CS2100 Computer Organization
Cache memory
Review: The Memory Hierarchy
Take advantage of the principle of locality to present the
user with as much memory as is available in the cheapest
technology at the speed offered by the fastest technology
Processor
4-8 bytes (word)
Increasing
distance
from the
processor in
access time
L1$
8-32 bytes (block)
L2$
1 to 4 blocks
Main Memory
Inclusive what
is in L1$ is a
subset of what
is in L2$ is a
subset of what
is in MM that is
a subset of is
in SM
1,024+ bytes (disk sector = page)
Secondary Memory
2011 Sem 1
memory at each level
(Relative) size ofCache
the memory
The Memory Hierarchy: Why Does it Work?
Temporal Locality (Locality in Time):
Keep most recently accessed data items closer to the processor
Spatial Locality (Locality in Space):
Move blocks consisting of contiguous words to the upper levels
To Processor Upper Level
Memory
Lower Level
Memory
Blk X
From Processor
2011 Sem 1
Blk Y
Cache memory
Locality in the flesh
Donald J. Hatfield, Jeanette Gerald: Program
Restructuring for Virtual Memory. IBM Systems
2011 Journal
Sem 1 10(3): 168-192 (1971)
Cache memory
The Memory Hierarchy: Terminology
Hit: data is in some block in the upper level (Blk X)
Hit Rate: the fraction of memory accesses found in the upper level
Hit Time: Time to access the upper level which consists of
RAM access time + Time to determine hit/miss
To Processor Upper Level
Memory
Lower Level
Memory
Blk X
From Processor
Blk Y
Miss: data is not in the upper level so needs to be retrieve
from a block in the lower level (Blk Y)
Miss Rate = 1 - (Hit Rate)
Miss Penalty: Time to replace a block in the upper level
+ Time to deliver the block the processor
2011
Sem 1
Hit Time << Miss PenaltyCache memory
How is the Hierarchy Managed?
registers memory
cache main memory
by compiler (programmer?)
by the cache controller hardware
main memory disks
by the operating system (virtual memory)
virtual to physical address mapping assisted by the hardware
(TLB)
by the programmer (files)
2011 Sem 1
Cache memory
Cache
Two questions to answer (in hardware):
Q1: How do we know if a data item is in the cache?
Q2: If it is, how do we find it?
2011 Sem 1
Cache memory
The two extremes
Fully associative caches
Just put a block into any empty location
To locate a block, search the entire cache
Very expensive!
Direct mapped
For each item of data at the lower level, there is exactly one
location in the cache where it might be - so lots of items at
the lower level must share locations in the upper level
Address mapping:
(block address) modulo (# of blocks in the cache)
2011 Sem 1
First consider block sizes of one word
Cache memory
The Fully Associative Cache
2011 Sem 1
Cache memory
Direct mapped caches
Cache
Index Valid Tag
Data
00
01
10
11
Q1: Is it there?
Compare the cache
tag to the high order 2
memory address bits
to tell if the memory
block is in the cache
2011 Sem 1
Main Memory
0000xx
0001xx Two low order bits
0010xx define the byte in the
word (32b words)
0011xx
0100xx
0101xx
0110xx
0111xx
1000xx Q2: How do we find it?
1001xx
1010xx Use next 2 low order
1011xx memory address bits
1100xx the index to
1101xx determine which
1110xx cache block (i.e.,
1111xx modulo the number of
blocks in the cache)
(blockCache
address)
modulo (# of blocks in the cache)
memory
11
Direct Mapped Cache
Consider the main memory word reference string
Start with an empty cache - all
blocks initially marked as not valid
0 miss
00
01
00
00
00
00
Mem(0)
4 miss
4
Mem(0)
Mem(1)
Mem(2)
Mem(3)
0 1 2 3 4 3 4 15
1 miss
00
Mem(0)
00
Mem(1)
2 miss
00
00
00
3 hit
01
00
00
00
Mem(0)
Mem(1)
Mem(2)
4
Mem(4)
Mem(1)
Mem(2)
Mem(3)
01
00
00
00
3 miss
00
00
00
00
hit
Mem(4)
Mem(1)
Mem(2)
Mem(3)
Mem(0)
Mem(1)
Mem(2)
Mem(3)
15 miss
01
00
00
11 00
Mem(4)
Mem(1)
Mem(2)
Mem(3)
15
8 requests, 6 misses
2011 Sem 1
Cache memory
13
MIPS Direct Mapped Cache Example
One word/block, cache size = 1K words
31 30
Hit
Tag
...
