lunes, 31 de diciembre de 2018

Programming PDF – Theme 3: The Memory System


Computer Architecture and Technology

Topic 3: The Memory System

4th Course in Telecommunications Engineering
Department of Computer Engineering and Telecommunications

Higher Polytechnic School

Contents

6.1 .- Memory hierarchy system.
6.2.- The cache memory.

Principles of operation. Organization.
Schemes of operation.
Substitution Algorithms.

6.3 The Virtual Memory System

Virtual Addressing: Paging and Segmentation
The Memory Management Unit (MMU): Functioning of the TLB

6.4.- Mechanisms to improve the memory system

Multilevel cache
Virtual cache vs real cache
Examples: The current microprocessor cache unit

Autonomous University of Madrid

The Memory System Topic 3 /

1

Technology of memory devices

Why should the memory system be improved?

I

O
T
N
E
M
D
N
E
R

I

103

102

10

1

0
8
9
1 [19659002] "Moore's Law"

CPU

μProc
60% / year
(2x / 1.5year)

Gap performance
Processor-Memory
(↑ 50% / year)[19659002] DRAM

DRAM
9% / year
(2x / 10 years)

2
8
9
1

4
8
9
1

6
8
9
1

8
8
9
1

0
9
9
1

2
9
9
1

4
9
9
1

6
9
9
1

8
9
9
1

0
0
0
2

Year

Autonomous University of Madrid

The System of Memory Topic 3 /

2

Technology of memory devices

How can a memory system be improved?
Organizing a memory hierarchy

Memory cache (level L1, L2, L3 )

Applying the concept of virtual memory

MMU and TLB

Designing more efficient CPU / MEM connection systems

with the aim of obtaining transfer data to
maximum speed

Autonomous University of Madrid

The Memory System Topic 3 /

3

Memory hierarchy system

Cache

KBytes
~ 2-5 ns
10-2-10-3 $ / bit

Main Memory

Capacity
Access Time
Cost

GPR

100's Bytes
~ 1 ns

MBytes
~ 10-20 ns
10-5-10-6 $ / bit

Disk6

GBytes
~ 5-10 ms
10-7- 10-8 $ / bit

Tape

TBytes
~ sec-min
10-10 $ / bit

Top level

Manager
Transfer unit.

GPR

Instr./Data

Cache

Blocks

Memory

Pages

Disk

Files

Tape

Lower Level

Prog./Comp.
] 1-8 Bytes

Control cache
8-128 Bytes

Sist. operative
512-4kBytes

User
MBytes

Fastest

Largest

Autonomous University of Madrid

The Memory System Topic 3 /

4

System memory hierarchy

What is a memory hierarchy based on?
The principle of Locality:

Programs access a relatively small portion of the

address space at any instant in time

] There are two different types of locations:

Temporary Location: Codes and data recently accessed,
have a high probability of being accessed again in the near future
.

Space Locality: If a code or data is referenced, codes or

data whose memory addresses are close to the previous one,
will tend to be referenced soon.

Universidad Autónoma de Madrid

The Memory System Topic 3 /

5

Yes Hierarchy of memory hierarchy

What should be taken into account when designing a hierarchy of

memory?
Where to place a block in the upper level?
How to find a block if it is at the top level?
Which block should be replaced in case of failure?
What strategy should be followed when modifying (writing) a block in

the upper level?

CPU

To the CPU [19659002] CPU

Memory

Top level

Blk X

Blk Z

Memory

Lower level

Blk X

Blk Y

Blk Z [19659002] Blk K

Autonomous University of Madrid

The Memory System Topic 3 /

6

The cache memory

Definitions:
Cache: Invisible memory, very fast and small in size the processor and the memory

H, hit rate:% of references to memory that are in the cache.
F, frequency of failures (Miss rate):% of references to memory that are not in the cache.

main.

F = 1-H

tc, Access time to the cache: time needed to access a cache data.
tB, penalty for failure: time needed to replace a block from the lower level ,

includes the time to take it to the processor.

The average access time to the memory system is:
A failure in the cache INCREASE the execution time
(
of the instruction involved

= NI

T

×

CPU

t

acc

1 (- + =

t

c

tH
) [19659002] B

CPI + CD M) T

×

CYCLE

CD M = A ccesses M emia

NI

×

Freq. failures

×

Penalty

Universidad Autónoma de Madrid

The Memory System Topic 3 /

7

The cache memory. Principles of operation

Address

CACHE

Directory

Blocks

CPU

U.Control

Data

MEMORIA
PRINCIPAL

Autonomous University of Madrid

] The Memory System Topic 3 /

8

The cache memory. Principles of operation

Directory cache

Carried out in associative memory or CAM memory (Content

AddressableMemory).

Indicates whether the data requested by the processor is in the

cache. If hit (match) it indicates the address of the block where
finds the data.

Cache blocks

Carried out in SRAM memory
Each block contains a certain number of bytes to which

is accessed in the case of a successful cache.

Control unit

Generates the control signals for the operation of the unit

cache

Autonomous University of Madrid

The Memory System Topic 3 /

9

The cache memory. Organization

Cache Completely Associative

LABEL BITS

BYTE EN
BLOCK

ADDRESS

DIRECTORY

BLOCKS

If Success,
data to CPU

= ?
=?
=?
=?
=?

