04 - Cache Memory PDF
04 - Cache Memory PDF
Computer Organization
and Architecture
8th Edition
Chapter
p 4
Cache Memory
Characteristics
• Location
• C
Capacity
• Unit of transfer
• Access method
• Performance
• Physical type
• Ph i l characteristics
Physical h i i
• Organisation
Location
• CPU
• Internall
• External
Capacity
p y
• Word size
—The
Th natural
t l unit
it off organisation
i ti
• Number of words
—or Bytes
Unit of Transfer
• Internal
—Usually
U ll governed
dbby data
d t bus
b width
idth
• External
—Usually a block which is much larger than a
word
• Addressable unit
—Smallest location which can be uniquely
addressed
—Word internally
—Cluster on M$ disks
Access Methods (1)
( )
• Sequential
—Start
St t att the
th bbeginning
i i and
d read
d through
th h in
i
order
—Access
Access time depends on location of data and
previous location
—e g tape
—e.g.
• Direct
—Individual
I di id l bl
blocks
k hhave unique
i address
dd
—Access is by jumping to vicinity plus
sequential search
—Access time depends on location and previous
location
—e.g. disk
Access Methods (2)
( )
• Random
—Individual
I di id l addresses
dd identify
id tif locations
l ti exactly
tl
—Access time is independent of location or
previous access
—e.g. RAM
• Associative
A i ti
—Data is located by a comparison with contents
off a portion
ti off the
th store
t
—Access time is independent of location or
previous access
—e.g. cache
Memory
y Hierarchy
y
• Registers
—In
I CPU
• Internal or Main memory
—May include one or more levels of cache
—“RAM”
• External memory
—Backing
g store
Memory
y Hierarchy
y - Diagram
g
Performance
• Access time
—Time
Ti between
b t presenting
ti the
th address
dd and
d
getting the valid data
• Memory
M C
Cycle
l time
ti
—Time may be required for the memory to
“ eco e ” before
“recover” befo e next
ne t access
—Cycle time is access + recovery
• Transfer
T f Rate
R
—Rate at which data can be moved
Physical
y Types
yp
• Semiconductor
—RAM
RAM
• Magnetic
—Disk & Tape
• Optical
p
—CD & DVD
• Others
—Bubble
—Hologram
Hologram
Physical
y Characteristics
• Decay
• Volatility
l l
• Erasable
• Power consumption
Organisation
g
• Physical arrangement of bits into words
• Not always
l obvious
b
• e.g. interleaved
The Bottom Line
• How much?
—Capacity
C it
• How fast?
—Time is money
• How expensive?
p
Hierarchy
y List
• Registers
• L1 Cache
C h
• L2 Cache
• Main memory
• Disk cache
• Disk
• O i l
Optical
• Tape
So y
you want fast?
• It is possible to build a computer which
uses only static RAM (see later)
• This would be very fast
• This would need no cache
—How can you cache cache?
• This would cost a very large amount
Locality
y of Reference
• During the course of the execution of a
program memory references tend to
program,
cluster
• e.g. loops
l
Cache
• Small amount of fast memory
• Sits
S between
b normall main memory andd
CPU
• May be located on CPU chip or module
Cache and Main Memory
y
Cache/Main Memory
y Structure
Cache operation
p – overview
• CPU requests contents of memory location
• Ch k cache
Check h for
f this
h data
d
• If present, get from cache (fast)
• If not present, read required block from
main
a memory
e o y to cache
cac e
• Then deliver from cache to CPU
• Cache includes tags to identify which
block of main memory is in each cache
slot
Cache Read Operation
p - Flowchart
Cache Design
g
• Addressing
• S
Size
• Mapping Function
• Replacement Algorithm
• Write Policy
• Block Size
• N
Number
b off Caches
C h
Cache Addressing
g
• Where does cache sit?