13 12 11
20
...
2 1 0
Byte
offset
Data
10
Index
Index Valid
Tag
Data
0
1
2
.
.
.
1021
1022
1023
20
32
What kind of localityCache
are memory
we taking advantage of?
2011 Sem 1
14
Handling Cache Hits
Read hits (I$ and D$)
this is what we want!
Write hits (D$ only)
allow cache and memory to be inconsistent
- write the data only into the cache block (write-back the cache
contents to the next level in the memory hierarchy when that cache
block is evicted)
- need a dirty bit for each data cache block to tell if it needs to be
written back to memory when it is evicted
require the cache and memory to be consistent
- always write the data into both the cache block and the next level in
the memory hierarchy (write-through) so dont need a dirty bit
- writes run at the speed of the next level in the memory hierarchy so
slow! or can use a write buffer, so only have to stall if the write
buffer is full
2011 Sem 1
Cache memory
15
Write Buffer for Write-Through Caching
Cache
Processor
DRAM
write buffer
Write buffer between the cache and main memory
Processor: writes data into the cache and the write buffer
Memory controller: writes contents of the write buffer to memory
The write buffer is just a FIFO
Typical number of entries: 4
Works fine if store frequency (w.r.t. time) << 1 / DRAM write cycle
Memory system designers nightmare
When the store frequency (w.r.t. time) 1 / DRAM write cycle leading to
write buffer saturation
- One solution is to use a write-back cache; another is to use an L2 cache
2011 Sem 1
Cache memory
16
Review: Why Pipeline? For Throughput!
To avoid a structural hazard need two caches on-chip:
one for instructions (I$) and one for data (D$)
Time (clock cycles)
Inst 4
2011 Sem 1
I$
Reg
D$
I$
Reg
D$
I$
Reg
D$
Reg
ALU
Inst 3
D$
ALU
Inst 2
Reg
ALU
Inst 1
I$
ALU
O
r
d
e
r
Inst 0
ALU
I
n
s
t
r.
To keep the
pipeline
running at its
maximum rate
both I$ and D$
need to satisfy
a request from
the datapath
every cycle.
Reg
I$
Reg
Cache memory
Reg
Reg
D$
What happens
when they
Reg
cant do that?
17
Another Reference String Mapping
Consider the main memory word reference string
Start with an empty cache - all
blocks initially marked as not valid
0 miss
00
00
01
Mem(0)
0 miss
Mem(4)
01
00
01
00
0 4 0 4 0 4 0 4
4 miss
Mem(0)
4 miss
Mem(0)
00
01
00
01
0 miss
0
01
00
0 miss
0
Mem(4)
01
00
Mem(4)
4 miss
Mem(0)4
4 miss
Mem(0)
8 requests, 8 misses
Ping pong effect due to conflict misses - two memory locations that
map into the same cache block
2011 Sem 1
Cache memory
19
The Three Cs of Caches
Compulsory (cold start or process migration, first
reference):
First access to a block, cold fact of life, not a whole lot you
can do about it
If you are going to run millions of instruction, compulsory
misses are insignificant
Conflict (collision):
Multiple memory locations mapped to the same cache location
Solution 1: increase cache size
Solution 2: increase associativity
Capacity:
Cache cannot contain all blocks accessed by the program
Solution: increase cache size
2011 Sem 1
Cache memory
20
Read misses (I$ and D$)
Handling Cache Misses
stall the entire pipeline, fetch the block from the next level in the
memory hierarchy, install it in the cache and send the requested
word to the processor, then let the pipeline resume
Write misses (D$ only)
stall the pipeline, fetch the block from next level in the memory
hierarchy, install it in the cache (which may involve having to evict
a dirty block if using a write-back cache), write the word from the
processor to the cache, then let the pipeline resume
or (normally used in write-back caches)
2. Write allocate just write the word into the cache updating both
the tag and data, no need to check for cache hit, no need to stall
or (normally used in write-through caches with a write buffer)
3. No-write allocate skip the cache write and just write the word to
the write buffer (and eventually to the next memory level), no need
to stall if the write buffer isnt full; must invalidate the cache block
since it will be inconsistent (now holding stale data)
1.
2011 Sem 1
Cache memory
21
Multiword Block Direct Mapped Cache
Four words/block, cache size = 1K words
31 30 . . .
Hit
Tag
13 12 11
20
...
4 32 10
Byte
offset
Data
Block offset
Index
Data
Index Valid Tag
0
1
2
.
.
.
253
254
255
20
32
What kind of localityCache
are memory
we taking advantage of?
