1 Core 2 Storage