— Between processor and virtual memory management
unit
— Between MMU and main memory
• Logical cache (virtual cache) stores data using
virtual addresses
— Processor accesses cache directly, not thorough physical
cache
— Cache access faster, before MMU address translation
— Virtual addresses use same address space for different
applications
– Must
M flush
fl h cache
h on each
h context switch
i h
• Physical cache stores data using main memory
physical addresses
Size does matter
• Cost
—More
M cache
h is
i expensive
i
• Speed
—More cache is faster (up to a point)
—Checking cache for data takes time
Typical
yp Cache Organization
g
Comparison
p of Cache Sizes
Year of
Processor Type L1 cache L2 cache L3 cache
Introduction
IBM 360/85 Mainframe 1968 16 to 32 KB — —
PDP-11/70 Minicomputer 1975 1 KB — —
VAX 11/780 Minicomputer 1978 16 KB — —
IBM 3033 Mainframe 1978 64 KB — —
IBM 3090 Mainframe 1985 128 to 256 KB — —
Intel 80486 PC 1989 8 KB — —
Pentium PC 1993 8 KB/8 KB 256 to 512 KB —
P
PowerPC
PC 601 PC 1993 32 KB — —
PowerPC 620 PC 1996 32 KB/32 KB — —
PowerPC G4 PC/server 1999 32 KB/32 KB 256 KB to 1 MB 2 MB
IBM S/390 G4 Mainframe 1997 32 KB 256 KB 2 MB
IBM S/390 G6 Mainframe 1999 256 KB 8 MB —
Pentium 4 PC/server 2000 8 KB/8 KB 256 KB —
High-end server/
IBM SP 2000 64 KB/32 KB 8 MB —
supercomputer
CRAY MTAb Supercomputer 2000 8 KB 2 MB —
Itanium PC/server 2001 16 KB/16 KB 96 KB 4 MB
SGI Origin 2001 High-end server 2001 32 KB/32 KB 4 MB —
Itanium 2 PC/server 2002 32 KB 256 KB 6 MB
IBM POWER5 High-end server 2003 64 KB 1.9 MB 36 MB
CRAY XD-1 Supercomputer 2004 64 KB/64 KB 1MB —
Mapping
pp g Function
• Cache of 64kByte
• Cache
C h blblock
k off 4 bytes
b
—i.e. cache is 16k (214) lines of 4 bytes
• 16MBytes main memory
• 24 bit address
—(224=16M)
Direct Mapping
pp g
• Each block of main memory maps to only
one cache line
—i.e. if a block is in cache, it must be in one
specific place
• Address is in two parts
• Least Significant w bits identify unique
word
• Most Significant s bits specify one
memory y block
• The MSBs are split into a cache line field r
and a tag of ss-rr (most significant)
Direct Mapping
Address Structure
• 24 bit address
• 2 bit word identifier (4 byte block)
• 22 bit block identifier
— 8 bit tag (=22-14)
— 14 bit slot or line
• No two blocks in the same line have the same Tag field
• Check contents of cache by finding line and checking Tag
Direct Mapping from Cache to Main Memory
Direct Mapping
Cache Line Table
C h line
Cache li M i M
Main Memory bl
blocks
k hheld
ld
0 0, m, 2m, 3m…2s-m
1 1,m+1,
, , 2m+1…2s-m+1
…
m-1 m-1, 2m-1,3m-1…2s-1
Direct Mapping
pp g Cache Organization
g
Direct
Mapping
Example
Direct Mapping
pp g Summary
y
• Address length = (s + w) bits
• Number
b off addressable
dd bl units = 2
2s+w
words or bytes
• Block size = line size = 2w words or bytes
u be of
• Number o blocks
b oc s in main
a memory
e o y = 2s+
s
w/2w = 2s
• Number of lines in cache = m = 2r
• Size of tag = (s – r) bits
Direct Mapping
pp g pros
p & cons
• Simple
• Inexpensive
• Fixed location for given block
—If a program accesses 2 blocks that map to
the same line repeatedly, cache misses are
very high
Victim Cache
• Lower miss penalty
• Remember
b whath was discarded
d d d
—Already fetched
—Use again with little penalty
• Fully
y associative
• 4 to 16 cache lines
• Between direct mapped L1 cache and next
memory level
Associative Mapping
pp g
• A main memory block can load into any
line of cache
• Memory address is interpreted as tag and
wordd
• Tag uniquely identifies block of memory
• Every line’s tag is examined for a match
• Cache searching gets expensive
Associative Mapping from
Cache to Main Memory
Fully
y Associative Cache Organization
g
Associative
Mapping
pp g
Example
Associative Mapping
Address Structure
Word
Tag 22 bit 2 bit
• 22 bit tag stored with each 32 bit block of data
• Compare
p tagg field with tag
g entry
y in cache to
check for hit
• Least significant 2 bits of address identify which
16 bit word is required from 32 bit data block
• e.g.