2011 Sem 1
22
Taking Advantage of Spatial Locality
Let cache block hold more than one word
Start with an empty cache - all
blocks initially marked as not valid
0 1 2 3 4 3 4 15
0 miss
00
Mem(1)
1 hit
Mem(0)
00
Mem(0)
Mem(2)
01
00
00
3 hit
00
00
Mem(1)
Mem(3)
Mem(1)
Mem(0)
4 miss
5
4
Mem(1) Mem(0)
Mem(3) Mem(2)
4 hit
01
00
2011 Sem 1
Mem(5)
Mem(3)
00
00
2 miss
Mem(1) Mem(0)
Mem(3) Mem(2)
3 hit
01
00
Mem(5)
Mem(3)
Mem(4)
Mem(2)
15 miss
Mem(4)
Mem(2)
1101
00
Mem(5) Mem(4)
15
14
Mem(3) Mem(2)
8 requests, 4 misses
Cache memory
24
Miss Rate vs Block Size vs Cache Size
Miss rate goes up if the block size becomes a significant
fraction of the cache size because the number of blocks
that can be held in the same size cache is smaller
(increasing capacity misses)
2011 Sem 1
Cache memory
25
Block Size Tradeoff
Larger block sizes take advantage of spatial locality but
If the block size is too big relative to the cache size, the miss rate
will go up
Larger block size means larger miss penalty
- Latency to first word in block + transfer time for remaining words
Miss Exploits Spatial Locality
Rate
Average
Access
Time
Miss
Penalty
Increased Miss
Penalty
& Miss Rate
Fewer blocks
compromises
Temporal Locality
Block Size
Block Size
Block Size
In general, Average Memory Access Time
2011 Sem 1
= Hit Time + Miss Penalty x Miss Rate
Cache memory
26
Multiword Block Considerations
Read misses (I$ and D$)
Processed the same as for single word blocks a miss
returns the entire block from memory
Miss penalty grows as block size grows
- Early restart datapath resumes execution as soon as the
requested word of the block is returned
- Requested word first requested word is transferred from the
memory to the cache (and datapath) first
Nonblocking cache allows the datapath to continue to
access the cache while the cache is handling an earlier miss
Write misses (D$)
Cant use write allocate or will end up with a garbled block in
the cache (e.g., for 4 word blocks, a new tag, one word of data
from the new block, and three words of data from the old
block), so must fetch the block from memory first and pay the
stall time
2011 Sem 1
Cache memory
27
Measuring Cache Performance
Assuming cache hit costs are included as part of the
normal CPU execution cycle, then
CPU time = IC CPI CC
= IC (CPIideal + Memory-stall cycles) CC
CPIstall
Memory-stall cycles come from cache misses (a sum of
read-stalls and write-stalls)
Read-stall cycles = reads/program read miss rate
read miss penalty
Write-stall cycles = (writes/program write miss rate
write miss penalty)
+ write buffer stalls
For write-through caches, we can simplify this to
Memory-stall cycles = miss rate miss penalty
2011 Sem 1
Cache memory
28
Reducing Cache Miss Rates #1
1.
Allow more flexible block placement
In a direct mapped cache a memory block maps to
exactly one cache block
At the other extreme, could allow a memory block to be
mapped to any cache block fully associative cache
A compromise is to divide the cache into sets each of
which consists of n ways (n-way set associative). A
memory block maps to a unique set (specified by the
index field) and can be placed in any way of that set (so
there are n choices)
(block address) modulo (# sets in the cache)
2011 Sem 1
Cache memory
29
The limits of direct mapped caches
Conflict!
2011 Sem 1
Cache memory
30
Set Associative Cache Example
Main Memory
0000xx
0001xx Two low order bits
0010xx define the byte in the
word (32-b words)
0011xx
One word blocks
0100xx
0101xx
0110xx
0111xx
1000xx Q2: How do we find it?
1001xx
1010xx Use next 1 low order
1011xx memory address bit to
1100xx determine which
1101xx cache set (i.e., modulo
1110xx the number of sets in
1111xx the cache)
Cache
Way Set V
0
0
1
0
1
Tag
Data
Q1: Is it there?
Compare all the cache
tags in the set to the
high order 3 memory
address bits to tell if
the memory block is in
the cache
2011 Sem 1
Cache memory
31
Another Reference String Mapping
Consider the main memory word reference string
Start with an empty cache - all
blocks initially marked as not valid
0 miss
000
Mem(0)
0 4 0 4 0 4 0 4
4 miss
0 hit
4 hit
000
Mem(0)
000
Mem(0)
000
Mem(0)
010
Mem(4)
010
Mem(4)
010
Mem(4)
8 requests, 2 misses
Solves the ping pong effect in a direct mapped cache
due to conflict misses since now two memory locations
that map into the same cache set can co-exist!