Right? ó ¿Fallo?

Universidad Autónoma de Madrid

The Memory System Topic 3 /

10

The cache memory. Organization

Example: Cache complet. associative 512 Bytes, 8By / block

29b

0045FF3 0

64 comp.

F

A

F

F

F

DIRECTORY
54445F0 0
0045FF3 0

0045FF3 1

6632FF3 1
F045FF3 0

Right? or Fail?

3b
100

ADDRESS (32 b)

0045FF34

BLOCKS

100

0F FA 00 FF 33 AA 0F 22

0FFA0 … ..A0F22

CPU
FF00FA0F

Autonomous University of Madrid

The Memory System Topic 3 /

11

The cache memory. Organization

Cache of direct correspondence (Associative of 1 way)

BITS OF LABEL

BITS OF INDICE

BYTE IN
BLOCK

ADDRESS

DIRECTORY

BLOCKS

=?

Right? or Fail?

Si Acierto,
data to CPU

Universidad Autónoma de Madrid

The Memory System Topic 3 /

12

The cache memory. Organization

Example: Direct correspondence cache of 4kB and 16 By / Bloq

20b
FA067

8b
04

4b
C

ADDRESS (32 b) [19659002] FA06704C

DIRECTORY
00

BLOCKS

C

A

FF

FA067

2F 3D 00 A0

FF

A0

25 7F

2F3D0 … 0257F

CPU
A0003D2F

Autonomous University of Madrid

The Memory System Topic 3 /

13

The cache memory. Organization

Associative cache by way (N tracks)

LABEL BITS

INDEX BITS

BYTE IN
BLOCK

VIA-1

DIRECTORY BLOCKS

VIA-N

DIRECTORY BLOCKS

ADDRESS

If Success, extracts
data from BL-1 or BL-N
and sends it to the CPU

BL-1

BL-N [19659002] =?

Right? or Fail?

=?

Autonomous University of Madrid

The Memory System Topic 3 /

14

The cache memory. Organization

Example: Cache of 4kB with 64 By / Block and 4-way associative

22b

AFA64 00

4b
1100

6b
00 8

ADDRESS ( 32 b)

AFA64308

VIA-1

VIA-2

VIA-3

VIA-4

3FAA ….

AFA64 00

257A .. ..

46FF ….

452DD3FF … 48 Bytes … 4356FFCD3D3F4F3D00A03420

001000

CPU

CDFF5643

Autonomous University of Madrid

The Memory System Theme 3 /

15

1100

0

F

The cache memory. Organization

Cache Correspondencia Directa VS Associative by N Vías

CD needs 1 comparator AV needs N comparators
CD can obtain the data before a success or failure occurs. AV
has the data only after the comparison, you need to insert a
multiplexer to select the data of the path in which the
success occurs.

What mission do the validity bits have? (BV)?

If there is 1 BV for each block ⇒ In case of success if BV = '0', block no

contains valid data.

If there is 1BV for each word in the block ⇒ If successful, the word

with BV to '0' does not contain valid data.

What mission has the modification or dirt bit (M)?

Always associated with a block. If M = '1' ⇒ there has been at least one

writing in the cache that has not been written in memory.

Universidad Autónoma de Madrid

The Memory System Topic 3 /

16

] The cache memory. Operating schemes

Capture operations

Demand capture: A block (whole or part) is loaded only when

references a piece of data contained in it and a fault occurs.

Anticipatory capture: A block (integer) is loaded before being

referenced.

Selective capture: A data whose reference has

originated a failure is not always loaded. Example: "non-cacheable" addresses

Memory write and update operations

Direct Write (ED) (WriteThrough): Writes to the time in the cache and

in the main memory.

Two Strategies: EDAE and EDSAE.

Post-Write (PE) (WriteBack): The main memory is updated, only

when replacing a block. Two modalities depending on the bit M.

Autonomous University of Madrid

The Memory System Topic 3 /

17

The cache memory. Schemes of operation

1.- ED and PE: Operation of CONCERT in reading

Address

CACHE

Directory

Blocks

CPU

U.Control

Data [19659002] MEMORY
MAIN

Cache⇒CPU

AL = (H) (1 w) tC
t acc

Autonomous University of Madrid

The Memory System Topic 3 /

18

The cache memory. Operation diagrams

2.- ED: Operation of FAULT in reading

Address

CACHE

Directory

Blocks

CPU

U.Control

Data

MEMORY
MAIN

FL = (1- H) (1 w) t C
t acc

Cache⇒CPU

MEM⇒Cache

+ ( 1- H) (1 w) t B

Autonomous University of Madrid

The Memory System Topic 3 /

19

The cache memory. Operating diagrams

2.- PE: Operation of FAILURE in reading

Address

CACHE

Directory

Blocks

CPU

U.Control

Data

MEMORY
MAIN

FL = (1- H) (1 w) t C
t acc

Cache⇒CPU

Cache⇒MEM

+ ( 1- H) (1 w) w tM B

MEM⇒Cache

+ (1- H) (1 w) t B

Autonomous University of Mad

.



Source link



from Nettech Post http://bit.ly/2LEiCD2

No hay comentarios:

Publicar un comentario

Slutty Japanese Babe Toyed And Creamed

Japanese hot babe with big tits gets toyed and creamed. Author: sexualbabe Added: 02/11/2021