— Address Tag Data Cache line
— FFFFFC FFFFFC24682468 3FFF
Associative Mapping
pp g Summary
y
• Address length = (s + w) bits
• Number
b off addressable
dd bl units = 2
2s+w
words or bytes
• Block size = line size = 2w words or bytes
u be of
• Number o blocks
b oc s in main
a memory
e o y = 2s+
s
w/2w = 2s
• Number of lines in cache = undetermined
• Size of tag = s bits
Set Associative Mapping
pp g
• Cache is divided into a number of sets
• Eachh set contains a number
b off lines
l
• A given block maps to any line in a given
set
—e.g. Block B can be in any line of set i
• e.g. 2 lines per set
—2
2 way associative mapping
—A given block can be in one of 2 lines in only
one set
Set Associative Mapping
Example
p
• 13 bit set number
• Block
l k numberb in main memory is modulo
d l
213
• 000000, 00A000, 00B000, 00C000 … map
to same set
Mapping From Main Memory to Cache:
v Associative
Mapping From Main Memory to Cache:
k-way
k way Associative
K-Way Set Associative Cache
Organization
g
Set Associative Mapping
Address Structure
Word
Tag 9 bit Set 13 bit 2 bit
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1k 2k 4k 8k 16k 32k 64k 128k 256k 512k 1M
Cache size (bytes)
direct
2-wayy
4-way
8-way
16-way
Replacement Algorithms (1)
Direct mapping
pp g
• No choice
• Each
h block
bl k onlyl maps to one lline
• Replace that line
Replacement Algorithms (2)
Associative & Set Associative
• Hardware implemented algorithm (speed)
• Least Recently
l usedd ((LRU))
• e.g. in 2 way set associative
—Which of the 2 block is lru?
• First in first out (FIFO)
—replace block that has been in cache longest
• Least frequently used
—replace block which has had fewest hits
• Random
R ndom
Write Policy
y
• Must not overwrite a cache block unless
main memory is up to date
• Multiple CPUs may have individual caches
• I/O may address main memory directly
Write through
g
• All writes go to main memory as well as
cache
• Multiple CPUs can monitor main memory
traffic
ffi to keep
k local
l l ((to C
CPU)) cache
h up to
date
• Lots of traffic
• S
Slows
o s down
do writes
s
Contention occurs when both the Instruction Prefetcher and Create separate data and instruction Pentium
the Execution Unit simultaneously require access to the caches.
cache. In that case, the Prefetcher is stalled while the
E
Execution
ti Unit’s
U it’ data
d t access takes
t k place.
l
Core Cache Cache Size (kB) Cache Line Size Associativity Location Write Buffer
Type (words) Size (words)
ARM1136 JF S
ARM1136-JF-S S lit
Split 4 64/4 64 D/I
4-64/4-64 8 4
4-way Ph i l
Physical 32
ARM Cache Organization
g
• Small FIFO write buffer
—Enhances
E h memory write it performance
f
—Between cache and main memory
—Small
S ll c.f.
f cache
h
—Data put in write buffer at processor clock
speed
—Processor continues execution
—External
E t l write
it iin parallel
ll l until
til empty
t
—If buffer full, processor stalls
—Data
D t in i write
it buffer
b ff nott available
il bl until
til written
itt
– So keep buffer small
ARM Cache and Write Buffer Organization
Internet Sources
• Manufacturer sites
—Intel
I t l
—ARM
• Search on cache