2011 Sem 1
Cache memory
33
Four-Way Set Associative Cache
28 = 256 sets each with four ways (each with one block)
31 30
...
13 12 11
...
22
Tag
Index
Index V Tag
0
1
2
.
.
.
253
254
255
V Tag
Data
0
1
2
.
.
.
253
254
255
V Tag
Data
Byte offset
2 1 0
V Tag
Data
0
1
2
.
.
.
253
254
255
Data
0
1
2
.
.
.
253
254
255
32
4x1 select
2011 Sem 1
Hit
Cache memory
Data
34
Range of Set Associative Caches
For a fixed size cache, each increase by a factor of two
in associativity doubles the number of blocks per set (i.e.,
the number or ways) and halves the number of sets
decreases the size of the index by 1 bit and increases
the size of the tag by 1 bit
Used for tag compare
Tag
Decreasing associativity
Selects the set
Index
Block offset Byte offset
Increasing associativity
Fully associative
(only one set)
Tag is all the bits except
block and byte offset
Direct mapped
(only one way)
Smaller tags
2011 Sem 1
Selects the word in the block
Cache memory
36
Costs of Set Associative Caches
When a miss occurs, which ways block do we pick for
replacement?
Least Recently Used (LRU): the block replaced is the one that
has been unused for the longest time
- Must have hardware to keep track of when each ways block was
used relative to the other blocks in the set
- For 2-way set associative, takes one bit per set set the bit when a
block is referenced (and reset the other ways bit)
N-way set associative cache costs
N comparators (delay and area)
MUX delay (set selection) before data is available
Data available after set selection (and Hit/Miss decision). In a
direct mapped cache, the cache block is available before the
Hit/Miss decision
- So its not possible to just assume a hit and continue and recover later
if it was a miss
2011 Sem 1
Cache memory
37
Benefits of Set Associative Caches
The choice of direct mapped or set associative depends
on the cost of a miss versus the cost of implementation
Data from Hennessy &
Patterson, Computer
Architecture, 2003
Largest gains are in going from direct mapped to 2-way
(20%+ reduction in miss rate)
2011 Sem 1
Cache memory
38
Reducing Cache Miss Rates #2
2.
Use multiple levels of caches
With advancing technology have more than enough
room on the die for bigger L1 caches or for a second
level of caches normally a unified L2 cache (i.e., it
holds both instructions and data) and in some cases
even a unified L3 cache
For our example, CPIideal of 2, 100 cycle miss penalty (to
main memory), 36% load/stores, a 2% (4%) L1I$ (D$)
miss rate, add a UL2$ that has a 25 cycle miss penalty
and a 0.5% miss rate
CPIstalls = 2 + .0225 + .36.0425 + .005100 +
.36.005100 = 3.54
(as compared to 5.44 with no
2011 Sem 1L2$)
Cache memory
39
Multilevel Cache Design Considerations
Design considerations for L1 and L2 caches are very
different
Primary cache should focus on minimizing hit time in support of
a shorter clock cycle
- Smaller with smaller block sizes
Secondary cache(s) should focus on reducing miss rate to
reduce the penalty of long main memory access times
- Larger with larger block sizes
The miss penalty of the L1 cache is significantly reduced
by the presence of an L2 cache so it can be smaller
(i.e., faster) but have a higher miss rate
For the L2 cache, hit time is less important than miss rate
The L2$ hit time determines L1$s miss penalty
2011 SemL2$
1
local miss rate >> than
global miss rate
Cachethe
memory
40
Key Cache Design Parameters
L1 typical
2011 Sem 1
L2 typical
Total size (blocks)
250 to 2000 4000 to
250,000
Total size (KB)
16 to 64
500 to 8000
Block size (B)
32 to 64
32 to 128
Miss penalty (clocks)
10 to 25
100 to 1000
Miss rates
(global for L2)
2% to 5%
0.1% to 2%
Cache memory
41
Two Machines Cache Parameters
Intel P4
AMD Opteron
L1 organization
Split I$ and D$
Split I$ and D$
L1 cache size
8KB for D$, 96KB for
trace cache (~I$)
64KB for each of I$ and D$
L1 block size
64 bytes
64 bytes
L1 associativity
4-way set assoc.
2-way set assoc.
