Scribd
Upload
142 views4 pages

CS 356 Cache Exercises: Redekopp Name: - Score

The document contains solutions to cache exercises involving calculating address field sizes for direct-mapped and set-associative caches with different configurations, analyzing cache hit/miss rates for different access patterns and block sizes, and generating address sequences to compare hit rates between direct-mapped and set-associative caches.

Uploaded by

Nguyễn Trọng Nhân
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
142 views4 pages

CS 356 Cache Exercises: Redekopp Name: - Score

The document contains solutions to cache exercises involving calculating address field sizes for direct-mapped and set-associative caches with different configurations, analyzing cache hit/miss rates for different access patterns and block sizes, and generating address sequences to compare hit rates between direct-mapped and set-associative caches.

Uploaded by

Nguyễn Trọng Nhân
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

CS 356 Cache Exercises

Redekopp
Name: ____Solutions________________________________
Score: ________
Note: Attach all work to receive full credit

1.) (18 pts.) A processor has a 32-bit memory address space (i.e. 32-bit addresses). The memory
is broken into blocks of 32 bytes each. The computer also has a cache capable of storing
16K bytes.
a.) How many blocks can the cache store?
16Kbytes / 32 bytes = 214 / 25 = 29 = 512 blocks
b.) Assuming the cache uses direct-mapping, how many bits are there in each of the
TAG, BLOCK, and BYTE OFFSET fields fields of the address. Show your
calculations.

Tag Block Byte Offset


A31-A14 A13-A5 A4-A0
18-bits 9-bits 5-bits
Remaining Bits 512 blocks = 29 32 bytes per block
=> 9 address bits

c.) Assuming the cache uses a 4-way set-associative mapping, how many bits are
there in each of the TAG, SET and BYTE OFFSET fields of the address. Show
your calculations.

Tag Set Byte Offset


A31-A12 A11-A5 A4-A0
20-bits 7-bits 5-bits
Remaining Bits 512 blocks / 4-ways= 32 bytes per block
27 blocks in a set
=> 7 address bits

2.) A processor has a 36-bit memory address space (i.e. 36-bit addresses). The memory is
broken into blocks of 64 bytes. The computer also has a cache capable of storing 1
Megabyte.
a.) How many blocks can the cache store?
1M / 64 bytes = 220 / 26 = 214 = 16K (16,384) blocks

b.) Assuming the cache uses direct-mapping, how many bits are there in each of the
TAG, BLOCK, and BYTE OFFSET fields of the address. Show your
calculations.
Tag Block Byte Offset
A35-A20 A19-A6 A5-A0
16-bits 14-bits 6-bits
Remaining Bits 16K blocks = 214 64 bytes per block
=> 14 address bits
c.) Assuming the cache uses an 8-way set-associative mapping, how many bits are
there in each of the TAG, SET and BYTE OFFSET fields of the address. Show
your calculations.
Tag Set Byte Offset
A35-A17 A16-A6 A5-A0
19-bits 11-bits 6-bits
Remaining Bits 16K blocks / 8-ways= 64 bytes per block
211 blocks in a set
=> 11 address bits

3.) Exercises 5.3.3 in CO&D, 4th Ed. page 550. Exercises 5.3.3 in CO&D, 4th Ed. page
550. You are asked to optimize a cache capable of storing 8 bytes total for the given
references. There are three direct-mapped cache designs possible by varying the block
size: C1 has one-byte blocks, C2 has two-byte blocks, and C3 has four-byte blocks. In
terms of miss rate, which cache design is best? If the miss stall time is 25 cycles, and C1
has an access time of 2 cycles, C2 takes 3 cycles, and C3 takes 5 cycles, which is the best
cache design? (Every access, hit or miss, requires an access to the cache)

Do this exercise only for access pattern a below. To make this easier for you, the address
sequence has been converted to binary below. First do the address bit breakdown, then
complete the table to answer the question.

Address Sequence a LS 8-bits of address


1 0000 0001
134 1000 0110
212 1101 0100
1 0000 0001
135 1000 0111
213 1101 0101
162 1010 0010
161 1010 0001
2 0000 0010
44 0010 1100
41 0010 1001
221 1101 1101

Cache Tag Block Byte Offset


C1 An-A3 A2-A0 None (no need)
C2 An-A3 A2-A1 A0
C3 An-A3 A2 A1-A0
Address LS 8-bits C1 Block H/M C2 Block H/M C3 Block H/M
Sequence of address
1 0000 0001 1 M 0 M 0 M
134 1000 0110 6 M 3 M 1 M
212 1101 0100 4 M 2 M 1 M
1 0000 0001 1 H 0 H 0 H
135 1000 0111 7 M 3 H 1 M
213 1101 0101 5 M 2 H 1 M
162 1010 0010 2 M 1 M 0 M
161 1010 0001 1 M 0 M 0 H
2 0000 0010 2 M 1 M 0 M
44 0010 1100 4 M 2 M 1 M
41 0010 1001 1 M 0 M 0 M
221 1101 1101 5 M 2 M 1 M
Miss Rate 11/12 9/12 10/12
Time 2*12 + 11*25 299 3*12+9*25 251 5*12+10*25 310

4.) Exercises 5.3.5 in CO&D, 4th Ed. page 551. Consider a direct-mapped cache of 64 KB
and a block size of 8 bytes. Generate a series of read requests (address sequence) that
will have a lower miss rate on a 2 KB 2-way set associative cache with the same block
size than this 64 KB direct mapped cache? Then generate a series of read requests
(address sequence) that will have a lower or equal miss rate in the 64KB direct-mapped
cache. To make answers more standard, use address that are as small as possible to make
your case (i.e. make unneeded tag bits of your address 0 [you should just need to change
a few of the LS tag bits] and make the byte offset portion [i.e. LSBs] 0 wherever
possible).

Use a sequence of 4 addresses for each answer. Write out your address to 20-bits. (i.e.
0x00800)

Cache Tag Block or Set Byte Offset


64 KB, direct, An-A16 A15-A3 A2-A0
2 w/block
2 KB, 2-way SA, An-A3 A9-A3 A2-A0
2 w/block

To have the 2KB, 2-way SA cache have a lower miss rate, we want to find addresses that will
map to the same block in the larger direct mapped cache but to different sets in the 2 KB
cache.

Address Block in 64 KB H/M Set in 2 KB H/M


0x00000 Block 0 M Set 0 – Way 0 M
0x10000 Block 0 M Set 0 – Way 1 M
0x00000 Block 0 M Set 0 – Way 0 H
To have the 64KB, direct mapped cache have a lower miss rate, we want to find addresses
that will map to different blocks in the larger direct mapped cache but to the same set in the 2
KB cache.

Address Block in 64 KB H/M Set in 2 KB H/M


0x00000 Block 0 M Set 0 – Way 0 M
0x00400 Block 0x80 = 128 M Set 0 – Way 1 M
0x00800 Block 0x100 = 256 M Set 0 – Way 0 M / evict 00000
0x00000 Block 0 H Set 0 M

You might also like