L1 replacement
~ LRU
LRU
L1 write policy
write-through
write-back
L2 organization
Unified
Unified
L2 cache size
512KB
1024KB (1MB)
L2 block size
128 bytes
64 bytes
L2 associativity
8-way set assoc.
16-way set assoc.
L2 replacement
~LRU
~LRU
L2 write policy
write-back
write-back
2011 Sem 1
Cache memory
42
4 Questions for the Memory Hierarchy
Q1: Where can a block be placed in the upper level?
(Block placement)
Q2: How is a block found if it is in the upper level?
(Block identification)
Q3: Which block should be replaced on a miss?
(Block replacement)
Q4: What happens on a write?
(Write strategy)
2011 Sem 1
Cache memory
43
Q1&Q2: Where can a block be placed/found?
# of sets
Direct mapped
# of blocks in cache
Set associative
(# of blocks in cache)/
associativity
Fully associative
Location method
Direct mapped
Index
Set associative
Index the set; compare
sets tags
1
Associativity (typically
2 to 16)
# of blocks in cache
# of comparisons
1
Fully associative Compare all blocks tags
2011 Sem 1
Blocks per set
Cache memory
Degree of
associativity
# of blocks
44
Q3: Which block should be replaced on a miss?
Easy for direct mapped only one choice
Set associative or fully associative
Random
LRU (Least Recently Used)
For a 2-way set associative cache, random
replacement has a miss rate about 1.1 times higher
than LRU.
LRU is too costly to implement for high levels of
associativity (> 4-way) since tracking the usage
information is costly
2011 Sem 1
Cache memory
45
Q4: What happens on a write?
Write-through The information is written to both the
block in the cache and to the block in the next lower level
of the memory hierarchy
Write-back The information is written only to the block in
the cache. The modified cache block is written to main
memory only when it is replaced.
Write-through is always combined with a write buffer so write
waits to lower level memory can be eliminated (as long as the
write buffer doesnt fill)
Need a dirty bit to keep track of whether the block is clean or dirty
Pros and cons of each?
Write-through: read misses dont result in writes (so are simpler
and cheaper)
Write-back: repeated writes require only one write to lower level
2011 Sem 1
Cache memory
46
Improving Cache Performance
0. Reduce the time to hit in the cache
smaller cache
direct mapped cache
smaller blocks
for writes
- no write allocate no hit on cache, just write to write buffer
- write allocate to avoid two cycles (first check for hit, then write)
pipeline writes via a delayed write buffer to cache
1. Reduce the miss rate
bigger cache
more flexible placement (increase associativity)
larger blocks (16 to 64 bytes typical)
victim cache small buffer holding most recently discarded blocks
2011 Sem 1
Cache memory
47
Improving Cache Performance
2. Reduce the miss penalty
smaller blocks
use a write buffer to hold dirty blocks being replaced so dont
have to wait for the write to complete before reading
check write buffer (and/or victim cache) on read miss may get
lucky
for large blocks fetch critical word first
use multiple cache levels L2 cache not tied to CPU clock rate
faster backing store/improved memory bandwidth
- wider buses
- memory interleaving, page mode DRAMs
2011 Sem 1
Cache memory
48
Caches in the real world
Intel Core 2 Extreme QX9650 Yorkfield 45nm (Nov 12, 2007)
CPU1
L1d: 32KB, 8 way
assoc, 3 cycles
access time
L2: 6MB, 24 way
assoc, 15 cycles
access time
6MB L2 cache
L1i: 32KB, 8 way
assoc
CPU2
2011 Sem 1
Cache memory
49
Cache Summary
The Principle of Locality:
Program likely to access a relatively small portion of the address
space at any instant of time
- Temporal Locality: Locality in Time
- Spatial Locality: Locality in Space
Three major categories of cache misses:
Compulsory misses: sad facts of life. Example: cold start misses
Conflict misses: increase cache size and/or associativity
Nightmare Scenario: ping pong effect!
Capacity misses: increase cache size
Cache design space
total size, block size, associativity (replacement policy)
write-hit policy (write-through, write-back)
write-miss policy (write allocate, write buffers)
2011 Sem 1
Cache memory
50
Summary: The Cache Design Space
Several interacting dimensions
cache size
block size
associativity
replacement policy
write-through vs write-back
write allocation
Associativity
Block Size
The optimal choice is a compromise
depends on access characteristics
- workload
- use (I-cache, D-cache, TLB)
Cache Size
depends on technology / cost
Good Factor A
Less
Simplicity often wins
2011 Sem 1
Bad
Cache memory
Factor B
More
51